INTRODUCTION
Various species and ecomorphotypes/subspecies of the genus Equus are recorded in several Early Pleistocene southern European local faunal assemblages (LFAs). Their taxonomy, phylogenetic relationships, and chronological distribution are, however, still a matter of debate (for different opinions on systematics and taxonomy, see for example Azzaroli, Reference Azzaroli1965, Reference Azzaroli1979, Reference Azzaroli1990, Reference Azzaroli1999; Boeuf, Reference Boeuf1986, Alberdi et al., Reference Alberdi, Caloi and Palombo1988, Reference Alberdi, Ortiz Jaureguizar and Prado1998, Reference Alberdi, Ortiz Jaureguizar and Prado2001; Musil, Reference Musil1992; Caloi, Reference Caloi1997; Forsten, Reference Forsten1999; Eisenmann, Reference Eisenmann2004, Reference Eisenmann2010; Alberdi and Palombo, Reference Alberdi and Palombo2013a, Reference Alberdi and Palombo2013b; and references in those papers). The horses most commonly recorded in the middle and early late Villafranchian European Land Mammal Ages (ELMAs; Gelasian and pre-Jaramillo Calabrian epochs) are those of “Equus stenonis group,” to which a number of alleged subspecies have been ascribed (e.g., among others E. stenonis livenzovensis, E. s. stenonis, E. s. vireti, E. s. guthi, E. s. olivolanus, E. s. senezensis, E. s mygdonoensis, E. s. perimensis). The E. stenonis group includes quite large and stout horses recorded in several Eurasian localities. Specimens ascribed to the group sometimes slightly differ in size and proportions, depending on ecogeographic and possibly chronological factors. During the first part of the Early Pleistocene (Gelasian epoch), horses larger or smaller than those belonging to the E. stenonis group are not frequent. Their relationships with taxa of the “E. stenonis group” have widely been discussed by authors (see e.g., Alberdi et al., Reference Alberdi, Ortiz Jaureguizar and Prado1998; Forsten, Reference Forsten1999; Eisenmann, Reference Eisenmann2006). In particular, authors disagree on the phylogeny and the taxonomic rank of the middle-sized horses identified as Equus stehlini, typically represented by specimens found in Italy (Azzaroli, Reference Azzaroli1965), and Equus senezensis, a species created for the specimens from the pre-Olduvai locality of Senèze (France) (Prat, Reference Prat1964). Morphometric multivariate analyses and the application of a “neighbour-joining-method” (Saitou and Nei, Reference Saitou and Nei1987) performed by Alberdi et al. (Reference Alberdi, Ortiz Jaureguizar and Prado1998) showed that E. senezensis is a bona fide species, which includes the nominal ecomorphotypes/subspecies from Senèze (E. senezensis senezensis) and E. senezensis stehlini, recorded in Italy in some slightly younger LFAs (late Villafranchian ELMA, post-Olduvai/pre-Jaramillo Calabrian) (e.g., Casa Frata, Upper Valdarno, and Farneta, Val di Chiana; Azzaroli, Reference Azzaroli1965, Reference Azzaroli1984; Caloi and Palombo, Reference Caloi and Palombo1982). According to the taxonomic assessment proposed by Alberdi et al. (Reference Alberdi, Ortiz Jaureguizar and Prado1998), six stenonoid species were present in Europe during the Early Pleistocene: Equus livenzovensis, Equus stenonis, Equus senezensis, Equus altidens, Equus suessenbornensis, and Equus major. The species belong to two monophyletic sister groups (the group of E. stenonis, E. senezensis, and E. altidens, and the group of E. major and E. suessenbornensis) possibly stemmed from E. livenzovensis.
During the Gelasian epoch, from 2.588±0.005 to 1.806±0.005Ma (middle and early late Villafranchian ELMA), in southern Europe E. livenzovensis is reported in the oldest LFAs (e.g., Montopoli, Italy and Huélago and El Rincón, Spain). E. major is recorded in various localities by a few remains, sometimes found together with more abundant remains of a smaller stenonoid horse, and E. senezensis is recorded only from Senèze. Various subspecies (ecomorphotypes) of E. stenonis are recorded in several LFAs ranging in age from about 2.4 to 1.6 Ma. E. s. stenonis is recorded, for instance, in Italy in various localities of the Upper Valdarno Basin and at Olivola (Val di Magra), E. s. vireti is reported in French middle Villafranchian LFAs (i.e., at Saint-Vallier), E. s. guthi is recorded in the French Chilhac and Spanish La Puebla de Valverde LFAs, and a large and fairly stout E. stenonis morphotype is present at Sésklo in Greece (see Athanassiou, Reference Athanassiou2001).
This article aims to increase knowledge on the European pre-Olduvai stenonoid horses by describing and discussing the equid remains from the middle Villafranchian (Gelasian) Italian site of Coste San Giacomo (CSG; Anagni) (Fig. 1). The sample includes a few specimens already identified as E. stenonis (Biddittu et al., Reference Biddittu, Cassoli, Radicati di Brozolo, Segre, Segre Naldini and Villa1979) and some bones found during recent field activities and excavations carried out in 2011 and 2013 by an Italian team of researchers led by Fabio Parenti (Italian Institute of Human Palaeontology [IsIPU]) and Raffaele Sardella (Earth Science Department of Sapienza University of Rome). The field activities (2011 drilling and 2013 excavations) confirmed that the CSG vertebrate assemblage comes from a single fossiliferous level (at 5 m depth in the drilled sequence). The accumulation of bones was possibly related to the presence of river channels in a fluviolacustrine environment (for a more detailed description of the site, see Bellucci et al., Reference Bellucci, Mazzini, Scardia, Bruni, Parenti, Segre, Naldini and Sardella2012). Bellucci et al. (Reference Bellucci, Bona, Corrado, Magri, Mazzini, Parenti, Scardia and Sardella2014) indicate for the CSG vertebrate fauna an approximate age of 2.1 Ma, based on magnetostratigraphy, pollen, and small mammals.
