ArticlePDF Available

A new mesenosaurine from the lower Permian of Germany and the postcrania of Mesenosaurus: implications for early amniote comparative osteology



Based on an exceptionally well-preserved articulated postcranial skeleton, the early amniote Cabarzia trostheidei gen. et sp. nov. is described. As the lack of the skull hampers its taxonomic assignment, a large sample of basal amniotes is included as part of an exhaustive comparison. Considering the slender, long-limbed proportions of the skeleton, several potential determinations are suggested in order to test for bolosaur, millerettid, araeoscelid, basal neodiapsid, or synapsid affinities. Numerous character conditions are re-evaluated regarding their distributions among early amniote subclades. The closest match to Cabarzia is found to be the middle Permian Mesenosaurus from Russia. The documentation for both genera provides the most complete postcranial descriptions of non-varanodontine varanopids. One of the main differences from Mesenosaurus is the curved ungual phalanx, indicating a use related to predatory behavior. An investigation of limb proportions, such as slender trunks, elongated hindlimbs, and relatively short forelimbs, may suggest the occurrence of facultative bipedalism in Mesenosaurus and Cabarzia. The oldest known mesenosaurine, C. trostheidei from the Asselian/Sakmarian, also pushes back the oldest evidence of bipedal locomotion by more than 15 Ma.
1 3
A new mesenosaurine fromthelower Permian ofGermany
andthepostcrania ofMesenosaurus: implications forearly amniote
comparative osteology
FrederikSpindler1 · RalfWerneburg2· JörgW.Schneider3,4
Received: 26 April 2018 / Accepted: 25 November 2018
© Paläontologische Gesellschaft 2019
Based on an exceptionally well-preserved articulated postcranial skeleton, the early amniote Cabarzia trostheidei gen. et sp.
nov. is described. As the lack of the skull hampers its taxonomic assignment, a large sample of basal amniotes is included
as part of an exhaustive comparison. Considering the slender, long-limbed proportions of the skeleton, several potential
determinations are suggested in order to test for bolosaur, millerettid, araeoscelid, basal neodiapsid, or synapsid affinities.
Numerous character conditions are re-evaluated regarding their distributions among early amniote subclades. The closest
match to Cabarzia is found to be the middle Permian Mesenosaurus from Russia. The documentation for both genera provides
the most complete postcranial descriptions of non-varanodontine varanopids. One of the main differences from Mesenosaurus
is the curved ungual phalanx, indicating a use related to predatory behavior. An investigation of limb proportions, such as
slender trunks, elongated hindlimbs, and relatively short forelimbs, may suggest the occurrence of facultative bipedalism
in Mesenosaurus and Cabarzia. The oldest known mesenosaurine, C. trostheidei from the Asselian/Sakmarian, also pushes
back the oldest evidence of bipedal locomotion by more than 15Ma.
Keywords Synapsida· Reptilia· Paleozoic· Postcrania· Bipedalism· Character optimization
The Thuringian Forest Basin (Fig.1) is one of the largest
and best-known late Carboniferous to late Permian continen-
tal basins in Europe (Schneider and Romer 2010; Lützner
etal. 2012). The first Permian plant fossils from this area
were described in the middle of the seventeenth century,
and von Schlotheim’s (1804) report on floras of the “Stein-
kohlen-Formation” and the “Rothliegend” of the Thuring-
ian Forest marked the beginning of scientific paleobotany
(e.g., Barthel and Rößler 1995; Barthel 2003). Because
of historical mining for coal and for sulfidic ores in black
shales, fossil sampling was traditionally focused on palus-
trine and lacustrine gray facies. Outcrops of red bed facies
are rare in this densely forested mountainous area and are
restricted to historical quarries for dimension stone. Con-
sequently, for a long time, reports of terrestrial tetrapods
were restricted to their tracks (for history and details see
Voigt 2005, 2012), until the discovery of the famous but thus
far unique Tambach locality, which produced skeletons of
twelve tetrapod taxa (e.g., Martens etal. 1981; Eberth etal.
2000; Voigt etal. 2007; Berman etal. 2014). At present,
Handling editor: Jörg Fröbisch.
Electronic supplementary material The online version of this
article (https :// 2-018-0439-z) contains
supplementary material, which is available to authorized users.
* Frederik Spindler
Ralf Werneburg
Jörg W. Schneider
1 Dinosaurier-Park Altmühltal, Dinopark 1,
85095Denkendorf, Germany
2 Naturhistorisches Museum Schloss Bertholdsburg,
Burgstraße 6, 98553Schleusingen, Germany
3 TU Bergakademie Freiberg, Geological Institute, Bernhard
von Cotta-Straße 2, 09599Freiberg, Germany
4 Kazan Federal University, 420008Kazan, Russia
F. Spindler etal.
1 3
only one active quarry for basaltic volcanites, the Cabarz
quarry near Tabarz, exposes fossiliferous sediments of flu-
vial and lacustrine gray and red facies at a uniquely large
scale in Central Europe.
After a blast in the Cabarz quarry, an incomplete but
articulated tetrapod skeleton was discovered by Frank Tros-
theide at the level of the 4th floor in 1989. The articulation,
presenting exceptionally intact mesopodia, implies that the
fossil was complete, but the skull and neck are missing in
the recovered block, as well as the distal portions of the
left extremities and the posterior portion of the tail. Prior
to preparation, only parts of the dorsal sequence were vis-
ible. A single piece containing the right forelimb was glued
onto the slab. After the preparation of the specimen by Tor-
sten Krause (Erfurt), it was obvious that the skeleton was
of an amniote, making it of outstanding significance given
the rarity of amniotes. Subsequently, this specimen gained
considerable publicity. Its informal assignment to the early
diapsid Araeoscelidia (Martens 1992) could not be tested
before performing broad postcranial comparisons.
Institutional abbreviations used in this article: FMNH
Field Museum of Natural History, Chicago, Illinois, USA;
MB—Museum für Naturkunde, Berlin, Germany; MCZ
Museum of Comparative Zoology, Harvard University,
Cambridge, Massachusetts, USA; MNG—Museum der
Natur, Gotha, Germany; NHMS—Naturhistorisches Museum
Schloss Bertholdsburg Schleusingen, Germany; NML
Naturkundemuseum Leipzig, Germany; PIN—Paleonto-
logical Institute, Russian Academy of Sciences, Moscow,
Russia; TMM—Texas Memorial Museum, University of
Texas, Austin, USA.
Anatomical abbreviations used in figures of this article: as
astragalus, ax axis spine, cal calcaneus, cau caudal elements,
ce centrale, cer cervical vertebrae, cl clavicle, dc distal car-
pals, dr dorsal ribs, dt distal tarsals, f femur, fibula, g gas-
tralia, h humerus, il ilium, im intermedium, mc metacarpals,
mt metatarsals, pi pisiform, pu pubis, ra radius, rad radiale,
sc scapula, sr sacral rib, ti tibia, ul ulna, ule ulnare.
Methodological approach
The informal determination of the newly described specimen
as an araeoscelid early diapsid (Martens 1992, 2010) was
based on its rough proportions and co-occurrence with the
lacertoid tracks of Dromopus. Because the cranium is lost,
the most significant characters for a phylogenetic assessment
cannot be scored. Based solely on the proportions of the
postcranium, a number of possible taxonomic assignments
may be considered: (1) an early diapsid eureptile, araeos-
celid, or basal neodiapsid; (2) an early parareptile, bolosau-
rid or millerettid; or (3) a varanopid synapsid.
Fig. 1 Geographical position
and geologic setting of the
Thuringian Forest Basin with
Paleozoic formations. The
asterisk marks the studied site at
Cabarz (city district of Tabarz)
T h u r i n g i a n B a s i n
200 km
G e r many
S o u t h G e r m a n S c a r p l a n d s
Oberhof Syncline
Crystalline Compl.
Oberhof Fm.
Goldlauter Fm.
Manebach Fm.
Ilmenau Fm./
Möhrenbach Fm./
Georgenthal Fm.
Rotterode Fm.
Tambach Fm.
Eisenach Fm./
Neuenhof Fm.
Höhenberg dolerite
10 km
Cabarz quarry
A new mesenosaurine from the lower Permian of Germany
1 3
There have been cases of certain genera that required
revision with respect to their taxonomic affiliation—sev-
eral varanopids that were formerly assigned to Archosau-
ria or ‘Eosuchia,’ meaning early neo-diapsid reptiles. This
was partially due to the revision of Mycterosaurus (Ber-
man and Reisz 1982), in which it was pointed out that this
pelycosaur-grade synapsid is a slender varanopid instead of
a ‘nitosaurid’ ancestor of the Caseidae (Romer and Price
1940). But even before that, Romer and Price (1940) had
noticed the varanopid affinities of Elliotsmithia and Mese-
nosaurus, which were in conflict with a later grouping of
these genera with Millerettidae (Watson 1957; see Romer
1976 (1st ed. 1956), p. 521: “?Familiy Mesenosauridae”).
Both genera were then confirmed to be relatives of the vara-
nopid Mycterosaurus (Reisz etal. 1998; Reisz and Berman
2001; left out by Reisz 1986). Heleosaurus was similarly
revised and considered to be a “mycterosaurine” by Reisz
and Modesto (2007; mesenosaurine according to Spindler
etal. 2018; Mycterosaurus and other non-varanodontine
genera are summarized in the following as “slender varano-
pids”). Most recently, Apsisaurus has been re-examined and
switched from a diapsid classification (Laurin 1991) to a var-
anopid one (Reisz etal. 2010), mainly triggered by the cor-
rect recognition of Archaeovenator as the first known basal
varanopid (Reisz and Dilkes 2003; preliminarily identified
as a reptile, Reisz and Modesto 2007). Reisz etal. (2010)
demonstrated this with a phylogenetic matrix comprising
two sets of characters that were previously designed to test
diapsids (Laurin 1991) and varanopids (Reisz and Dilkes
2003), respectively.
This study examines a wide range of early amniote taxa
to identify the branch to which the new specimen can be
assigned. Given the previous cases of varanopid–diapsid
confusion, the best template for such an analysis is the com-
bined phylogenetic matrix of Reisz etal. (2010). However,
since there are numerous errors in their data set [e.g., incor-
rect codings, coding of unknown elements, undefined state
(2) in character 89], the evaluation of the few postcranial
characters applicable to the new specimen is incorporated
into the descriptive part. Furthermore, the list of taxa given
previously does not fully cover the possible determinations
expected for the new specimen.
Data on proportional comparisons of limb elements have
been added to the character list of Reisz etal. (2010). To
facilitate this, the limb elements of several terrestrial amni-
otes that could reveal a closer relationship to the newly
described form were graphically reconstructed. Certain
ratios have been collected to examine their variation among
basal amniotes and possibly score their phylogenetic signifi-
cance. In Online Resource 1 of the Electronic supplementary
material (ESM), ratios are offered instead of basic measure-
ments in order to avoid suggesting citable precise values.
Because most of the published reports provide drawings but
only some primary measurements, such values would lead
to a bias towards various sources (projection angles, scaling
errors, degree of reconstruction, taphonomy). Given that the
global rarity of many small Paleozoic amniotes does not
reflect any representative distribution of potential disparity,
no boxplots were produced. Due to limited comparability,
the highly autapomorphic arboreal varanopid Ascendonanus
(Spindler etal. 2018) has been left out of the present dis-
cussion of proportions that aims to identify similarly built
ground-dwelling amniotes. As this represents a new level in
the comprehensive anatomical comparison of small early
amniotes, new insights and determinations were possible.
The present study also provides an example of how to
deal with incomplete specimens. Previous determinations
have largely focused on skull material, meaning that the
amount of cranial phylogenetic information is disproportion-
ately large. The need for comparisons at a high taxonomic
level cannot be satisfied by published cladistic analyses.
Using this integrated comparison of osteological details
and extremity proportions, the new form was confidently
determined as a varanopid basal synapsid. The anatomical
comparison also addresses Mesenosaurus from the mid-
dle Permian of Russia, the cranium of which is well known
(Reisz and Berman 2001), whereas a detailed postcranial
description is still missing. A subadult and an adult skeleton
were used to examine the postcranial morphology. The new
taxon and its closest known relative, Mesenosaurus, were
included in a detailed phylogenetic analysis of varanopid
synapsids (Spindler etal. 2018).
Geological andstratigraphic setting
The Thuringian Forest Basin, an approximately 80km
long and 40–60km wide depression oriented NE–SW, is
exposed in the late Mesozoic/early Cenozoic horst struc-
ture of the Thuringian Forest (Fig.1). It originated in the
late Pennsylvanian (Gzhelian) during the post-Variscian
extensional phase as a rift-graben structure (Andreas 1988,
2014; Lützner etal. 2012). Sedimentation started on deeply
weathered and eroded metamorphites and magmatites of the
former Variscian orogen (e.g., Schneider and Romer 2010).
The up to 4,500m thick (stacked maximal thickness) pure
continental basin fill comprises more than 2,000m of vol-
canites in the late Pennsylvanian and early Permian part. In
the middle Permian, sedimentation was interrupted by long
gaps and expired in the early late Permian with only deci-
meter-thick alluvial fan deposits of the Neuenhof Forma-
tion, immediately before the marine Zechstein transgression
started in the Wuchiapingian (Fig.2; Lützner etal. 2012;
Legler and Schneider 2008). The Gehren and Rennsteig
subgroups are biostratigraphically well subdivided and cor-
related to other European basins by insect and amphibian
F. Spindler etal.
1 3
Fig. 2 Stratigraphy of the Thur-
ingian Forest Basin. The Cabarz
quarry is part of the Gold-
lauter Formation, one of three
formations that have produced
amniote skeletal remains
bituminous black shale
insect remains
fish remains
amphibian remains
amniote remains
tetrapod footprints
arthropod ichnia
folded basement
Medusina limnica
Medusina atava
Lower Rotliegend
Steinkohlen roupG
b Möhrenbach
a Georgenthal
max. thickness [m]
Upper Rotliegend I
n. n.
++ + +
+ +
++ ++
+ +
Gzhelian Asselian Sakmarian Artinskian
Kungurian -
A new mesenosaurine from the lower Permian of Germany
1 3
biozones (Werneburg and Schneider 2006; Schneider and
Werneburg 2012) and conchostracan biostratigraphy (Sch-
neider and Scholze 2016), and are partially connected to the
marine Standard Global Chronostratigraphic Scale (SGCS)
by co-occurrences of insects with conodonts and fusulinids
in mixed marine-continental sections in North America and
the East European Donets Basin (e.g., Lucas etal. 2013;
Schneider etal. 2013, 2017). This correlation has improved
our understanding of the architecture of the facies, especially
the distributions of gray and red facies and the respective
fossil contents in the section of the Thuringian Forest Basin,
in the context of Euramerican climatic development dur-
ing the late Carboniferous and the Permian. The increas-
ing aridization that occurred in particular during the Per-
mian ran cyclically with interchanging wet and dry phases,
wherein each successive wet phase was somewhat drier than
the one before (Roscher and Schneider 2006; Montañez etal.
2007; DiMichele etal. 2010). In this regard, the Ilmenau
and Manebach formations with basin-wide lacustrine and
palustrine gray facies belong to the Gzhelian–Asselian wet
phase C of Roscher and Schneider (2006). The Goldlauter
Formation, from which the tetrapod remains discussed here
originate, belong to the transition to a relatively drier phase
of the Sakmarian with spreading wet red beds in the sense
of Schneider etal. (2010). The subsequent Oberhof Forma-
tion and its correlatives in the European basins (Roscher and
Schneider 2006; for updated chronostratigraphy see Schnei-
der and Lucas 2015: fig.2) belong to the final Sakmarian/
early Artinskian wet phase, D, in which the last perennial
lakes occur in all European basins. After wet phase E at the
Artinskian/Kungurian transition, to which the famous Tam-
bach tetrapod locality belongs, red beds change to dry playa
red beds up to the marine Zechstein transgression (Schneider
etal. 2016).
