Carpus and tarsus of Temnospondyli

The carpus of Eryops megacephalus and tarsus of Acheloma cumminsi known from complete and articulated individuals have provided the standard anatomy of these skeletal regions for temnospondyls. Restudy of the carpus of Eryops confirms the presence of only four digits, but refutes evidence for a prepollex, postminimus, and distal carpal 5. The supposed contact surface on centrale 1 for a prepollex is reinterpreted as part of the articulation for metacarpal 1 that includes distal carpal 1. Contrary to previous interpretations, a notch on the intermedium does not fit against the lateral corner of the radius. An articular surface on the distal end of the ulna thought previously to contact an absent postminimus fits against the ulnare. Preparation of the tarsus of the type specimen of Trematops milleri (junior synonym of Acheloma cumminsi ) and a previously undescribed crus and pes of Eryops finds no evidence for a pretarsale in either genus. Centrale 4 of the tarsus shares a similar rectangular shape with a wide contact for the tibiale among several temnospondyls whether terrestrial or aquatic. Limited flexibility of the carpus of Eryops and a strong palmar arch are probably weight-bearing features. A proximal-distal line of flexibility is present along the tibial side of the tarsus between the tibiale and centrale 4 and between centrale 2 and centrale 1. A phylogenetic analysis of Temnospondyli including new characters of the carpus and tarsus reveals considerable instability, highlighting the significance of Dendrerpeton acadianum , Balanerpeton woodi , Capetus palustris , and Iberospondylus schultzei .


INTRODUCTION
The autopodium consisting of the carpus (wrist) or tarsus (ankle) and the digits is the hallmark feature of the tetrapod limb. Bones of the carpus and tarsus articulate with each other, proximally with bones of the zeugopodium (radius and ulna/ tibia and fibula), and distally with the digits. In this location between the digits and zeugopodium, the carpus and tarsus in a quadruped often provide flexibility for flexion and extension and support of the body during locomotion. The evolutionary origin of the carpus and tarsus is not coincident with the origin of digits despite the evidence for homology between digits and distal radials in the extinct sarcopterygian fish Panderichthys (Boisvert et al. 2008), extant lungfish (Johanson et al. 2007), and conserved activity of enhancers of Hox genes for development of fins in fish and limbs in tetrapods (Gehrke et al. 2015). Proximal elements of the wrist (ulnare and intermedium) and ankle (fibulare and intermedium) are present in the more fish-like tetrapodomorphs (Andrews and Westoll 1970;Boisvert et al. 2008; Davis et al. 2004;Shubin et al. 2006) predating the appearance of tetrapod digits. A single element in the tarsus of Tiktaalik is a possible intermedium (Shubin et al. 2014). Only a single ossified bone identified as the intermedium is known for the carpus of the stem tetrapod Acanthostega (Coates 1996), but a second stem tetrapod Tulerpeton has an ossified ulnare and intermedium (Lebedev and Coates 1995). The carpus of Tulerpeton includes the radiale, another standard proximal carpal bone. The manus is largely unknown for Ichthyostega although one specimen has some apparent metacarpals (Callier et al. 2009). Acanthostega, Ichthyostega, and Tulerpeton have a fibulare, intermedium and a third proximal tarsal bone the tibiale (Coates 1996;Lebedev and Coates 1995). The more distal carpal and tarsal bones next to the metacarpals and metatarsals, respectively, are also present in early tetrapods. However, there is far more variability in the number of central bones typically known as the centralia, and their evolution evidently lagged behind that of the remainder of the carpus and the digits (Johanson et al. 2007). Subsequent evolution of the carpus and tarsus involved addition of centralia to a stable number of four. The presumably more structurally stabilized carpus and tarsus are unknown in many extinct post-Devonian tetrapods, either because these regions were not recovered or identified in fossils, or were cartilaginous at the time of death. Even if present, carpals and tarsals may be small, poorly ossified shapeless bones. The cartilaginous or incompletely ossified state of the carpus and tarsus is a consequence of delayed ossification. There is a proximal-distal sequence of ossification in the limbs, and the carpals and tarsals are some of the last regions to ossify, if they ossify at all (Fröbisch et al. 2010). The tarsus sometimes ossified before the carpus. Greererpeton burkemorani has a well-ossified tarsus in which individual bones can be recognized ) but the carpus, although ossified, is disarticulated and the unusual shapes of the elements severely limits meaningful comparisons with carpi of other tetrapods. Tarsals are present, but unidentified, in the stem-tetrapod Eucritta melanolimnetes, and there are no ossified carpals (Clack 2001). Fossils of the amphibamid Eoscopus have a complete tarsus that can be compared to the tarsus of Acheloma, but the carpus is incomplete, and many bones cannot be identified with certainty (Daly 1994). A well-ossified tarsus consisting of the fibulare, intermedium, tibiale, four centralia, and the first distal tarsal is known for the stem-amniote Proterogyrinus scheelei, but only two distal carpals are ossified (Holmes 1984). The fibulare, intermedium, tibiale, and two centralia are known for the tarsus of Archeria, but of the carpus, only the radiale is known (Romer 1957). Five tarsals including an intermedium and fibulare, but no carpals, are described for the stem-amniote Silvanrepeton miripedes (Ruta and Clack 2006). Specimens of Gephyrostegus bohemicus have a pair of carpals (possibly the radiale and intermedium) and a modified tarsus consisting of a partially fused tibiale and intermedium, fibulare, three centralia, and five distal tarsals (Carroll 1970;Rieppel 1993). Several tarsals are also known for Westlothiana lizziae (Smithson et al. 1994). Only a single tarsal, probably a fibulare, is known for the stem tetrapod Ossinodus pueri from the Early Carboniferous of Australia (Warren and Turner 2004). These examples demonstrate that while a wide diversity of early tetrapods have preserved carpal and tarsal elements, the tarsus is invariably better represented, and their precise morphologies and homologies are often uncertain. The best-known and indeed archetypical carpal and tarsal anatomies for early tetrapods are exhibited in the carpus of Eryops megacephalus and the tarsus of Acheloma cumminsi. Eryops has been the subject of numerous publications over the past hundred and fifty years covering aspects such as cranial anatomy (Sawin 1941), vertebrae and ribs (Moulton 1974), and the limbs and girdles (Gregory et al. 1923;Miner 1925;Pawley and Warren 2006). Of particular significance is a specimen of the forelimb of Eryops with a complete carpus described and best illustrated by Gregory et al. (1923) and Miner (1925). This single specimen (AMNH FARB 4186) consisting of an ulnare, intermedium, radiale, four centralia, and four distal carpals, established the general morphology of the carpus for temnospondyls. Similarly, a single complete specimen of the tarsus of the trematopid Trematops milleri, now recog-nized as a junior synonym of Acheloma cumminsi , laid the foundation for our understanding of the temnospondyl ankle (Schaeffer 1941) which became the generalized non-amniote tarsal pattern and the starting point for the origin of the amniote tarsus (Meyer and Anderson 2013;O'Keefe et al. 2006;Peabody 1951). Description of this specimen showed the temnospondyl tarsus to consist of a tibiale, intermedium, fibulare, four centralia, and five distal tarsals (Schaeffer 1941). The carpus and tarsus have provided little data for phylogenies of temnospondyls and other early tetrapods because they typically do not preserve or only a few poorly ossified elements are present. In order to address this issue, the following material will be reexamined: 1, the carpus of Eryops derived from the specimen described in Gregory et al. (1923); 2, the tarsus of Acheloma from the specimen described by Schaeffer (1941); and 3, the carpal and tarsal bones of Dissorophus multicinctus described briefly in . For the first time, the tarsus in the holotype of Trematops milleri, and isolated carpal and tarsal bones of Cacops aspidephorus will be described. New phylogenetic characters of the carpus and tarsus will be included in a phylogenetic analysis of Temnospondyli using the data matrix of . The node based definition of Temnospondyli given by  as the least inclusive clade of Edops and Mastodonsaurus will be followed.

MATERIALS
Directional Terms: Locations of articular facets on individual bones and sides of articulated limbs will be described using standard directional terms. Dorsal (extensor) refers to the front (upper) side of the manus and pes facing the vertebral column. Ventral (flexor) is the side facing the ground and refers to the palmar (manus) or plantar (pes) side. Proximal is closer to the attachment of the limb to the body and distal further away from this attachment. Medial (preaxial) and lateral (postaxial) refer to the sides with the first and last digit, respectively.

