Equids from the late Middle Pleistocene to Early Holocene of the Apulia Peninsula (southern Italy): reassessment of their taxonomy and biochronology

ABSTRACT The Apulian Peninsula represents a key-region for the study of climatic changes and paleoenvironmental dynamics during the Quaternary. Both large and small-sized horses are well documented in this region during the Pleistocene and are frequently found associated. The caballoid horses from Middle to Late Pleistocene of Europe show a large intraspecific ecomorphological variation, emphasizing a strong link between skeletal adaptations and specific aspects of the palaeoenvironment. This large variability led to an extended debate concerning the taxonomy of equids and their evolutionary history. In the Italian Peninsula, fossils from the Middle to the Late Pleistocene have been historically referred to several species (or even subspecies), emphasizing the uncertainty of the taxonomic attribution. Here, a large craniodental sample of Equidae fossils from late Middle Pleistocene to the Early Holocene localities of the Apulia Peninsula are described for the first time. The comparison of the protocone index allowed us to determine the first occurrences of Equus mosbachensis von Reichenau, 1903 in the Apulian Peninsula, from a few localities chronologically referred to late Middle Pleistocene. Most of the material from the late Middle to the end of the Late Pleistocene is instead attributed to Equus ferus Boddaert, 1785. The results of this work thus add novel information on the biochronology of Italian fossil equids and their evolutionary history through the Middle and Late Pleistocene.


