Multiple tooth-rowed captorhinids from the Early Permian fissure fills of the Bally Mountain Locality of Oklahoma

Captorhinids were Paleozoic eureptiles that originated in the Late Pennsylvanian in Laurasia and dispersed across the major landmasses of Pangaea by the Late Permian. Their evolutionary success as omnivorous and herbivorous members of Permian terrestrial communities has been attributed to the evolution of multiple marginal tooth rows. Multiple tooth rows evolved at least twice within Captorhinidae: once in the omnivorous Captorhinus aguti and again in the diverse subfamily of herbivorous moradisaurines. The earliest known moradisaurines co-occured with C. aguti in Lower Permian strata of Texas; however C. aguti is also known from much older fissure fills in the famous Dolese Brothers quarry near Richards Spur, Oklahoma, suggesting that C. aguti preceded any other multiple-rowed captorhinid. Here we report on new material of multiple-rowed captorhinids from the Lower Permian fissure fills of the Bally Mountain locality in Oklahoma, only 35 miles from Richards Spur. Some of this material is referrable to Captorhinikos valensis , which was previously only known from younger strata in Texas, making this species the geologically and phylogenetically oldest moradisaurine. Furthermore, we determined that Ca. valensis co-existed with C. aguti at Bally Mountain and we explore the potential for niche partitioning in these early captorhinids. Lastly, we assess the potential temporal and environmental differences between Bally Mountain and Richards Spur, in order to explain the abundance of herbivorous moradisaurines at Bally Mountain and the complete lack of moradisaurines at the neighbouring Richards Spur locality.


INTRODUCTION
Captorhinids are Paleozoic amniotes that represent the first major radiation of terrestrial eureptiles (Modesto et al. 2014) much of it in the shadow of the more diverse and larger synapsids (Reisz 1997). Their initial diversification occurred in Laurasia in the Late Pennsylvanian and early Permian and included mainly carnivorous and omnivorous taxa with snout-vent lengths between 15-70 cm (Heaton 1979;Müller and Reisz 2005;Modesto et al. 2007). By the late Permian, herbivorous captorhinids had become much more dominant forms in terrestrial communities, with some taxa, including Moradisaurus from the upper Permian of Niger, having an estimated snout-vent length in excess of 2 meters (de Ricqlès and Taquet 1982;O'Keefe et al. 2005). Captorhinids in general are characterized by a stocky body and short limbs, and a number of interesting dental innovations that include a downturned premaxilla and multiple rows of marginal teeth, which formed from the interactions between tooth development and novel patterns of jaw growth (de Ricqlès and Bolt 1983;LeBlanc and Reisz 2015). One of the most interesting aspects of their evolutionary history is the apparent shift from carnivory to herbivory, with accompanying changes in dentition, and substantial increases in body size. Their success as herbivorous and carnivorous members of Paleozoic communities has been attributed to their unique dentitions, which included the earliest record of multiple marginal tooth rows (Bolt and Demar 1975;de Ricqlès and Bolt 1983;LeBlanc and Reisz 2015). The emerging consensus is that at least two lineages of captorhinids independently evolved dental batteries composed of multiple tooth rows: once within the genus Captorhinus and again in the subfamily Moradisaurinae (Reisz et al. 2011;Modesto et al. 2014). Captorhinus aguti is the only multiple toothrowed species of the genus and its remains are known from the middle Clear Fork Group (sensu Hentz 1988) Multiple tooth-rowed captorhinids from the early Permian fissure fills of the Bally Mountain Locality of Oklahoma (Vale Formation, sensu Lucas 2006) of Texas and the cave deposits of the geologically older Dolese Brothers Quarry near Richards Spur, Oklahoma, U. S. A. (hereafter referred to as Richards Spur), both of which are early Permian in age (Olson 1954;Woodhead et al. 2010;Modesto et al. 2014). By comparison, the moradisaurines extended from the late early Permian (Kungurian) into the late Permian, were geographically widespread, and much more diverse (Reisz et al. 2011). The success of the moradisaurines may have been related to the evolution of high-fiber herbivory within the clade, given that even the earliest known moradisaurines possessed grinding batteries of teeth (Reisz 2006;Modesto et al. 2014). Records from the Vale Formation of Knox