Nuevas evidencias de la coespecificidad entre Hipparion primigenium melendezi Alberdi, 1974 de Los Valles de Fuentidueña (Segovia) e Hipparion concudense concudense Pirlot, 1956 de Concud (Teruel) España ; New evidence of conspecificity between Hipparion primigenium melendezi Alberdi, 1974 from Lo

This paper reports a morphological and biometric comparison of Hipparion primigenium melendezi from Los Valles de Fuentidueña (Segovia) and Hipparion concudense concudense from Concud (Teruel) to investigate their possible conspecificity. Comparisons with other Spanish (La Roma 2, Puente Minero, Milagros, Las Casiones and Venta del Moro) and Eurasian (Höwenegg, Germany; Pavlodar, Kazakhstan [small form]) material were also made. Based on their morphology, and on bivariate and multivariate analysis to establish size similarities, the Hipparion remains from Los Valles de Fuentidueña and Concud were both determined to belong to Hipparion concudense. However, differences in both dental morphology and limb slenderness could suggest differences at the subspecific level.

tion of the medium-sized species H. mediterraneum and H. concudense, among others, led to the appearance of very small taxa such as H. matthewi and H. periafricanum.
The abundant Eurasian record of Vallesian and Turolian Hipparion gave rise to a great proliferation of taxonomic names that Forsten (1968) tried to synthesize using statistical methods.Later, using other statistical methods that combined morphological and biometric features, Alberdi (1974) set limits to Forsten's reduced systematics.Alberdi (1989) then went on to establish six morphotypes sensu lato in an attempt to unify the Eurasian and African Hipparion species appearing from the early Vallesian to the early Villafranchian.A number of authors have recently proposed these species be assigned to different genera and subgenera.Based on Hipparion remains found in China, Qiu et al. (1987) proposed grouping a number of Chinese species into different subgenera.Bernor et al. (1996, p. 333) placed all species of the genus Hipparion into the Hipparionini tribe, and grouped them into several genera or group complexes.Following Bernor et al. (1996) these include four supra-specific evolutionary complexes: the Hippotherium primigenium-Complex (with 13 taxa), the Hipparion s.s.-Group (six taxa), the Cremohipparion-Group (seven taxa), and the "Plesiohipparion"-Group (six taxa) belonging to the "Sivalhippus" Complex.
The nomenclature of Hipparion species is, however, becoming more and more complicated, a likely consequence of the philosophical differences in taxonomic meaning employed when comparing genera and species.The examination of the morphological features of many remains of Hipparion could, however, provide a global framework for defining the morphotype of Hipparion sensu lato (see Pesquero et al., 2006Pesquero et al., , 2007)).
The Turolian mammal site of Concud, in the northeast of the Teruel Province, Spain, is home to accumulations of vertebrate bones, the richness, diversity, and special features of which render the faunal assemblages recorded a reference for studies on Eurasian Neogene mammals.The site is the type-locality of Canis cipio (the oldest canid -Caninae sensu stricto -of the fossil record), of the bovid Hispanodorcas torrubiae, of Turiacemas concudensis (one of the oldest three-tined deer), and of the tridactyl equid Hipparion concudense.The fossiliferous level was deposited during the progressive transition between alluvial and shallow lake envi-ronments.The mammal assemblages, similar to that found at Los Mansuetos locality, typify the middle Turolian (MN12) (Alcalá, 1998;Alcalá et al., 1999;Pesquero et al., 2010).
The Vallesian mammal site of Los Valles de Fuentidueña (Segovia Province), located in the upper part of the Duero Basin, reflects a pond system, the bone material of which was provided via a small alluvial fan coming from a Cretaceous calcareous palaeorelief (Hoyos et al., 1981).This site contains the remains of a diversified fauna of micro-and macromammals, and is highly important for the characterization of the early Vallesian of the Duero Basin as it allows correlations with faunal assemblages of the same age in other basins.The analysis of this fauna indicates that, during the late Aragonian and early Vallesian, the Duero Basin fauna was similar to that of the Calatayud-Teruel Basin.In fact, both can be included within a single biogeographic unit.This unit, however, shows clear differences to the faunal composition of the Vallés-Penedés Basin, which is more similar to German and French basins of the same age (Alberdi et al., 1981).
The main aim of this paper is to review the Los Valles de Fuentidueña and Concud remains of Hipparion, checking against the work of Forsten (1982), who indicated a conspecific relationship between Hipparion concudense (Pirlot, 1956) from Concud and Hipparion primigenium melendezi (Alberdi, 1974) from Los Valles de Fuentidueña, referred to by Eisenmann (1995) as Hipparion melendezi.