The Goldlauter Formation has a mean thickness of roughly
500m and a maximum thickness of 800m. It was deposited
after a strong relief rejuvenation accompanied by redeposition
of previously accumulated fluvial coarse clastics and newly
eroded material, which form a significant basal conglomerate
of this unit. Generally, the formation consists of interfingering
conglomeratic to fanglomeratic alluvial-fan wet red beds origi-
nating from the basin border with basin-central alluvial-plain
wet red bed fine clastics and lacustrine gray facies. Intercalated
pyroclastic beds are used for the lithostratigraphic subdivision
of the formation and the identification and basin-wide correla-
tion of lake horizons (Andreas and Haubold 1975; Andreas
2014). The deeper lakes at the center of the basin were fish-
dominated (Acanthodes–actinopterygian–xenacanthid lakes)
and alternated with branchiosaurid amphibian-dominated shal-
low phases. Littoral areas of the lakes as well as shallow lakes
and ponds close to the basin borders were mostly lacking in
fishes but rich in branchiosaurs. Situated close to a NW–SE
stretching paleo-swell (Lützner etal. 2012: fig.9), the Cabarz
quarry exposes the interfingering of wet red bed alluvial-plain
clastics with littoral greenish to black lake sediments com-
pletely dominated by branchiosaurid amphibians at the level
of the 2nd to 3rd quarry floors (Figs.2, 3). Higher up in the
quarry, at the level of the 4th and 5th floors (Fig.3a, b), only
alluvial wet red beds of the Scoyenia facies with common
tetrapod tracks are exposed (Fig.3c). The tetrapod remains
described herein originate from these deposits at the 4th floor.
The near-complete skeleton is preserved in a reddish brown,
unbedded, massive clayish siltstone containing randomly dis-
tributed clay chips 5mm in size. A low pedogenic overprint
is indicated by greenish-pink leached spots 2cm in diameter
as well as by single calcium carbonate glaebules 1–4mm in
size. The sediment was most likely deposited as a mudflow,
which may have been the reason that the complete carcass was
suddenly embedded. Some other isolated bones originate from
a pedogenic, homogenized, immature soil horizon intercalated
in fluvial silty fine sandstones and overbank siltstones.
In the last 30years, diverse early Permian fauna and flora
have been collected from the Goldlauter Formation. The flora
is dominated by meso- to xerophilous elements belonging to
about 50 species (Barthel and Brauner 2015). Invertebrates
are represented by freshwater hydromedusae, bivalves, con-
chostracans, triopsids, diplopods, arachnids, hundreds of
different insects, and some invertebrate ichnia. Fishes are
represented by actinopterygians and xenacanthids. Tetrapod
ichnogenera comprise at least three cotylosaurs (Ichniotherium
cottae, Dromopus lacertoides, Dimetropus leisnerianus; prob-
ably Tambachichnium), a seymouriamorph (Amphisauropus
latus), and a temnospondyl (Batrachichnus salamandroides),
which were fully terrestrially adapted (Voigt 2005). Tetra-
pod skeletal remains comprise (1) the aquatic branchiosaurid
Apateon dracyi (see Werneburg and Brauner 2014 for nomen-
clatural correction), with about 5,000 specimens (which make
up 99% of the entire tetrapod fauna from the Cabarz quarry),
including some mass deaths (Werneburg 1988, 2001, 2002);
(2) the aquatic branchiosaurid Schoenfelderpeton prescheri,
from only two skeletal remains (Werneburg 1988); (3) the
aquatic micromelerpetontid Branchierpeton reinholdi, with
the holotype skeleton (Werneburg 1988) and three additional
remains; (4) four undescribed specimens of the aquatic stem-
stereospondyl ?Glanochthon (compare to Schoch and Witz-
mann 2009); and (5) amniotes, which are represented by the
articulated specimen of interest here as well as by fragmentary
elements described in the following section.
Permian amniotes fromtheThuringian
Forest Basin
Amniote skeletons are rare in Paleozoic assemblages outside
of North America, Russia, and South Africa. There are only
a few specimens from the early Permian of Western and
F. Spindler etal.
1 3
Central Europe, including three pelycosaur-grade synapsids
from Kenilworth, England (Paton 1974; Spindler 2015), a
few reptiles and synapsids from Autun, France (Heyler 1969;
Falconnet 2013; Spindler etal. 2016), as well as those from
localities with a rich assemblage in the Döhlen Basin, Ger-
many (e.g., Werneburg 1991; Schneider 1994; Werneburg
and Schneider 2001; Spindler 2013, 2016), an arboreal
varanopid from the Petrified Forest of the Chemnitz Basin,
Germany (Spindler etal. 2018), and two synapsids from the
Palatinate Saar–Nahe Basin, Germany (Fröbisch etal. 2011;
Voigt etal. 2014; Spindler etal., in prep.). The early Permian
amniote diversity of the Thuringian Forest Basin, Germany,
however, far surpasses those of the other assemblages.
Throughout the Thuringian Forest, numerous localities
have produced an enormous number of tetrapod tracks,
providing the richest and chronologically most complete
record of those ichnia from the late Carboniferous up to
the middle Permian (Haubold 1970, 1998, 2000; Voigt and
Haubold 2000; Voigt 2005). Within these tetrapod track
assemblages, most ichnotaxa can be assigned to potential
trackmakers known from the osteological record. For exam-
ple, the Bromacker locality of the Tambach Formation has
yielded a rich, diadectid-dominated assemblage based on
tracks (Haubold 1998; Voigt 2005) and skeletal remains
(Berman etal. 1998, 2000a, b, 2004). This locality also
produced the specimens on which a highly remarkable case
study on track–trackmaker correlation is based, namely the
species-level identification of distinct diadectids and their
corresponding trace fossils (Voigt etal. 2007).
Apart from the Tambach Formation, amniotes are mainly
represented by ichnia from the less intensively sampled
terrestrial facies, including Varanopus, Dimetropus, and
Fig. 3 Exposure of sediments of the Goldlauter Formation with
andesite intrusions in the Cabarz quarry at the mountain named
“Großer Inselsberg” in the Thuringian Forest Mountains, Cabarz/
Tabarz, district of Gotha, Germany. After Hübner (2014). a Panorama
of the southern part of the quarry in 2012 with the levels of floors 2A
to 5; frames indicate the positions of the close-ups in b and c. b Out-
crop of the lower Goldlauter Formation with lacustrine gray facies
and the transition into fluvial wet reds. (i) Fine-bedded gray profun-
dal siltstones with 10cm thick finely laminated silty claystone con-
taining the biostratigraphically indicative branchiosaurs and insects.
(ii) Interbeddings of grayish siltstone and fine sandstone of the lit-
toral at cm to dm scale, deposited during the silting-up phase of the
lake. (i) and (ii) combined are 15m thick. (iii) Fluvial main-channel
sandstones roughly 8m thick (a subsequent unit of about 15m thick
grayish-green alluvial plain sediments with cm- to dm-thick fossilif-
erous lacustrine black shales is tectonically sheared in this area). (iv)
Sandstone–siltstone interbeddings of alluvial-plain to floodplain wet
red beds at the 4th and 5th floors. Humans are shown for scale (the
exposed sediments are about 30 m in height). c Typical fine sand-
stone–siltstone interbeddings of floodplain wet red beds in Scoyenia
facies with common tetrapod tracks at the 5th floor. A 2-m folding
ruler is shown for scale
A new mesenosaurine from the lower Permian of Germany
1 3
Tambachichnium (Haubold 1998; Voigt 2005). The lacertoid
Dromopus is the most abundant amniote trace fossil; it has
been found in densities ranging from isolated tracks to mass
occurrences. It occurs in all sedimentary facies that have
preserved ichnia, including lacustrine deposits (Voigt 2005).
More specifically, Dromopus is associated with lake envi-
ronments that contained fish (S. Voigt, pers. comm. 2017).
Interestingly, whereas other amniote ichnogenera can be at
least roughly correlated to a taxon that is represented by
skeletal remains from the Bromacker locality, there is no
evidence of the trackmaker of the common Dromopus so far.
Dromopus is missing from the Tambach Formation (Voigt
2005). Usually, a basal diapsid reptile is assumed to be the
best match with Dromopus. In contrast to the Euramerican-
wide abundance of Dromopus, basal diapsid skeletons are
rare in early Permian assemblages, which include the well-
known Araeoscelis from Texas (Williston 1914; Vaughn
1955; Reisz etal. 1984), the fragmentary Zarcasaurus from
New Mexico (Brinkman etal. 1984), as well as two par-
tial skeletons each of Aphelosaurus from France (Falconnet
2007) and Kadaliosaurus from Saxony (Credner 1889).
Prior to our study, amniote skeletons from the Thuring-
ian Forest Basin included only the Artinskian/Kungurian
Bromacker assemblage (Berman etal. 2014) and a single
synapsid humerus from the Asselian Manebach locality
(Werneburg 1999; Spindler 2015). The latest amniote to be
discovered in the Bromacker collection (Tambach Forma-
tion) is MNG-12516, which we have identified as a varano-
dontine (Fig.4). Its interclavicle bears a trapezoidal plate,
strongly resembling the condition in Varanops MCZ 1926.
In accordance with Berman etal. (2014), MNG-12516 is
tentatively assigned to Tambacarnifex.
Additionally, some isolated bones from the fluvial wet
red beds of the Goldlauter Formation at the 4th floor of the
Cabarz quarry (Fig.3) indicate the presence of the largest
vertebrates found so far in this locality (Fig.5). NHMS-
WP 10577 is apparently a dermal skull element—probably
a frontal or postfrontal. On the concave ventral side, there
is a straight edge. The curvature of the plane implies that
this element contributed to the orbit. NHMS-WP 10578
represents a humerus with orthogonally twisted terminal
regions. Its degree of ossification is remarkably low. The
assumed body size and slender shape match with those of
early amniotes (synapsids or large captorhinids). NHMS-
WP 10579 contains a thinner long bone, most likely a tibia
with a cnemial crest stepped by a vertical trough. Although
heavily damaged, it appears to be an amniote as well. A
flat and twisted element is preserved in NHMS-WP 10580,
possibly representing an angular lamina, the dorsal tip of a
cleithrum, or most likely a clavicle ventral plate. Finally,
NHMS-WP 10581 preserves three slender ribs. Smaller frag-
ments labeled NHMS-WP 10582 could not be determined.
In total, the remains are tentatively interpreted as being at
least one type of pelycosaur-grade synapsid more derived
than Varanopidae.
Systematic paleontology
(unranked) Synapsida Osborn, 1903
Family Varanopidae Romer and Price, 1940
Subfamily Mesenosaurinae Spindler, Werneburg, Schnei-
der, Luthardt, Annacker and Rößler, 2018
Cabarzia trostheidei gen. et sp. nov.
Figures6, 7, 8, 9, 12, and 18
Etymology. Genus designation reflecting the locality of the
holotype; epithet honoring the collector Frank Trostheide
(Wolmirstedt), who generously provided the type specimen
for science.
Material. Holotype: NML-G2017/001 (Figs.6, 7, 8), the
majority of an articulated postcranial skeleton recovered
Fig. 4 Varanodontine interclavicles. a Varanodontine varanopid,
possibly Tambacarnifex, MNG 12516, from the Bromacker local-
ity, Tambach Formation (Thuringia). b Varanops brevirostris, MCZ
1926, Cacops Bonebed, Arroyo Formation (Texas), curved element
shown from right lateroventral. c Silhouettes of MNG 12516 com-
pared to that of Varanops (underlaying pale gray and dark shading
for comparison at same scale), the latter taken from Romer and Price
(1940: fig.21). The scale bars measure 1cm each
F. Spindler etal.
1 3
without the cervical column, parts of the left extremities,
and the distal tail.
Locality and horizon. Cabarz quarry at the mountain
“Großer Inselsberg,” Cabarz/Tabarz, district of Gotha,
Thuringia, Germany: wet red beds at the 4th quarry floor;
Goldlauter Formation, Asselian/Sakmarian (Schneider etal.
2013, 2014).
Diagnosis. Slenderly built varanopid with slightly swollen
neural arches, dorsal ribs probably holocephalous, sacral
ribs with knob-like scars on dorsal surface, pubic tubercle
large and dorsally directed, entepicondylar foramen located
distally on humerus, forelimb short and stout, radius straight,
ulnare broader than long, phalanges slender, ungula phalan-
ges long and curved.
Comparative description
Cabarzia trostheidei gen. et sp. nov. is to be compared with
several basal amniotes of similar body proportions, espe-
cially Mesenosaurus romeri Efremov, 1938, for which the
first postcranial description is provided. Regarding the diag-
nosis of the latter, the latest revision (Reisz and Berman
2001) is reliably based on cranial characters. A compara-
tive diagnosis is subject to the cladistic analysis provided by
Fig. 5 Isolated amniote bones from the Goldlauter Formation, Cabarz
quarry, Thuringia. a Dermal skull element, NHMS-WP 10577. b
Dermal shoulder or mandible element, NHMS-WP 10580. c Juve-
nile humerus, NHMS-WP 10578. d Questionable tibia, NHMS-WP
10579. Scale bars measure 1cm
A new mesenosaurine from the lower Permian of Germany
1 3
Fig. 6 Articulated postcranium of NML-G2017/001, holotype of Cabarzia trostheidei gen. et sp. nov. Scale bar measures 5cm
F. Spindler etal.
1 3
A new mesenosaurine from the lower Permian of Germany
1 3
Spindler etal. (2018). Comments addressing the differential
determination are given in the following section.
For the present study, two largely complete skeletons of
Mesenosaurus have been examined (Figs.10, 11). One of
them is a juvenile or subadult individual, PIN 3717/1 (erro-
neously labeled 3713/1 by Spindler etal. 2018: fig.33d;
the label found in 2015 offered the catalogue number PIN
4659/16, which was then corrected by the curator). One
larger skeleton has been mentioned by Reisz and Modesto
(2007: p. 736) as PIN 3706/4, matching anatomical details.
Actually, this catalogue number belongs to a skull of the
contemporary procolophonomorph Nyctiphruretus (V. Gol-
ubev, pers. comm. 2016). In a photograph provided by D.
Scott (pers. comm.), this specimen is labeled PIN 1580/1,
although this could not be confirmed by Ivakhnenko (2008).
We can only conclude that this specimen is the subject of a
cataloging error.
Ontogenetic status—As far as preserved, none of the termi-
nal regions of the long bones in the holotype of Cabarzia
show the unfinished condition indicative of early ontogenetic
stages. The neurocentral suture is almost invisible in all pre-
served vertebrae. Along with the fully ossified mesopodials,
this implies an adult individual.
In comparison, different ontogenetic stages of Meseno-
saurus are known. An early juvenile skeleton, PIN 162/60,
is relatively short-snouted and bears broad neural arches
(Ivakhnenko 2008: fig.1). The skull of the largest skeleton,
PIN 3706/4 (Fig.10d), ranks among the largest known iso-
lated skulls. No indication of unfinished ossification could
be observed. The juvenile PIN 3717/1 (Fig.10a) is interme-
diate in size.
Presacral axial skeleton—Unfortunately, no cervical verte-
brae belonging to the holotype of Cabarzia could be recov-
ered. Elongated cervicals are characteristic of araeoscelid
diapsids (Reisz etal. 1984) and could therefore easily have
been used to test the taxonomic affiliation of Cabarzia. The
cervicals of Mesenosaurus are of about the same size as the
dorsals, which is typical of varanodontids except for derived
varanodontines (Pelletier 2014) and Apsisaurus (Laurin
1991). There might be up to seven cervicals, indicated by the
articulated position of the shoulder girdle as well as the poor
remains of straight cervical ribs (Fig.11). Microvaranops
(Botha-Brink and Modesto 2009) is also described as bear-
ing seven cervicals, which is slightly more than the number
of cervicals in most basal synapsids (Romer and Price 1940).
The axis spine is low (Fig.11). Some cervical neural arches
show a weak lateral excavation. On the left side, a single
well-preserved rib is slightly broadened and appears to be
dichocephalous, as estimated from the head, which exceeds
the diapophysial dimensions by far. For Cabarzia, a few pos-
terior cervical ribs are present as fragments of the left side,
recognizable due to their broader sections.
Since the anterior dorsal vertebrae in the holotype of
Cabarzia were partially exposed at the time of discovery,
they are eroded. What is preserved in the posterior trunk
compares well to all suggested amniote subclades. Regard-
ing basal parareptiles, only a few characters have been
defined that can be used to resolve their interrelationships
with respect to vertebral morphology (Tsuji 2010). In the
phylogenetic analysis of early saurians by Gottmann-Que-
sada and Sander (2009), who included araeoscelids as an
outgroup, Neodiapsida and further nodes are mostly based
on cranial support. Among non-neodiapsid forms, Cabarzia
shares similarities with a wide spectrum, for example miller-
ettids, captorhinids, and araeoscelids (Sumida 1997: fig.3).