Ossification of Carpals and Tarsals: Bones compared
for any taxonomic, phylogenetic, or biomechanical interpretations should be at similar degrees of ossification. The typically unossified or marginally ossified carpus and tarsus of stereospondyls  likely correlated with a predominantly aquatic lifestyle severely restricting comparisons with other tetrapods. Larger and presumably fully mature individuals perhaps capable of some terrestrial locomotion have a variable number of ossified carpals and tarsals (Boy 1988;. These individuals offer the best chance for meaningful comparisons assuming the preserved bones reflect accurate relative size differences and accurate shapes.

Carpus
Ulnare: The ulnare of Eryops is a proximo-distally elongate bone contacting medially the ulna and the intermedium and distally distal carpal 4 (Figs. 1, 2). Its flattened proximal end projects laterally away from the carpus. A single slightly deformed right ulnare in a specimen of Dissorophus multicinctus (Figs. 2C, D, 3) has a similar shape as the ulnare of Eryops and similarly placed articular surfaces for the ulna, intermedium, and distal carpal 4. There is a slight rim along the dorsal edge of the contact surface between the ulnare and intermedium in Dissorophus. The ulnare of Acheloma cumminsi (Fig. 4) is very similar to that of Dissorophus. There are two distinct articular surfaces along the medial edge: one for the ulna and a second for the intermedium. The rim along the dorsal side of the contact surface for the intermedium is raised. The same edge on the ulnare of Dissorophus is also raised, but only slightly. Distal to the contact surface for the intermedium, the edge of the ulnare is concave as in Dissorophus.
Intermedium: The intermedium of Eryops has a deep V-shaped notch (Fig. 1A, B) interpreted by Gregory et al. (1923) as fitting against the lateral corner of the distal end of the radius, and as a consequence displacing the ulna proximally relative to the radius. This configuration requires the radius to have both a laterally facing articular surface and a distal articular surface for the intermedium. However, the radius of Eryops has only a distal articular surface for the intermedium (Pawley and Warren 2006). Furthermore, placement of the proximal carpal series into their natural position shifts the notch further from the corner of the radius, not closer ( Fig. 2A). While the biological significance of this notch is uncertain, it did not articulate with the radius.
Proximally, the intermedium contacts the distal end of the ulna and, contrary to Pawley and Warren (2006), has an extensive rather than narrow contact with the distal end of the radius. Medially, it contacts centrale 4. A narrow gap separating the intermedium and distal carpal 4 lies in approximately the same location as the opening restored between the intermedium and ulnare by Gregory et al. (1923). When the ulnare and intermedium articulate, the lateral side of the ulnare is oriented towards the caudal end of the body contributing to a pronounced palmar concavity of the carpus (Fig. 2B). The distal end of the ulna with its separate medial and lateral surfaces for the intermedium

55
and ulnare, respectively (Pawley and Warren 2006) also indicate a more caudal position for the ulnare. An intermedium could not be identified in specimens of Dissorophus (Fig. 3). The intermedium in Acheloma (Fig.  4) is a square bone wedged between the radius and ulna contacting, in addition to the radius and ulna, centrale 4 and the ulnare. It is a small bone comprising approximately one-half the length of the ulnare whereas the intermedium and ulnare are approximately equal in length in Eryops. In accordance with the relatively smaller size of the intermedium in Acheloma, there is a wide separation between the intermedium and distal carpal 4.

Centrale 4: Centrale 4 of
Eryops has six distinct articular facets (Fig. 2C). The largest is the proximal contact with the radius. Medially, centrale 4 meets the radiale and laterally the intermedium. Distally, a trio of smaller surfaces contact, in medial to lateral sequence, centrale 2, centrale 3, and distal carpal 4. The distal surfaces for centralia 2 and 3 are flat. Centrale 4 has a rectangular lateral extension that underlaps the ventral surface of distal carpal 4 ( Fig. 2A, C). Two surfaces of this lateral extension contact distal carpal 4: a convex surface on the lateral and ventral sides and a flat dorsal surface. Contact between centrale 4 and the radius crosses a sharply angled edge on the distal end of the radius that is a continuation of the radial flexor ridge (Pawley and Warren 2006). This edge divides the distal surface of the radius into two distinct facets: one for the radiale and most of centrale 4 and a second for the remainder of centrale 4 and the intermedium. These two articular regions are set apart by an angle of approximately 130° with the articular surface for the intermedium directed posteriorly as well as laterally. In order to fit across this edge, centrale 4 has a notched proximal surface. Centrale 4 in Dissorophus (Figs. 2D, 3) has clearly visible contacts with the ulna, radiale, centrale 2, centrale 3, and the intermedium as in Eryops, but a different association with distal carpal 4. The lateral extension of centrale 4 that underlies distal carpal 4 in Eryops is absent in centrale 4 of Dissorophus and there is no facet for distal carpal 4. A slightly concave non-articular surface on centrale 4 of Dissorophus between the contacts for the intermedium and centrale 3 is not seen in Eryops. This non-articular surface faces distal carpal 4, but is clearly separated from it. One other notable difference is the absence in centrale 4 of Dissorophus of the strong ventral curvature in Eryops. Centrale 4 in Acheloma is the largest of the centrale bones lying along the distal end of the radius between the radiale and intermedium (Fig. 4). Its distal margin is curved with a pair of distinct facets for centrale 2 and centrale 3 as in Eryops and Dissorophus.

Radiale:
The radiale is similar in Eryops (Figs. 1, 2A, B), Dissorophus (Figs. 2D, 3), and Acheloma (Fig. 4). It has a square outline with a slightly concave medial edge. The cross-sectional profile of the radiale of Eryops is wedgeshaped and tapers to the medial edge. The extensive contact with the radius in Eryops occupies most of the distal surface medial to the radial flexor ridge. Centrale 2: Centrale 2 in Eryops is a square block-shaped bone that contacts centrale 4 proximally, centrale 1 and the radiale medially, distal carpals 1 and 2 distally, and centrale 3 and distal carpal 4 laterally (Figs. 1, 2A). The dorsal surface is flat, but ventrally the surface has a deep concavity separating the rounded medial and lateral ends of the bone. This concavity is confluent with the sulcus on the intermedium. Unlike Eryops, centrale 2 of Dissorophus (Figs. 2D, 3) lacks separate facets for distal carpals 2 and 3. Centrale 2 in Acheloma has the same triangular shape (Fig. 4) as present in Dissorophus. The base of the triangle contacts centrale 1 and the radiale, and the tapering laterally portion contacts centrale 4 proximally and centrale 3 distally and laterally.

Centrale 3: Centrale 3 is the smallest of the carpals in
Eryops wedged between centralia 2 and 4 and distal carpals 2 and 3 (Figs. 1, 2A). Its dorsal face is rectangular with a longer medio-lateral dimension, contrary to the restoration of Gregory et al. (1923) in which the bone is shown as proximo-distally elongate. Centrale 3 could not be identified in either specimen of Dissorophus (Fig. 3). Centrale 3 in Acheloma is a diamond-shaped bone (Fig. 4) and a relatively larger carpal bone than in Eryops. In the former, centrale 3 is the same size as centrale 2 whereas in the latter it is much smaller.

Distal carpal 1: This is a rectangular bone in Eryops
with a tapering medio-distal corner (Figs. 1, 2A) and shares with centrale 1 a distal contact with metacarpal 1. Its proximal contact with centrale 2 is broad medio-laterally and dorso-ventrally. Distal carpal 1 is similar in Dissorophus (Figs. 2D, 3) and Eryops. Plaster obscures the lateral side of distal carpal 1 in Acheloma, but the bone is clearly tall and rectangular (Fig. 4).
Distal carpal 2: Distal carpal 2 in Eryops has a flat dorsal side and strongly convex ventral side (Figs. 1, 2A). The distal contact surface for metacarpal 2 is smaller than the proximal surface for centrale 2. A distal carpal 2 cannot be identified in specimens of Dissorophus (Fig. 3). Distal carpal 2 in Acheloma has a rectangular shape (Fig. 4).

Distal carpal 3: Distal carpal 3 in
Eryops has a suite of contact points for distal carpal 4 laterally, centralia 2 and 3 proximally, distal carpal 2 medially, and the third metacarpal distally. Its dorsal side is slightly concave and the ventral side strongly convex. A polygon-shaped bone next to distal carpal 4 in MCZ 4173 of Dissorophus (Fig. 3C,

Distal carpal 4: Distal carpal 4 shares a similar shape in
Eryops, Dissorophus, and Acheloma. It has a wider medial side contacting centrale 3 and centrale 4 and a tapering lateral end contacting the ulnare proximally and metacarpal 4 distally (Figs. 1-4). The dorsal side is concave in these genera and ventral side convex in Eryops.