INTRODUCTION
The Italian Peninsula is considered a crossroads in the Mediterranean area, representing a key area for the study of climatic changes and paleoenvironmental dynamics during the Quaternary. The Apulian region, or Apulian Peninsula (AP; southern Italy), is a "peninsula within a peninsula" due to its isolated position which extends mostly along the latitudinal axis (Fig. 1). The AP maintained this conformation through most of the Pleistocene Epoch and is rich in localities with mammal fossil remains frequently associated to lithic artefacts referred to both the Middle (Middle to Late Pleistocene) and the Upper Palaeolithic (Late Pleistocene to Early Holocene) (Berto et al. 2017;Spinapolice 2018;Sardella et al. 2018Sardella et al. , 2019Zanchetta et al. 2018). The abundance of Middle Pleistocene and Early Holocene sites makes this region a key-territory to study the evolution of the mammal fauna and of the climatic changes that took place during the Quaternary in the Mediterranean area (e.g., Mecozzi & Bartolini Lucenti 2018;Iannucci et al. 2020;Iurino et al. 2020).
The fossil record often only faunal lists are reported for unheated in archeological sites of the AP. With the exception of the material from Tana delle Iene (Conti et al. 2010) and Grotta del Cavallo (Sarti et al. 1998, 2002 horse fossil materials usually are not formally studied and described. Moreover, due to the probable exploitation of the carcasses by human populations (Sarti et al. 1998;Boscato et al. 2006), equid fossil remains are often highly fragmented. Consequently, fossil
material from many archeological localities of the AP is rarely studied from a paleontological perspective. Nevertheless, the isolated and fragmentary craniodental remains can represent an important source of information for the study of the evolution of equids in the AP through the Middle Pleistocene and Early Holocene. Several preliminary studies were conducted on the material from Cardamone (Rustioni 1998), Cava Spagnulo , Grotta del Cavallo (Sarti et al. 1998, 2002, Grotta Paglicci (Boscato 1994), Melpignano (Rustioni 1998) and Tana delle Iene (Conti et al. 2010) where the material were referred to Equus ferus. The material from San Sidero was attributed to the large-sized Equus chosaricus by Eisenmann (1991b) and Equus ferus by De Giuli (1983) and Rustioni (1998). The taxonomic attribution of other material from the AP was mainly based on the chronology of deposits and therefore frequently ascribed to Equus ferus (Table 1).
Equids were considered an important and a very common component of European large vertebrate fauna during the Middle and Late Pleistocene, but their taxonomy remains controversial (Eisenmann 1991b;Forstén 1991;van Asperen 2012; Boulbes & van Asperen 2019). The lack of clear diagnostic morphological characters has led the proliferation of species and subspecies mainly due to the notable size variation observed in the Middle Pleistocene Equus (van Asperen 2012; Boulbes & van Asperen 2019).
For Pleistocene equid taxa, a useful character for taxonomic determination is the protocone index (length of protocone × 100/length of tooth). Caloi & Palombo (1987) effectively utilized this index for differentiating late Early to early Middle Pleistocene Equus material from different Italian localities. Caloi (1997)  This taxon has been instituted by von Reichenau (1915) based on the material from the fossiliferous sites of Mosbach (about 0.5 Ma) (Germany). Although the Mosbach horse (E. mosbachensis) shared several features with the wild horse (E. ferus), it possesses some "archaic" morphological features (presence of the tendon insertion of the anterior brachialis muscle on the inner edge of the diaphysis of the radius, strong supraarticular tuberosities on metapodials) (Bonifay 1980;Eisenmann 1991a;Guadelli & Prat 1995;Hadjouis 1998;Langlois 2005;Boulbes & van Asperen 2019). The affinity of the morphological features between E. ferus and E. mosbachensis suggests a close phyletic relationship between them, supported by the lack of clear craniodental diagnostic characters. This homogeneous overall morphology led some authors (van Asperen 2012) to refer the European Middle Pleistocene sample to a singles species, E. ferus ssp., where the slight morphological and biometric variability is considered to reflect the adaptive responses to local climatic and palaeoenvironmental conditions. These differences could otherwise be considered to be due to normal population variability. However, there is a substantial consensus considering Equus mosbachensis as a distinct taxon because of its large-sized and robust build (Eisenmann et al. 1985;  In Italy, the first occurrence of large-sized and robust caballoid horses is reported in the Faunal Unit of Slivia (~ 0.9 Ma; Epivillafranchian Mammal Age) (Sala et al. 1992;Rustioni 1998;Conti et al. 2010;Bellucci et al. 2014). Palombo & Alberdi (2017) suggested that the caballoid horses first appeared in the Ponte Galeria Unit (Cesi, about 0.7 Ma;Ficcarelli et al. 1997). Moreover, Conti et al. (2010) suggests that the body size of this taxon decreased from the late Middle Pleistocene onwards (e.g., Malagrotta, Torre in Pietra). Other fossils from the Mid-dle Pleistocene of Italy (0.8-0.3 Ma) have been referred Equus caballus ssp., E. cf. E. mosbachensis, Equus caballus malatestai Caloi, 1997 and Equus sp., highlighting the uncertain of the taxonomic attribution and the difficult to reconstruct the phylogeny of this group (Caloi & Palombo 1987;Caloi 1997;Palombo & Alberdi 2017;Strani et al. 2018Strani et al. , 2019. In addition, Berzi (1972)  and Early Holocene (MIS 1). This dataset was also subject to a test for normal distribution verification using the Shapiro test. We evaluated differences in upper and lower teeth in the late Middle Pleistocene to Early Holocene samples, using a linear model (Anova) with corner point parameterization testing the null hypothesis of no-differences between the mean of the MIS 9-8 and the other samples (p.value > 0.05). The statistical analysis was performed using the R software (Team 2000). Very worn teeth were excluded by statistical analysis. In addition, in order to evaluate the dental variability in the considered samples, the protocone index [(length of protocone * 100)/length of teeth] and postflexid index [(length of postflexid * 100)/length of teeth] were compared. The protocone and postlfexid index is calculated as the mean of the upper and lower teeth respectively when the sample includes more than one specimen. The pattern of the protocone index assumes a taxonomical value for the identification of Middle Pleistocene to Early Holocene equids. In E. mosbachensis the protocone index displays an M 3 value higher than M 1-2 one and P 3-4 value higher or sub-equal than M 1-2 one. In E. ferus, the M 1-2 value is higher than both P 3-4 and M 3 . Finally, in E. hydruntinus the values of P 3-4 and M 1-2 are much lower than both E. mosbachensis and E. ferus, whereas those of P 2 and M 3 fall in the variability of the caballoid horse ones.
Finally, the ratio of the length and the breadth of the muzzle could represent an adaptation to climate, where the short and large muzzle should be found in specimens from sites attributed to glacial stage and vice versa (see Crégut-Bonnoure et al. 2018 for discussion). In order to investigate this adaptations we created standard bivariate plots of muzzle breadth at the posterior borders of I 3 (MB) and muzzle length (ML).