The vertebrae of slender varanopids are also affected by this
deficiency in diagnostic features, suggesting that functional
morphology is the main reason for early amniote vertebral
structures, which hampers phylogenetic assessment.
The total vertebral count cannot be determined with
certainty. Considering the articulation of all documented
skeletons, the trunk comprises about 19 vertebrae between
the girdles of Mesenosaurus and Cabarzia, whereas Arae-
oscelidia can be estimated to have 20–21 within the same
topological range. None of the mentioned taxa equal Eudi-
bamus (Berman etal. 2000b: fig.1), in which the vertebrae
are markedly shorter, as inferred from neural arches.
Only limited observations can be used to describe the
vertebral centra of Mesenosaurus and Cabarzia. The dor-
sal centra are elongated in Cabarzia, which is not seen in
Mesenosaurus but is observed in all non-varanodontine
varanopids (Berman and Reisz 1982; Laurin 1991; Reisz
and Dilkes 2003; Botha-Brink and Modesto 2009). While
this characteristic is rare in non-varanopid basal synapsids,
it is common in early reptiles of all included subclades, so
it cannot be used to pin down the relationships of Cabarzia.
Nothing can be said about the ventral surface of the cen-
tra in both of the mesenosaurine genera examined. Reisz
etal. (1984) stated that strongly developed ventral keels on
the cervical and anterior dorsal centra are a synapomorphy
of araeoscelid diapsids. Again, this is a homoplasy shared
not only with several basal synapsids but also with some
non-diapsid eureptiles (Fox and Bowman 1966; Heaton and
Reisz 1986).
Although the proportions of the entire postcranium of
Cabarzia roughly resemble those of Eudibamus, these gen-
era can easily be distinguished based on their vertebrae. In
Fig. 7 Close-up photographs of NML-G2017/001, holotype of
Cabarzia trostheidei gen. et sp. nov. a Trunk and left forearm; b right
dorsolateral aspect of hip region with sacrals left of the center; c
straight aspect on right hindlimb; d frontal aspect of right pes; e right
forearm; f posteriormost preserved caudals in lateral aspect. Scale
bars measure 1cm
F. Spindler etal.
1 3
Fig. 8 Interpretative drawing of NML-G2017/001, holotype of Cabarzia trostheidei gen. et sp. nov. Scale bar measures 5cm
A new mesenosaurine from the lower Permian of Germany
1 3
Eudibamus (Berman etal. 2000b), the neural arches are
broad and short, as is characteristic of most parareptiles.
Nonetheless, this morphology does not represent a typi-
cal morphology of bolosaurid vertebrae. As demonstrated
by Watson (1954: fig. 5), Bolosaurus has less broadened
neural arches. In dorsal aspect, there is an overall similar-
ity in the dorsal vertebrae of, for example, Mesenosaurus,
Cabarzia, Bolosaurus, Erpetonyx (Modesto etal. 2015),
millerettids (Watson 1957; Gow 1972), Youngina (Watson
1957), and araeoscelids (Reisz etal. 1984). This similar-
ity is due to symplesiomorphy, interpreting the widely
spaced postzygapophyses (Benson 2012: character 161) as
a trivial observation in most of the included taxa. Unlike
Erpetonyx (pers. obs. F.S.) and millerettids, bolosaurids and
derived parareptiles have greatly swollen posterior neural
arches which form a convex eminence that emerges from
the low dorsal spine and reaches across the postzygapophy-
sis, regardless of a possible lateral excavation of the neural
arch (as, for example, in Nyctiphruretus, PIN 158/4). This
contrasts with the rather slender neural arches of eureptiles
and synapsids. A laterally constricted neural arch is typical
of araeoscelids (Reisz 1981: fig.14J; Reisz etal. 1984). In
contrast to parareptiles, the prezygapophyses are pronounced
in these reptiles.
Swollen neural arches are listed as a synapomorphy
of diapsids (Müller and Reisz 2006, which included only
Petrolacosaurus and Araeoscelis). Depending on refer-
ences, this swelling either relates to the overall dimensions
of the neural arch, or, as in the present description, implies
a bloated appearance resulting from highly convex surfaces.
Basal diapsids often show slightly more vertebral swelling
than the moderate swelling present in Mesenosaurus. In
the holotype of Cabarzia, the second, third, and sixth neu-
ral arches of the preserved sequence have similar shapes,
whereas most of the other vertebrae are affected by erosion.
A slight swelling occurs in the lumbar region. None of the
dorsal neural arches of Cabarzia and Mesenosaurus show
any trace of lateral excavation. Although there is a sharp
notch between the prezygapophysis and the diapophysis in
Cabarzia and Mesenosaurus, this does not compare to the
distinct lateral pockets observed in all known araeoscelids
with well-preserved neural arches (Reisz 1981; Brinkman
etal. 1984; Reisz etal. 1984). A few shallow excavations
on the lateral surface of the neural arch can be seen in Mese-
nosaurus, where they are restricted to the cervicals (Fig.6;
Reisz and Modesto 2007). A similar pattern is present in
Heleosaurus (Reisz and Modesto 2007). Very weak exca-
vations occur, for example, on the dorsals of the varanopid
Apsisaurus (Laurin 1991) and the parareptile Erpetonyx
(Modesto etal. 2015). No excavation is present in the vara-
nopid Archaeovenator (Reisz and Dilkes 2003). The basal
procolophonomorph parareptile Australothyris shows weak
pockets in its cervicals, whereas dorsal vertebrae are com-
pletely unknown (Modesto etal. 2009). As a result, there
is much variation in the neural arch morphology. Cabarzia
does not match the pattern of basal diapsids.
There is no trace of a mammillary process in Meseno-
saurus and Cabarzia. This process is present on the poste-
rior cervical and anterior dorsal neural arches lateral to the
spine in Araeoscelis (Vaughn 1955) and Petrolacosaurus
(Reisz 1981) as well as in a non-diapsid eureptile (Heaton
and Reisz 1980). It has been recognized as a synapomorphy
of a clade comprising both genera (Reisz etal. 1984, 2010:
character 76). In fact, it is absent from the fragmentarily
known araeoscelid Zarcasaurus (Brinkman etal. 1984). This
character was not observed in Spinoaequalis and Aphelo-
saurus (deBraga and Reisz 1995; Falconnet 2007). Addi-
tional observations of Kadaliosaurus, which is currently
being redescribed (J. Müller, pers. comm.), may help to
elucidate the distribution of this rare synapomorphy among
basal diapsids. Unfortunately, the drawing of the only known
skeleton provided by Credner (1889) does not indicate well-
preserved neural arches. His documentation, carried out by
Franz Etzold, usually ranks among the best of its time, as
Fig. 9 Reconstruction of preserved autopodia of NML-G2017/001,
holotype of Cabarzia trostheidei gen. et sp. nov. Scale bar measures
F. Spindler etal.
1 3
Fig. 10 Articulated skeletal material of Mesenosaurus romeri Efre-
mov, 1938. a PIN 3717/1, immature individual; b close-up of the hip
region of the latter; c right scapulocoracoid of the same skeleton in
lateral aspect; d PIN 3706/4, adult individual; e–h close-ups of the
latter; e left humerus, rib cage, and trunk column; f left forearm; g
left lower leg; h hip and right hindlimb. Scale bars measure 5cm for
whole skeletons and 1cm for details
A new mesenosaurine from the lower Permian of Germany
1 3
Fig. 11 Interpretative drawings of articulated Mesenosaurus romeri Efremov, 1938; PIN 3717/1 (left) and PIN 3706/4 (right), with chosen
details in lateral aspect (shaded). Scale bar measures 5cm
F. Spindler etal.
1 3
proven by the redescription of Palaeohatteria (Credner
1888; Spindler 2016).
Regarding the overall dorsal spine shape, most basal
amniotes share a broad, subrectangular profile. In Cabar-
zia, the spines are largely eroded but present a blade-like
shape (Fig.7b). With the minor exceptions of Varanosaurus
(Sumida 1989) and sail-backed sphenacomorphs (Edapho-
sauridae, Dimetrodon), this applies to all pelycosaur-grade
synapsids. In contrast, a subtriangular spine shape is found
in certain reptiles, along with a partial tendency to develop
pointed dorsal tips. Reisz etal. (2010: character 72) listed
Captorhinus, Paleothyris, Araeoscelis, Coelurosauravus,
and some lepidosauromorphs as presenting this condition.
The spine shape is indistinctly preserved in Thuringothyris
(Boy and Martens 1991), but Protocaptorhinus (Clark and
Carroll 1973: fig.16) resembles Captorhinus. Additionally,
millerettids show the subtriangular spine type, at least in
their dorsals (Watson 1957: fig.3E; Gow 1972: fig.15d).
In Bolosauria (Modesto etal. 2015; Berman etal. 2000b;
Watson 1954), the spine is very low, thus hampering the
appropriate determination of a certain spine type. However,
an intermediate type is present in Petrolacosaurus (Reisz
1981: fig.15), suggesting that the subtriangular type is com-
mon in araeoscelids. It cannot be scored for Spinoaequalis
(deBraga and Reisz 1995), but is present in Zarcasaurus
(Brinkman etal. 1984). According to current knowledge,
the neural spines of araeoscelids and captorhinids provide
diagnostic features, which contrasts with the presence of
subrectangular blade-like spines in Protorothyrididae and
Brouffia (Brough and Brough 1967; Carroll and Baird 1972;
Clark and Carroll 1973).
The dorsal spines are low in Cabarzia, as they are in all
varanopids except for Mycterosaurus (Berman and Reisz
1982) and Varanodontinae (Campione and Reisz 2010).
Based on the taxa list of Reisz etal. (2010: character 73),
high spines are rarely found in some early archosauromorphs
and lepidosauromorphs. Among early reptiles, Petrolaco-
saurus is the only genus known to have higher dorsal spines.
Alternation in dorsal spine height has been demonstrated
for various reptiles, but is best documented in araeoscelids
and captorhinids (Heaton and Reisz 1980; Reisz 1981; Reisz
etal. 1984). A functional discussion of this assumably ple-
siomorphic cotylosaur feature is provided by Sumida (1997).
Further evidence for a certain functional constraint might
come from initial examples in ‘microsaurian’ lepospondyls
(Carroll and Gaskill 1978: figs.47, 58, 61). Varanosaurus,
an exception among the synapsids, shows alternation and
an overall reptile-like vertebral anatomy (Sumida 1989). No
reliable observation is possible in Cabarzia and Meseno-
saurus, except to say that there is no trace of alternation in
the preserved spine sockets. The sacral spines of Cabarzia
are well preserved, resembling the dorsals in their blade-like,
longitudinally stretched shape. As alternation is apparently
associated with the subtriangular spine type as well as to a
horizontal orientation of zygapophyses (Sumida 1997), an
affinity of Cabarzia for araeoscelids can be rejected.
In the ‘lumbar’ regions of some varanopids, including
the basal form Archaeovenator (Reisz and Dilkes 2003),
the dorsal spines are inclined anteriorly (Benson 2012:
character 165). This condition is not clearly indicated in
Cabarzia. The last presacral spine probably displays an
anterior inclination, but is imperfectly preserved (Fig.7b).
In PIN 3717/1 of Mesenosaurus, the posterior dorsal ver-
tebrae are tilted anteriorly, hampering reliable observa-
tion. The larger PIN 3706/4 is indistinctly preserved in
this area too, but shows no indication of inclined spines.
In Mycterosaurus, the inclination is weak and applies to
one or two ‘lumbars’ (Benson 2012) less distinctly than in
varanodontines (Benson 2012: character 165, fig. A4A).
The diapophyses are short and placed anteriorly in the
neural arch in Cabarzia and Mesenosaurus, as also found
in all smaller varanopids and early reptiles. A greater lat-
eral extent of the transverse process is characteristic of
certain larger synapsids and of Archosauromorpha (Reisz
etal. 2010: character 75). A few pronounced diapophyses
can be found in the anteriormost dorsals of Araeoscelis
(Vaughn 1955: fig.6A, C; Reisz etal. 1984: fig.1).
The dorsal ribs in Cabarzia and especially in Meseno-
saurus are short, producing markedly slim trunks. They
rapidly decrease in size in the ‘lumbar’ region. Some
amniotes tend to evolve a holocephalous articulation of the
proximal rib to the associate vertebra. In accordance with
Reisz etal. (2010: character 79), this is a synapomorphy of
Acerosodontosaurus and more derived Neodiapsida. Previ-
ous studies also found this to be present in smaller varano-
pids (Berman and Reisz 1982; Botha-Brink and Modesto
2009; Modesto etal. 2011). Moreover, holocephalous
ribs are present in millerettids (Watson 1957; Gow 1972;
Thommasen and Carroll 1981) and probably in Erpetonyx
(pers. obs. F.S.), whereas this is unclear for Eudibamus
due to a limited osteological description (Berman etal.
2000b). In contrast to other araeoscelids, Spinoaequalis
is described as showing holocephalous ribs in its dor-
sal sequence (deBraga and Reisz 1995). Therefore, this
character shows considerable variability. In Cabarzia, the
smallish rib heads are indistinctly preserved. An offset
tuberculum might be erroneously interpreted from dam-
aged terminal regions. The tuberculum is clearly visible
in Mesenosaurus PIN 3706/4 (Fig.11), but the rib head
resembles the triangular expansion seen in Mycterosau-
rus (Berman and Reisz 1982). The smallest skeleton, PIN
162/60, exposes several rib heads in which the tubercular
and capitular articulation are visible as separate facets,
but at an even level. This implies that the holocephalous
appearance of varanopid dorsal ribs is either a special con-
dition of primary dichocephalous articulation, resembling
A new mesenosaurine from the lower Permian of Germany
1 3
the condition in diadectomorphs (Moss 1972; Berman
and Sumida 1990), or was achieved during ontogeny. As
estimated from the position relative to the parapophysis,
Mesenosaurus and likely also Cabarzia are functionally
Sacrum and caudal column—Preliminary discussion
included the supposed presence of four sacrals in NML-
G2017/001 (J. Müller, pers. comm.), suggesting a parareptil-
ian affinity. This apomorphy is seen only in derived, often
robustly built Parareptilia (Tsuji etal. 2012). In fact, the
holotype of Cabarzia is illusive when evaluated from docu-
mentation focusing on the dorsal aspect. A damaged area
across the right pubis leaves two bright areas of a similar
size to the sacral ribs. Because the sacral ribs are distinctly
elevated from the ventral pelvic level, the skeleton unques-
tionably displays the plesiomorphic condition of two sacrals.
The proximal dorsal surfaces of both the first and second
sacral ribs of Cabarzia bear a small knob-like scar, which is
also weakly present on the proximal caudal ribs. Distally, the
first sacral rib flares to about twice the longitudinal width of
the stout shaft. The second sacral rib is broken and ambigu-
ously shows the same outline. It is not narrower in its shaft
and does not contact the first sacral rib with its distal ter-
minus, so the sacral morphology greatly resembles that of
Mesenosaurus (Figs.8, 11) with two equally sized flaring
sacral ribs on each side. This matches the supposed apomor-
phic pattern of Archaeovenator (Reisz and Dilkes 2003) and
is apparently plesiomorphic in mesenosaurines, though it
is not preserved in any other genus. First and second sacral
ribs with equal dimensions are found in more taxa than pre-
viously recognized (Reisz etal. 2010: character 80), such
as Casea (LeBlanc and Reisz 2014), Paleothyris (Carroll
1969), and Protorothyris (Clark and Carroll 1973), as well as
in millerettids (Gow 1972; Thommasen and Carroll 1981),
Erpetonyx (regardless of the presence of a tiny third sacral
rib; Modesto etal. 2015), and neodiapsids (Carroll 1976b:
fig.1; Harris and Carroll 1977: fig.2). Contradicting Reisz
etal. (2010), Apsisaurus cannot be confirmed based on what
has been described by Laurin (1991). In general, the sacrum
of Cabarzia matches the condition of several clades, but
is clearly distinct from that of araeoscelids. In Araeoscelis
and Petrolacosaurus, the sacral ribs are strongly uneven,
with the delicate second one lacking any distal expansion
but articulating with the first sacral rib distally (Reisz etal.
1984). Furthermore, the anterior sacral rib is incised distally,
contacting the ilium with a concave distal terminus (Vaughn
1955: fig.6H–J; Reisz 1981). The same sacral features are
seemingly present in Aphelosaurus (Falconnet 2007).