Metacarpals
Two brief points regarding metacarpal 1 of Eryops are not apparent from the description or illustration in Gregory et al. (1923). First, metacarpal 1 is asymmetric along its proximo-distal axis (Fig. 1). The proximal end has a large flattened medial corner and a more typical slender rounded lateral corner. The distal end may also be asymmetric, but the medial corner is incomplete. Second, the proximal and distal articular surfaces are not parallel, but diverge at a small angle of approximately 10°. In conjunction with the more medially facing facets on centrale 1 and distal carpal 1 for metacarpal 1, there is a medial divergence of digit 1 whereas digits 2-4 either point anteriorly or antero-laterally.

Tarsus
Tibiale: The tibiale of Acheloma is rectangular with a slightly concave lateral edge (Fig. 5) rather than a relatively squat bone with a strongly concave lateral margin (Schaeffer 1941). The proximal and distal contact surfaces for the tibia and centrale 1, respectively, are extensive with the proximal surface inclined towards the center of the tarsus. The distal articular surface for centrale 1 is transverse. The contact surface on the tibiale for centrale 4 is concave, as observed by Schaeffer. A bone identified herein as the tibiale of Eryops is no longer embedded in matrix (Fig. 6C), but held in position by plaster. Hence, its identification is less certain than those in articulation. However, its shape and size relative to the tibia, centrale 4, and centrale 1 matches that of the tibiale in Acheloma, although its proximo-distal length relative to the width is smaller than the tibiale of Acheloma, and the length of the unfinished side facing centrale 4 matches the length of the corresponding side of centrale 4. The tibiale of Dissorophus is an elongate transversely narrow bone with a concave lateral side (Fig. 7).

Intermedium:
The proximo-distally elongate intermedium of Acheloma has a wedge-shaped cross-section with the thick side next to the tibia and the tapered edge next to the fibula (Fig. 5). Constriction of the central portion relative to the proximal and distal ends forms a depression on the dorsal side noted by Schaeffer (1941) and a ventral concavity. The ventral concavity is part of the plantar arch of the tarsus. There is a large articular surface on the intermedium that contacted the tibia. The lateral and medial edges of the intermedium are concave with the more pronounced of the two concavities on the lateral side. A gap between the intermedium and fibulare probably included a passageway for a perforating artery. The intermedium of Eryops (Fig. 6), like that of Acheloma, has a concave medial side, contact surfaces with the tibia and fibula separated by a non-articular surface, and a tapered lateral side contacting the fibulare. As in Acheloma, a perforating foramen, formed by complementary grooves on the medial edge of the intermedium and lateral edge of the fibulare, passes through the tarsus. A large facet on the tibial side of the intermedium articulates with the distal end of the tibia (Pawley and Warren 2006). A ventral depression, continuous with the concavity on the ventral surface of centrale 1, is shallow closest to the border with the fibula and deeper along the distal contact with centrale 4. The ventral surface of the fibulare bears a pair of ridges meeting distally to form a V flanked laterally by a third ridge. As in Acheloma, the fibular articulation occupies only the medial portion of its medial edge. The remaining free lateral projection of the fibulare bears a concavity that probably served as the attachment site for one or more of the flexor accessorius lateralis and medialis, extensor cruris et tarsi fibularis, and abductor digit minimi (Diogo and Tanaka 2014). Although the intermedium in specimen MCZ 4169 of Dissorophus is damaged (Fig. 7), it clearly has the same shape as the intermedium in Acheloma and Eryops including a notch in the side contacting the fibulare. This notch leads to a groove along the ventral side of the intermedium.

Fibulare:
The fibulare of Acheloma has a convex lateral margin ( Fig. 5) rather than flattened or slightly concave as drawn by Schaeffer (1941). The proximal end has a small contact surface for the fibula. A larger medially directed free edge forms part of the rounded medial edge. Medial and lateral edges are raised above the central region of the bone. Distally, the fibulare is V-shaped with distinct facets for distal tarsals 4 and 5. The ventral surface of the fibulare in Eryops (Fig. 6) has a pair of ridges meeting distally to form a V flanked laterally by a third ridge. As in Acheloma,

59
the fibular articulation occupies only the medial portion of its medial edge. The remaining free lateral projection of the fibulare bears a concavity that probably served as the attachment site for one or more of the flexor accessorius lateralis and medialis, extensor cruris et tarsi fibularis, and abductor digit minimi (Diogo and Tanaka 2014). The fibulare of Dissorophus (Fig. 7) is an elongate bone with raised margins bordering a concave ventral surface similar to that of Acheloma. It lacks the ridges found on the ventral side of the fibulare in Eryops. A shallow notch near the proximal corner next to the intermedium matches a similar notch in the fibulare in Eryops and Acheloma.

Centrale 4: Centrale 4 in
Acheloma is a rectangular bone with its greatest length along the transverse axis (Fig. 5). The lateral edge contacting the tibiale is wider proximo-distally and dorso-ventrally than the medial edge contacting 60 the fibulare. The articular surface for the tibiale continues onto the ventral side as a convex ridge. The lateral articular surface for the fibulare faces slightly ventrally and is also raised although less than that of the tibial side. Thus, the ventral curvature of the intermedium and the ventrally facing facets for the tibiale and fibulare form a deep transverse arch in the tarsus. The distal side of centrale 4 is V-shaped as noted by Schaeffer (1941), but the tip of the V is located closer to the lateral edge such that the size of the contact surface for centrale 2 is smaller than the other side of the V for centrale 3 and distal tarsal 4.
Proximally, centrale 4 in Acheloma has an extensive contact with the intermedium. The dorsal edge of this proximal contact is raised and there is a large dorsal exposure of the contact surface along the corner next to the tibia. The dorsal margin of the distal contact for distal tarsal 4 is also raised with a small dorsal exposure of the contact surface. The rounded ventral margin of centrale 4 for the tibiale continues dorsally suggesting a potentially large range of dorsoventral flexion along this joint. The shape of centrale 4 in Eryops (Fig. 6) matches that in Acheloma including the broadly rounded articular end for the tibiale (crushed in MCZ 7555), a ventral concavity that is continuous with the depression on the ventral side of the intermedium, narrower proximo-distal width along fibulare side than tibiale side, and a separation distally between a surface for distal tarsal 4 and centrale 3 and another surface for centrale 2. Centrale 4 in Dissorophus (Fig. 7) shares the same basic morphology of centrale 4 in Eryops and Acheloma. It is a

61
transversely rectangular bone with a concave ventral side formed by raised medial and lateral sides. The medial side for the tibiale is wide and the articular surface continues onto the ventral side. The proximo-distal width decreases towards the opposite end that contacts the fibulare. There is little distinction on the distal side of centrale 4 between the contact surface for centrale 2 and the common surface for centrale 3 and distal tarsal 4.
Centrale 1: Centrale 1 in Acheloma (Fig. 5) is proximo-distally elongated (Schaeffer 1941), but FMNH UC 640 does not show any evidence that it is incomplete medially as suggested by Schaeffer, who reconstructed a cartilaginous extension. Medial edge of centrale 1 contacts the entire lateral edge of centrale 2 and the proximal section of the lateral edge of distal tarsal 1. The distal end of centrale 1 is rounded. Schaeffer described a pretarsale in both FMNH UC 1756 and 640. A small bone next to centrale 1 in FMNH UC 640, possibly the bone mentioned by Schaeffer, appears to be a rib fragment. Centrale 1 has a similar shape in Eryops and Acheloma, but in contrast to the relatively smaller centrale 1 in Acheloma this tarsal is approximately the same size as the tibiale in Eryops (Fig.  6). Centrale 1 could not be identified in any specimen of Dissorophus (Fig. 7).
Centrale 2: Centrale 2 in Acheloma is second in size to centrale 4 (Fig. 5). It is more rectangular than shown by Schaeffer (1941) with a larger and straighter medial edge contacting centrale 3. The lateral edge for centrale 1 is dorso-ventrally convex and proximo-distally concave. Centrale 2 has a clearly defined articular surface for the tibiale. Centrale 2 in Eryops (Fig. 6) is rectangular with the proximo-distal length slightly greater than the medio-lateral length. As in Acheloma, the proximal end is expanded dorso-ventrally where it contacts the similarly expanded tibiale end of centrale 4. The medial-distal corner is extended between centrale 1 and distal tarsal 1. In articulation, centrale 2 is partially covered in dorsal view by the centrale 1 and distal tarsal 1. Centrale 2 in Dissorophus (Fig. 7) is a simple block of bone identified primarily by its location next to centrale 4.