Methodology
The morphological and morphometric traits (fig.2) proposed by Alberdi (1974) were examined.The hypsodonty index was calculated for unworn or only slightly worn teeth.This index is defined as the ratio between the mesio-distal length (2) and the height of the unworn teeth (1) (Eisenmann et al., 1988: Figs. 6-8).The dimensions of the teeth are presented in a bivariate plot using the customary P3-4, M1-2, p3-4, and m1-2 teeth.Two bivariate plots were generated to analyse the robustness or gra- Estudios Geológicos, 68(2), 247-260, julio-diciembre 2012 ).Bivariate and multivariate analyses were performed with the numerical data to establish size similarities.Metapodials, astragali, calcanei and first phalanges were analysed using principal component analysis (PCA) to evaluate size similarities and/or differences among the different Hipparion remains.Groups were identified based on the outcome of PCA, and discriminant analysis (DA) then preformed using the Mahalanobis method with the aim of maximizing the separation between them.This determines whether the centroids differ significantly or not, and often identifies specimens not included in the original PCA.Body masses were estimated using the first phalanx, the Ph5 measurement (r = 0.991), and McIII and MCIII13 measurements (r = 0.982), according to Alberdi et al. (1995).Calculations were made using SPSS 15.0.
The nomenclature and measurements follow the recommendations of the "Hipparion Conference", New York, November 1981 (Eisenmann et al., 1988).All dimensions are expressed in millimetres.
Description: The protocone shape is oval in premolars and molars, sometimes connected to the protoloph in stage of wear II in P2 and in stage of wear III in P3-4 and M1-2.In M1-2, the fossette folds show medium folding varying from 0 to 7 at Apre, 2-11 at Dpre, 2-7 at Apof, and 0-3 at Dpof.The modes are 2 plus 3, 7, 5, and 1 respectively.In P3-4, the fossette folds show medium folding varying from 0 to 3 at Apre, 3-10 at Dpre, 1-8 at Apof, and 0-3 at Dpof.The modes are 1, 6, 3, and 1 respectively.The pli caballin varies between 1 and 4, while the mode for premolars is 2 and for molars is 1.Hypsodonty index 2.1 in P3-4 and 2.5 in M1-2.In the lower cheek teeth the enamel is complex.Protostylid isolated on occlusal surface of unworn teeth, only joined to protoconid in stages of wear III.Hypsodonty index 2.2 in p3-4 and 2.4 in m1-2.
Postcranial remains are medium-sized and slender.Metapodials are elongated and slender.In McIII proximal articulation, the angle between the magnum and unciform facets varies from 124º to 134º.Proximal articulation surface of MtIII shows a well developed cuboid facet.A small cuneiform facet is present in two specimens (four in total).Gracility index of MtIII1/MtIII3 varies from 9.4 to10.8 (X = 10.3); that of MtIII1/MtIII11 varies from 13.9 to 14.8 (X = 14.4).Pirlot, 1956   Type locality: Concud, Teruel, Spain.
Lower cheek teeth are medium-sized.Occlusal morphology not complicated.Molars with a protostylid isolated on occlusal surface in unworn teeth, only joined to protoconid in stage of wear I in molars and II in premolars.Hypsodonty index 2.4 in p3-4 and 2.6 in m1-2.