The proximal caudal spines are more delicate in Mese-
nosaurus than in Cabarzia. Reisz etal. (1984) mentioned
elongate midcaudals as a synapomorphy of Araeoscelidia.
They are elongate in Cabarzia and Mesenosaurus as well,
especially in the juvenile PIN 3717/1. The adult skeletons
show massive vertebrae with about the same dimensions as
the presacrals, producing a greater partial mass percentage
than in other basal synapsids. As indicated by Microvara-
nops (Botha-Brink and Modesto 2009: fig.1), massive and
elongated caudals appear to be common in Mesenosaurinae.
Proximal caudal ribs are typically hook-like in basal
amniotes. In the holotype of Cabarzia, these elements are
incompletely preserved but apparently robust. They resem-
ble the corresponding ribs of Aerosaurus in possessing a
robust and strongly bent proximal portion and a slender,
nearly straight distal portion (Pelletier 2014: fig.4.3e).
Overall, there is only a limited comparison to be made with
other varanopids. Nonetheless, this contrasts with the condi-
tion in araeoscelids. Their caudal ribs are hook-like too, but
delicate, as seen in Araeoscelis (Reisz etal. 1984), Kadali-
osaurus (Credner 1889), and Petrolacosaurus (pers. obs.
F.S.). Neodiapsids bear straight, laterally pointing caudal
ribs (Carroll 1976b; Gow 1975; Harris and Carroll 1977).
Although not described, the restored skeleton of Milleretta
seems roughly similar, with broad, laterally directed caudal
ribs (Gow 1972: fig.1), as confirmed by Broomia (Thom-
masen and Carroll 1981). However, this does not represent
a common pattern in early parareptiles, as moderately slen-
der, hook-shaped caudal ribs are observed, for example, in
Erpetonyx (Modesto etal. 2015), Eudibamus (Berman etal.
2000b), and Nyctiphruretus (pers. obs.). Straight, laterally
pointing caudal ribs are present as a unique structure among
basal synapsids in Mesenosaurus. While incompletely pre-
served in PIN 3706/4, the juvenile PIN 3717/1 displays
a complete set of proximal caudals. This specimen is an
unquestionable varanopid, showing a synapsid temporal
region and a maxillary dorsal lobe that contacts the prefron-
tal as well as a prominent canine (Reisz and Berman 2001),
which is why there can be no confusion with the contempo-
rary neodiapsid Lanthanolania (Modesto and Reisz 2002).
Pectoral girdle—Anterior to the right scapula of NML-
G2017/001, there is a thickened rod-like element with
unsharp outline. Its position may indicate that this is the
cleithrum, whereas its broader ventral portion suggests a
clavicle. In PIN 3706/4 of Mesenosaurus, the cleithrum
is distally swollen but much narrower overall than the
smoothly curved clavicle (Fig.11). The cleithrum is present
in all basal synapsids and non-saurian reptiles (Reisz etal.
2010: character 85).
The pectoral girdle of Cabarzia is shifted dorsally due to
compaction, making it difficult to estimate the vertical level
at which the scapular fragment is broken. The preserved
section reveals a strong scapula that is longitudinally nar-
row. Considering that the presumably still articulated gle-
noid—marked by the vertical level of the humerus—is about
1cm below the broken area, the tip of the scapula remains
F. Spindler etal.
1 3
Carroll 1964
Brough and
Brough 1967;
Carroll and
Baird 1972
Carroll 1969
(Heaton and
Reisz 1986)
Modesto et al.
Milleretta rubidgei
Gow 1972
Broomia perplexa
Thommasen and
Carroll 1981
Milleropsis pricei
Watson 1957;
Gow 1972;
Thommasen and
Carroll 1981
Berman et al.
Mycterosaurus longiceps
Berman and Reisz 1982;
Reisz et al. 1997
Cabarzia trostheidei
gen. et sp. nov.
Mesenosaurus romeri
PIN 3717/1, PIN 3706/4
Microvaranops parentis
Botha-Brink and Modesto 2009
slender Neovaranopsia
lineage Bolosauria
2 cm
2 cm
2 cm
2 cm
A new mesenosaurine from the lower Permian of Germany
1 3
unknown. There is no opportunity to state whether a flaring
dorsal area was present or not.
The shoulder girdle of Mesenosaurus forms a narrow
bowl. In lateral view, the scapular blade of PIN 3717/1 is
round throughout the strengthened anterior margin. This
profile is also observed in several basal synapsids, such as
Heleosaurus (Carroll 1976a: fig.1), Mycterosaurus (Reisz
etal. 1997: fig.3), and Oedaleops (Langston 1965; Sumida
etal. 2014). Among reptiles, a fully convex anterior scapula
occurs in Araeoscelidia and certain Sauria (deBraga and
Reisz 1995; Falconnet 2007; Reisz etal. 2010: character
86), as well as in Millerettidae (Gow 1972; Thommasen and
Carroll 1981). In PIN 3706/4, the dorsal tip of the scapula
is offset by a shallow embayment, resembling another indi-
vidual of Mycterosaurus (Berman and Reisz 1982: fig.10A)
and thus implying an ontogenetic effect.
Pelvic girdle—The pelvis of every described specimen is
affected by moderate compaction, which implies that the
hip was even narrower in life. The puboischiadic plate has
possibly undergone some disarticulation in the holotype of
Cabarzia, which would conflict somewhat with the supposed
adult status. So far, there is no indication of a loss of coos-
sifications in the puboischiadic plate, as has recently been
reported for the specialized Eudibamus (Sumida etal. 2013).
As seen in Mesenosaurus, and to a lesser extent in Cabar-
zia, the pelvic rim forms an anterodorsally directed convex-
ity around the anterior end of the pubis-ilium suture. This
bulge is shared with Heleosaurus (Carroll 1976a: fig.1) and
Mycterosaurus (Berman and Reisz 1982: fig.9A).
Both Mesenosaurus and Cabarzia bear a strong, posteri-
orly directed iliac dorsal process that is typically basal syn-
apsid in outline. Although there are superficial similarities
with the ilium of araeoscelids (Reisz 1981: fig.18), the large
size of the dorsal process precludes the genera in question
from being in any early reptilian clade. Although the ilia are
damaged in the holotype of Cabarzia, it should be noted that
they are rather long posteriorly and narrow in cross-section.
This may suggest an initial sharp dorsal blade.
The development of iliac anterodorsal blades is com-
monly seen in varanopids, although they are far weaker
than the characteristic blades of Caseasauria and Sphenaco-
morpha. Some reptiles even show a tiny, anteriorly directed
initial process (Watson 1957: fig.8B; Reisz 1981: fig.18).
Varanodontines expose an intermediate blade along the
dorsal edge of the ilium (Maddin etal. 2006: fig.5G, H;
Pelletier 2014: fig.4.9). There is only limited information on
mesenosaurines, including a strong resemblance of the out-
line of the ilium in varanodontines to that of Microvaranops
(Botha-Brink and Modesto 2009: fig.1: SAM 2), similar
to the juvenile PIN 3717/1 and Cabarzia and likewise to
Eocasea (Reisz and Fröbisch 2014: fig.3). A unique struc-
ture is observed in PIN 3706/4, where an anteriorly directed
tubercle marks the point at which the iliac dorsal process
bends rearward. Posterodorsally to this, a distinct lobe forms
a blade-like eminence. In the right ilium, this lobe is set off
from the iliac dorsal process by a sharp notch. Most likely,
this autapomorphic structure is involved with modifications
of the limb musculature. There is no dorsal trough. Medially,
the ilium of Mesenosaurus and Cabarzia is smooth, lacking
an anteroventrally directed buttress above the ventral margin
as well as a medial shelf. The latter is reported from Araeos-
celidia (Vaughn 1955: fig.11B; Reisz 1981).
A small pubic tubercle is visible in Cabarzia and proba-
bly in PIN 3717/1 of Mesenosaurus, but is apparently absent
from PIN 3706/4 and Heleosaurus (Carroll 1976a: fig.1).
The tubercle dimensions seen in Cabarzia and Mycterosau-
rus (Berman and Reisz 1982: fig.9A) are similar and far
smaller than the pronounced tubercle of Ophiacodontidae
(Romer and Price 1940) and Araeoscelidia (Reisz 1981;
Reisz etal. 1984, 2010: character 99). The pubic tubercle is
absent from known basal parareptiles (Watson 1957; Gow
1972). No ischium is exposed in Mesenosaurus, and only a
small portion can be identified in Cabarzia.
General limb proportions—Although there is a large com-
ponent of functional morphology, it can be assumed that
evolutionary trends are stabilized by phylogenetic niche con-
servatism and can therefore yield phylogenetic information.
Figures12, 13, 14, 15, and 16 represent a compilation of
basal amniote extremities that bear comparison to the holo-
type of Cabarzia. The following comparison of the appen-
dicular skeleton is largely based on these graphical recon-
structions. For cited references, see the captions of Figs.12,
13, 14, 15, and 16. In order to demonstrate proportional
deviations, several ratios are given in Fig.17 and Online
Resource 1 of the ESM. Standardized limb reconstructions
as well as their phylogenetic relationships are provided in
Online Resource 2 of the ESM.
The hindlimb to trunk ratio was included by Reisz etal.
(2010: character 101), who coded ratios of less than 100%
for Casea, Ophiacodon, Mesenosaurus, and Paleothyris. In
the present study, trunk lengths were estimated as the dis-
tance between the glenoidea and the acetabulum (Fig.17a).
Based on the given compilation, a shorter hindlimb is com-
mon in all non-diapsid reptiles except for the specially
adapted Eudibamus. Araeoscelidia show a value of around
100%. Araeoscelis is known from controversial reports,
one based on disarticulated materials (Vaughn 1955) and
Fig. 12 Compilation of basal amniote limb reconstructions: slender
neovaranopsians (formerly Mycterosaurinae), early non-captorhinid
Eureptilia, and Parareptilia. Dorsal vertebrae outlines at the upper
right of each limb set are provided in order to get some insight into
the total-body dimensions. For a size-standardized compilation and
phylogenetic relationships, see Online Resource 2 in the ESM
F. Spindler etal.
1 3
Langston and
Reisz 1981;
Pelletier 2014
Tambacarnifex unguifalcatus
Berman et al. 2014
Watongia meieri
Olson 1974;
Reisz and Laurin 2004
Varanops brevirostris
Romer and Price 1940;
Reisz and Tsuji 2006;
Campione and Reisz 2010;
pers. obs. TMM 43628, MCZ 1926
Varanodon agilis
Olson 1965;
pers. obs.
Ruthiromia elcobriensis
Eberth and Brinkman 1983;
Spielmann and Lucas 2010;
pers. obs. MCZ 3150
early Varanopidae
Reisz and Dilkes 2003
Apsisaurus witteri
Laurin 1991
Ascendonanus nestleri
Spindler et al. 2018
10 cm
2 cm
Fig. 13 Remaining varanopid limb material (note the different scaling for Varanodontinae). See caption of Fig.12 for general remarks
A new mesenosaurine from the lower Permian of Germany
1 3
Romeria prima
Clark and Carroll
Clark and Carroll
Boy and Martens 1991;
Müller et al. 2006
Captorhinus spp. / Eocaptorhinus
Fox and Bowman 1966;
Heaton and Reisz 1980, 1986;
Holmes 2003
Reisz 1972, 1975
Reisz 1972
Case 1907;
Romer and Price 1940 Milosaurus mccordi
deMar 1970
Eocasea martini
Reisz and Fröbisch
captorhinid lineage
5 cm
2 cm
Fig. 14 Limb reconstructions of further Synapsida, Captorhinidae, and Thuringothyris. See caption of Fig.12 for general remarks
F. Spindler etal.
1 3
Leonardi 1959
MB 1901.1379
Carroll and
Baird 1972
Spinoaequalis schultzei
deBraga and Reisz 1995
Credner 1889
Araeoscelis gracilis
Reisz et al. 1984 Vaughn 1955
Aphelosaurus lutevensis
Falconnet 2007; Steyer 2012
Brinkman et al.
Petrolacosaurus kansensis
Reisz 1981
early Diapsida (Araeoscelidia)
2 cm
early amniotes
2 cm
Fig. 15 Limb reconstructions of early Diapsida and enigmatic amniotes. See caption of Fig.12 for general remarks
A new mesenosaurine from the lower Permian of Germany
1 3
Carroll 1976b
composite early
Carroll 1977
Youngina capensis
Gow 1975; Carroll 1976b;
pedal phalanges
from juveniles:
Smith and Evans 1996
Clark and Carroll
Harris and Carroll 1977
Carroll and Baird
Carroll and Baird
1972; Reisz and
Baird 1983
Stein et al. 2008
and Sander 2009
Carroll 1964
5 cm
early Neodiapsida
2 cm
2 cm
Fig. 16 Limb reconstructions of Protorothyrididae and Neodiapsida. See caption of Fig.12 for general remarks
F. Spindler etal.
1 3
hindlimb : trunk
forelimb : hindlimbhumerus width : length ulnare width : lengthradius : humerusfemur : humerus metacarpal IV : radius
ab dfec g
MB 1901.1379
early lepidosaur
basal pararaptiles
captorhinid lineag
basal diapsid
various early synapsid
slender varanopids
basal neodiapsid
romeriid-like reptiles
(not prejudiced)
Cabarzia, gen. nov.
60 70 80 90 100 110 120 130 140 150 %
60 70 80 90 100 110 120 130 140 150 %
50 60 70 80 90 %
50 60 70 80 90 %20 30 40 50 %
20 30 40 50 % 50 60 70 80 90 100 110 %
190 200
50 60 70 80 90 100 110 %
50 60 70 80 90 %
50 60 70 80 90 %
100 110 120 130 140 150 %
100 110 120 130 140 150 %
30 40 50 60 %
30 40 50 60 %
tibia+astragalus : toe IV [incl. metatarsal]tibia : femur toe I : IV and V : IV [incl. metatarsals]
metacarpal V : metacarpal IV
metatarsal IV : phalangesmetatarsal V : metatarsal IV
femur width : length calceneus width : length
hkj m on
40 50 60 70 80 90 100 110 120 %
40 50 60 70 80 90 100 110 120 %
50 60 70 80 90 %
50 60 70 80 90 %
20 30 40 50 60 70 %
20 30 40 50 60 70 %
40 50 60 70 80 90
40 50 60 70 80 90
% 20 30 40 50 60 70 %
20 30 40 50 60 70 %
30 40 50 60 70 80 %
30 40 50 60 70 80 %
10 20 30 %
10 20 30 %
70 80 90 100 110 120 130 %
70 80 90 100 110 120 130 %
A new mesenosaurine from the lower Permian of Germany
1 3
a more reliable one based on articulated skeletons (Reisz
etal. 1984). Kadaliosaurus is incompletely known (Credner
1889), but the preserved elements imply that the hindlimb
most likely did not exceed the trunk length. Synapsids usu-
ally have shorter limbs, with the exception of known mese-
nosaurines and possibly some basal caseasaurians (Spindler
etal. 2016). In this context, Cabarzia lies between certain
Araeoscelidia and Mesenosaurinae, close to Kenyasaurus,
Protorosaurus, and the enigmatic Tridentinosaurus.
The extremities are slender in Cabarzia and numerous
other amniotes. Slender limbs are considered a synapo-
morphy of Araeoscelidia (Reisz etal. 1984; deBraga and
Reisz 1995) for early reptiles. Based on the included list
of taxa, slender long bones were also used to support the
monophyly of non-captorhinid eureptiles (Müller and Reisz
2006). This group, comprising Brouffia and Romeriida (the
latter is defined as Paleothyris plus Diapsida according to
Hill 2005), is also characterized by long and slender auto-
podial complexes. As far as known, this also applies to basal
varanopids and slender varanopids, as well as to basal para-
reptiles. The tibia and fibula of Cabarzia are less slender
than in Petrolacosaurus and derived Araeoscelidia, whereas
their shape perfectly matches the known mesenosaurines.
The anterior zypopodials of Cabarzia are more robust than
in any araeoscelid. With respect to toe slenderness, Cabarzia
surpasses the range shown by millerettids.