Figure 1 (colour online) The Coste San Giacomo site: Geographic location and view of the excavation area during the 2013 campaign. The core is modified from Bellucci et al. (Reference Bellucci, Mazzini, Scardia, Bruni, Parenti, Segre, Naldini and Sardella2012).
METHODS
The studied material from the CSG LFA belongs to IsIPU collections stored at Anagni (Frosinone, Italy). The sample mainly consists of isolated teeth (33 upper and 46 lower cheek teeth, more or less complete, and several fragments) and a few (29) long and short limb bones, mostly either incomplete or badly preserved (Supplementary Tables 1a, 1b, and 2). Neither skulls nor jaws have been found thus far.
The morphology and dimensions of the CSG horse were compared with those of the most significant samples of stenonoid horses found in various European LFAs (Spain, France, and Germany), dating from the Early to the early Middle Pleistocene. The compared material belongs to different middle and early late Villafranchian ecomorphotypes/subspecies of E. stenonis (E. s. vireti from Saint-Vallier [SV], E. s. guthi from Chillac [CH] and La Puebla de Valverde [LPV]) and E. senezensis (E. sen. senezensis from Senèze [SE], E. sen. stehlini from Upper Valdarno [UV] and Casa Frata [CF]). The sample from CSG was also compared with samples of E. altidens, a latest Villafranchian-Galerian species sharing with the CSG horse the slightly small dimensions and the quite slender phalanges. In particular, the CSG sample was compared with E. altidens from the early Middle Pleistocene type locality of Süssenborn (SUE; Germany) (Musil, Reference Musil1969, Reference Musil2001), E. altidens from Pirro Nord (PN) (Alberdi and Palombo, Reference Alberdi and Palombo2013a, Reference Alberdi and Palombo2013b), and E. altidens ecomorphotypes/subspecies recorded in the Guadix-Baza basin (Spain) (E. a. granatensis from Venta Micena [VM] [Alberdi and Ruiz Bustos, Reference Alberdi and Ruiz Bustos1985; Alberdi et al., Reference Alberdi, Ortiz Jaureguizar and Prado1998], Barranco León-5 [BL-5], and Fuente Nueva-3 [FN-3] [Alberdi, Reference Alberdi2010] and E. a. altidens from Huéscar-1 [HU-1] and Cúllar de Baza-1 [CB-1] [Alberdi and Ruiz Bustos, Reference Alberdi and Ruiz Bustos1989; Alberdi et al., Reference Alberdi, Ortiz Jaureguizar and Prado1998]), as well as with the French equids from Solilhac (SOL) where Prat (Reference Prat1980) identified E. suessenbornensis and possibly E. altidens. E. major and E. suessenbornensis, although initially included in the statistical analysis, were successively excluded from the comparison because both have a much larger size and more robust proportions in comparison with those of the CSG horse.
Comparative data were collected by two of us (MTA and MRP) in several museums and institutions in Italy (Museo Preistorico-Etnografico “L. Pigorini,” Rome; Museum of Geology and Palaeontology, University of Florence; Earth Science Department, University of Turin; Museum of Palaeontology, Sapienza University, Rome; Natural History Museum, Verona) (MTA and MRP, unpublished data), France (Musée Guimet, Lyon; University of Poitiers; Musée Préhistorique of Monaco), Spain (Museo Nacional de Ciencias Naturales, Madrid; Museo de Prehistoria J. Gibert, Orce, Granada), and Germany (Research Station of Quaternary Palaeontology, Weimar) (MTA, unpublished data).
Nomenclature and measurements follow the recommendations of the “Hipparion Conference,” New York, November 1981 (Eisenmann et al., Reference Eisenmann, Alberdi, De Giuli and Staesche1988; Supplementary Figures 1 and 2). All dimensions are expressed in millimetres.
Morphological and morphometric analyses and comparisons are mainly based on teeth, astragali, and proximal phalanges (1PHIII) because of the lack in the CSG sample of skulls, mandibles, and well-preserved long bones. The hypsodonty indices were calculated on unworn or only slightly worn teeth (P3-4/p3-4 and M1-2/m1-2) as H/L ratios, where H is the maximal height of the tooth and L is the maximal length at 1 cm of the base of the crown. The robustness and/or slenderness indices of third metacarpals (MCIII) and metatarsals (MTIII) were not calculated because of the lack of any complete metapodial.
To estimate variation and differences in size and proportions in cheek teeth, four groups were analysed: P3-4, M1-2, p3-4, and m1-2. Length versus width dimensions (measured at 1 cm above the base of the crown) and ratios of tooth length versus protocone length in upper teeth and tooth length versus postflexid length in lower teeth were plotted in bivariate diagrams. To estimate variation and differences in size and proportions among the astragali, proximal phalanges of the third digit, and proximal and distal epiphyses of MCIII of the CSG horse and stenonoid horses, we performed multivariate analyses using as variables the measurements obtainable for the CSG specimens and those of samples from key Early to early Middle Pleistocene European sites (MTA and MRP, unpublished data). A principal component analysis (PCA) was performed to assess relationships among the Equus remains from CSG and horses from the other localities analysed herein (SV, CH, LPV, SE, UV, CF, SUE, PN, VM, BL-5, FN-3, HU-1, CB-1, and SOL LFAs). The PCA analysis enabled us to separate specimens into groups according to their affinity in dimensions/proportions. The variables that have the highest weight in the PCA components represent the characters that mainly support the aggregation of samples in groups as well as the separation among groups. A discriminate analysis (DA) was then performed to evaluate whether centroids of the groups previously identified by PCA differ significantly from each other. DA may also suggest how to identify specimens that were not included in the groups resulting from the PCA analysis. The PCA analysis enables us to group the remains on the basis of their similarities/differences in size, while DA shows how significant such differences are (Marcus, Reference Marcus1990; Reyment, Reference Reyment1991). Calculations were made using the software SPSS 15.0.