Centrale 3: Centrale 3 is the smallest of the centralia in
Acheloma (Fig. 5) and Eryops (Fig. 6). Rather than triangular as described by Schaeffer (1941), centrale 3 in Acheloma (FMNH UC 640) has a square outline. Its medial and lateral contact surfaces with distal tarsal 4 and centrale 2, respectively, are proximo-distally straight with a slight convexity on the side with distal tarsal 4. The distal contact with distal tarsal 3 is transverse and the proximal contact with centrale 4 is slightly convex. Centrale 3 in Eryops, wedged between centrale 2 and distal tarsal 4, is more proximodistally elongate than in Acheloma. Centrale 3 could not be identified in any specimen of Dissorophus (Fig. 7).

Distal Tarsals:
Of the distal tarsals in Acheloma (Fig. 5), the fifth is the smallest, and the fourth is the largest; the remaining distal tarsals are of approximately equal size. The facet on distal tarsal 4 for centrale 3 is convex in FMNH UC 640 rather than concave as described by Schaeffer (1941). There is a distinct proximal facet that fits into the dorsally concave distal edge of centrale 4. Distal tarsal 5 is a wedge-shaped bone fitting between the fibulare and metatarsal 5. Preserved distal tarsals in MCZ 7555 of Eryops (Fig. 6) differ little from those in Acheloma. No distal tarsals can be found in specimens of Dissorophus (Fig. 7).

Carpal and Tarsal Bones of Cacops aspidephorus
FMNH UC 930 includes several isolated phalanges, carpals, and tarsals (Fig. 8). Their identification is less certain given a lack of association with a front limb or hind limb. A left fibulare (Fig. 8A) is identified by its close overall resemblance to the fibulare in Acheloma, Eryops, and Dissorophus. In each genus, the fibulare has an oval shape with separate articular surfaces for the intermedium and centrale 4. However, this separation between articular surfaces on the fibulare of Cacops occupies most of the medial margin whereas in Eryops, Dissorophus, and probably Acheloma there is a narrow gap in the form of a groove between these articular surfaces. A slender bone with one side consisting of largely smooth bone surface and the opposite with a roughened surface and a deep longitudinal groove is possibly an intermedium of the tarsus (Fig. 8B). It is wedge-shaped with the wider side covered by finished bone except for a region of unfinished bone continuous with an unfinished end. The proximal and distal ends are slightly expanded along the wider side. Assuming this bone is an intermedium, the edge with finished bone is the medial margin that faces the tibia and the region of unfinished bone at one end of the side with finished bone is a facet for the tibia. These features are shared with the tarsal intermedium of Eryops, Dissorophus, and Acheloma. An oval bone with one concave surface is identified tentatively as a left ulnare (Fig. 8C) based on similarities with the ulnare of Dissorophus. In both cases, the dorsal surface of the ulnare is smooth, slightly convex, and has a deep concavity along the medial edge. The ulnare of FMNH UC 930 has a broad concavity on its ventral surface. Unfortunately, this side of the ulnare is not exposed in specimen MCZ 4173 of Dissorophus, so cannot be compared. A small disc-shaped bone with concavities on both sides (Fig. 8D) is possibly a centrale 2 or 3 of the tarsus. It also has some resemblance to distal carpal 1 of Acheloma. Similarly, a pair of smaller circular bones (Fig. 8E) is likely two smaller centralia of either the carpus or tarsus.

Historical Importance and Previous Interpretations of AMNH FARB 4186
Eryops megacephalus was prominent in discussions during the first half of the twentieth century on the evolution of the manus in early tetrapods primarily because a single specimen, AMNH FARB 4186 (Fig. 1), was the best example of a nearly complete and articulated front limb in a Permo-Carboniferous tetrapod. The Devonian sarcopterygian fish Eusthenopteron and Sauripterus served as the ancestral forms closest to tetrapods in numerous theoretical discussions (e.g., Gregory 1915Gregory , 1935Gregory , 1949Gregory et al. 1923; Gregory and Raven 1941). Developmental studies on modern frogs and salamanders such as those by Steiner (1921Steiner ( , 1922 were also important in hypotheses of the origin of digits and the construction of the early tetrapod hand because early stages held clues for the construction of the first limbs. As noted by Clack (2009), different assumptions heavily influenced these hypotheses. One common assumption was the presence of an axis of bones in the fin or limb known as the metapterygial axis from which other elements branched. A specific set of carpals and a pentadactyl manus were also assumed to be present in the first limbs. The existence of additional digits (prepollex and postminimus) was assumed given their presence in the manus of frogs and salamanders (Huene 1922;Gregory et al. 1923;Steiner 1922). The key difficulties for hypotheses of early tetrapod limb evolution were where to place the metapterygial axis among the digits and the pattern of branches from this axis to form carpals and digits. Understanding the nature of the deformation to AMNH FARB 4186 is critical for proper interpretation of the number of digits and how the digits articulate with the carpals and the carpals with each other. Digits of AMNH FARB 4186 are twisted towards the radius and flattening has removed most of the palmar arch to the carpus (Gregory 63 et al. 1923). This flattening has displaced the ulnare from the ulna. Additional forces not described by Gregory et al. (1923) apparently acted on this specimen. As a result, the radiale, centrale 4, and intermedium are shifted laterally, indicated in ventral view by a separation of articular surfaces between the radiale and centrale 4 and the radius (Fig. 1 C, D). Centralia 1, 2, and 3 and distal carpal 4 have been rotated to expose portions of their proximal articular surfaces on the dorsal side of the carpus (Fig. 1A, B). This rotation is clear for centralia 1 and 2 and distal carpal 4, but not for centrale 3. The dorsally visible surface interpreted in this paper as the proximal articular surface was evidently interpreted by Gregory et al. (1923) as the dorsal non-articular side, and central 3 was drawn accordingly as a proximo-distally elongate bone. Distal carpal 3 is shifted dorsally, but not rotated, to expose its proximal surface. A medial shift of the digits, in particular the most medial preserved digit, is key to an interpretation of the number of digits in Eryops. In Cope's (1888) original description of AMNH FARB 4186, a gap between two digits distal to an element identified as carpale 2 (distal carpal 1 in this paper) was interpreted as indicating a missing digit 2 and a total of five digits, a conclusion endorsed by Gregory (1915). In a different interpretation offered by Huene (1922), the first digit was displaced towards the radius away from its expected contact with distal carpal 1 to lie against a carpal named the carpale praepollicis by Huene, mediale 1 by Gregory et al. (1923), and in this paper centrale 1. The remaining digits preserved their correct positions against the other distal carpals. Gregory et al. (1923), while criticizing other aspects of the interpretation of AMNH FARB 4186 by Huene (1922), agreed that the first digit was moved postmortem away from its proper articulation with distal carpal 1 and that Eryops has four rather than five digits in the manus.
Shifting placement of the first digit from the distal end of centrale 1 (Cope, 1888)  It is true that the distal end of centrale 1 is too small to accommodate the proximal end of metacarpal 1, but the metacarpal need not be restricted to a single carpal bone. In fact, the distal articular surface on centrale 1 observed by Gregory et al. (1923) is continuous with the adjacent articular surface on distal carpal 1 and their combined length equals the proximal width of the first metacarpal.
Thus, metacarpal 1 articulates with centrale 1 and the adjacent half of distal carpal 1 ( Fig. 2A), and there is no need to assume postmortem lateral displacement of metacarpal 1 as hypothesized by Huene (1922) and Gregory et al. (1923). Since metacarpal 1 occupies the entire articular facet on centrale 1, there is no room to accommodate a prepollex in Eryops.
Despite the presence of only four digits in the manus of Eryops, embryological data (e.g., Steiner 1922) on extant salamanders and frogs indicated pentadactyly as primitive for tetrapods. A flattened section of the lateral side of the ulnare of AMNH FARB 4186 supposedly provided evidence of an evolutionarily lost fifth digit. This hypothesized missing fifth digit of Eryops was initially represented by a distal carpal 5 (Gregory et al. 1923;Miner 1925), but in later publications (e.g., Gregory 1935: Gregory and Raven 1941), a small digit with two phalanges was restored in this position. Thus, Eryops with its 'lost' fifth digit provided indirect paleontological support of pentadactyly as ancestral for tetrapods, but direct evidence came from a specimen of the embolomere Diplovertebron (Watson 1926) with five preserved digits. Embolomeres were thought originally to be the earliest group of labyrinthodonts and ancestral to the rhachitomes that included Eryops (Watson 1926), but in more recent phylogenetic analyses (e.g., Ruta et al. 2003) they are recovered as stem amniotes. Polydactyly in Devonian tetrapods (Coates 1996;Coates and Clack 1990;Lebedev 1984;Lebedev and Coates 1995) suggests that a count of five digits is not plesiomorphic for tetrapods, but is a derived state that likely evolved once (Laurin 1998). The point within tetrapod evolution where digits were reduced from more than five to only five is ambiguous and there is no evidence that any clade including Temnospondyli was diagnosed by possession of five digits on the manus (Laurin 1998; Ruta et al. 2003). Consequently, five digits may be transitional (Coates 1996;Ruta et al. 2003), and there is little reason to necessarily interpret a facet on the ulnare as evidence of an ancestral fifth digit without considering other explanations. The classic reconstruction of the manus of Eryops (Gregory et al. 1923;Miner 1925) has the proximal end of the ulnare in full articulation with the distal end of the ulna, sharing the distal end of the latter with the intermedium. A laterally facing articular surface at the distal end of the ulna was interpreted as evidence for a postminimus digit that also articulated with the proximal end of the ulnare (Gregory et al. 1923;Miner 1925;Romer 1933Romer , 1945. If this interpretation is correct, there should be, in addition to the facet for a postminimus, distinct distal articular surfaces on the ulna for the intermedium and ulnare. However, the ulna of Eryops has only two distal surfaces, one facing medially for the intermedium and another facing laterally (the putative articulation for a postminimus) for the ulnare (Pawley and Warren 2006). Consequently, the flattened proximal end of the ulnare projects laterally away from the carpus ( Fig. 2A) and there is no evidence of a postminimus. Muscle attachment is an alternative explanation for the surfaces on the ulnare interpreted as contact for the postminimus and distal carpal 5. This is supported by the presence of roughened bone surface similar to the rugose attachment site of the pectoralis muscle on the humerus (Fig. 1C, D). Contact surfaces between carpals or between carpals and the antebrachium are much smoother. The proximal end of the ulnare is the attachment site for the extensor antebrachii et carpi ulnaris and flexor antebrachii et carpi ulnaris in Ambystoma mexicanum which is homologous to the extensor carpi ulnaris and flexor carpi ulnaris in other tetrapods (Diogo and Tanaka 2012). Miner (1925) restored the extensor carpi ulnaris and flexor carpi ulnaris with attachments to the ulnare in Eryops. The distal surface on the ulnare for distal carpal 5 (Gregory et al. 1923) is likely for the abductor digit minimi that is also homologous across other amphibians and amniotes.