One measurement of stylopodial relative slenderness
is defined as the ratio of femur to humerus shaft diameter
(Reisz etal. 2010: character 105). A femur shaft that is
nearly 150% as thick as the humerus was coded for Mycter-
osaurus, Ophiacodon, Apsisaurus, Paleothyris, Petrolaco-
saurus, and Araeoscelis. This is difficult to evaluate from
incompletely recovered, sometimes flattened fossils. How-
ever, such a ratio is unquestionably present in Araeosce-
lidia, is possibly also seen in Anthracodromeus, Milleretta,
Apsisaurus, and Microvaranops, but is absent in Cabarzia.
The relative limb slenderness is also affected by interde-
pendencies with relative lengths. Forelimb to hindlimb ratios
are shown in Fig.17b. Cabarzia, with its shortened forelimb,
is at the lower edge of the range seen for basal parareptilians.
In araeoscelidians, the forelimb is usually not greatly short-
ened, in stark contrast to Araeoscelis itself (Reisz etal. 1984;
unknown based on Vaughn 1955). The closest match with
Cabarzia is seen in Eudibamus, Milleropsis, Varanosaurus,
and Mesenosurus. Relative to its shortness, the forelimb of
Cabarzia is stouter than in any basal diapsid.
Additional information from the femur to humerus length
ratio is considered (Fig.17c; Reisz etal. 2010: character
104, erroneously coded for Varanodon). A critical value of
120% separates Mesenosaurus and Aerosaurus from the rest
of the Varanopidae. Araeoscelis was coded by Reisz etal.
(2010), based on Reisz etal. (1984), resulting in an outlier
in Fig.17c. When Vaughn (1955) is considered, Araeoscelis
is found to fall within the range exhibited by other araeos-
celids. In general, there does not appear to be a significant
gap in the plotted ratios. Cabarzia is right at the mean, such
that it resembles some basal reptiles, most araeoscelidians,
Mesenosaurus, and Aerosaurus.
Forelimb—The humerus of Cabarzia is massively built in
comparison to Romeriida (Sumida 1997: fig.7), but slender
when compared to other synapsids. In its overall shape, there
is some resemblance to captorhinids. A precise description
is hampered by the eroded condition of the terminal regions.
The humerus distal width to length ratio (Reisz etal. 2010:
character 89) is about 40%. Based on the original wording,
which implies a limit of 30%, it would code as a broad form.
This is common in synapsids, captorhinids, and neodiapsids
more derived than Youngina. Slender distal humeri are
found, for example, in Eudibamus, Broomia, Araeoscelidia,
Apsisaurus, and Microvaranops. When the ratio is estimated
from Fig.17d, Cabarzia most closely resembles Milleretta,
Protoclepsydrops, Spinoaequalis, and Mesenosaurus.
There is a weak ectepicondylar ridge and an inconspicu-
ous ectepicondylar groove in the heavily damaged distal
humerus of Cabarzia. It is not clear whether an ectepicon-
dyle foramen was present or not. Araeoscelis, Coelurosaura-
vus, and Lepidosauromorpha have a bridged ectepicondylar
groove (Reisz etal. 2010: character 92). Recently, an ect-
epicondylar foramen was identified in the remarkably slen-
der humerus of Microvaranops (Botha-Brink and Modesto
A large entepicondyle was found by Reisz etal. (2010:
character 90) for Captorhinus, Acerosodontosaurus, Youn-
gina (questionable, see Gow 1975: fig.9), Ophiacodonti-
dae, and Casea, but—erroneously—not in Varanodontinae.
Certain smaller varanopids show a reduced entepicondyle
(Apsisaurus, Microvaranops), whereas Mycterosaurus and
Mesenosaurus retain the plesiomorphic condition. Basal
parareptiles and eureptiles also have narrow entepicondyles,
which are greatly reduced in Bolosauria, derived Miller-
ettidae, Araeoscelidia, and most of the basal Neodiapsida.
The entepicondyle of Cabarzia does not resemble any early
reptile with the exception of the captorhinid branch and
Protorosaurus. The presence of an entepicondylar foramen
precludes Cabarzia from Archosauromorpha (Reisz etal.
2010: character 91). This large, oval foramen has a rather
distal position, contrasting with all reptiles but Milleretta
and all known basal synapsids aside from Mesenosaurus.
Fig. 17 Compilation of selected ratios of limbs and limb elements in
early amniotes. Histograms indicating the occurences of certain val-
ues are shown. See text for a description of the methodical approach
used and Figs. 1215 for references. Ratios are listed in Online
Resource 1 of the ESM
F. Spindler etal.
1 3
Since the humerus is incomplete in the holotype of
Cabarzia, the radius to humerus length ratio could only be
estimated as a maximum of 56%. Reisz etal. (2010: char-
acter 93) defined three states. Cabarzia would share a ratio
of no higher than 68% with Hyperodapedon, Captorhi-
nus, Paleothyris, and Casea. As illustrated in Fig. 17e,
this also applies to derived millerettids, all non-diapsid
eureptiles, Varanosaurus, Mesenosaurus, and Watongia.
An intermediate range of between 68% and 82% is exhib-
ited by Bolosauria, Milleretta, Cephalerpeton, some early
diapsids, and most varanopids. Among the latter, the ratio
for Varanodon rather uncertain, as addressed by a com-
ment below regarding its forearm reconstruction. Reisz
etal. (2010) found araeoscelids and Varanops to have
a radius to humerus ratio of more than 82%. This was
rejected following the revised reconstruction of Varanops
(Fig.13). Kadaliosaurus and Aphelosaurus contribute
lower ratios to the range for araeoscelids, which supports
to some extent the synapomorphy suggested by Reisz etal.
(1984), referring to the outlier ratios of Petrolacosaurus
and Araeoscelis. Upon introducing a boundary at around
60%, which is the only significant gap within the varanopid
spectrum, a few taxa fall within the range corresponding
to a short radius. This condition is shared by Milleropsis,
Thuringothyris, Captorhinus (but not Protocaptorhinus),
most Palaeothyrididae, a synoptic basal lepidosaur (Car-
roll 1977), Mesenosaurus, and Cabarzia. In contrast, the
lowest ratio among basal diapsids is seen for Aphelosau-
rus, with about 73%.
Reisz etal. (2010: character 94) noted a twisted radius
in Youngina and Acerosodontosaurus. A sigmoidal shape
is also seen in certain varanodontines, Mesenosaurus, and
probably in Microvaranops and Galesphyrus. It is not seen
in Cabarzia. Apsisaurus indicates that a straight radius is
present in basal varanopids.
The ulnar olecranon processus was previously coded
to be low in all varanopids including Apsisaurus, in Ace-
rosodontosaurus, and in Archosauromorpha (Reisz etal.
2010: character 95). This has to be re-evaluated, as a pro-
nounced olecranon is present in Protorosaurus, Watongia,
and Mesenosaurus. In the holotype of Cabarzia, the proxi-
mal ulna is partially broken or unexposed but appears to
lack a strongly developed olecranon. If correct, this would
contrast with derived millerettids and better evidence for
derived araeoscelids.
Both the carpus and the tarsus in the holotype of Cabar-
zia show excellent articulation. With the exception of Gales-
phyrus, the carpus does not closely resemble any neodiapsid.
Instead, the entire carpus is similarly built in Aerosaurus,
Mesenosaurus, Petrolacosaurus, and basal parareptiles,
though somewhat elongated in Eudibamus. The radiale,
intermedium, and centralia are of a similar size in the carpus
of Cabarzia and Mesenosaurus, which most likely represents
a plesiomorphic condition shared with Paleothyris, Miller-
ettidae, Erpetonyx, and Neodiapsida. In Microvaranops, the
carpals are all smallish, including an elongated intermedium
(Botha-Brink and Modesto 2009: fig.6). If unaffected by
damage, this carpus displays a unique morphology. Cabar-
zia and Mesenosaurus show a broad intermedium. This
contrasts with the tall, slender intermedium that occurs
in association with a tall ulnare in Araeoscelidia and MB
1901.1379 and independently in Captorhinidae.
The width to length ratio of the ulnare was estimated
(Fig.17f). A wide range of values were found in basal amni-
otes, with the narrowest elements found in Eudibamus, an
unnamed early amniote (MB 1901.1379; Carroll and Baird
1972: fig.8), and Araeoscelis (lower than 60%), followed by
other basal non-captorhinid eureptiles, Petrolacosaurus, and
Galesphyrus (up to 70%). Although not strictly diagnostic,
this ratio of the ulnare is probably associated with an adap-
tation for improved cursorial ability; possibly for increased
effect during ankle stretching. The majority of the included
taxa have an ulnare width that is approximately equal to its
length, including several eureptiles and all varanopids except
for Mesenosaurus. In this genus, the ratio is around 80%—
an unusual value that is shared with certain early reptiles.
Cabarzia is an outlier with an autapomorphic ratio of more
than 110%. Among the depicted taxa (Figs.12, 13, 14, 15,
16), Youngina shows the highest ratio, with an ulnare that is
about twice as broad as it is long. Similar ratios are common
in tangasaurids (Carroll 1981; Currie 1981: fig.26; Currie
1982; Currie and Carroll 1984), or rather a grade of early
Neodiapsida that includes Acerosodontosaurus (Bickelmann
etal. 2009: fig.3A). Because early neodiapsids frequently
show aquatic adaptations, their apomorphic broad ulnares
could be a potential indicator of swimming locomotion.
A medial centrale carpi is present in all basal synap-
sids and non-archosauromorph reptiles (Reisz etal. 2010:
character 96). The lateral centrale carpi is lost in Sau-
ria (Reisz etal. 2010: character 97). Both of these state-
ments have been debated with respect to Protorosaurus
(Gottmann-Quesada and Sander 2009). Moreover, only
one centrale was recorded in Mesenosaurus, leaving some
doubt regarding its true anatomical nature. In fact, there
is an absence of the lateral centrale in Varanodontinae.
Although they coded all varanopids in the same manner,
Reisz etal. (2010) explicitly remarked that Varanodon has
a large lateral centrale carpi. This observation is based on
the problematic preliminary description by Olson (1965),
who stated this to be a unique feature of Varanodon. A
revised reconstruction is provided in Fig.13, based on the
fact that the distal portion of the radius is incompletely
preserved (pers. obs. F.S.). The main portion of the radius
is obviously rotated, as indicated by its distal and medial
curvature as well as a thin anterior crest, resembling other
varanodontines (Reisz and Laurin 2004: fig.5B; Maddin
A new mesenosaurine from the lower Permian of Germany
1 3
etal. 2006: fig.4H–K). A large, roughly quadratic por-
tion has previously been interpreted as the radiale (Olson
1965). When compared to the short trapezoidal radiale
of varanodontines (Reisz and Laurin 2004), this shape
appears to be rather atypical. In contrast, this portion
is interpreted herein as the thickened distal terminus of
the broken radius, still articulated to the carpus. This
is confirmed by the observation that the element previ-
ously interpreted as the lateral centrale in fact matches the
shape of a typical varanodontine radiale. The carpus would
then have a normal position relative to the intermedium.
According to a preliminary examination, the holotype of
Varanodon has no lateral centrale in its articulated carpus.
In this, Varanodon resembles Watongia (Fig.13), its clos-
est relative (Reisz and Laurin 2004; Benson 2012). In the
redescription of Watongia (Reisz and Laurin 2004), the
centrale is not included in the text. However, both manus
are articulated and clearly reveal the absence of the lat-
eral centrale through contact of the radiale with the first
distal carpal (Reisz and Laurin 2004: fig.5A, B). There
is some doubt due to the resulting length of the radius in
Varanodon, which is probably caused by an overlooked
region of overlap between the radius pieces. In Varanops,
the presence of the lateral centrale is unknown, although
an articulated partial manus (Campione and Reisz 2010)
appears to show a gap for a cartilaginous element (Olson
1965). Both centralia are present in Tambacarnifex and
Aerosaurus. Final proof is expected from the redescription
of Varanodon (R. Reisz, pers. comm.).
The preservation of the carpus of Cabarzia is exceptional;
even a tiny pisiform in articulation with the remaining car-
pus can be seen. In PIN 3706/4, the large element in the
posterior carpus is the ulnare, as indicated by the articular
margin for the ulna. Thus, Cabarzia yields the only known
carpal pisiform of a mesenosaurine. Its small size is shared
only with Erpetonyx, Hylonomus, and Galesphyrus. A mid-
sized pisiform is preserved in specimens of Paleothyris,
Araeoscelis, Petrolacosaurus, Milleropsis, and Broomia.
In contrast, all varanodontines that preserve this element
bear a greatly enlarged pisiform (Fig.13). This is a com-
mon feature of pelycosaur-grade synapsids (Romer and Price
1940: fig.40), shared with only one other terrestrial clade:
the lineage of Thuringothyris and Captorhinidae (Fig.14).
Considering the shortened forelimb of mesenosaurines (and
Varanosaurus), a secondarily reduced pisiform would not
be surprising if it was involved with functionally induced
modifications. Interestingly, a large pisiform is present in the
enigmatic MB 1901.1379 (Carroll and Baird 1972), which
is commented on below.
One suitable measurement for the relative length of the
manus is the ratio of the fourth metacarpal to the radius
(Fig.17g), because this ratio is independent of the elonga-
tion or incomplete preservation of phalanges. Most of the
taxa included here show a ratio of about 47%, as seen in
some basal reptiles, Varanops, Tambacarnifex, and Cabar-
zia. All araeoscelids show lower ratios. Except for Micro-
varanops, all varanopids show higher ratios. Overall, this
character shows great variation, reflecting different preferred
locomotion styles or applications of the limbs. The varia-
tion is greatest in slender varanopids, intermediate in early
parareptiles, but narrow in known araeoscelids. Therefore,
the latter do not provide a ready affiliation for Cabarzia with
respect to the manus length.
A special feature of Petrolacosaurus and Araeoscelis is
that the first metacarpals do not constantly increase in size—
the second metacarpal is closer to the length of the third than
to the first. A similar pattern is found in Protorosaurus, as
well as in Mesenosaurus and Cabarzia, albeit only slightly
so, enhanced by the relative shortening of the first meta-
carpal. Future analyses might include this observation in
functional considerations.
The third and fourth metacarpals are of a similar length in
Mesenosaurus and Cabarzia, a feature shared only with the
captorhinid branch, early neodiapsids, known millerettids,
and Varanops. In Cabarzia, the fifth string (metacarpal and
associated phalanges) is intermediate in length, between that
of the second and the third strings of the manus. This char-
acter is difficult to evaluate due to incomplete knowledge. A
fifth string that is shorter than the second can be observed in
Captorhinus, Petrolacosaurus, several neodiapsids, varano-
dontines, and probably Mesenosaurus. In general, there is
much variation in these ratios.
Potential resolution is expected to be achieved by examin-
ing the ratio of the fifth to the fourth metacarpal (Fig.17h).
There is a huge gap between 68% and 82%, with Erpetonyx,
Brouffia, and Cephalerpeton being the only outliers. The
majority of the taxa present ratios of around 60%, grouping
Cabarzia with basal parareptiles, Thuringothyris, Aphelo-
saurus, and derived varanodontines. A group with lower
ratios of up to 50%, separated by the intermediate Aero-
saurus, apparently comprises cursorial forms such as MB
1901.1379, Araeoscelidia, early Sauria, and the two known
The phalangeal rows are incomplete but shorter than
for the hind toes. The third digit of Cabarzia is apparently
compressed longitudinally. This is tentatively interpreted as
preserving the ungual phalanx. The fourth finger can be esti-
mated from the ranges of the other, more complete strings.
A comparison of the phalanges is provided along with the
hind foot.
Hindlimb—For the femoral distal width to total length ratio,
Reisz etal. (2010: character 103) distinguished their termi-
nal taxa by applying a boundary at 25%, which equals the
estimated value for Cabarzia. This would mark the minimal
value needed to join a group comprising Mycterosaurus,
F. Spindler etal.
1 3
Archaeovenator, Varanodontinae, Paleothyris, Araeoscelis,
and nearly all Neodiapsidians, but not Casea, Ophiacodonti-
dae, Mesenosaurus, Captorhinus, Petrolacosaurus, and
Apsisaurus. These codings (Reisz etal. 2010) do not reflect
the revised reconstructions. Moreover, the previous defini-
tion of states does not relate to a significant distinction of
conditions in the current spectrum shown in Fig.17i. Cabar-
zia is among the mean of the entire taxon list, slightly lower
than the ratio found in small varanopids, and very similar
to the relatively narrow range of early diapsids. Employ-
ing a critical value of 30% (Reisz etal. 1998: character
21), the stout condition applies to half of the non-diapsid
reptiles, all non-varanopid synapsids, Mycterosaurus, and
Few details can be observed on the proximal femur. As
in other slender varanopids, the scar for the M. puboischi-
ofemoralis internus is inconspicuous. Cabarzia exposes a
distinct intertrochanteric fossa and a well-developed inter-
nal trochanter (Fig.7b, c). The latter is a narrow blade, not
as rounded as in Mycterosaurus and Varanodontinae (Reisz
etal. 1997: fig.6B; Campione and Reisz 2010: fig.14). Its
elongated edge reaches to about the middle length of the
femur, similar to most slender basal amniotes (Vaughn 1955;
Reisz 1981; Thommasen and Carroll 1981; Berman etal.