The measurements included as variables in the PCA and DA analyses are those of the CSG specimens and of 99 proximal articulations of third metacarpal+length, 135 distal articulations of third metacarpal, 90 astragali, and 112 proximal phalanges of Equus specimens from the localities mentioned previously.
Morphological description
Biddittu et al. (Reference Biddittu, Cassoli, Radicati di Brozolo, Segre, Segre Naldini and Villa1979) based their identification of Equus remains from CSG as E. stenonis, mainly on the geologic age inferred from the stratigraphic succession cropping out at the site and stenonoid characters of teeth. The teeth of the CSG horse actually show some stenonoid morphological traits. The dimensions, however, are slightly smaller than those of E. stenonis sensu stricto (Supplementary Tables 1a,b; 2).
Premolars and molars
The occlusal surface of upper chewing teeth (Fig. 2a–g and Supplementary Table 1a) shows a protocone nearly triangular in shape, flattened or somewhat lingually indented; the distal part is slightly more extended than in E. stenonis, especially in molars. A single pli caballin is often present, both in premolars and molars. The enamel of fossettes is moderately folded. Parastyle and mesostyle are wide in premolars and narrow in molars. The hypocone is generally open and triangular in shape, with marked distal and slightly evident lingual grooves. The grooves are more pronounced in the less worn teeth. The hypocone of some M3 is isolated inside the loph. Premolars are bigger than molars; both are high crowned. The hypsodonty index of molars ranges from 2.52 to 2.58 in M1-2, whereas it is not measurable in premolars because the wear is too advanced.
Figure 2 (colour online) Selected upper and lower teeth: right P3-4 89186 (a); right P3-4 13-481 (b); right P3-4 13-411 (c); left M1-2 11-24 (d); right M1-2 13-221 (e); right M1-2 13-276 (f); left M3 98175 (g); right p2 56674 (h); right p2 98176 (i); right p2 98178 (j); right p3-4 18-68 (k); right p3-4 91188 (l); right p3-4 98199 (m); right p3-4 11-28/1 (n); left m1-2 13-381 (o); and right m1-2 981101 (p). Scale bars 10 mm.
The occlusal morphology of lower teeth (Fig. 2h–p and Supplementary Table 1b) is of a “stenonian” type, characterised by a double knot with rounded metaconids, pointed metastylids, and “V-shaped” linguaflexids. The ectoflexid of premolars does not enter the isthmus, whereas in molars it is deep and penetrates the isthmus contacting the linguaflexid. The ectoflexid shows a small pli, more marked in premolars than in molars and more evident on the little-worn teeth. As in upper teeth, premolars are bigger than molars and high crowned. It was possible to estimate the hypsodonty index only on a p3-4 (≈2.33) and an m1-2 (≈2.32).
Limb bones
The dimensions of the few remains of the appendicular skeleton (Fig. 3 and Supplementary Table 2), in particular those of astragalus, phalanges, magnum, and triquetum indicate that the CSG equid was a middle-size horse with no particularly slender limbs, as suggested by the proportions of a nearly complete metacarpal (CSG-981106) (Fig. 3a). In the MCIII the distal articulation is broken, but the supra-articular tuberosities are clearly visible, making it possible to infer the approximate length and proportion (gracility index ≈ 18).
Figure 3 (colour online) Selected postcranial bones. (a) Left third metacarpal bone 981106: proximal view (a1), cranial view (a2), and caudal view (a3). (b) Proximal phalanx III 13-386: proximal view (b1), cranial view (b2), and caudal view (b3). (c) Proximal phalanx III CSG198-56679: proximal view (c1), cranial view (c2), and caudal view (c3). (d) Left astragalus 981109: cranial view (d1) and caudal view (d2). (e) Left astragalus 11-97: cranial view (e1) and caudal view (e2). (f) Distal phalanx III 11-41: proximal view (f1) and lateral view (f2). Scale bars 50 mm.
In the astragalus (Fig. 3d and e), the articular surface for the calcaneus shows two facets delimited by an edge but confluent each other, while a third facet is isolated. The latter is polished and less defined in CSG-78/981109 and clearly delimited in CSG-11/97 (Fig. 3d2 and e2). The proximal phalanges (Fig. 3b and c) have a wide insertion for the trigonum phalangis, more marked in the posterior CSG-198/56679 than in the anterior CSG-13/386 phalanx (identified following Prat [Reference Prat1957]). The surface of the proximal articulation is more quadrangular in the posterior phalanx (CSG-198/56679) and more oval in the anterior (CSG-13/386) (Fig. 3b1–3 and c1–3). The dimensions of the fragments of humerus, radius, and tibia confirm the relatively small size of the CSG horse (Supplementary Table 2).
The morphology and dimensions of teeth and limb bones (though largely fragmented) are consistent with each other, suggesting the presence at CSG of only one middle-sized species showing various stenonian morphological traits.
Statistical bivariate and multivariate analysis
A statistical analysis was performed to compare the CSG sample with nearly all the Early to early Middle Pleistocene stenonoid horse samples from the most important European sites (E. livenzovensis, E. s. stenonis, E. s. vireti, E. s. guthi, E. sen. senezensis, E. sen. stehlini, E. a. altidens, E. a. granatensis, E. major, E. suessenbornensis). Results obtained stress the difficulty of separating samples attributed by some authors to separate subspecies, as well as the large variation characterising most of the samples. The range of variation of some samples of middle-sized horses partially overlaps each other. This is possibly attributable to the presence in the analysis of samples of large-sized horses (i.e., E. major, E. suessenbornensis) that hampers any clear separation among groups having a similar size. Therefore, to better investigate the affinity (if any) of the CSG horse with one or another species/ecomorphotype, the CSG specimens were compared with samples of horses selected because they are close in age or in size/proportions to the CSG ones—that is, subspecies/ecomorphotypes of E. altidens (E. a. altidens and E. a. granatensis), E. senezensis (E. sen. senezensis and E. sen. stehlini), and E. stenonis (E. s. stenonis from UV, E. s. vireti from SV and LPV, and E. s. guthi from CH).