Comparisons Among Temnospondyls
Carpus: Few comparisons are possible between taxa described in this paper (Acheloma, Eryops, Dissorophus, and Cacops) and other temnospondyls because the carpus is either absent or if present then consists of simple circular or oval pieces of bone identified solely by their position relative to each other and the antebrachium. Furthermore, reference to more distantly related taxa such as Tulerpeton (Lebedev and Coates 1995) does not help because they have fewer carpals relative to temnospondyls and their homologies are uncertain. One specimen of Dendrerpeton acadianum has a set of 10 carpals in articulation with the radius and ulna (Holmes et al. 1998). The bone identified as the ulnare is unusual in its possession of a large notch on the lateral side and square rather than rectangular shape. However, given the articulated state of the forelimb and carpals and the position of this bone next to the ulna, this identification is reasonably certain. The intermedium is a pentagon unlike the elongate rectangular intermedium in Eryops, Acheloma, and Dissorophus; however, its location between the radius and ulna supports this identification. Some of the other carpals in this specimen may be interpreted differently. A carpal next to the intermedium is identified as the radiale fused to centrale 2. However, in temnospondyls, the radiale does not articulate with the intermedium. Instead, centrale 4 separates the radiale and intermedium. This carpal is more likely centrale 4 and the radiale is missing. There are two possible interpretations of the much smaller portion thought to be a fused carpal. It may be a fused centrale and most likely centrale 3 because its smaller size relative to the carpal reinterpreted as centrale 4 and its lateral location next to the intermedium matches the size differences and relative locations of centralia 3 and 4 in Eryops. Alternatively, the supposed fused carpal may actually be the lateral extension of centrale 4 observed on centrale 4 of Eryops. The carpal identified as centrale 3 is more likely distal carpal 4 because distal carpal 4 is the only carpal to contact both the ulnare and intermedium in Eryops and Acheloma. Assuming this new identification is correct, distal carpal 4 of Dendrerpeton is similar to that of other temnospondyls in that the medial side has two separate facets for centrale 3 and distal carpal 3, and a tapering lateral side contacting the ulnare.
Identification of the remaining carpals of Dendrerpeton is problematic. A tiny bone next to those originally interpreted as a fused radiale and centrale 2 is identified as centrale 1 (Holmes et al. 1998). However, it is probably not centrale 1 because centrale 1 is larger relative to the other carpals in Eryops, Dissorophus, and Acheloma. If the larger proximal bone is actually centrale 4 with a lateral extension as in Eryops, then this bone would most likely be centrale 3. On the other hand, if the larger bone is a fused centrale 3 and centrale 4, then the small carpal is perhaps distal carpal 1. Only the pair of larger bones identified as distal carpals are probably correctly identified if the bone distal to the ulnare is distal carpal 4 and the tiny bone is distal carpal 1. The identity of the two smaller bones distal to the reinterpreted distal carpal 4 is uncertain; they are perhaps not part of the carpus. According to the various new interpretations presented, the radiale and centralia 1 and 2 are absent in this specimen and if the tiny carpal is centrale 3, distal carpal 1 is also missing. A specimen tentatively assigned to Archegosaurus decheni  Tarsus: The fibulare in Acheloma (Fig. 5), Eryops (Fig. 6), Dissorophus (Fig. 7), and Eoscopus lockardi (Daly 1994) is consistent in shape. It is a proximo-distally oval bone with a similar pattern of proximal facet for the fibula, medial facets for the intermedium and centrale 4, and distal facets for distal tarsals 4 and 5. The distal end tapers to a V-shape clearly separating the facets for distal tarsals 4 and 5. A notch or concavity is present along the medial edge in contact with the intermedium. Shape of the fibulare in Balanerpeton woodi varies from an irregular or circular outline to more proximo-distally elongate, but in all instances has a V-shaped distal end (Milner and Sequeira 1994). A partial pes of the dissorophoid Ecolsonia cutlerensis has an elongate fibulare with a slight V-shaped distal end . In contrast, the fibulare in stereospondyls such as Sclerocephalus haeuseri, S. nobilis, and a specimen of either Archegosaurus decheni or Cheliderpeton is broader with a transverse dimension equal to or greater than the proximo-distal length, lacks the distal V-shape, has little distinction between the facets for distal tarsals 4 and 5, and lacks any evidence of a notch or depression on the side contacting the intermedium (Boy 1988;). The fibulare of tetrapods outside of Temnospondyli is very similar to the fibulare of stereospondyls. For example, in the colosteid Greererpeton burkemorani, the fibulare is transversely oval without distinct facets for distal tarsals 4 and 5 ). The intermedium of Eryops, Acheloma, and Dissorophus is proximo-distally elongate with a concave medial edge of finished bone and a lateral edge with a groove or concavity that fits against a similar groove or concavity on the fibulare. The medial edge is wider than the lateral edge and there is a depression on the ventral side that lines up with a depression on centrale 4. Distal contact with the tibia faces medially and the proximal contact with the fibula faces proximally and laterally in all three taxa. The same morphology is present in the intermedium of Eoscopus, a specimen of Tersomius cf. T. texensis (Daly 1994), and Balanerpeton (Milner and Sequeira 1994;Fig. 14). The bone identified as a possible intermedium in Cacops has facets for the tibia and fibula in the same positions as on the intermedium of Eryops, Acheloma, and Dissorophus, but is more slender. The intermedium of the stereospondyls Sclerocephalus haeuseri, S. nobilis, and the tarsus attributed to either Archegosaurus decheni or Cheliderpeton is also proximo-distally elongate with similar facets for the tibia and fibula (Boy 1988;). The medial edge between the tibia and fibula is concave. An intermedium in the holotype of Cheliderpeton vranyi is a simple elongate bone between the fibula and tibia with only a small free edge (Werneberg and Steyer 2002). It is uncertain if the dorso-ventral width changes in a medial to lateral direction in these stereospondyls as in Eryops, Acheloma, and Dissorophus. Only a poorly preserved fibulare is known for Ecolsonia cutlerensis . It is elongate with a V-shaped distal end. The elongate diamond-shaped intermedium of Greererpeton has only a small free edge between the tibia and fibula ) unlike the larger concave free margin in temnospondyls. The tibiale of Eryops, Acheloma, Dissorophus, Eoscopus (Daly 1994), and Balanerpeton (Milner and Sequeira 1994) is proximo-distally elongate with a straight or concave medial margin. Length of the side in contact with centrale 4 is equal to the length of the corresponding contact side of centrale 4. Cheliderpeton vranyi may be an exception with a squat tibiale that is comparable in size to centrale 4 (Wernberg and Steyer 2002). In contrast, contact side of the tibiale for centrale 4 in stereospondyls is between one-half and three-quarters of the length of centrale 4 (Boy 1988;. The relatively shorter tibiale of stereospondyls would preclude contact between the tibiale and centrale 2, whereas in Eryops, Acheloma, Dissorophus, and Eoscopus this contact is present. The tibiale of non-temnospondyl tetrapods such as Greererpeton  and Proterogyrinus scheelei (Holmes 1984) is more similar to the tibiale in non-stereospondyl temnospondyls. Centrale 4 is the largest centrale in the tarsus of temnospondyls and has a consistent shape throughout this clade including the typically incompletely ossified tarsus of stereospondyls with the possible exception of Cheliderpeton. The length of the contact side for the tibiale is greater than the length of the contact side for the fibulare. Ventrally, centrale 4 is concave and formed a portion of a transverse arch of the pes. An element labeled as a proximal centrale in a partial pes of Ecolsonia is probably centrale 4 because it articulates with the intermedium and fibulare : fig. 12G). The illustration of this bone shows a curved side opposite the contact with the fibulare. This rounded side matches the shape of the side contacting the tibiale in other temnospondyls. Outside of Temnospondyli, centrale 4 in taxa such as Greererpeton and Proterogyrinus has a diamond rather than rectangular shape with only a shallow ventral concavity and lacks an enlarged rounded contact edge for the tibiale (Holmes 1984;. Centrale 1 is an elongate and narrow bone between the tibiale and distal tarsal 1 in Acheloma, Eryops, and Eoscopus (Daly 1994). In contrast, it has a square outline in stereospondyls (Boy 1988;). Centrale 2 deviates from a square outline by a small extension of the medial-distal corner between centrale 1 and distal tarsal 1 in Acheloma, Eoscopus (Daly 1994), and Balanerpeton (Milner and Sequeira 1994). The morphology of this corner of centrale 2 is uncertain in Dissorophus. There is an articular facet at the proximal-medial corner of centrale 2 for the tibiale in Eryops, Acheloma, and Eoscopus (Daly 1994 ; Fig. 15B). Centrale 3 has few details other than being the smallest of the centralia. A bone identified as a distal centrale in Ecolsonia ) is probably centrale 3 in view of its small size and its location between distal tarsal 4 and the large bone identified as centrale 4.
Distal tarsal 4 is the most distinctive of the five distal tarsals. It is the largest and contacts distal tarsals 3 and 5, centralia 3 and 4, and the fibulare in temnospondyls such as Acheloma, Eryops, Eoscopus (Daly 1994), Ecolsonia , and Sclerocephalus (Boy 1988; Schoch and Witzmann 2009). One apparent exception to this temnospondyl pattern is Balanerpeton in which distal tarsal 4 is reduced in size, and centrale 3 completely separates distal tarsal 4 and centrale 4 (Milner and Sequeira 1994). Similarly, a partial pes of either Archegosaurus decheni or Chelerpeton shows a large gap between distal tarsal 4 and centrale 2 mostly occupied by centrale 3 . Few details are known for distal tarsal 4 in non-temnospondyl tetrapods. However, in Proterogyrinus and Greererpeton centrale 3 is situated between distal tarsal 4 and centrale 2 precluding the contact present in the tarsus of many temnospondyls (Holmes 1984;). The pretarsale described by Schaeffer (1941) is absent in reasonably complete and articulated tarsi of Acheloma, Eryops, Eoscopus (Daly 1994), Sclerocephalus (Schoch and Witzmann 2009), and Archegosaurus , suggesting that, contra Schaeffer (1941), this element does not occur in any temnospondyl.