2000b), but significantly longer than in Apsisaurus (Lau-
rin 1991: fig.10). The proximoventral end of the internal
trochanter in Cabarzia is more angular than in compared
taxa. A fragmentary femur of Bolosaurus (Watson 1954:
fig.6) does not yield any differences from the early amni-
ote spectrum. More distally, the femoral shaft shows a sig-
moidal curvature in Mesenosaurus that is even stronger in
Cabarzia along with early diapsids, e.g., Kadaliosaurus and
Both condyles on the distal femur of Mesenosaurus and
Cabarzia end at about the same longitudinal level. This con-
tradicts Reisz etal. (2010: character 102), who coded this
apomorphic condition for Mycterosaurus, Aerosaurus, and
Varanops, whereas Mesenosaurus was considered to show
a posterior condyle that significantly projects over the tibial
condyle. The revised compilation demonstrates that this
plesiomorphic condition is present in varanodontines and
Mycterosaurus (Fig.12, 13). In this respect, Mesenosaurus,
Cabarzia, and the basal varanopid Apsisaurus show similar
condyles to those of Neodiapsida and Milleropsis.
The dorsal surface of the posterior distal condyle (fibular
condyle) presents a longitudinal trough in derived ophiaco-
dontids (introduced by Benson 2012: character 234). A very
similar structure on the posterior side of the femur is seen in
Varanops (Campione and Reisz 2010: fig.14). Additionally,
a shallow trough is present in Mycterosaurus (Berman and
Reisz 1982: fig.9) and Mesenosaurus, but is absent from the
basal varanopids Archaeovenator (Reisz and Dilkes 2003)
and Apsisaurus (Laurin 1991), as well as Cabarzia. As the
trough is present in both condyles of the femur of Petrola-
cosaurus (Reisz 1981: fig.22A), the considerable diversity
in this character should be considered in future analyses.
Müller and Reisz (2006) stated that equal tibia and fibula
lengths are indicative of Diapsida. However, there is no
significant deviation in any of the included taxa (Figs.12,
13, 14, 15, 16). The authors also refer to the tibia to femur
ratio, which is higher than 80% in almost all diapsids except
for Zarcasaurus (Fig.17j). Reisz etal. (1984) reported that
near-equal tibia and femur lengths are a synapomorphy of
Araeoscelidia, as confirmed by ratios > 90% in Petrolaco-
saurus, Araeoscelis, and Spinoaequalis. In general, the 80%
boundary is significant in reptiles, as all of the included non-
diapsid taxa present values < 70%, two millerettids have
ratios of nearly 80%, and the specialized Eudibamus is an
outlier. Extremely shortened hind zygopods are observed in
Milleropsis, Hylonomus, and Anthracodromeus, apparently
indicating that they were not fast runners. Except for Archae-
ovenator and Ruthiromia, all varanopids exhibit ratios that
resemble those of diapsids. Cabarzia most closely resembles
Aphelosaurus, Kenyasaurus, Protorosaurus, Aerosaurus,
Microvaranops, and Mesenosaurus.
None of the tibiae of Cabarzia or Mesenosaurus display a
distinct cnemial crest, which is otherwise present, for exam-
ple, in Captorhinus (Fox and Bowman 1966: fig.34) and
Araeoscelis (Vaughn 1955: fig.13) but absent in Petrolaco-
saurus (Reisz 1981) and all known varanopids.
The ratio of the tibia plus astragalus to the fourth metatar-
sal plus associated phalangeal string is a good characteristic
value for the functional pes length that is included in the data
set of Reisz etal. (2010: character 106). In Cabarzia, this
ratio is about 75%, which would group it with Casea, Vara-
nodontinae, Archaeovenator, Captorhinus, Paleothyris, and
most neodiapsidians, according to the original wording. In
the revised compilation (Fig.17k), the lower outliers Pale-
othyris, Hylonomus, and Anthracodromeus imply that the
ratios of Fig.17j refer solely to shortened zygopods, whereas
Milleropsis shows an elongation of the femur, the pes, or
both. Apart from the outliers, the pes length (Fig. 17k)
shows a significant gap at around 95%. Higher ratios apply
only to Captorhinus, Petrolacosaurus, Araeoscelis, Triden-
tinosaurus, and Varanodontinae. The only direct match for
Cabarzia is Mesenosaurus, further resembling few basal
reptiles, Spinoaequalis, Aphelosaurus, Kenyasaurus, and
Archaeovenator. Müller and Reisz (2006) stated that a short
fourth metatarsal relative to the tibia is a synapomorphy of
Diapsida. This is debated, as basal diapsids range from about
62% to 90%, whereas Müller and Reisz (2006) included only
Petrolacosaurus and Araeoscelis in their analysis. Both gen-
era appear specialized, even in an araeoscelid context.
The tarsus of Cabarzia shares similarities with neodi-
apsid taxa, but lacks a good match with any certain genus.
None of the suggested relatives of Cabarzia shows a
A new mesenosaurine from the lower Permian of Germany
1 3
lepidosauromorph ankle joint or the archosauromorph fea-
tures of the lateral calcaneus tubercle or a hooked fifth meta-
tarsal (Reisz etal. 2010: characters 108, 109, 113). Another
major distinction is presented by Eudibamus, in which the
tarsus is unusually broad and the tibia articulation is offset
from the fibula articulation with the tarsus.
Due to minor crushing, the astragalus of Cabarzia is bro-
ken, but it apparently shares its compact shape with Mese-
nosaurus, Microvaranops, and Millerettidae. This type of
astragalus contrasts with the reversed trapezoidal or some-
what ‘heart-shaped’ type in which the dorsal neck is weak
but offset from the tibial articulation. This type is found in
Mycterosaurus, Varanodontinae, Echinerpeton, Paleothy-
ris, Captorhinus, Araeoscelis, and Neodiapsida. A modi-
fied condition is seen in taxa that tend to show an elongated
astragalus neck, such as Archaeovenator, Ruthiromia, Vara-
nosaurus, Milosaurus, and Petrolacosaurus. A more trans-
verse and tall astragalus is rare, being present in Hylonomus,
Eudibamus, and (questionably so) in Erpetonyx.
Basal amniote clades share the plesiomorphic condi-
tion of a flat astragalus–calcaneus articulation (Reisz etal.
2010: character 107), with different diagnostic conditions
evolved by Archosauromorpha and Lepidosauromorpha.
Therefore, the condition in Cabarzia cannot be used to
determine whether it is a non-saurian reptile or a synapsid.
However, Reisz etal. (2010) introduced a new condition for
Varanodontinae, described as an expanded astragalus–cal-
caneus articulation with the perforating foramen visible on
the calcaneus. Such a condition is known from Aerosaurus
(Langston and Reisz 1981: fig.16; Pelletier 2014: fig.4.14),
Varanops (Campione and Reisz 2010: fig.15; even stronger
in Maddin etal. 2006: fig.6), Tambacarnifex (Berman etal.
2014: fig.5.7b), and is indicated in Ruthiromia (pers. obs.
F.S.). In Mesenosaurus, the calcaneus is flat with a slightly
expanded medial articulation surface. The calcaneus of
Cabarzia shows a weakly concave contribution to the per-
forating foramen. This is lacking in Mycterosaurus accord-
ing to Berman and Reisz (1982: fig.10C), but is present in
an ankle tentatively assigned to this genus (Davis 2013: p.
165). A calcaneus that contributes to the perforating foramen
is seen in Archaeovenator (Reisz and Dilkes 2003) and also
in many other synapsids and eureptiles. No foramen could
be identified in Microvaranops (Botha-Brink and Modesto
The range of calcaneus width to length ratios is com-
piled in Fig.17l. Cabarzia plots within a range of 90–100%,
which is shared only by Mesenosaurus, Mycterosaurus,
Aerosaurus, and probably Hylonomus, Galesphyrus, and
Protorosaurus with about 100%. The remaining romeriids,
diapsids, and varanopids are spread around this narrow field.
Parareptiles, captorhinids, and non-varanopid synapsids are
all below 90%. In Mesenosaurus, the calcaneus–fibular
articulation is a flat surface, suggesting a rather stiff contact.
The single centrale tarsi of Cabarzia is affected by minor
damage. Nonetheless, it represents the best evidence thus
far of such an element in varanopids. There is no aspect of
its shape that clearly distinguishes it from any basal amniote
subclade except from the enlarged centrale tarsi in derived
millerettids and the lack of this element in Eudibamus.
The fourth distal tarsal is disproportionally enlarged in
most basal amniotes, except for basal romeriids and pro-
torothyrids. The fusion of the fourth and fifth distal tarsals
to form a large cuboid, as seen in Araeoscelis (Vaughn
1955) and Milleropsis (Thommasen and Carroll 1981), is
an isolated phenomenon in each corresponding lineage. Its
separate ossification in Cabarzia precludes it from the clade
Sauria according to Reisz etal. (2010: character 112). In
contrast, the isolated fifth distal tarsal is lost in Kenyasaurus,
Milleropsis, and Eudibamus, and questionably so in Tambac-
arnifex, but not in Varanops.
With the exceptions of Casea, Captorhinus, Paleothyris,
and one lepidosauromorph (Reisz etal. 2010: character 110),
all basal amniotes show proximally overlapping metapodials.
Datheosaurus probably also lacks this feature (Spindler etal.
2016: fig.4), whereas it is present in Eocasea (Reisz and
Fröbisch 2014: fig.3). Overlapping metapodials are difficult
to confirm in Cabarzia. The metacarpals and metatarsals
might have been pushed apart after burial, as suggested by
the initial dislocation of the carpals and tarsals. The fourth
metatarsal is expanded proximally, providing an articula-
tion surface for the fifth metatarsal, as in Petrolacosaurus
(Peabody 1952: fig.2). A similar structure is preserved in
Mesenosaurus, where the overlap is enhanced by a proximal
flaring in the fifth metatarsal (Fig.10g, h).
Based on the excellent articulation and lifelike postures
of the described skeletons, it can be stated that Meseno-
saurus and Cabarzia show an angled fifth toe incorporat-
ing the fifth metatarsal. This pattern is also found in Vara-
nodontinae (Pelletier 2014: fig.4.14.; Berman etal. 2014:
fig.5.7b), Araeoscelidia (deBraga and Reisz 1995: fig.4;
Peabody 1952: fig.4; Reisz etal. 1984: fig.1), certain basal
Neodiapsida (Harris and Carroll 1977), and some ichnotaxa
such as Dromopus, and is retained in extant Lepidosauria. In
contrast, an angled fifth toe is apparently absent, for exam-
ple, from the feet of Paleothyris (Carroll 1969: fig.1A),
Millerettidae (Thommasen and Carroll 1981), Milosaurus
(deMar 1970: fig.6), and Sphenacodontia (Spindler 2015:
figs.4.16, 5.23, 5.32; Berman etal. 2004: fig.3A; Henrici
etal. 2005: fig.2).
In order to examine relative toe lengths, the ratio of string
I to IV and that of V to IV are plotted in a single diagram
(Fig.17m). Each primary measurement includes the meta-
tarsal plus associated phalangeal row. Both ratios are highly
variable and subject to specific adaptations. Interestingly,
there is a stable limit at around 47%, which is in no case
exceeded by string I relative to IV. At the same time, string V
F. Spindler etal.
1 3
never falls below 47% the length of string IV. There might be
further interdependencies apparently representing functional
constraints of a superior, plesiomorphic type of locomotion.
Taxa that are close to this boundary (Captorhinus and Eoca-
sea: toe I; Protorosaurus: toe I and V; Kuehneosuchus and
Araeoscelis: toe V) probably evolved until they reached a
certain limit on adaptation within a feasible morphospace.
Unfortunately, the fifth toe is incomplete in the holotype
of Cabarzia. What can be estimated are the ratios seen in
varanopids, Petrolacosaurus, and some basal reptiles. The
relative length of the first toe equals that in Aphelosaurus,
Milosaurus, and Mesenosaurus, marking the lower edge of
both the diapsid and the synapsid range. Highly dissimi-
lar toes indicate a special adaptation, except in the extreme
shortening of the first toe or elongation of the middle toes in
Eudibamus and Anthracodromeus, respectively. In total, the
ranges of certain amniote branches are similar. The relatively
short fifth string, which probably provides a varanopid syna-
pomorphy, is weakly supported but needs further confirma-
tion through the inclusion of a broader synapsid context. The
fifth string is not shorter than the second in early amniote
pedes except for Protorosaurus.
Müller and Reisz (2006) listed a short fifth metatarsal rel-
ative to the fourth as being a diagnostic feature of Diapsida.
They referred solely to Petrolacosaurus and Araeoscelis,
which do indeed occur around the lower edge of the range of
values seen in the majority of early amniotes (Fig.17n). The
araeoscelid mean of about 60% is shared only by Broomia,
Mesenosaurus, and Cabarzia. Lower outliers are represented
only by Sauria. The varanodontine ratios resemble basal rep-
tiles, revealing that the shortened fifth string is mainly an
indicator of the phalangeal string rather than the metatarsal.
Similar to the manus, Varanosaurus and Milosaurus show
some resemblance to Varanodontinae regarding their toe
A boundary of 40% for the relative length of the fourth
metatarsal compared to the toe length (Reisz etal. 2010:
character 111) was previously defined to distinguish cer-
tain conditions. A higher ratio was considered to be a clear
synapomorphy of both Varanodontinae (erroneously coded
for Varanodon by Reisz etal. 2010) and Archosauromor-
pha. Presumably, this wording referred to the toe includ-
ing the metatarsal. The revised spectrum shown in Fig.17o
was obtained by calculating the ratio of metatarsal IV to the
associated phalangeal sequence. With ratios of about 60%,
Cabarzia and Thuringothyris bear relatively long fourth
metatarsals that are only surpassed by Aphelosaurus, Sauria,
and Mesenosaurus as the only other known mesenosaurine.
The majority of the parareptile, remaining diapsid, varano-
dontine, and non-varanopid basal synapsid taxa are spread
over the same range.
Overall, the hind feet of Mesenosaurus and Cabarzia
fall within the known range for Varanopidae. Whereas
Cabarzia has retained some similarity to the slender feet of
Archaeovenator and basal reptiles, Mesenosaurus displays
a stout autopod that resembles varanodontines to a remark-
able extent. This includes the presence of broadened pha-
langes. As already stated for Aerosaurus (Langston and
Reisz 1981; Pelletier 2014), this character is affected by
ontogeny, with subadult individuals showing a somewhat
intermediate condition. Unfortunately, the phalanges of
other mesenosaurines are largely unknown. Undescribed
material from Oklahoma, tentatively assigned to Myctero-
saurus (Davis 2013), contains moderately broadened pha-
langes. In Microvaranops, the unquestionably adult stage
shows a few phalanges that are intermediate in terms of
width (Botha-Brink and Modesto 2009: fig.6). An unam-
biguous phylogenetic history of phalangeal width in vara-
nopids is yet to be defined.
One of the major differences between Cabarzia and
Mesenosaurus is that the latter bears long but barely
curved unguals (Figs.7d, g, 18). Its claw curvature is at the
lower end of the spectrum documented in pelycosaur-grade
synapsids (Maddin and Reisz 2007; Spindler etal. 2018).
The flexor tubercle is weakly developed, about as low as
in Varanops. In contrast, Cabarzia has strongly curved
terminal phalanges with a well-developed flexor tubercle.