In all these analyses, teeth and bones from CSG fall in the range of the specimens belonging to small/middle-sized horses.
In the bivariate dispersion diagrams of the length versus the width of upper (P3-4 and M1-2) and lower (p3-4 and m1-2) cheek teeth, most of the CSG teeth generally fall in the variation range of specimens having small-sized teeth, close to the E. sen. stehlini specimens from UV and CF, whereas the largest CSG teeth fall in the range of the smallest premolars and molars of the E. altidens (Fig. 4). Results indicate some variation in the tooth sample of the CSG horse (Fig. 4); for example, the square is higher than the diagrams where protocone and postflexid lengths (measured on the occlusal surface) are plotted against the tooth length. It is difficult to ascertain the actual meaning of such a variation. It is important to note, however, that the measure of the protocone and postflexid length has to be regarded as indicative because it may depend on the degree of wear of the tooth.
Figure 4 Bivariate diagrams of the length versus breadth at 1 cm base of P3-4, M1-2, p3-4, and m1-2 of the equids from Coste San Giacomo (CSG; Italy), compared with several stenonoid horses from Italy, France, Spain, and Germany. Abbreviations: BB/bb, breadth at base; LB/lb, length at base; capital letters, upper teeth; lowercase letters, lower teeth.
In the PCA loading plots of principal component 1 (PC1) and principal component 2 (PC2) obtained by using as variables some selected measurements of limb bones, the few remains from CSG fall in the same quadrant including most of the specimens of small/middle size (i.e., E. senezensis and E. altidens).
In the PCA using dimensions of the proximal epiphysis of MCIII and its approximate length, the variance accumulated by the three first components reached 86.33%. The variables with more weight in the first component are of the MCIII5, MCIII7, and MCIII4 dimensions of the proximal epiphysis; in the second component, the variable is MCIII1; and in the third, they are MCIII6 and MCIII3 (Fig. 5 and Supplementary Table 3). Results obtained point to a fair separation between two groups that include, respectively, E. altidens and E. senezensis. It is worth noting that the MCIII CSG-1979/981106, the only metacarpal for which an inferred length is available, generally falls in the space between the groups including E. altidens and E. senezensis specimens.
Figure 5 Principal component analysis diagrams of the skeleton remains from Coste San Giacomo (CSG, Italy) compared with several stenonoid horses from Italy, France, Spain, and Germany. Abbreviations: MCIII prox.+L, proximal articulation of the metacarpal plus maximal length; MCIII distal, distal articulation of the metacarpal; 1PHIII, proximal phalanx of the central digit.
The multivariate analysis of measurements of MCIII distal epiphyses did not yield results completely consistent with those obtained from the analysis of MCIII proximal epiphysis. These data, however, have to be considered with caution because some mistakes may have occurred in the identification of some fragments. It is objectively difficult to separate the distal epiphyses of metacarpals from those of metatarsals when the metapodials are not complete, as occurs in the CSG sample. In the PCA of distal MCIII, the first three components account for 94.44% of the variance. The variables that have higher variances and more weight in the analysis are MCIII12 and MCIII11 measurements for the first component, MCIII13 and MCIII10 for the second component, and MCIII14 for the third component (Fig. 5 and Supplementary Table 3). In the PCA loading plot of PC1 versus PC2, specimens belonging to middle-sized species (E. altidens and E. senezensis) fall in the same space, whereas E. stenonis specimens are quite separate. The CSG specimens fall in the E. stenonis group. As noted previously, the result has, however, to be considered with caution because the sample of E. stenonis consists of complete bones, whereas only broken MCIII were found at CSG.
In the PCA of astragalus, the three first components accumulated a variance of 92.03%. The first component is mainly influenced by the variables corresponding to AST2, AST7, AST4, and AST5 measurements; in the second component, the main variable is AST3; and in the third component, it is AST1 (Fig. 5 and Supplementary Table 3). In the PCA loading plot of PC1 versus PC2, the groups including E. altidens, E. senezensis, and specimens from CSG are fairly separate from that of E. stenonis. In the PCA obtained for the proximal phalanx, the three first components accumulated a variance of 85.77%. The variables that have a higher variance and more weight in the analysis are 1PHIII6, 1PHIII7, and 1PHIII4 measurements for the first component; 1PHIII2, 1PHIII1, and 1PHIII8 for the second component; and 1PHIII9 for the third component (Fig. 5 and Supplementary Table 3). In PCA loading plot of PC1 versus PC2, most of E. altidens specimens group together mainly falling in the first quadrant, whereas those of the large and stout E. stenonis fall in the fourth quadrant. Specimens from CSG fall in the group gathering E. senezensis and smaller and slender specimens from Chilhac, close to the limit of the range of E. altidens (Fig. 5).
The results of PCA performed by comparing the CSG sample to those of stenonoid species as identified by some authors (cf. Alberdi et al., Reference Alberdi, Ortiz Jaureguizar and Prado1998; Alberdi, Reference Alberdi2010; Alberdi and Palombo, Reference Alberdi and Palombo2013a, Reference Alberdi and Palombo2013b) suggest that the CSG horse shared with E. sen. stehlini, E. a. altidens, and E. a. granatensis at least the size and some proportions of the astragalus and proximal phalanges.