Functional Morphology of the Carpus and Tarsus in Temnospondyls
Carpus: As noted by Gregory et al. (1923) and Miner (1925, the carpus in Eryops is arched through a combination of a pronounced ventral curvature of centrale 4, a slight ventral curvature of the contact surfaces between the radiale, centrale 4, intermedium, and the radius, and the strongly angled distal surfaces on the ulna for the intermedium and ulnare (Figs. 1C, D, 2B). The arc formed by the four digits is not symmetrical along the midline of the antebrachium and carpus. Instead, there is a lateral orientation to the manus (Fig. 2B). Third and fourth digits are directed laterally, the second digit points anteriorly, and the first digit is directed medially. Movement between the antebrachium and proximal carpals in Eryops is limited. The flat contact surfaces between the radius and the radiale and centrale 4 fit closely. Any possible movement between the radius and centrale 4 is further constrained by the edge (represented by a dashed line in Fig.  2B) along the distal surface of the radius extending from the radial flexor ridge on the ventral side (Pawley and Warren 2006) to the point of contact between centrale 4 and the intermedium at the dorsal side. Although the intermedium does not lock against the corner of the radius as depicted by Gregory et al. (1923), the medially facing surface for the radius and laterally facing surface for the ulna preclude any significant movement between the antebrachium and intermedium. Furthermore, the curvature of the carpus places the articular surface on the intermedium for the ulna in a more ventral location relative to the radius (Fig. 2B), shifting the ulnare ventral to the intermedium. Thus, the morphology of the proximal carpus locked the intermedium to the radius and ulna. Proximal carpals are little more than distal extensions of the antebrachium. Limited movement is also indicated by the morphology of contacts between other carpals in Eryops. The proximal surfaces of centrale 2 and 3 and distal surface of centrale 4 are flat and fit closely, although a slight convexity to the distal surface of the radiale contacting the slightly concave proximal side of centrale 1 suggests some movement. Proximal sides of distal carpals 1-3 are flat as are the distal and lateral sides of centrale 2 that contacted these distal carpals. The distal side of the ulnare is convex and articulates with a slightly concave surface on distal carpal 4. Distal carpal 4 has complex contacts with distal carpal 3, centrale 4, the intermedium, and the ulnare indicating the possibility of movement between these bones. In particular, flexion is possible between distal carpal 4 and the lateral projection of centrale 4 with its convex dorsal, lateral, and ventral articular surfaces. The line of flexion between distal carpal 4 and centrale 4 continues distally between distal carpal 4 and distal carpal 3 where the curved side of distal carpal 4 lies against a curved side of distal carpal 3. Flexion is also possible between the rounded surface of centrale 3 and distal carpal 3. The pattern of these articular surfaces and their movement suggest greater movement on the ulnar than radial side of the carpus. The lateral projection of the ulnare suggests larger lever arms for the flexor antebrachii et carpi ulnaris and extensor antebrachii et carpi ulnaris to increase flexion and extension. Digit 4 may be capable of a greater range of flexion and extension than the other three digits because there is a transverse plane of movement between metacarpal 4 and distal carpal 4 and another between distal carpal 4 and the intermedium. This greater mobility of digit 4 may include abduction by the abductor digiti minimi that given the more ventral position of the ulnare could produce posterior and lateral movement of the digit (Diogo and Tanaka 2012). Significant flexion and extension of the first three digits was probably limited to the carpal-metacarpal joint because the carpals on the radial side fit more closely. It's also clear that there is no transverse line of flexion across the carpus because any attempt to draw a transverse line would produce an undulating line between carpals and include some flat and closely fitted articular surfaces. The only clear line of flexure that transverses the width of the carpus is the distal carpal-metacarpal joints.
Convex articular surfaces of the carpals and distal ends of the radius and ulna of Dissorophus suggest more movement between carpals and between the antebrachium and carpus than in Eryops. The different shapes of centrale 4 in Eryops and Dissorophus and absence of the lateral projection on centrale 4 of Eryops also imply a greater range of carpal flexure. One point of similarity between the carpi of Eryops and Dissorophus is the curvature between the antebrachium and proximal carpals that would contribute to a ventrally concave carpus. However, two observations indicate a shallower ventral curvature in Dissorophus. Centrale 4 of Dissorophus lacks the pronounced ventral concavity that in centrale 4 of Eryops contributes to the carpal arch. In addition, the radial flexor ridge of the radius is less prominent in Dissorophus than Eryops indicating a flatter articulation between the radius and the proximal carpals. Lacking a full complement of carpal bones, the degree of curvature is uncertain but clearly less than in Eryops. The strong proximal curvature of the carpus in Eryops can be interpreted as an adaptation for supporting the body mass of this much larger temnospondyl. A tubular-shaped object, even if only a partial tube as in the arched carpus of Eryops, is better able to support compressive stress than a bar because it has a larger second moment of area (Wainwright et al. 1982).
Tarsus: There is one striking difference between the carpus and tarsus of temnospondyls as exemplified by Eryops. The distal articular surfaces of the radius and ulna are oriented along a largely transverse line despite the pronounced curvature. However, the distal articular surfaces on the tibia for the tibiale and intermedium are aligned in a dorso-ventral direction (Pawley and Warren, 2006) with the facet for the tibiale displaced towards the plantar side of the pes relative to the transverse articular facets at the distal end of the fibula for the intermedium and fibulare. As a consequence, the tarsus of Eryops, like the carpus, has a pronounced ventral arch. However, rather than an arch equally spanning the width as in the carpus, the plantar arch of the tarsus is greatest on the tibial side along a proximal-distal line separating the tibiale and centrale 1 on the medial side from centralia 2 and 4 on the lateral side (Fig. 6E). The arch is continued towards the fibular side by the ventral curvature of centrale 4 and the intermedium. Articular surfaces on centrale 4 and centrale 2 for the tibiale and centrale 1, respectively, extend onto the ventral surfaces and indicate a line of potentially significant movement along this proximo-distal line. Dorso-ventral flexion of the tibiale against the tibia is possible given the relatively larger articular surface on the tibia for the proximal end of the tibiale. The tibiale and centrale 1 may have acted as a unit to move along their contacts with centralia 2 and 4. Schaeffer (1941) identified two transverse planes of flexure in the tarsus of Acheloma: one between the fibulare, cen-trale 4, and tibiale proximally and centralia 1-3 and distal tarsals 4 and 5 distally, and the other between the distal tarsals and the metatarsals. In Schaeffer's (1941: fig. 1) reconstruction, the V-shaped distal side of centrale 4 would limit flexure along this first plane because it could force centralia 2 and 3 and distal tarsal 4 together. Some movement would be possible between the fibulare and distal tarsal 4 and 5 and between the tibiale and centrale 1. However, in the new reconstruction (Fig. 5E), the distal surface on centrale 4 is flatter, producing a straighter transverse line though the tarsus that suggests little or no resistance to flexion. Convex articular surfaces of the tarsals along this plane such as between centrale 1 and the tibiale, centrale 2 and centrale 4, and distal tarsal 4 and the fibulare support flexure and extension along this hinge-like line. Loosely fitting tarsals with rounded articular surfaces implies movement between many individual tarsals. Schaeffer (1941) commented that the tarsal bones of Eryops have a more cuboidal shape and a more direct contact with each other unlike the tarsals of the smaller Acheloma. This difference was interpreted as greater flexibility in the tarsus of Acheloma than in Eryops. A more consolidated tarsus in a larger temnospondyl such as Eryops could be a weight-bearing feature. Reduction in flexibility with increasing body size may be true for the carpus when one compares Eryops and the smaller Dissorophus and Acheloma, but not for the tarsus where the shapes of tarsals and extent of their articulations are similar regardless of body size and suggest roughly equal ranges of flexibility. The more distally positioned end of the tibia relative to the fibula precludes a transverse plane of flexure at the crural-tarsal junction (Schaeffer 1941), even if the tibiale, intermedium, and fibulare formed a transverse series, although these proximal tarsals could move individually against the crus. This individual movement was greatest between the tibiale and tibia and least between the intermedium and fibula. Both genera share a proximo-distal line of flexure on the fibular side of the tarsus with extensive ventral articular surfaces on centralia 2 and 4 for centrale 1 and the tibiale, respectively. Both genera also share a transverse distal tarsal-metatarsal joint.