County, Texas indicate that early moradisaurines and Captorhinus aguti may have had some geographical and temporal overlap (Olson 1952;Modesto et al. 2014). Their co-occurrence in the lower Permian deposits of Texas suggest that C. aguti and the moradisaurines had adapted to different ecological niches that would have eased pressures of competition between the two related taxa. Interestingly, moradisaurines have never been found in the geologically older fissure fills of the Richards Spur locality in Oklahoma despite intense sampling at the site over many decades and the recovery of thousands of specimens of C. aguti and at least three single-rowed captorhinid taxa (Olson 1954;Fox and Bowman 1966;Heaton 1979;Modesto 1996Modesto , 1998Reisz et al. 2015). This suggests that C. aguti may have had a broader temporal and geographic range compared to the earliest moradisaurines during the early Permian in North America. The earliest moradisaurines may thus have had to compete with an already established population of small-bodied captorhinids with multiple-rowed dentitions, which may have promoted their radiation into their alternative niche as high-fiber herbivores. These types of hypotheses, however, are sensitive to earliest occurrence data and whether the appearance of Captorhinus aguti preceded that of the first moradisaurines in the fossil record. Here we present the first description of abundant multiple-rowed captorhinids from lower Permian fissure fills of the Bally Mountain locality (previously referred to as the "South Carnegie Site" by Olson 1967) in southwestern Oklahoma, which some consider to be equivalent in age to the Richards Spur site 55 kilometers to the southeast (Olson 1967;Donovan 1987), and assign some of this material to the early moradisaurine Captorhinikos valensis. This new occurrence extends the geographic and possibly the temporal ranges of the oldest moradisaurines, but also permits a more complete description of the dentition of this small and poorly known taxon. Furthermore, we identify tooth-bearing elements of Captorhinus aguti from the Bally Mountain locality, thus potentially extending the Captorhinus aguti-Captorhinikos valensis species overlap by nearly 20 million years, from the Sakmarian of Oklahoma to the Kungurian of Texas, assuming that the two fissure fill sites in Oklahoma are equivalent in age (Olson 1967). Given the abundance of tooth-bearing elements of Captorhinikos valensis from the Bally Mountain locality and the complete absence of any moradisaurine from the nearby fissure fills of Richards Spur, we examine the potential for differences in the stratigraphic positions and palaeoenvironments between these neighbouring fissure fill sites in Oklahoma.

MATERIAL & METHODS
All of the material presented in this study was collected by W. May and was recovered from fissure fills of the lower Permian Bally Mountain locality ( Table 1). The locality represents an abandoned limestone quarry on Bally Mountain in Kiowa County, Oklahoma (Donovan 1987;Busbey 1990). The fissures occur in the carbonate units of the Ordovician Kindblade Formation and the fissure fills were originally thought to be Late Pennsylvanian in age, however the presence of Captorhinus aguti at the site suggested an early Permian (Artinskian) age (Heaton 1979;Simpson 1979;Donovan 1987). According to preliminary reports, the fossils were recovered from sinkholes and collapsed fissures and presumably represent the washed-in remains of vertebrates that inhabited the area near the caves (Donovan 1987;Busbey 1990). This interpretation is supported by the fact that many of the specimens recovered from the locality show evidence of tumbling and water transport and are all disarticulated remains (Busbey 1990). Similar to Richards Spur, some specimens are completely impregnated with hydrocarbons, which stain the fossils black, whereas others appear beige or grey and show no evidence of invasive hydrocarbons.
All of the material was photographed by D. Scott using a Canon EOS 40D digital SLR camera with a Macro 1000 mm lens. Scanning Electron Microscopy was undertaken using a NeoScope ACM-5000 Scanning Electron Microscope (SEM). Specimens that were examined in the SEM were uncoated. Illustrations of OMNH 55796, 77533, and 77534 were prepared by N. Wong Ken. Thin sections of ROM 67608 and 73638 were prepared following the methodology outlined by LeBlanc and Reisz (2015). The sections were photographed using a Nikon DS-Fi2 camera mounted to a Nikon AZ-100 microscope and NIS Elements (Basic Research) imaging software.