To validate this hypothesis, a DA was performed. DA, a statistical analysis to predict a categorical dependent variable (grouping variable) by one or more continuous binary independent variables (predictor variables), is indeed commonly used to test the reliability of groups that are known a priori (see “Methods”). The results of DA indicate the following: with regard to MCIII, the proximal and distal epiphyses were correctly identified in 76.5% and 64.4% of cases, respectively, with a cross validation in 69.4% and 62.2% of cases (Fig. 6); with regard to the astragalus (less significant than other limb bones to separate species because of its quite conservative morphology), the identification was correct in 72.2% of cases, with a cross validation in 57.8% of cases (Fig. 6); with regard to the 1PHIII, the identification was correct in 75.0% of cases, with a cross validation in 59.8% of cases. It is worth noting that the DA of 1PHIII stresses the separation between anterior and posterior phalanges (Fig. 6). The results of DA confirm some affinity of the CSG horse with middle-sized equids, in particular with E. senezensis, but highlight the difficulties of a sound species identification if the studied sample includes scanty and poorly preserved specimens. This may account for the apparently inconsistent results obtained for the CSG sample. The nearly complete metacarpal (CSG-981106 left), indeed, falls in the group of E. sen. stehlini, whereas two distal portions of MCIII cluster with E. sen. senezensis (CSG-13/343 and CSG-91.2) and two others with E. s. guthi (CSG-1980/56678.1 and 1979.3/981237), possibly because of their slightly large dimensions. Among astragali, one (CSG-11/97) is close to E. sen. senezensis, and two others (CSG-78/981109 and CSG-99/56677) are close to E. sen. stehlini. The two 1PHIII group in one case (CSG-198/56679) with E. a. altidens, and in the other (CSG-13/386) with E. sen. stehlini.
Figure 6 Discriminate analysis diagrams of the skeleton remains from Coste San Giacomo (CSG; Italy) compared with several stenonoid horses from Italy, France, Spain, and Germany. Abbreviations as in Figure 5.
DISCUSSION
Except for large-sized equids such as E. major, the middle Villafranchian (pre-Olduvai) European species show a rather large variation in morphology and dimensions, perhaps related to the characteristics of the environments, which accounts for the differences characterising local populations/ecomorphotypes across space and during time. It is difficult to establish whether the differences among samples from different localities may depend on genetic/taxonomic, environmental, and/or chronological factors. As a result, the systematics and taxonomy of the middle-sized Early Pleistocene horses is controversial. This increases the difficulty of any identification when the studied sample is not statistically significant and some diagnostic elements such as skull, mandible, tooth series, and complete metapodial bones are lacking. The sample from CSG consists of isolated teeth and incomplete bones (i.e., all metapodials); therefore, the taxonomic identification is challenging. The teeth of the different stenonian species show a considerable uniformity of some morphological features (e.g., in the upper teeth the morphology of the protocone and the double knot in the lower ones). Therefore, the stenonoid morphology of CSG teeth is not so useful for a specific identification, whereas the size, slightly less than that of E. stenonis sensu stricto, suggests some affinities with middle-sized stenonoid species as supported by the bivariate diagrams (Fig. 4). In addition, the morphology of the complete limb bones, astragalus and phalanges, is not particularly diagnostic in identifying stenonoid horses. The astragali of the stenonian species show differences in size more than in their overall morphology. Moreover, the distinction between anterior and posterior phalanges may be poorly convincing if the sample is scarce (as the CSG is), and the actual range of variation of the shape of the proximal articular surface is unknown. This, in turn, makes it difficult to evaluate the actual gracility/robustness of phalanges. The complete metapodials are the most diagnostic bones; the broken proximal epiphysis may provide some valuable information, whereas the distal ones are less or no more informative because of the difficulty of disentangling between metacarpal and metatarsal distal epiphyses.
Taking into account these limitations and considering the indication given by the qualitative, quantitative, and statistical analyses, in particular the multivariate ones, the CSG remains seem to be closer to E. sen. stehlini and E. altidens than to other stenonian horses. Given the middle Villafranchian (MN 17) age of the CSG mammal assemblage (e.g., Biddittu et al., Reference Biddittu, Cassoli, Radicati di Brozolo, Segre, Segre Naldini and Villa1979; Gliozzi et al., Reference Gliozzi, Abbazi, Ambrosetti, Argenti, Azzaroli, Caloi and Capasso Barbato1997; Palombo et al., Reference Palombo, Azanza and Alberdi2003; Palombo, Reference Palombo2009; Bellucci et al., Reference Bellucci, Mazzini, Scardia, Bruni, Parenti, Segre, Naldini and Sardella2012, Reference Bellucci, Bona, Corrado, Magri, Mazzini, Parenti, Scardia and Sardella2014; and references therein), the identification of the CSG horse as E. altidens, a species that is recorded in Europe from the post-Olduvai Early to the early Middle Pleistocene (Alberdi and Palombo, Reference Alberdi and Palombo2013a, Reference Alberdi and Palombo2013b; Palombo, Reference Palombo2014, Reference Palombo2016a, Palombo, 2016b), seems to be quite unreasonable. The hypothesis that the CSG horse may belong to a species different from E. altidens is also supported by the dimensions of the MCIII (CSG-981106), whose inferred length is lower than the minimum recorded for E. altidens.