Morphological and univariate analysis
Morphological analysis revealed similarities in skeletal structure between the Hipparion remains from LVF and CD.Both show similar occlusal tooth morphology, although the enamel is more complicated in the LVF lower cheek teeth.Hipparion from LVF differs from H. primigenium from Nombrevilla in Spain and HO in Central Europe (Pesquero et al., 2006), which tend to have a more complicated enamel pattern on the occlusal surface (Alberdi, 1974).
The hypsodonty index indicates that, in general, the teeth of the CD form are somewhat more hypsodont than those from LVF (table 1), and that both CD and LVF teeth are more hypsodont than those of either H. primigenium from Nombrevilla (Pesquero et al., 2006) 4).The McIII from LVF and CD are longer than those of H. matthewi from KS, ML, VM, and PAV, and similar to that of H. primigenium from HO.Nevertheless, the diaphysis breadth of the LVF and CD forms is larger than that of H. matthewi from VM and smaller than that of H. primigenium from HO, this is, the McIII from LVF and CD are slender than those from HO.Only one McIII from CD is similar in length and breadth to those from HO.The MtIII of LVF and CD forms clearly lie outside the distribution of the H. matthewi and H. laromae remains.In both Spanish forms the length of the MtIII is similar, but the breadth is wider in CD metatarsals than in the LVF samples and closer to that of H. primigenium from HO, although the latter has greater length.
The gracility index indicates the LVF specimens to be more slender than those of CD.Both forms are more slender than H. primigenium from HO and more robust than H. matthewi from KS, ML, VM, and PAV (table 2).
The analysis of body mass following the methodology of Alberdi et al. (1995) indicated the LVF and CD forms to be of similar weight, while both are considerably smaller than H. laromae from RO2 and PM or H. primigenium from HO, but larger than H. matthewi from KS, ML, and VM.

Multivariate analysis
Principal component analysis of the skeletal measurements identified four main groups based on size   9.5 9.5 9.5 10.6 10.9 11.3 12.9 12.9 13 14.3 14.6 14.9 121    3).In the first component the most influential character was size.The most influential variables were the maximal diameter of the articular facet for the third tarsal (MtIII7), the maximal distal supra-articular breadth (MtIII10), and the maximal distal articular breadth (MtIII11).In the second component the most important variables were the diameters of the articular facets for the second and fourth tarsals (MtIII 8 and MtIII9) and the maximal lengths (MtIII2 and MtIII1).In this analysis the large PM specimen fell among the values for the HO remains, those from LVF grouped with those from CD, and those from VM, ML, and PAV were clearly separated.
In the PCA for the astragalus, the distributions were similar to those recorded in the analysis of MtIII (fig.5, table 3).All variables influenced the first component, emphasizing the size of the different specimens.The breadth of the trochlea (AS3) carried the most weight in the second component.
In the PCA for the calcaneus the distribution was again similar (fig.5, table 3).All variables influenced the first component; the minimal breadth (CAL3) carried the most weight in the second component.
In the PCA for the first phalanx the most influential variable in the first component was skeletal size (fig.6, table 3).The maximal length was the most important variable (1PhIII1) in the second component.In figure 6, the results for the 1PhIII PCA are similar to those for the calcaneus and astragalus in general terms, but the PCA for 1PhIII indicated large differences between the anterior and posterior phalanges.
Discriminant analysis was performed on the PCA results to maximize the separation among the groups produced.The first group included the largest remains, i.e., those from RO2 and PM; the second included the large sized remains from HO, the third consisted of the relatively middle remains from LVF and CD, and the fourth included the small-sized remains from VM, KS, ML, and PAV.
The DA for McIII clearly separated the LVF and CD group from the samples of H. primigenium from HO, and indeed from other sites (fig.7, table 4).The results indicate this group to be well classified in 91.7% of cases, with only one of 12 specimens overlapping with some of H. primigenium from HO.The first and the second group were always correctly identified.The DA for MtIII reproduced these results.Only one specimen out of 13 (7.1%)from the third group overlapps with some specimens of H. primigenium from HO (fig.7