For the sake of brevity, Captorhinikos has been abbreviated to Ca. throughout the manuscript. Emended Diagnosis: Multiple-rowed moradisaurine distinguished by the presence of radiating rows of teeth on both maxilla and dentary. Teeth increase in size posteriorly in the multiple-rowed region, with the smallest teeth along the entire dentition being in the anterior positions in the multiple-rowed region. Tooth crowns of the multiple-rowed region possess a ring of raised enamel surrounding the occlusal tip of each tooth. Differs from other moradisaurines by the presence of two enlarged anterior maxillary teeth and a more densely denticulate pterygoid.
Locality and Horizon: All previously described specimens were recovered from the Leonardian (Kungurian) Vale Formation, Clear Fork Group in Knox County, Texas, U. S. A. (Modesto et al. 2014). The Bally Mountain locality represents an inactive limestone quarry at the northwest-most end of the Slick Hills in southwestern Oklahoma, approximately 55 kilometers northwest of the Richards Spur locality (Busbey 1990). The site represents an ancient cave system with extensive sediment infillings that were previously described as Late Pennsylvanian and later as Leonardian (Kungurian) in age. Preliminary reports suggested an abundance of disarticulated and worn material of Paleozoic tetrapods, including early synapsids, captorhinids, and the anamniote Doleserpeton, which rendered the site similar in faunal composition to the Richards Spur locality (Bolt 1969;Donovan 1987;Busbey 1990;Sullivan and Reisz 2002).

Premaxillae
Several captorhinid premaxillae have been collected from the Bally Mountain locality. These isolated elements can be assigned to two types. Many of these appear to belong to a species of Captorhinus, whereas others appear to represent a different captorhinid ( Fig. 1A, C, E). The premaxillae of Captorhinus have four or five teeth with the mesial tooth being the largest (Fox and Bowman 1966;Heaton 1979;de Ricqlès and Bolt 1983). The second tooth is typically much smaller and the subsequent teeth steadily decrease in size distally. The second type of premaxilla also possesses four or five teeth, but differs from that of Captorhinus in that the second premaxillary tooth is as large as, or larger than the first tooth. The enlarged second premaxillary tooth also contributes to the more broadly arcuate shape of the tooth row in ventral view when compared to the premaxilla of Captorhinus (Fig. 1). The second tooth is also slightly procumbent, such that the tips of the first and second premaxillary teeth reach equally far mesially. In the premaxillae of Captorhinus, the tip of the second tooth is distal to the tip of the first (Fox and Bowman 1966;de Ricqlès and Bolt 1983) (Fig. 1). The shapes of the tooth crowns in both types of premaxillae are identical: the crowns taper to flat, spatulate tips and possess a pair of faint cutting edges on the lingual surfaces. Both types of premaxillae bear three processes: a dorsal nasal process, a medial vomerine process, and a lateral maxillary process. The dorsal extremities of the nasal processes in the premaxillae from Bally Mountain are often broken, but on one specimen of the new type of premaxilla (OMNH 77528), the maxillary and vomerine processes are complete (Fig. 1). The vomerine and maxillary processes both extend posteriorly, but the vomerine processes projects more dorsally than does the maxillary process. This condition also occurs in premaxillae of Captorhinus and indicates that this second type of premaxilla was downturned, as in all other captorhinids (Fox and Bowman 1966;Modesto 1998). The anterior surface of the second premaxilla type is similar to Captorhinus in that it is perforated by numerous foramina (Fig. 1). Unfortunately, the second type of premaxilla cannot be confidently assigned to a taxon, given that they are always found disarticulated. However, we would argue that the differences in the size and orientation of the second premaxillary tooth indicate that these premaxillae do not belong to a species of Captorhinus, given that all species within the genus have a larger anterior premaxillary tooth (Fox and Bowman 1966;Heaton 1979;de Ricqlès and Bolt 1983).