Bellucci et al. (Reference Bellucci, Bona, Corrado, Magri, Mazzini, Parenti, Scardia and Sardella2014) suggest an age around 2.1 Ma for the CSG faunal assemblage. Therefore, the age may be close to that of the French site of Senèze, the type locality of E. senezensis, dated to about 2.093 Ma±10 ka to 2.206 Ma±21 ka (Nomade et al., Reference Nomade, Pastre, Guillou, Faure, Guérin, Delson, Debard, Voinchet and Messager2014). The Senèze fauna shares with CSG a number of species commonly recorded in the middle Villafranchian (e.g., among others Anancus avernensis and Gazella borbonica) and the presence of a Canis representative, while the giant hyaena Pachycrocuta, seen as typical of the late Villafranchian faunal assemblages, is absent at both sites. Thus far, in Italy E. senezensis has never been recorded in any middle Villafranchian LFA. The apparently Italian endemic subspecies E. sen. stehlini has been recorded only in faunal assemblages younger than the Olduvai submagnetochron (Napoleone et al., Reference Napoleone, Albianelli, Azzaroli, Bertini, Magi and Mazzini2003), such as Casa Frata and other localities of Upper Valdarno (Borselli et al., Reference Borselli, De Giuli, Ficcarelli and Mazzini1980; Caloi and Palombo, Reference Caloi and Palombo1982; Azzaroli, 1990; Caloi, Reference Caloi1997; Alberdi and Palombo, Reference Alberdi and Palombo2013a). Representatives of the E. senezensis lineage are reported from a few French and Italian localities, though small horses resembling the European ones found at Casa Frata, Valdarno, and Senèze were present in eastern Africa (e.g., Garba IV-Melka-Kunturé, Ethiopia, dated at about 1.5 Ma) (Geraads et al., Reference Geraads, Eisenmann and Petter2004). The European record of E. senezensis ranges in age from about 2.1 to 1.5 Ma, a period during which E. stenonis was widely spread in Eurasia (see, e.g., Azzaroli, Reference Azzaroli1987; Alberdi et al., Reference Alberdi, Ortiz Jaureguizar and Prado1998; Forsten, Reference Forsten1999). Whether the rarity of E. senezensis may be related to environmental factors, perhaps more favourable to heavier and stouter horses, is difficult to ascertain.
If the aim is to verify to what extent (if any) the presence/diffusion of E. stenonis and/or E. senezensis may have been affected by local ecological conditions, some hints should come from a comparison among the faunal structure and the environmental characteristics of sites where these stenonoid horses are recorded. The available data useful for inferring the ecological structure of LFAs from some key Early and early Middle Pleistocene sites (Fig. 7) (Saint-Vallier, CSG, and Chilhac II referred to V3 Faunal Complex [FC] by Palombo [Reference Palombo2014, Reference Palombo2015]; and La Puebla de Valverde [?V3 FC]; Fonelas and Senèze [?V3-V4]; Olivola, Poggio Rosso, Valdarno 2, and Casa Frata [V4 FC]; and Selvella, Venta Micena, and Pirro Nord [V5a FC]) bearing stenonoid horse remains (i.e., Equus stenonis, Equus senezensis, and Equus altidens) were compared to infer what conditions may have promoted/hampered the diffusion of different middle-sized stenonoid species throughout southwest Europe.
The palaeoenvironment of the middle Villafranchian site of Saint-Vallier, where equid remains have been ascribed to E. s. vireti, was characterized by some development of wooded areas and tree vegetation as indicated, for instance, by the presence of the cercopithecoid Macaca, and by the dietary behaviour of some herbivores (Palombo and Valli, Reference Palombo and Valli2005; Valli and Palombo, Reference Valli and Palombo2008). The palynological analysis at the oldest level LD3 indicates a partially wooded steppe environment close to a water body (Guérin et al., Reference Guérin, Faure, Argant, Argant, Crégut-Bonnoure, Debard and Delson2004). The horse remains from this level were ascribed to a robust middle-sized horse (average weight of 400 kg), whereas the remains from the oldest level belong to a less stout species, with morphological features intermediate between the horse from the Spanish site of El Rincón (V2 FC) and the most recent Saint-Vallier horses, with an average weight of 325 kg (Eisenmann, Reference Eisenmann2004) and 466 kg (Alberdi et al., Reference Alberdi, Prado and Ortiz‐Jaureguizar1995), respectively. Also, in the oldest SV level scanty remains of a smaller and more slender horse, sharing some morphological traits with morphotypes inhabiting arid environments, are present (cf. Eisenmann, Reference Eisenmann2004). The autoecological analysis performed by Alcalde (Reference Alcalde2013) on ungulates from Saint-Vallier and La Puebla de Valverde (Spain) indicates a mosaic landscape for both sites, with a predominance of arid plains, but dotted with forest patches and steep areas. These forested areas are not very extensive but with well-developed shrubs would probably be situated in the lower parts of the rough zones, where climate tends to be warmer and more humid because of the geographic barriers. In spite of the environmental similarity of SV and LPV, the equid from LPV (E. s. guthi according to Alberdi et al. [Reference Alberdi, Ortiz Jaureguizar and Prado1998]), which shares a number of morphological features with E. s. guthi from Chilhac II, is more robust than E. s. stenonis but slightly smaller and more slender than E. s. vireti. The overall morphology and the size of limb bones of the LVP horse, however, are more similar to those of E. s. vireti than E. s. stenonis.
Besides those from CSG, remains of a smaller horse are reported shortly before the beginning of the Olduvai chron at the French site of Senèze, in levels dated to about 2.21 and 2.09±0.02 Ma (Nomade et al., Reference Nomade, Pastre, Guillou, Faure, Guérin, Delson, Debard, Voinchet and Messager2014). Palynological data suggest that at the time of the accumulation of Senèze LFA, open environments with patches of open wood, and wooded areas with a mesothermophilic character were present in the surrounding area (Delson et al., Reference Delson, Faure, Guérin, Aprile, Argant, Blackwell and Debard2006). E. senezensis from Senèze shows more slender and smaller metapodials than in E. stenonis (the slenderness is halfway between E. stenonis and E. altidens). The particular features of metapodial bones of E. senezensis may be related to the reduction of forests in favour of open environments recorded especially during glacial cycles in the Mediterranean region by the end of the Gelasian (e.g., Suc and Popescu, Reference Suc and Popescu2005; Bertini, Reference Bertini2010). Conversely, in E. sen. stehlini recorded in the Italian post-Olduvai Casa Frata LFA (about 1.8–1.75 Ma in Napoleone et al. [Reference Napoleone, Albianelli, Azzaroli, Bertini, Magi and Mazzini2003]), the metapodials are shorter suggesting a more robust horse. The accompanying fauna (counting among others Praeovibos mediterraneus along with Lynx issiodorensis) (Borselli et al., Reference Borselli, De Giuli, Ficcarelli and Mazzini1980) suggests the presence of a mosaic environment including both grasslands and forest.