Implications of Carpal and Tarsal Anatomies for Phylogeny of Temnospondyli
With few exceptions (e.g., Ruta and Clack 2006), the manus and pes, in particular the carpus and tarsus, have contributed little to phylogenetic analyses of temnospondyls and other early tetrapods and tetrapodomorphs (Coates 1996; Laurin and Soler-Gijón 2006; Ruta et al. 2003;. Given the scant data in the literature on the temnospondyl carpus and tarsus limited largely to the classic papers on Eryops (Gregory et al. 1923) and Acheloma (Schaeffer 1941), it has been difficult to identify potential synapomorphies. New data on the carpus and tarsus of Eryops, Acheloma, and Dissorophus in this paper and recent descriptions of other temnospondyls such as Eoscopus (Daly 1994) and stereospondyls  provide an opportunity to address this deficiency.

Analysis and Results:
The full data set in Schoch (2013) of 72 taxa and 212 characters was modified by adding Eoscopus, combining characters 33 and 34, rescoring of several entries, and addition of nine new characters (Fig.  9). Details of these modifications are found in Appendix 1. A list of all characters used in both analyses is provided and Sangaia lavina. Many of the eliminated taxa are stereospondyls for which carpal and tarsal characters cannot be scored presently, and include some of the problematic taxa removed by Schoch in his small dataset. Schoch also eliminated Greererpeton, Capetus, and Iberospondylus in his small dataset, but these taxa are retained in my analysis because the first taxon can be scored for several of the new carpal and tarsal characters and the latter two taxa are potential early temnospondyls. The data matrix has been submitted to Morphobank [http://morphobank.org/index. php/Projects/ProjectOverview/project_id/2289]. The matrix was analyzed in PAUP* 4.0b10 (Swofford 2002). Characters 66,74,96,109,157,169, and 181 were ordered in both analyses because the states in each character formed a progressive sequence. A heuristic search was used with parsimony as the optimality criterion, multistate taxa viewed as polymorphism, stepwise addition for starting trees, random addition sequence with 1000 replicates, and tree-bisection-reconnection selected as the branch-swapping algorithm. ACCTRAN was chosen for character state optimization. The search with the full data matrix of 73 taxa and 220 characters found 40 equally parsimonious trees of 717 steps with a CI of 0.3389 and RI of 0.7902. The strict consensus tree is shown in Figure 10. Analysis of the smaller data matrix of 48 taxa recovered eight most parsimonious trees of 542 steps with a CI of 0.4225 and RI of 0.7709. The strict consensus tree is shown in Figure 11A. Results from both data sets are broadly comparable to the results of Schoch (2013) with the notable exception of less resolution at the base of Temnospondyli. In both results, Iberospondylus, Capetus, the clade of Dendrerpeton and Balanerpeton, and Edopoidea form part of a polytomy. In the results from the larger data set, there is an additional polytomy of the remaining individual taxa and major groups such as Dissorophoidea, Zatracheidae, Dvinosauria, Rhinosuchidae, a clade of stereospondyls minus rhinosuchids, and Eryopidae. Resolution is improved with the smaller data set although there remains a polytomy of Iberospondylus, Capetus, the clade of Dendrerpeton and Balanerpeton, Edopoidea, and similar to the results of Schoch (2013), Eryopiformes, and the clade of Dvinosauria, Zatracheidae and Dissorophoidea. A third analysis used the data matrix from the second analysis and also removed Iberospondylus and Capetus. Parameters of the analysis were the same as those in the first two analyses. A strict consensus tree of four most parsimonious trees (Fig. 11B) shows improved resolution. The trees have a length of 534 steps, CI of 0.4288 and RI of 0.7734. The strict consensus tree has the same basic relationships found in  with the exception of Amphibamidae. These differences extend beyond the inclusion of Eoscopus that, as in Sigurdsen and Bolt (2010), is the most basal amphibamid in this study.  found Doleserpeton and Gerobatrachus to form a clade with Amphibamus, Platyrhinops, and Micropholis successively more distantly related sister taxa. In sharp contrast, this study recovered a pair of monophyletic groupings consisting of Amphibamus + Doleserpeton and Gerobatrachus + Micropholis that in turn form a monophyletic group with Platyrhinops and Eoscopus as successive sister taxa to this group. However, support for the alternative relationships in this paper is weak with a bootstrap value of greater than 50% only found in one node and Bremer support values of only one or two. Support for Amphibamidae is slightly stronger than in Schoch (2013) with a pair of unambiguous characters (89 and 128) of which one (128 -palatine and ectopterygoid as narrow as maxilla) is a synapomorphy. Presence of pedicellate teeth is a synapomorphy of the clade of Amphibamus and Doleserpeton only due to the scoring of pedicely in Gerobatrachus as unknown following Marjanović and Lauren (2008).