Discussion
The present results show that the remains of Hipparion from LVF and CD have similarities that clearly differentiate them from the materials from the other studied sites.Morphological affinities between them were confirmed in the bivariate and multivariate analyses, although some differences in dental and postcranial morphology were apparent.The hypsodonty index of the LVF specimens is lower than that of the CD specimens, and the occlusal morphology of the lower cheek teeth is more complicated in the LVF specimens (fig.8, table 1).These characters, hypsodonty and plication, would be related to different paleoenvironment conditions and consequently to the diet.The higher hipsodonty and the more complicated enamel plication are related with the drier conditions of the LVF environment.With respect to the postcranial skeleton, the metapods of both forms are slender, although the gracility index indicates the LVF specimens to be more slender than those from CD (the former are somewhat longer and the width of the diaphysis is smaller) (fig.8, table 2).These kinds the analyses confirmed the remains from LVF and CD to be very similar.Only a few morphological differences exist in the postcranial bones, PCA analysis showed them to overlap, and they merged in the DA analysis.The LVF and CD forms are clearly distinguished from the large forms from HO, PM, and RO2.The gracility indices also show that the LVF postcranial skeletons more closely resemble those from CD than those of H. primigenium from HO (table 2), although the MtIII, astragalus, calcaneus, and 1PhIII from LVF are more slender than the homologous bones from CD. Nonetheless, the body masses of the LVF and CD Hipparion taxa are similar (table 2).It is important to note the differences between the LVF and CD taxa and other Spanish Hipparion (Ortiz Jaureguizar & Alberdi, 2003).The mean body mass of Hipparion from LVF was 170 kg with 1PhIII5, and 186 kg with McIII13.This is close to the mean weight of 160 kg with 1PhIII5 and 178 kg with McIII13 for the CD form (table 2).These similarities in weight, morphological features, and the results of the bivariate and multivariate analyses suggest, as Forsten pointed out (1982) that the Hipparion from LVF and CD should be included within Hipparion concudense, although as belonging to different subspecies.In accordance with the priority rules (ICZN, 2000) the name of Hipparion concudense is prioritary.Thus, the taxonomical separation between H. c. melendezi, from the Duero Basin, and H. c. concudense, from the Calatayud-Teruel Basin, would be consequence of different environmental conditions between both Basins, an idea supported by the presence in LVF of a faunal assemblage which suggests drier environment than that from CD.

Conclusions
The present results show the similarities between H. concudense from Concud and H. p. melendezi from LVF, to be sufficient to assign them to the same species, H. concudense, agreeing with previous Forsten's (1982) proposal.However, these taxa are different enough to represent two subspecies: H. concudense melendezi Alberdi, 1974, andH. concudense concudense Pirlot, 1956, by the morphological and biometrical differences.
The dental morphology of H. concudense melendezi is somewhat more complicated, and the distal parts of its limbs (McIII,MtIII,AST,CAL,and 1PHIII) are more slender than those of H. concudense concudense.

Figure 2 .
Figure 2.-Morphological characteristics of the upper (top) and lower (bottom) cheek teeth of Hipparion.

(
figs. 5 and 6).The first clustered the specimens belonging to the large PM and RO2 forms, while the second contained the specimens from HO.The third group included the LVF and CD forms, and the fourth group contained the relatively small specimens of H. matthewi from VM, KS, PAV, and ML.In the PCA for McIII, all variables influenced the first component with the skeletal size carrying the greatest weight (fig.5, table 3).The most influential traits were the proximal articular breadth (McIII5), the proximal articular depth (McIII6), and the maximal distal depth of the medial condyle (McIII14).

Figure 6 .
Figure 6.-Principal component analysis for the first phalanx III and discriminant analysis based on PCA results.

of Spanish Hipparion. N = number of specimens; Sd = standard deviation; X = mean
clearly fall within the distribution of the CD remains.In both forms the length and breadth of the lower and upper cheek teeth are similar.The teeth of Hipparion from CD and LVF are larger than those of H. matthewi from KS, ML, VM, and PAV and clearly rest outside of the distribution of H. primigenium from HO and H. laromae from RO2 and PM (fig.3).The bivariate plots for McIII and MtIII indicate differences between the LVF and CD Hipparion remains and those from the other sites (fig.

table 4 )
. The DA for the astragalus confirmed the separation observed in PCA (fig.7,table 4).The remains from LVF and CD were clearly separated from those from the other sites.Only one of the 57