A similar landscape possibly characterised the Anagni basin at the time of the accumulation of the CSG LFA, possibly deposited in an abandoned fluvial channel (Bellucci et al., Reference Bellucci, Bona, Corrado, Magri, Mazzini, Parenti, Scardia and Sardella2014). The pollen record indicates a forested landscape, though the discontinuous nature of the record makes it difficult to ascertain whether it corresponds to a single forest phase or to multiple expansions. The faunal assemblage and the feeding behaviour of large herbivores indicate a mosaic environment including both savannah and woodland/wetlands (Strani et al., Reference Strani, DeMiguel, Sardella and Bellucci2015). The temperature range, estimated from ostracods, suggests the occurrence of mild climatic conditions in an alluvial environment with clear, running waters, sensitive to the precipitation regime. Whether such a mosaic is composed of different penecontemporaneous environments is, however, difficult to say because of the unknown time averaging of accumulation and the discontinuity of the pollen record. The palaeoenvironmental reconstruction of CSG is reasonably consistent with the vegetation scenario in central Italy during the latest Gelasian, as documented, for instance, at Poggio Rosso (Upper Valdarno), a site sharing a number of taxa with the slightly older site of Olivola (Val di Magra) (cf. Napoleone et al., Reference Napoleone, Albianelli, Azzaroli and Mazzini2001, Reference Napoleone, Albianelli, Azzaroli, Bertini, Magi and Mazzini2003). At Poggio Rosso, the palynological and sedimentological analyses reveal two major climatic fluctuations that occurred just before the Olduvai-Matuyama palaeomagnetic inversion. The fluctuations were correlated to a warmer interglacial phase characterised by a climate with mean annual temperature and precipitation higher than the present day and a glacial phase with low precipitation that reached a minimum value in the level above the fossiliferous bed. The presence in the fauna of forest dwellers, such as Felis silvestris lunensis and L. issiodorensis indicate that woodlands were still present (Bertini et al., Reference Bertini, Magi, Mazza and Fauquette2010). In the Poggio Rosso and Olivola LFAs, only E. stenonis is recorded. The equid of Olivola, type of the species E. stenonis, is a large-sized horse (Alberdi et al., Reference Alberdi, Ortiz Jaureguizar and Prado1998). The long bones are smaller than those of E. livenzovensis but larger than in other subspecies of E. stenonis, though more slender than those of E. s. vireti. The hypothesis that the reduction in size from E. livenzovensis to the late Villafranchian E. stenonis is related to an average increase of the woodlands is partially supported by the morphology and proportion of the acropodium of the Olivola horse. The metatarsals being less slender and hooves less high, with the dorsal “margin” less convex and the dorsal surface less steep than in E. sen. stehlini from Casa Frata (Caloi, Reference Caloi1997), would suggest that E. stenonis had a less pronounced adaptation to hard ground that E. sen. stehlini. This hypothesis, however, needs to be better substantiated.
Some additional clues may come from a comparative analysis of the structure of the key Gelasian and Early Calabrian (middle and early late Villafranchian ELMA) southwest European sites recording stenonoid horses. The functional diversity spectra (FDS) obtained following the methodology and using data as in Palombo (Reference Palombo2016a) show some structural differences between the CSG fauna and the middle Villafranchian ones selected for the analysis (Fig. 8A). Divergences are mainly related to the dominance at CSG of species inhabiting open environments, most of which are mixed feeders, and to some different proportions of predators and prey. Such differences are much more marked when comparing CSG fauna with latest Gelasian–Early Calabrian LFAs yielding E. stenonis remains. These LFAs indicate a higher number of predators and forest dwellers than the CSG LFA, whereas the number of groups based on the mammal size is rather similar.
Figure 7 Biochronological setting and chronological scheme for the Pleistocene mammalian record from SW Europe. VDPB = Vienna Pee Dee Belemnite standard value. (Modified from Palombo, Reference Palombo2016a).
Figure 8 (colour online) Comparison among the functional diversity spectra of Coste San Giacomo fauna and those of selected middle Villafranchian (A) and late Villafranchian local faunal assemblages where either Equus stenonis (B), Equus senezensis (C), and Equus altidens (D) are recorded.
Conversely, the CSG spectrum shows some similarities with the LFAs recording middle-sized horses of the E. senezensis group, in spite of the still higher percentage of open environment dwellers. The similarity is particularly high with the Senèze LFA, close in age to the CSG LFA, while the post-Olduvai LFAs are characterized, once more, by a different prey/predator ratio (Fig. 8B). The same happens with regard to the post-Olduvai/pre-Jaramillo LFAs recording the slender, more advanced horse E. altidens. The spectra obtained for these LFAs share with that of CSG the same pattern as regards the ecological group more related to the vegetation cover. The spectrum of CSG, however, shows a higher percentage of browsers, maybe because the structure of the CSG ecosystem was on average less dry and open and characterised by an extension of forest/woodlands higher than that of the post-Olduvai LFAs selected for the comparison. In addition, the relative percentage of ecological groups related to body mass is more variable as well as the prey/predator ratio (Fig. 8C).
All things considered, the ecological structure of the CSG LFA is more similar to the structure of LFAs including middle-sized, slender horses, roughly independently from their age and geographic distribution of the sites (Fig. 8D). As a result, the hypothesis that some relationships exist among the presence of only one middle-sized or large equid or even more horse species with different size, the environment, and the structure of the accompanying mammalian fauna seems to be the most reasonable. Results obtained by FD analysis, however, have to be considered with caution because of the different depositional context of the analysed sites and the lack of taphonomic studies at some sites, including CSG.