Discussion:
Despite the reduction in number of polytomies between the full data set of 73 taxa and the reduced set of 48 in the second data set, both results demonstrate considerable uncertainty among basal temnospondyls. Iberospondylus and Capetus are wild card taxa, as also identified by , and in this analysis are responsible for the failure to resolve the base of Temnospondyli. Of the eight most parsimonious trees found using 48 taxa, Capetus and Iberospondylus are edopoids in two and eryopids in two others. In another pair of trees, Capetus is an edopoid and Iberospondylus is an eryopid. Finally, in the last two trees, Iberospondylus is an eryopid and Capetus falls outside of Temnospondyli along with dendrerpetontids. In the majority of trees, Iberospondylus is within Eryopidae a place higher in the cladogram than postulated by . Capetus is more likely either an edopoid (four of the eight trees) or eryopid (two of eight trees), but the existence of two trees where Capetus along with Dendrerpeton and Balanerpeton fall outside of Temnospondyli reinforces the instability of basal temnospondyl relationships. The tree derived from the third analysis is the basis for discussion of the new characters of the carpus and tarsus. Only two of the nine characters, both in the tarsus, unambiguously diagnosis clades (Fig. 12). The derived state of character 215, unequal lengths of mutual contact surfaces between the tibiale and centrale 4 reflecting the relatively smaller size of the tibiale (Fig. 9F), is diagnostic of Stereospondylomorpha (sensu Schoch 2013). It may be argued that the smaller size of the tibiale is merely evidence of the expected limited ossification of a stereospondyl tarsus, but it leaves unanswered why the other tarsals are more fully ossified. For example, shapes of centrale 4 and intermedium are similar in Archegosaurus, Sclerocephalus, Acheloma, and Eryops. Distal tarsal 4 is larger relative to distal tarsal 3 in these same genera. The close proximity of many tarsals, in particular Sclerocephalus (Boy 1988; Schoch and Witzmann 2009), suggests little room for cartilaginous portions. For these reasons, it is concluded that the relative size of the tibiale as preserved is accurate. The derived state of character 217, fibulare with latero-medial width less than proximo-distal height (Fig. 9E), is diagnostic of Rhachitomi (sensu Schoch 2013). There is a reversal to the plesiomorphic state in Archegosaurus and Uranocentrodon.

CONCLUSIONS
Considerable anatomical detail is present in the individual bones of the temnospondyl carpus and tarsus providing data for functional and phylogenetic analyses. However, unlike many other bones of the vertebrate skeleton that can often be identified in isolation, few carpals and tarsals can be identified outside of an articulated specimen. Bones such as the ulnare, fibulare, and the intermedium and centralia 1 and 4 of the tarsus have distinctive and consistent shapes, but others such as the smaller centralia and distal carpals and tarsals have more variable shapes (Figs. 5,6). In light of the relative rarity of finding a preserved autopodium with an articulated carpus or tarsus in an advanced state of ossification, it's not surprising that most of the new carpal and tarsal characters introduced in this study provide only ambiguous support for any clade and do not alter any basic relationships established in . Would discovery of additional specimens provide the often-cited solution to missing data? A developmental pathway producing robust terrestrial adults is key because the carpus and tarsus are more likely to be ossified in terrestrial than in fully aquatic taxa (Schoch 2009). Dissorophids, trematopids, and amphibamids have steeper growth trajectories than aquatic taxa, and can be expected to furnish future discoveries of well-ossified carpals and tarsals. Smaller amphibamids such as Doleserpeton annectens may be problematic because their carpals and tarsals, while ossified, lack the distinctive shapes seen in dissorophids and trematopids . Other amphibamids such as Eoscopus lockardi have at least the tarsus sufficiently ossified to reveal details in individual bones for comparison with other temnospondyls (Daly 1994). Even among aquatic taxa, developmental plasticity can produce an altered pathway forming more terrestrial adults (Schoch 2014). Populations of Sclerocephalus in environmentally different lake systems can develop into large or small fully aquatic adults with poorly ossified postcranial skeletons or occasionally large terrestrial adults with robust postcranial skeletons including well-formed carpus and tarsus. In contrast, most stereospondyls remained aquatic throughout their lives and either lacked any trace of an ossified carpus or tarsus or only formed a few bones. These taxa are less likely to reveal more details of their autopodia. To better understand the origin of Temnospondyli, defined as the least inclusive clade containing Edops and Mastodonsaurus , focus should be directed on the relationships of Dendrerpeton, Balanerpeton, Capetus, and Iberospondylus to edopoids and other temnospondyls and whether any of these four genera are actually temnospondyls. Knowledge of the carpus and tarsus in these taxa would also help to clarify the plesiomorphic anatomy of the wrist and ankle in temnospondyls that, in turn, could help determine whether temnospondyls are plesiomorphically amphibious or aquatic. Figure 11B to show distribution of new carpal and tarsal characters.

APPENDIX 1
Modifications to data matrix of Schoch (2013) with original numbering sequence of that paper. Character 26. Replaced by new character. The old character concerned the location of the orbits relative to the midline of the skull as an expression of the width of the jugals. Wide jugals for laterally placed orbits and narrow jugals for medially placed orbits. However, there is no quantification of jugal height to allow determination as either wide or narrow. The new character expresses height of the jugal relative to the height of the maxilla beneath the orbit. A jugal with a height greater than that of the maxilla would be equivalent to a wide jugal in the original character. Character 10 for Acroplous changed from 1 to 0 because Englehorn et al. (2008) state that the alary process is absent. It is scored as present in data matrix of Englehorn et al. (2008), but scored as ? by . Character 13 wording changed by deletion of "separating the two alary processes" because presence of alary processes not required (e.g., Acroplous). This character is changed for Acroplous from 0 to 1 because Englehorn et al. (2008) show a large foramen identified as an anterior dorsal fenestra along the midline between the premaxillae. Character 13 changed for Dissorophus from 0 to 1 because DeMar (1968) is incorrect. MCZ 4170 has a small fenestra. Character 13 changed for Acheloma from 0 to 1 because the fenestra is present (Dilkes and Reisz 1987; Polley and Reisz 2011). Character 28 changed for Balanerpeton to 0 and1 according to . Character 28 changed for Dissorophus from ? to 1 because palpebral ossifications are present in MCZ 4186. Character 29 changed from 1 to 0 for Dissorophus. All skulls have some distortion, but the type FMNH UC 648 has an interorbital width equal to orbital width. If Schoch (2013) is relying on DeMar (1968: text- Fig. 2), then this drawing is incorrect. The skull is crushed which has reduced the width of the orbits and the upper orbital border for the right orbit is incomplete. Character 31 changed from 0 to 1 for Cacops based on Cacops morrisi. Character 33 is deleted and character 34 rewritten to original wording in character 32 of Englehorn et al. (2008). Character 34 of Schoch (2013) does not differentiate in state 0 between those taxa with sulci and those lacking sulci. Once this distinction is made, character 33 becomes redundant. Also, resolves the problem that numerous taxa are coded as 0 in character 35 that lack sulci. The old character 34 is now character 33. Character 33 changed from absence to presence for Nigerpeton because Steyer et al. (2006) describe lateral line sulci as present in adults. Character 33 changed from absence to presence for Iberospondylus because Lauren and Soler-Gijón (2006) describe lateral line sulci as present in adults. Character 33 changed from 1 to ? for Gerobatrachus following scoring in Anderson et al. (2008) where the only known specimen is exposed only in ventral view. Figure 13. Strict consensus tree presented in Figure 11B with labeled nodes. Unambiguous characters diagnostic of each labeled node are listed in Appendix 3 .