The overall shape resembles that reported from the manus
of Tambacarnifex (Berman etal. 2014: fig.5.5a), mak-
ing it an exception compared to other known varanodon-
tines. The fact that most early reptiles have short terminal
Fig. 18a–d Terminal phalanges of varanopid synapsids. a Tambac-
arnifex unguifalcatus (redrawn from Berman etal. 2014: fig. 5.5a).
b Varanodon agilis, first manual digit, FMNH UR 986. c Second
(above) and fourth pedal ungual of NML-G2017/001, holotype of
Cabarzia trostheidei gen. et sp. nov. d Mesenosaurus romeri, dorsal
and lateral aspects. Not to scale
A new mesenosaurine from the lower Permian of Germany
1 3
phalanges supports the assignment of Cabarzia to basal
Synapsida. The phylogenetic history of ungual dimensions
is rather complicated due to the observation that Archaeo-
venator resembles araeoscelids and other early reptiles in
tall but shortened terminal phalanges. Several varanopid
claw types are present in the diverse assemblage from the
Dolese quarry, Oklahoma (Davis 2013).
Undetermined element—An indeterminate element is posi-
tioned next to the distal left humerus of NML-G2017/001.
It appears to be the impression of a distinctly shaped bone,
with the outermost bone substance left as a whitish coat.
An oval foramen is preserved but cannot be assigned to
the humerus. Its affiliation to the described skeleton is
Dierential determination
In conclusion, the basal amniote comparisons yield rather
clear results for Cabarzia, including rejections of reptil-
ian affiliations. Cabarzia trostheidei is well supported as a
member of the Mesenosaurinae, as it shows strong similar-
ity to Mesenosaurus but is distinguishable from it. Due to
the disparity of Varanopidae and our limited knowledge of
Mesenosaurinae, the affiliation of Cabarzia trostheidei to
this subgroup is more reliable than its affiliation to Varano-
pidae as a whole. According to the analysis of Spindler etal.
(2018, without including diapsids in the outgroup), features
that support a varanopid relationship within Synapsida are
(1) a large ilium with a shallow but distinct dorsal blade,
(2) a slender humerus, (3) long terminal phalanges, and (4)
a probable anterior inclination of the last presacral neural
spine. Due to imperfect preservation, the latter observa-
tion should not be used to drive the assignment of Cabarzia
trostheidei to the Varanopidae, but it is the only character
that does not homoplastically occur in certain reptiles. This
underlines the natural potential for confusion among slender
early amniotes and the need for broad comparisons. Further-
more, Cabarzia shows mesenosaurine affinities such as (1)
slightly swollen neural arches, in common with Meseno-
saurus, (2) probably holocephalous dorsal ribs, (3) the pres-
ence of a bulge on the anterior terminus of the pubis–ilium
suture, and (4) the distal position of the entepicondylar
foramen, which is similar to that in Mesenosaurus. Specific
autapomorphies of Cabarzia include (1) knob-like scars on
the dorsal surfaces of the sacral ribs, (2) a straight radius, (3)
an ulnare that is broader than long, and (4) long and curved
terminal phalanges. Within Mesenosaurinae, Mesenosaurus
shows at least two postcranial autapomorphies: (1) an offset
lobe on the ilium dorsal blade and (2) straight caudal ribs
that lack a posteriorly oriented distal portion.
Cabarzia is clearly distinct from Eudibamus. In terms of
common postcranial conditions of Bolosauria, little material
is known. However, Erpetonyx and Eudibamus, for example,
lack the rectangular dorsal spines and a broad entepicondyle,
categorically precluding Cabarzia from this clade. A miller-
ettid determination for Cabarzia can also be excluded on the
basis of several specific features, as well as observations that
conflict with the entire clade Millerettidae: (1) rectangular
dorsal neural spines; (2) a large ilium; (3) the presence of a
pubic tubercle; (4) slender toes; (5) the broad entepicondyle;
and (6) the tiny pisiform. Moreover, millerettids appear to
have been endemic to South Africa during the late Permian.
The spatial and stratigraphic distance from Cabarzia was not
used as an a priori reason for preclusion, given the overall
lizard-like proportions and the long ghost lineage resulting
from the basal position within Parareptilia (Ruta etal. 2011:
fig.1; Modesto etal. 2015: fig.2c). Cabarzia is undoubt-
edly distinct from all parareptilian clades. Based on current
knowledge, millerettids remain endemic.
Contrary to the preliminary determination (Martens
1992), an overwhelming number of arguments indicate
that Cabarzia is distinct from Araeoscelidia. Irrespective
of specific features, Cabarzia can be excluded from basal,
non-neodiapsid diapsids in general based on the following
postcervical observations: (1) dorsal neural arches that lack
lateral excavations; (2) a complete absence of mammillary
processes; (3) rectangular dorsal neural spines; (4) a lack of
alternation in spine height; (5) similar dimensions of the first
and second sacral ribs; (6) massive caudal ribs; (7) a large
ilium lacking a medial shelf; (8) a moderate pubic tubercle;
(9) a stout and relatively short forearm; (10) a stout humeral
shaft; (11) a broader distal portion of the humerus with a
pronounced entepicondyle; (12) the short ulnare; (13) a tiny
pisiform; (14) the square-cut shape of the intermedium car-
pale; (15) similar femoral condyles; (16) a compact astra-
galus; and (17) large ungual phalanges.
With respect to the body shape, we also considered a neo-
diapsid determination for Cabarzia. From an osteological
perspective, this hypothesis can be ruled out based on the
following features of the new specimen: (1) hook-shaped
sacral ribs; (2) a large ilium; (3) a broad entepicondyle; (4)
short forearm proportions; (5) a compact astragalus; (6) a
stout calcaneus; and (7) elongated terminal phalanges. The
oldest known Neodiapsida are based on cranial material
(Reisz etal. 2011; Modesto and Reisz 2002) that unfortu-
nately cannot be compared with Cabarzia.
Comments onfurther taxa
The compilation of early amniote limb elements (Figs.12,
13, 14, 15, 16) revealed a greater morphological diversity
than previously suggested by published comparisons that
based on only few representatives of certain subclades,
all the more due to a limited attention on postcranial
F. Spindler etal.
1 3
information. Future investigations may identify certain
ecotypes. At this point, we can highlight some taxonomic
considerations resulting from the above comparisons.
As already stated, bolosaurids have evolved high verte-
brae with broad neural arches, mainly constructed from
expanded, horizontal zygapophysial articulation facets, as
is also typical of more derived parareptiles (barely reflected
in the character list of Tsuji etal. 2012). This leads to a
superficial resemblance to diadectomorphs, from which
bolosaurid vertebrae can be distinguished by the low dor-
sal spines. Enigmatic seymourimorphs or diadectomorphs
from the middle Permian of Russia, such as Gnorhimosu-
chus and Timanosaurus—recently proposed to be large-
sized bolosaurids (Falconnet 2012)—remain questionable
(pers. obs. F.S.). However, the late occurrence of diadectids
raises doubt over the former determination. In comparison,
the supposed latest known diadectid Alveusdectes (Liu and
Bever 2015) appears to be a therapsid, based on the over-
looked presence of lower canines.
An unnamed protorothyridid has been described from the
Fort Sill locality in Oklahoma (Reisz 1980). Several well-
preserved long bones of disarticulated material cannot be
assigned to a single individual. Nonetheless, the size range
of these bones is apparently narrow. Based on proportions
and body size, its form resembles Anthracodromeus. This
contributes to the spectrum of amniotes within the range of
other protorothyridids discussed herein.
Protoclepsydrops can be considered a eureptile. Originally
suggested to be a representative of the earliest synapsids
(Carroll 1964; Reisz 1972), this idea was thrown into doubt
as knowledge of early amniote osteology increased, lead-
ing to an uncertain or “captorhinomorph” affiliation (Reisz
1986; Reisz and Modesto 1996). According to current
knowledge, there is no reason to assume a synapsid affili-
ation. Its ilium shows a typical rod-like dorsal process, in
which it resembles none of the known synapsids (Carroll
1964: fig.13). The limb proportions resemble those of Hylo-
nomus and other early Romeriida. The anterior position of
the small pineal foramen is seen in Protorothyrididae and
Captorhinidae (Carroll and Baird 1972; Clark and Carroll
1973) but not in synapsids, except for the markedly larger
foramen in Ianthodon (Kissel and Reisz 2004) and Casea-
sauria (Reisz etal. 2009). Its blade-like dorsal spines con-
trast with the triangular spines in Captorhinids, Paleothyris,
and Hylonomus (Carroll 1964, 1969), implying that Protoc-
lepsydrops is a protorothyridid. If correct, Protoclepsydrops
represents the oldest evidence of this group, which is a
trivial statement due to the fact that it is the oldest known
amniote, along with Hylonomus (Reisz and Modesto 1996:
fig.3). Its supposed phylogenetic position is not surpris-
ing, as Paleothyris from the younger Florence assemblage is
more basal than the Protorothyrididae but more derived than
Hylonomus; or, alternatively, Hylonomus and Paleothyris are
basal to the captorhinid branch, whereas Protorothyrididae
is a sister subgroup to Diapsida (Müller and Reisz 2006).
Early Diapsida
Two distinct species of Araeoscelis have been named but
have been bundled within the revised descriptions of Vaughn
(1955) and Reisz etal. (1984). Therefore, the type species A.
gracilis is shown in Fig.15. As there are minor deviations
in the limb material presented by these studies, ratios were
estimated for both sets of documentation (Online Resource 1
of the ESM and Fig.15). The comparison reveals an evolu-
tionary trend if the reconstruction following Vaughn (1955)
represents the stratigraphically older A. casei, which could
confirm its validity. A compilation of the limb elements pre-
sented by Vaughn (1955) indicates an intermediary condi-
tion between Petrolacosaurus (Peabody 1952; Reisz 1981)
and the Araeoscelis material depicted by Reisz etal. (1984)
with respect to the relative shortening of the humerus and
the size of the astragalus. In contrast, the fourth metatar-
sal is greatly elongated (Vaughn 1955: fig.14), as are the
tibia and fibula. In the latter character, this reconstruction
of Araeoscelis is the only basal amniote to show a longer
zygopod than its associated stylopod element. The material
described by Reisz etal. (1984) has an extreme femur to
humerus length ratio—even higher than in Youngina and
Kuehneosuchus. Perhaps two distinct forms of Araeoscelis
evolved distinctly adapted autapomorphies. A revision of A.
casei is recommended to address its taxonomic status and
relation to published reports.
Some basal diapsid taxa are based on single specimens
such as Aphelosaurus. Kadaliosaurus is considered to be
an araeoscelid, based on the short but strong terminal pha-
langes, the short manus, and the extremely elongated ante-
rior zygopods. Material assigned to the supposed araeosce-
lidian Dictybolos (Olson 1970) shows some resemblance
with respect to long bone slenderness, pelvis anatomy, and
dorsal spine alternation, but this classification was debted
(Evans 1988). Unfortunately, the Carboniferous Spinoaequa-
lis (deBraga and Reisz 1995) cannot be compared in many
aspects of the postcranium. Although some elements (such
as its interclavicle) resemble those of certain Araeoscelidia,
the holocephalous dorsal ribs suggest a neodiapsid affinity.
As already shown, the latter character has ungergone some
convergent evolution among early amniotes.
A new mesenosaurine from the lower Permian of Germany
1 3
Enigmatic forms
Tridentinosaurus from the lower Permian of Italy is cur-
rently under redescription (Leonardi 1959; Bernardi etal.
2014). Its general proportions resemble those of several
clades, such as basal diapsids, but also slender varanopids.
The presence of a considerable neck, although short, is a
feature shared, for example, with Mesenosaurus. As shown
for Cabarzia, imperfect specimens of a superficial lacertoid
appearance are generally affected by the mesenosaurine–rep-
tilian convergence. With respect to limb proportions, par-
ticularly the large manus, Tridentinosaurus exhibits a close
resemblance to Petrolacosaurus (Fig.15).
Carroll and Baird (1972) documented a large amni-
ote from the Carboniferous of the Czech Republic, MB
1901.1379 (note that Carroll and Baird 1972 give an incor-
rect specimen number in the caption for their fig.8; they
confused it with Gephyrostegus from Carroll 1970). The
authors noticed its synapsid size range but could not iden-
tify any diagnostic feature to support a pelycosaur-grade
determination. Based on the comparisons provided herein,
the elongated ungual phalanges combined with an enlarged
pisiform unequivocally suggest a basal synapsid. The over-
all structure of the carpus is surprisingly similar to that of
Ophiacodon (Romer and Price 1940: fig.52). Additionally,
there is an unnamed synapsid manus from the Pennsylvanian
of Linton, Ohio, that closely matches the pattern seen in MB
1901.1379 (Reisz 1975: fig.6).
The enigmatic synapsid Milosaurus (deMar 1970) was
introduced as a new varanopid, but was later rejected (Reisz
1986). Here, we suggest an ophiacodontid affinity, based
on extensive proportional similarities with Varanosaurus
(Fig.14; see Spindler etal. 2018 for a cladistic analysis).
As stated above, the articulated metatarsus (deMar 1970:
fig.6) provides further distinction from Varanopidae.
Evolutionary history oftheVaranopidae
Varanopidae have consistently been considered to repre-
sent a less-derived branch of pelycosaur-grade synapsids in
phylogenetic analyses (from Reisz 1986 to Benson 2012),
and first became known for the large varanodontines Vara-
nops (Romer and Price 1940; Campione and Reisz 2010),
Aerosaurus (Romer 1937; Langston and Reisz 1981), and
Varanodon (Olson 1965). Along with later assignments,
these potential apex predators are mainly known from the
Kungurian (Campione and Reisz 2010: fig.16), except for
the basalmost member Aerosaurus (from the Asselian).
Varanodontines form the sister group to mesenosaurines,
of which the late Cisuralian Mycterosaurus shares some
similarity with Aerosaurus, including the dentition pattern
and dorsal spine height. Other even more lightly built forms
come from the middle Permian of Russia (Mesenosaurus)
and South Africa (Afrothyra). Along with certain caseids,
mesenosaurines are the only basal synapsids that contribute
to the known therapsid communities (Kemp 2006: fig.3).
Varanopids show the longest temporal range and widest
geographic distribution of all early synapsids. Interestingly,
there is a ghost lineage from the Gzhelian basal varanopid
Archaeovenator (Reisz and Dilkes 2003) and the Asselian
varanodontine Aerosaurus to the Artinskian Mycterosau-
rus. Moreover, varanopids were found in North America,
South Africa, and Russia (Mesenosaurus and the basal form
Pyozia, Anderson and Reisz 2004). Recently, a new varano-
dontine from Thuringia has been described (Berman etal.
2014), which is the first varanopid from the early Permian
of Europe.
Cabarzia closes a gap in the spatiotemporal distribution
of known Varanopidae. It is the first mesenosaurine from
Central Europe, and at the same time the oldest known rep-
resentative. No other varanopid from the Asselian/Sakmar-
ian stages is known so far. Overall, neither its stratigraphic
position nor its locality are surprising. However, Cabarzia
pushes back the oldest known occurrence of low-spined,
slender varanopids with shortened forelimbs. Mycterosaurus
is incompletely known, but resembles the Varanodontinae in
its high dorsal spines, whereas most other features push the
genus closer to Mesenosaurus (such as shortened lacrimals).
Mycterosaurus may most strongly reflect the initial morphol-
ogy of neovaranopsians. Due to its overall resemblance to
Mesenosaurus, Cabarzia is expected to share some features
of its skull. A phylogenetic analysis of varanopid synapsids
confirmed that Cabarzia and Mesenosaurus form a clade
(Spindler etal. 2018).
The exclusion of Mycterosaurus from the former Mycter-
osaurinae, which was thus renamed the Mesenosaurinae, is
a hypothesis of Spindler etal. (2018). Since Mycterosaurus
apparently holds a key position in this issue, undescribed
new specimens from Oklahoma (Davis 2013; R. Reisz, pers.
comm.) are expected to have a major impact.
Relation totheichnological record
The interpretation of Cabarzia as a putative trackmaker is
subject to current investigation. The ichnogenus Dromopus
is usually linked to a reptilian producer, although doubt
over the commonly accepted araeoscelid nature has been
expressed (Voigt 2005; the author added the varanopid
Aerosaurus as well as the then-undescribed Cabarzia to a
list of potential lacertoid trackmakers). If this is partially
revised on the basis of diapsid–mesenosaurine similarities,
the subordinate role of Late Paleozoic reptiles in contrast
to the dominant contemporary basal synapsids (Kissel and
F. Spindler etal.
1 3
Reisz 2004) would gain further support. However, this most
significant impact of Cabarzia requires a detailed morpho-
metric analysis (Voigt etal., in prep.).