CONCLUSIONS
The horse remains from CSG are scanty and poorly preserved, and the most diagnostic skeletal bones are lacking. Nonetheless, the morphological traits, dimensions, and proportions of teeth and some bones of the CSG horse are sufficient to confidently suggest some affinities with middle-sized stenonoid horses, in particular with E. senezensis. The affinity is partially confirmed by results obtained by statistical bivariate and multivariate analyses according to which CSG astragalus and phalanges group more with E. sen. stehlini and E altidens than with E. stenonis specimens. Moreover, the shape and proportion inferred for the nearly complete MCIII (CSG-981106) suggest a quite stout and robust bone, shorter than E. sen. senezensis and similar to E. sen. stehlini.
All things considered, we may identify the horse from CSG as E. senezensis aff. E. sen. stehlini, though the identification has to be regarded as provisional, pending a confirmation by more compelling material. The presence in the middle Villafranchian of Italy of a horse slightly smaller than the typical E. stenonis found in late Gelasian–Early Calabrian Early Pleistocene Italian LFAs (i.e., Olivola, Matassino, and Poggio Rosso) and strictly related to “E. stehlini” reopens the question as to the taxonomic rank and phylogenetic relationship of this small stenonoid horse with rather stout metapodials and suggests new hypotheses as regards the phylogeny, taxonomy, and systematics of some middle Villafranchian horses. Alberdi et al. (Reference Alberdi, Ortiz Jaureguizar and Prado1998) hypothesised that E. sen. senezensis originated from E. stenonis and that E. sen. stehlini would include populations originated from the Senèze horse. Conversely, some scholars (e.g., Azzaroli, Reference Azzaroli1992 and references therein) suggested that E. stenonis gave origin to E. stehlini (regarded as a bona fide species) by a cladogenetic process. Some others claimed that at the beginning of the Quaternary two horses differing in size were present in western Europe (i.e., at Huélago, Spain) (e.g., Eisenmann, Reference Eisenmann2002). This hypothesis has to be discarded because the apparently different size of the Huélago horse depends on the different presentation status of bones that alters the actual dimensions of the alleged smaller bones.
Available data suggest that three different horse lineages were present in western Europe during the late middle Villafranchian: E. stenonis, the most widespread; Equus major, larger and less abundant; and middle-sized slender horses, recorded by a large sample at Senèze and by scanty, less diagnostic remains in Italy (i.e., CSG) and Greece (e.g., Vatera [Eisenmann, Reference Eisenmann2002; Lyras and van der Geer, Reference Lyras and van der Geer2007] and Gerakorou [Koufos, 1992]).
From this evidence, some alternative hypotheses arise to explain the origin of pre-Olduvai middle-sized horses and their relationship with the early post-Olduvai ones. A first hypothesis would assume that in the early middle Villafranchian two stenonoid lineages stemmed from the European E. livenzovensis group. One lineage would include quite large and ecologically flexible horses, belonging to the E. stenonis group, whose ecomorphotypes/subspecies were recorded in western Europe until the early late Villafranchian. The second lineage would include smaller, more slender and specialized horses, inhabiting rather dry environments with reduced forest cover (E. senezensis group). These horses were present, maybe with different ecomorphotypes, in the pre-Olduvai middle Villafranchian in France (Senèze), and Greece (Vatera, Gerakorou) and Italy (CSG), where they survived till the early late Villafranchian giving origin to a slightly robust ecomorphotype/subspecies (i.e., E. sen. stehlini). As a corollary of the hypothesis, it would be presumed that E. altidens stemmed from E. senezensis (cf. Alberdi et al., Reference Alberdi, Ortiz Jaureguizar and Prado1998).
A second hypothesis would presume that early smaller, more slender and specialized different species (i.e., E. senezensis, E. stehlini, and perhaps Equus sp. from Vatera and “E. s. mygdoniensis” from Gerakorou) independently originated by cladogenic processes from local populations of E. stenonis. Following this hypothesis, the lineage would share similar size and proportion because it is adapted to similar environments. Moreover, these alleged species would have disappeared in France and Greece before the end of the Olduvai subchron, while a species persisted in Italy that served as a refuge area. In this case, the issue of the supposed presence of “E. stehlini” in other European regions (Forsten, Reference Forsten1999) remains unsolved.
Finally, it would be hypothesised that E. altidens was derived from E. stenonis, replacing shortly before the Jaramillo submagnetochron (before the “Epivillafranchian” ELMA according to Bellucci et al. [Reference Bellucci, Sardella and Rook2015], but see Palombo [Reference Palombo2016]), the horses of medium size already present since the middle Villafranchian.
All things considered, available data fail to completely support one or the other alternative hypothesis. Nonetheless, the analysis already performed by Alberdi et al. (Reference Alberdi, Ortiz Jaureguizar and Prado1998) seems to suggest that E. altidens did not originate from E. stenonis but may have stemmed from a slender horse. As a result, questions about the taxonomic rank and actual phylogenetic relationship of the middle-sized, slender middle Villafranchian horses cannot be firmly answered. Moreover, it is difficult to disentangle the complex relationships that undoubtedly existed among environment, the structure of the accompanying mammalian fauna, and the presence of only one middle-sized or large equid or even of horse species with different body size and structure.
Acknowledgments
This research was funded by the Dirección General de Investigación Científica y Técnica of Spain (Project CGL2010-19116/BOS; principal investigator, M.T. Alberdi) and the Italian Ministry of Education, University and Research (MIUR) (Sapienza University 2014 project C26A14BNRM, M.R. Palombo; Sapienza University 2011 project C26A11SNA3, R. Sardella; Sapienza University 2012 project C26A12PZA2, R. Sardella). The Soprintendenza Archeologica per il Lazio is acknowledged for the field activity and research permission. The fieldwork at CSG was supported by MIUR 6/2000 grants (respectively R. Sardella and L. Bellucci), by IsIPU and BancAnagni. We thank Luciano Bruni and all the participants who worked in the excavations and in the field activities.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/qua.2017.6