Maxillae
At least two types of multiple-rowed maxillae are found at Bally Mountain. The first type possesses an anterior single-rowed region and a posterior multiple-rowed region with diagonally oriented tooth rows (Fig. 2). This type is readily referable to Captorhinus aguti based on the orientations of the rows of teeth and the morphology of the crowns in the multiple-rowed region (Fox and Bowman 1966;de Ricqlès and Bolt 1983;Modesto 1998). The teeth of the multiple-rowed region are ogival: they are laterally compressed, possess crenulated enamel apices, and the crowns terminate in mesiodistally aligned enamel crests (Fig. 2). The second type of maxilla clearly belongs to a moradisaurine, based on the parallel alignment of the multiple rows of teeth. Although we were unable to recover a fragment with the anterior tip of the maxilla, the tapering anterior portions of the preserved maxillae all indicate the presence of an acuminate premaxillary process, which is found in all captorhinids (Modesto et al. 2014). The single-rowed region consists of at least five chisel-shaped teeth, although the full count is unknown. The multiple-rowed region is composed of five rows of small, bullet-shaped teeth with circular cross-sections (Fig. 2). The teeth of the single-rowed region possess short mesial and distal cutting edges, similar to those in both types of premaxillae from Bally Mountain. The preserved single-rowed regions all show an increase in tooth size distally, with the distal two teeth being the largest, forming a short "caniniform" region, similar to Captorhinus aguti, but unlike other moradisaurines (de Ricqlès and Bolt 1983;Dodick and Modesto 1995). The subsequent mesial-most teeth of the multiple-rowed area are significantly smaller and lack cutting edges (Fig. 2). The teeth in the multiple-rowed area are all similarly bullet-shaped and possess unusual crown ornamentation. Scanning Electron Microscopy (SEM) analysis revealed that each tooth consists of a pointed enamel apex, surrounded by a ring of enamel (Fig. 2K). This ring represents a low ridge that also forms the borders of shallow basins near the tip of the tooth crown on the lingual and labial sides. The tooth rows of the multiple-rowed region are nearly parallel to the long axis of the bone and diverge from each other distally due to a steady increase in the sizes of the posterior teeth in each row (Fig. 2). In other moradisaurines, the teeth in the multiple-rowed region of the maxillae exhibit a more modest increase in size posteriorly (Dodick and Modesto 1995;Reisz et al. 2011). In C. aguti, the maxillary teeth tend to be largest along the middle of the multiple-rowed region and decrease in size posteriorly (Modesto 1998). The teeth along the labial-most rows of the maxilla in the Bally Mountain moradisaurine have the smallest teeth and the teeth in the lingual rows are the largest (Fig. 2I, J, L, M).

Dentaries
As was the case with the maxillae from Bally Mountain, we were also able to identify two types of multiple-rowed captorhinid dentaries. The first type of dentary possesses three or more diagonally arranged rows of teeth in the multiple-rowed region. The six teeth in the single-rowed area bear mesial and distal cutting edges, whereas the distal, multiple-rowed teeth are ogival, indicating that these dentaries are referable to Captorhinus aguti (Modesto 1998). The second dentary type again belonged to a moradisaurine, based on the parallel rows of teeth in the multiple-rowed region (Dodick and Modesto 1995;Reisz et al. 2011). The mesial-most dentary tooth is small, procumbent, and mesiodistally compressed similar to the first dentary tooth in Captorhinus aguti (Bolt and Demar 1975). The third tooth is the largest in the single-rowed area and is followed by three or four more teeth distally (Fig. 3). These teeth all bear mesial and distal cutting edges, similar to the anterior teeth of the maxillae. The teeth in the single-rowed area gradually decrease in size distal to the second dentary tooth. Unlike in C. aguti, the boundary between the single and multiple-rowed areas in the dentaries of the moradisaurine from Bally Mountain is well defined by a marked decrease in tooth diameter (Figs. 2, 3). The distal tooth of the single-rowed area is often preceded by three rows of very small, peg-shaped teeth. As in the maxilla of the moradisaurine, the dentary teeth in each row steadily increase in size, causing each row to widen distally. Each tooth is similar in crown morphology to those of the maxilla. At most, the dentaries possess four rows of teeth that radiate distally.