Preliminary results include a match between Cabarzia
and a body assigned to Dromopus from the Goldlauter For-
mation of the Cabarz quarry, briefly described by Martens
(1991). The steep increase in the lengths of toes I to IV, as
well as the angled toe V, fit with early diapsid reptiles and
mesenosaurine synapsids, whereas several basal reptiles can
be excluded with certainty (Figs.12, 13, 14, 15, 16). Addi-
tionally, the preserved skin imprint (Martens 1991: figs.1, 2)
resembles the epidermal scale pattern preserved in tiny vara-
nopids from the Petrified Forest of Chemnitz (Spindler etal.
2018). In conclusion, it is most likely that Dromopus reflects
a considerable component of mesenosaurine trackmakers.
Dromopus is recorded throughout the Permian (and Gzhe-
lian), raising the question of whether early diapsids and
Neodiapsida can be distinguished on an ichnological basis.
Both early diapsids and slender varanopids originated during
the Pennsylvanian, regardless of evidence from body fossils.
Therefore, no biostratigraphical support can be given for a
synpasid trackmaker of Dromopus. In contrast, the known
range of Tambachichnium tracks correlates perfectly with
that of Varanopidae, their widely accepted producers. How-
ever, the toes of Mesenosaurus resemble those of Varano-
dontinae in some aspects, whereas the rare evidence on other
slender varanopids, which has not been taken into account
up to now, implies a different ichnological expression.
Paleoecological andpaleobiological
Mesenosaurines have slender and strongly curved teeth.
Although they do not show the greatly enlarged temporal
openings seen in varanodontines, which apparently had an
extremely strong bite, the relatively large teeth of Meseno-
saurus (Reisz and Berman 2001: fig.2) could cause severe
wounds when piercing its prey. The overall proportions sug-
gest that Mesenosaurus and Cabarzia represent a guild of
highly agile subordinate predators in their communities.
In comparison to other pelycosaur-grade taxa, the pineal
foramen of slender Varanopidae is relatively large, in com-
mon with Aerosaurus, Caseasauria, and a few reptiles. The
large parietal eye probably implies either a strong day–night
rhythm or some kind of ‘warning system’ due to their eco-
logical role as potential prey for apex predators. Besides
these considerations, it is proposed that an adaptation for
scotopic vision and thus nocturnality is common among
carnivorous basal synapsids (Angielczyk and Schmitz
2014). This was concluded based on scleral elements in the
very large orbits of Microvaranops and Aerosaurus, which
additionally resemble those of Mesenosaurus (Reisz and
Berman 2001: fig.5) and probably Mycterosaurus (Berman
and Reisz 1982: fig.2). To summarize, meseosaurines were
highly adapted and successful secondary or tertiary con-
sumers of the lower to middle size range in early to middle
Permian terrestrial ecosystems.
Considering their different ungual shapes (Fig.18),
Cabarzia and Mesenosaurus appear to have occupied dis-
tinct ecological niches. Cabarzia and Tambacarnifex (Ber-
man etal. 2014) are the only varanopids with elongated but
also strongly curved terminal phalanges. These may have
been used to hold down prey or to cause additional wounds.
In contrast, the rather unbent and flattened unguals of Vara-
nops (Maddin and Reisz 2007: fig.4B) and Mesenosaurus
are only exceeded by those of derived caseids. So far, the
best functional interpretation for them is a digging adap-
tation, resembling the extant Tachyglossidae, Oryctero-
podidae, or certain mustelids, although it remains unclear
whether this allowed them to spend their lives burrowing or
whether digging was only performed when looking for food.
Maddin and Reisz (2007) recommended that this functional
interpretation of stretched claws should be tested by search-
ing for evidence based on muscle attachments. The well-
developed olecranon of Mesenosaurus, the strongest known
in varanopids, yields the insertation for the triceps and the
anconeus muscle (Haines 1939, 1950; Romer 1976). It is ini-
tially speculated that there might have been a powerful fore-
arm extension, which, along with the small body size, might
support the idea of a burrowing lifestyle. Admittedly, a large
olecranon is also found, for example, in ophiacodontids, in
combination with a rather different general morphology.
In stark contrast to the rather consistent morphology of
varanodontines, mesenosaurines vary greatly in their limb
proportions and their humeral or autopodial morphology,
implying a spectrum of different locomotive adaptations.
The broad ulnare of Cabarzia is unique among varanopids,
but is shared with aquatic to semiaquatic nonsaurian neo-
diapsids. So far, no Paleozoic lacertoid swimming tracks
have been found, apart from an isolated pes imprint from
cf. Dromopus in lacustrine black shales of the upper Ober-
hof Formation, near Tambach-Dietharz, Thuringian Forest
Basin (Fig.19). In any case, basal synapsids seem to have
been fully terrestrial based on their postcranial osteology
(Felice and Angielczyk 2014). Moreover, the proportions of
Cabarzia and Mesenosaurus present some superficial simi-
larities to those of the parareptiles Milleropsis and Eudiba-
mus (Fig.17). Since the latter was shown to display bipedal
adaptations (Berman etal. 2000b), the potential bipedal
abilities of other slender, cursorial amniotes are worth dis-
cussing. As a sidenote, vertical climbers and leapers can
show similar proportions, but they require woodland habi-
tats and curved claws as well as very small body sizes or
compensation via grasping autopodia. Therefore, this is not
A new mesenosaurine from the lower Permian of Germany
1 3
discussed as a specific adaptation here. In general, climbing
does not conflict with bipedal locomotion (e.g., in the extant
The functional indication of bipedalism in Eudibamus is
well supported (Berman etal. 2000b; Sumida etal. 2013).
Several of its adaptations are comparable to the osteologies
of certain mesenosaurines. (1) Mesenosaurinae resemble
certain bipeds in their general proportions, including long
hindlimbs, short forelimbs, slender trunk, and a long tail.
Proportions of extant bipedal reptiles listed by Berman
etal. (2000b) show that Cabarzia and Mesenosaurus are
well within their ranges of limb ratios (Fig.17). Cabarzia
even exceeds their hindlimb to trunk ratio, although the
measured distance may not be based on the same topology.
(2) Decreased asymmetry of the hindlimb is seen in basal
varanopids and mesenosaurines. This includes the above
comparison of femur distal slenderness (Fig.17i) and even
distal condyles. A less twisted plane of limb movement dur-
ing stride is also achived with a proximally slender fibula,
which is found at least in Mesenosaurus. (3) The loss of a
perforating foramen in the ankle, shared with Microvaranops
(Botha-Brink and Modesto 2009). (4) The tail, although not
known in distal portions, appears slender and low, with com-
pact vertebrae. This body portion was possibly well adapted
to act as a balancer during bipedal phases.
There are structures that clearly conflict with the more
differentiated bipedal adaptations of Eudibamus. For exam-
ple, the angled fifth toe of Cabarzia and Mesenosaurus is
typical of a ‘sprawling’ posture, as reported from Dromopus
for extant reptiles. The tarsus is much less compact than in
Eudibamus. On the other hand, Eudibamus shows a large,
proximally located internal trochanter that allowed for a
wide angle of femoral movement. This implies that bipedal-
ism does not necessarily require a full set of feasible derived
characters and can therefore evolve through convergent line-
ages along different morphological paths.
The late Permian early lepidosauromorph Lacertulus
was considered to be one of the oldest bipeds (Carroll
and Thompson 1982). Its femur to humerus ratio is 156%,
which does not resemble the corresponding ratio of any of
the taxa included here except for the later specimen of the
early Permian (Artinskian/Kungurian) Araeoscelis (Reisz
etal. 1984), and close to the range of basal neodiapsids,
mesenosaurines, and parareptiles, including the early Per-
mian (Artinskian/Kungurian) Eudibamus. The forelimb of
Lacertulus is roughly half the length of its hindlimb, most
closely resembling the same basal amniote branches. How-
ever, while some mesenosaurines permit an interpretation
similar to that for Lacertulus, applying this approach to iden-
tifying bipedalism would lead to a circular argument.
In terms of functional morphology, further support may
be supplied by the slender trunks of mesenosaurines. The
autapomorphic lobe on the ilium of the middle Permian
(Kazanian, ?Roadian) Mesenosaurus is probably shared with
Lacertulus (Carroll and Thompson 1982: fig.1), indicating
improved limb musculature. Along with the modified cau-
dal ribs of Mesenosaurus, this is poorly understood. If such
apomorphic structures are associated with modifications of
the locomotion style, the stratigraphically older early Per-
mian (Asselian/Sakmarian) Cabarzia would appear to be
less derived than Mesenosaurus.
Bipedalism—which is used as a model to discuss the
functional morphology of mesenosaurine postcrania—is
found in extant lepidosaurs. While few varanids show the
ability to stand on their hind extremities, locomotive biped-
alism can be observed in some Iguania, mainly Agamidae,
which bear relatively shortened forelimbs like Cabarzia and
Mesenosaurus. Clemente (2014) provided some biome-
chanical background to describe this passive bipedal model,
where the front of the body is lifted during acceleration.
Basically, apart from limb proportions, the only other con-
straint that can be examined in fossils is a rearward shift of
the center of body mass as a typical adaptation or exaptation
for bipedal locomotion (Clemente 2014: fig.3A). Meseno-
saurus and Cabarzia, and partially the late middle Permian
Microvaranops (Botha-Brink and Modesto 2009), are the
only Paleozoic synapsids that show, to some degree, this
rearward shift of the center of body mass as a requirement
for passive bipedalism. This is achieved or maintained by
the presence of (1) a short neck, as generally observed in the
included non-saurian taxa; (2) a relatively short and slender
trunk; (3) differently sized anterior and posterior extremities;
Fig. 19 Counter slabs of
NHMS-WP 3753, cf. Dromo-
pus, isolated pes imprint with
digits and probably weak
sole pad; lacustrine black
shales from Mösewegswiese,
Tambach-Dietharz, Thuringia
(upper Oberhof Formation),
collected in 2007. Scale bar
measures 1cm
F. Spindler etal.
1 3
and (4) a long and robust tail. In contrast to the latter aspect,
Eudibamus has remarkably narrow caudal vertebrae; this
may indicate that it evolved active bipedalism, facilitating
slow bipedal locomotion, which is beyond the interpretation
of Berman etal. (2000b). Alternatively, it could imply a
deviant postural mode during bipedal phases or a narrowing
of the caudals in favor of increased muscle volume.
The adaptive nature of bipedalism is not fully understood.
Clemente etal. (2008) raised the question of how optional
passive bipedalism can be. This distinction does not address
the question of the adaptive nature of bipedalism in general.
In any discussion of bipedalism, we recommend making a
clear distinction of terms. First, active bipedal locomotion
also allows for slow walking on the hindlegs, for which there
is no clear representative older than the Triassic. Passive
bipedalism, however, refers to an accelerative effect in the
stated taxa. On the other hand, the distinction between fac-
ultative (optional) and obligatory bipedalism is independent
of whether it is an active or passive mode, although there is
functional overlap.
Australian agamids have been investigated with respect to
their locomotive attributes. Clemente etal. (2008) concluded
that bipedalism has evolved as a consequence of accelera-
tion, although they observed that some lizards actively
attempt to run bipedally. Compared to quadrupedal locomo-
tion, there is no advantage such as increased speed or endur-
ance, suggesting an unknown advantage (Clemente et al.
2008). This may partially be explained by the observation
Fig. 20 Life restorations of the oldest known facultative bipeds, from
the stratigraphically oldest on the right to the youngest on the left:
Cabarzia trostheidei gen. et sp. nov. (Asselian/Sakmarian, Thur-
ingia); Eudibamus cursoris Berman etal., 2000b (Artinskian/Kungu-
rian, Thuringia); Lacertulus bipes Carroll and Thompson, 1982 (late
Permian, South Africa). Scale bar measures 5cm
A new mesenosaurine from the lower Permian of Germany
1 3
that three out of four species that were included in the men-
tioned study achieved significantly greater acceleration dur-
ing bipedal strides. Berman etal. (2000b) considered biped-
alism to be adaptive, stating that increased limb length and
therefore stride length permit the direct advantage of higher
speed. Although the forelimbs do in fact keep up with the
hindlimbs (Clemente 2014), one valid reason for biped loco-
motion might be the prevention of forelimb–hindlimb inter-
actions when maneuvering. Bipedalism may be an indirect
advantage of possessing elongated hindlimbs in combination
with a short trunk, as these proportions facilitate increased
velocity. Another adaptive pattern may be present in teiid
Scincomorpha, which have long trunks but perform bipe-
dalism during obstacle negotiation (Olberding etal. 2012).
Although speculative, active hunters of flying insects could
have gained some advantage through bipedalism, as it would
have allowed them to bypass limitations imposed by their
height and maneuverability. There are various valid reasons
why passive bipedalism is at least partially adaptive rather
than a pure exaptation or side effect, including the elevated
level of sight permitted by the raised trunk.
In conclusion, based on the comparisons provided above,
bipedalism can be assumed to have evolved in several early
amniote branches (Fig.20). None of the numerous known
lacertoid tracks have ever indicated bipedal locomotion
(a supposed specimen in the MNG in fact shows intersec-
tioning Dromopus imprints; S. Voigt, pers. comm. 2016).
Regardless of the intensive documentation of Permian
tracks, we reflect that bipedalism could be hidden from the
ichnological record when not performed on moist, track-pre-
serving grounds. Moreover, the assumed type of bipedalism
is a rarely used escape behavior that is not even preserved
in the rich track assemblages of the Tambach Formation
which have produced the bipedal and functionally tridactylic
Eudibamus. Likewise, ichnological support for bipedalism
in squamates in the fossil record is rather scarce, although it
has been demonstrated to be widespread in early representa-
tives (Lee etal. 2018). Taking into account that mesenosau-
rine body proportions markedly contrast with the trends and
ranges seen for other Paleozoic synapsids, an initial stage
of bipedal locomotion can be concluded. The oldest known
mesenosaurine Cabarzia trostheidei is also a candidate for
the oldest known facultative bipedal vertebrate on earth. No
other synapsids evolved facultative bipedalism until the ori-
gin of mammals (bipedally hopping Mesozoic mammalian
ichnia were described by Kim etal. 2017 and Stanford etal.
Acknowledgments We owe a great debt of gratitude to Frank Tros-
theide (Wolmirstedt and Magdeburg) for transporting, loaning, and
facilitating the public cataloging of the newly described specimen,
as well as for valuable discussions. The documenting carried out in
Moscow was greatly facilitated by Ilja Kogan (Freiberg), Evgenia
Sytchevskaya, Lev S. Schatenstein, and Valeriy Golubev (Moscow),
who provided travel assistance, accommodation, supplies, and collec-
tion management. We are grateful for our constructive exchanges with
Johannes Müller (Berlin), Michael Buchwitz (Magdeburg), Sebastian
Voigt (Thallichtenberg), Thomas Martens (Gotha), and Diane Scott and
Robert Reisz (Mississauga). Nico Schendel (Freiberg) kindly assisted
with the literature search for the limb plates. Marcel Hübner (Freiberg)
and Steffen Trümper (Chemnitz) provided images for the geological
map and outcrop documentation, respectively. Stephan Brauner (Frie-
drichroda) helped by contributing his immense and detailed knowledge
regarding local mining history, geology, and re-identifying the horizon
of the new holotype. Furthermore, we greatly appreciate the English
language help given by Eleonore Horlacher as well as the general fund-
ing provided by Michael Völker and Raimund Albersdörfer (Dino-
saurier Museum Altmühltal, Denkendorf). The manuscript profited
considerably from the kind reviews of Stuart Sumida, Neil Brockle-
hurst, and one anonymous expert. This research was supported by the
Deutsche Forschungsgemeinschaft (DFG grants SCHN 408/12, 20, 21,
and 22 to J.W.S.) as well as by the Russian Government as part of the
program ‘Competitive Growth of Kazan Federal University Among
World’s Leading Academic Centers.’ This work represents a contri-
bution toward the tasks of the IGCP Working Group on Nonmarine-
marine late Carboniferous, Permian and early Triassic correlations.
Anderson, J.S., and R.R. Reisz. 2004. Pyozia mesenensis, a new, small
varanopid (Synapsida, Eupelycosauria) from Russia: “Pelyco-
saur” diversity in the Middle Permian. Journal of Vertebr