Referral of Bally Mountain Moradisaurine to Captorhinikos valensis
Whereas some of the captorhinid material from Bally Mountain is readily referable to Captorhinus aguti, the other multiple-rowed taxon is clearly a moradisaurine.
The parallel rows of marginal teeth indicate moradisaurine affinities, whereas the increasing diameter of the teeth along each row and the radiating tooth rows suggest that these remains are specifically referable to Captorhinikos valensis, known previously from potentially younger strata in Texas (Olson 1954;Modesto et al. 2014). The number of tooth rows in the maxilla and dentary of the Bally Mountain moradisaurine also match the number of rows in the holotype and referred materials of Ca. valensis (Modesto et al. 2014). Furthermore, the dentaries and maxillae from Bally Mountain bear teeth that have identical crown morphologies to those of the holotype dentary (FMNH UR 101). The ring of enamel surrounding a pointed crown had not previously been described for Ca. valensis, but it is clearly present on the teeth of the multiple-rowed region in FMNH UR 101 (Fig. 4). Thin sections of two partial dentaries (ROM 67608 and 73638) of Ca. valensis from Bally Mountain revealed that the enamel was not unusually thick when compared to that of other captorhinids from Richards Spur (see LeBlanc and Reisz 2015: figs. 4b, e, 6b;Reisz et al. 2015: fig. 1c). The enamel ridge in Ca. valensis was also only clearly visible in parasagittal sections through the tooth crowns and was formed primarily by a slight thickening of enamel and a minor occlusal extension of the underlying dentine (Fig. 4E, F). The change in shape of the dentine-enamel junction provides evidence for a developmental shift in the shape of the tooth prior to deposition of hard tissues, whereas enamel thickening alone would have indicated that the ridge had formed through an increase in the localized deposition of enamel only (Brink et al. 2015).
Unfortunately, no articulated material was recovered from our Bally Mountain sample and therefore we are uncertain as to the identity of the bearer of the second type of captorhinid premaxilla described here. Nevertheless, the abundant material of Captorhinikos valensis from Bally Mountain permits some minor additions to the knowledge of the dental  anatomy and mode of tooth replacement and tooth addition in this diminuative moradisaurine. First, the mesial-most dentary tooth is a small "symphyseal" tooth, which is preceded by two large, chisel-shaped teeth, all three of which are procumbent (Fig. 3). This morphology is reminiscient of the anterior dentary teeth of Captorhinus aguti, however the third dentary tooth in Ca. valensis is proportionally much larger (Fig. 3A). Second, the anterior single-rowed area in the maxillae is remarkably heterodont, more so than in other moradisaurines. The teeth increase in size distally forming a "caniniform" region, similar to C. aguti (de Ricqlès and Bolt 1983). Third, the abundant remains of Ca. valensis allow us to conclude that the number of tooth rows in this taxon is not solely related to body size, given that we have found relatively large and small individuals with the same number of maxillary and dentary rows (Fig. 5). Previous authors have also found no correlation between the number of tooth rows and the size of individuals in Captorhinus aguti (Bolt and Demar 1975;de Ricqlès and Bolt 1983;LeBlanc and Reisz 2015). Finally, the presence of replacement pits lingual to the functional teeth in the single-rowed region (Fig. 3), coupled with the presence of emplacement pits (sensu LeBlanc and Reisz 2015) along the distal ends of each row suggest that teeth were replaced regularly in the mesial regions of the jaws and new teeth were added posteriorly along each row and with a relatively high frequency in Ca. valensis (Fig. 5).
Older teeth in the multiple-rowed area were shed along the labial margins of the jaws, as indicated by the shallow pits of

PHYLOGENETIC ANALYSIS
Given the phylogenetic position of Ca. valensis as the earliest moradisaurine (Modesto et al. 2014) the anatomy and precise phylogenetic relationships of this taxon are important for understanding the timing of the evolution of multiple tooth rows and high-fiber herbivory within Captorhinidae. Although the remains presented in this paper are fragmentary, some of the material of Captorhinikos valensis from Bally Mountain permitted the coding of additional characters from the original analysis by Modesto et al. (2014) to which we also added the minor character state changes for Captorhinus magnus and Labidosaurus as suggested by Reisz et al. (2015). Three additional characters were coded for Ca. valensis (characters 4, 29, and 57) and one character was recoded (character 8) (Appendix 1). We then subjected this updated matrix to the same branch-and-bound search under maximum parsimony as Modesto et al. (2014) using PAUP* (Swofford 2002) in order to determine if this resulted in a different topology. The resulting single most parsimonious tree is identical to that of Modesto et al. (2014) and still recovers Ca. valensis as the phylogenetically oldest moradisaurine ( Fig. 6; TL = 147; CI = 0.633; RI = 0.779; RCI = 0.493). The addition of these characters yielded slightly lower bootstrap support values after 1000 replicates for all of the nodes, but the clade supporting Ca. valensis as a moradisaurine was still strongly supported (Fig. 6).
confirm that these two captorhinids co-existed during the early Permian in Oklahoma and Texas. This increases the geographical extent of the overlapping ranges of the two multiple-rowed captorhinids. Furthermore, the occurrence of Ca. valensis at Bally Mountain is now the geologically oldest occurrence of a moradisaurine, given that the locality is considered equivalent in age to the famous Richards Spur locality in Oklahoma (mid-Sakmarian, roughly 289 MYA; see Woodhead et al. 2010). If these dates are correct, then the two multiple-rowed captorhinids co-existed for approximately 20 million years (Modesto et al. 2014), suggesting that Captorhinus aguti and the earliest moradisaurines had diverged in diet and ecology prior to this time. Dental wear patterns can help determine dietary prefer-

Patterns of Tooth Wear and Palaeoecology of Two Co-occurring, Multiple Tooth-rowed Captorhinids
The new material of Captorhinikos valensis and the presence of Captorhinus aguti from the Bally Mountain locality ence in fossil taxa (Reisz 2006) and these patterns differ significantly between the two taxa (Fig. 7). Individual teeth in Captorhinus aguti are either worn only along the mesial and distal ends of the enamel crest, or along the labial or lingual face of the crown (Fig. 7A-E). Across the rows, the lingual-most teeth of C. aguti are always unworn, whereas most of the labial teeth show evidence of light to moderate tooth wear. The predominant gradient of tooth wear is therefore oriented labio-lingually. Overall, tooth wear is not as extensive in C. aguti as it is in Captorhinikos valensis (Fig.  7F-L). Furthermore, the tooth wear in C. aguti is restricted to either the labial or lingual surfaces of the tooth crowns, suggesting that only one upper and one lower tooth sheared past one another. The mesio-distally oriented enamel crests on the teeth of C. aguti would have assisted in shearing food as well. By comparison, many of the teeth of Ca. valensis are worn flat and were clearly forming grinding surfaces along the mesial portions of the multiple-rowed regions of the jaws. The teeth of Ca. valensis also show a drastically different wear pattern along and across the tooth rows compared to Captorhinus aguti (Fig. 7). Lightly worn teeth are primarily restricted to the distal ends of each row, moderately worn teeth further mesially, and the heaviest wear is associated with the mesial-most teeth. This dental wear pattern is nearly identical to that of the Triassic rhynchosaurs (Benton 1984), where the gradient of tooth wear is mainly oriented mesio-distally. A slight labio-lingual wear gradient is also present, but it is much less pronounced.
Despite being disarticulated elements, a survey of a large number of specimens shows that tooth wear is much more pronounced on the maxillae than on the dentaries in Ca. valensis, however dental wear is still most extensive on the  (OMNH 77538). Note that unworn teeth (blue) tend to be restricted to the distal ends of each tooth row; I, dental wear patterns along the multiple-rowed area of a partial left maxilla (OMNH 77532). The heaviest tooth wear (green) is concentrated along the mesial ends of each tooth row; I, SEM image of very lightly worn maxillary tooth with lingual and labial wear facets; J, SEM image of lightly worn maxillary tooth with a single large wear facet extending to the tip of the crown; K, heavily worn tooth; L, striations along the labial surface of a lightly worn tooth crown indicate fore and aft motion between upper and lower jaws. mesial teeth of the multiple-rowed regions of the dentaries (Fig. 7). Furthermore, the presence of two wear facets on the lingual and labial sides of the lightly worn teeth show that the upper teeth and lower teeth contacted in an interlocked fashion, and that food grinding was accomplished by propalineal movements of the lower jaws across the maxillary dentition (Fig. 7L). Even the lightly worn teeth of the distal and lingual MR regions show evidence of tooth wear on both the labial and lingual sides of the crowns. The ridges and small basins of enamel at the tips of the crowns would have been worn down first, followed by the underlying dentine. Heavily worn teeth possessed no enamel, but did not have exposed pulp cavities. From these data, it seems that even the earliest moradisaurines were adapted to grinding food material and show clear evidence of tooth occlusion. Despite having independently evolved multiple-rowed dentitions, the earliest moradisaurines and Captorhinus aguti used these dentitions in entirely different manners. Captorhinus aguti was probably more omnivorous and did not exhibit a true grinding dental battery, because the teeth sheared past their counterparts, whereas the moradisaurines exhibited extensive dental occlusion and tooth wear, presumably to pulverize fibrous plant material.

Significance of the Presence of Captorhinikos valensis at Bally Mountain
The presence of an herbivorous moradisaurine from the early Permian fissure fills of Oklahoma raises intriguing questions about the temporal and geographical distribution of early Permian captorhinids across North America. The Bally Mountain locality remains largely undescribed, but has already yielded a striking difference in faunal composition when compared to the well-known fissure fills of the Richards Spur locality only 55 kilometers away (Donovan 1987). Decades of descriptions of captorhinids from Richards Spur have yielded new species of single-and multiple-rowed captorhinids (Modesto 1996;May and Cifelli 1998;Kissel et al. 2002), thousands of specimens of Captorhinus aguti (Bolt and Demar 1975;Modesto 1998), but no moradisaurines (MacDougall andReisz 2012). By comparison, moradisaurines were clearly the most abundant taxon at Bally Mountain (W. May pers. obs.) with Captorhinus aguti being slightly less common. This difference can only be explained by two alternative, but not mutually exclusive hypotheses: (1) the Bally Mountain and Richards Spur fissure fills are not equivalent in age; or (2) the fissure fills at Richards Spur were deposited in a different environment compared to Bally Mountain. Based on the initial faunal descriptions of Bally Mountain, it is generally accepted that the Bally Mountain fissure fills are equivalent in age to those at Richards Spur (Olson 1967;Donovan 1987;Busbey 1990). They were deposited within caves carved out of the same Ordovician limestone of the Arbuckle Group, they overlie the same oil deposits, they exhibit similar modes of fossil preservation, and preliminary reports suggested they contained very similar early Permian faunas (Donovan 1987;Busbey 1990;Sullivan and Reisz 2002). However, at present it is impossible to determine with any confidence if the Bally Mountain locality is temporally equivalent to the Richards Spur locality without absolute dating (Woodhead et al. 2010). Vertebrate biostratigraphy is the only measure that provides any direct comparisons between the two localities, but the resolution is quite coarse (Olson 1952;Simpson 1979;Woodhead et al. 2010). Interestingly, the vertebrate fauna at Bally Mountain bears similarity to both the Richards Spur locality in Oklahoma and the Vale Formation in Texas. Concerning the former locality, Bally Mountain has yielded the remains of the anamniote Doleserpeton, which is known elsewhere only from Richards Spur (Olson 1967;Bolt 1969). However, Bally Mountain preserves a captorhinid fauna that is apparently more similar to the geologically younger Vale Formation (sensu Lucas 2006) of central Texas than it is to Richards Spur, given that the Vale Formation preserves the only other occurrence of Captorhinikos valensis (Olson 1952(Olson , 1954Modesto et al. 2014). It is possible that Bally Mountain is significantly younger than Richards Spur, possibly by tens of millions of years. However, a more thorough faunal analysis and absolute dating of speleothems would have to be undertaken at the Bally Mountain locality in order to provide a clearer understanding of the biostratigraphy and age of the site.
If the two fissure fill localities in Oklahoma are time-equivalent, then the absence of moradisaurines at Richards Spur may best be explained by environmental differences between the two localities. As a high-fiber herbivore, Captorhinikos valensis may have been more restricted in its distribution when compared to the omnivorous Captorhinus aguti. Herbivores are represented by fragmentary remains of diadectids, caseids, and bolosaurids at Richards Spur; however, they appear to have been rare members of the faunal assemblage (Reisz and Sutherland 2001;Reisz et al. 2002;Reisz 2005). Reisz and Sutherland (2001) suggested that the rarity of diadectid material at this site indicated that Richards Spur was an arid, upland locality with little vegetation to support large populations of herbivorous tetrapods. This would help explain the abundance of Captorhinus aguti at Richards Spur, which would have been more of a generalist, and the lack of any herbivorous moradisaurines. The vegetation at Bally Mountain may therefore have been less sparse and sufficient to support small-bodied herbivores like Captorhinikos valensis in great abundance.
In addition to possible environmental differences between Bally Mountain and Richards Spur, the presence of moradisaurines at a locality that is of equivalent age to the fissure fills of Richards Spur has major implications for the evolution of high-fiber herbivory in Captorhinidae (Fig. 8) and amniotes in general. If moradisaurine origins can be dated to approximately 289 MYA, then this coincides with the earliest occurrence of the other multiple-rowed captorhinid, Captorhinus aguti. Their co-occurrence suggests that the early evolution of moradisaurines may have been influenced by niche partitioning, given that both taxa were roughly equivalent in body size, co-occurred in time and space, and had multiple rows of teeth. The two groups probably evolved multiple rows of teeth independently (Reisz et al. 2011;Modesto et al. 2014) suggesting that competition would have been pivotal in the establishment of moradisaurines as high-fiber herbivores and C. aguti as an omnivore. Pinpointing the precise age of the Bally Mountain locality is thus incredibly important for testing this hypothesis and determining if interspecific competition between captorhinids could have promoted the evolution of high-fiber herbivory, which was clearly a key innovation that promoted the subsequent dispersal and diversification of captorhinids throughout the Permian (Reisz 2006;Modesto et al. 2014). In addition, the presence of a small herbivorous reptile this early in the evolution of the terrestrial vertebrate community would indicate that niche partitioning among herbivores was well underway by the mid-Sakmarian, with Captorhinikos valensis, Edaphosaurus, and Diadectes using extensive oral processing and tooth on tooth grinding of plant materials at different body sizes, with Ca. valensis clearly being the smallest known herbivore of its time (Reisz and Fröbisch 2014).

CONCLUSIONS
The fauna of the early Permian fissure fills of the Bally Mountain locality in southwestern Oklahoma remains largely undescribed, but has already yielded a wealth of information on the early evolutionary history of captorhinids, one of the first amniote clades to diversify as high-fiber herbivores (Reisz 2006;Modesto et al. 2014). These first formal descriptions of the multiple-rowed captorhinids from the early Permian fissure fills of the Bally Mountain locality in Oklahoma, U. S. A., provide compelling evidence that Captorhinus aguti, a well known form from other early Permian localities in Oklahoma and Texas, and the earliest moradisaurines co-existed across a wider geographic distribution than previously thought. Furthermore, the description of new material of the early moradisaurine Captorhinkos valensis from Bally Mountain may also extend the temporal range of the earliest herbivorous captorhinids, making it time-equivalent to the earliest occurrence of Captorhinus aguti. The co-occurrence of C. aguti and Ca. valensis, both small-bodied, multiple-rowed captorhinids, provides a window into the possible drivers of the origin and diversification of the herbivorous moradisaurines, but more data must first be collected from the locality to glean better resolution of the timing of this diversification. The Bally Mountain locality may thus play an important role in our future understanding of the structure of early Permian terrestrial communities, the evolution of herbivory (Reisz and Fröbisch 2014), and the diversification of early amniotes.