IGF
Museum of Natural History of the

MelpIgnano
The karst infilling deposits of Melpignano, locally known as "ventarole", are located in the area of the village of Maglie. These karst deposits were firstly described by Mirigliano (1941), since then several Institutions, as the IsIPU Italian Peninsula, to E. mosbachensis, and the small-sized ones from the Late Pleistocene to E. ferus germanicus.
In this scenario, the taxonomic attribution of the European caballoid horse from Middle to Late Pleistocene (300-12 Ka) is still a controversial topic. The proliferation of the subspecies reflects the homogeneous overall morphology and the lack of clear diagnostic characters, which should allow the identification of the different taxa.

MATERIAL AND METHODS
The sample of Equus studied in this work are part of several collections stored in different Italian Institutions and Museums ( Table 1).
The skull and dental features of Middle to Late Pleistocene Equus has been poorly investigated, which prevents a large morphological comparison of the studied material. A morphological description of the studied crania from San Sidero and Cardamone has been reported in Appendix 1.
Following Eisenmann (1981), 5 cranial variables have been considered to be important for discriminating species of Equus: basilar length (BL), muzzle length (ML), length of the check teeth (P 2 M 3 LL), facial length (FaL), frontal length (FrL), muzzle breadth at the posterior borders of I 3 (MB) (Appendix 1). Following Eisenmann (1980), we also measured 24 for upper and lower teeth: length, breadth and length of protocone of the upper teeth (P 2 L, P 2 B, P 2 Pr, P 3-4 L, P 3-4 B, P 3-4 Pr, M 1-2 L, M 1-2 B, M 1-2 Pr, M 3 L, M 3 B, M 3 Pr) and length, breadth and length of postflexid of the lower teeth (P 2 L, P 2 B, P 2 Pf, P 3-4 L, P 3-4 B, P 3-4 Pf, M 1-2 L, M 1-2 B, M 1-2 Pf, M 3 L, M 3 B, M 3 Pf ). The measurements were taken in occlusal view to the nearest 0.1 mm with a digital caliper. We used literature data on fossil horses from the Late Pleistocene to Early Holocene of the AP (Table 1).
We further assessed the degree of affinity between late Middle to Late Pleistocene Equus from Apulia by conducting a statistical analysis. First, in order to explore the affinity among the samples from different fossiliferous sites, we compared the upper and lower teeth from San Sidero to those from the other deposits. Specifically, we considered the length of the upper and lower premolars (P 2 , P 3-4 , P 2 and P 3-4 ) and molars (M 1-2 , M 3 , M 1-2 and M 3 ). The dataset was subject to normality distribution verification using Shapiro test. We evaluated differences in length of the upper and lower premolars (P 2 , P 3-4 , P 2 and P 3-4 ) and molars (M 1-2 , M 3 , M 1-2 and M 3 ) in the late Middle Pleistocene and Early Holocene, using linear model (Anova) with corner point parameterization testing the null hypothesis of no-differences between the mean of the San Sidero (SS) and the other samples (p.value > 0.05).
Taking into account the number of sites studied and the small sample size of some of the localities (i.e, Grotta Mario Bernardini -VI, Grotta Mario Bernardini -III, Grotta Mario Bernanrdini -II, Grotta Santa Croce, Grotta Uluzzo -IV, Grotta Uluzzo -II, Grotta Zinzulusa), the dataset has been assembled according to the MIS chronology. We considered the length of the upper and lower teeth of specimens Mecozzi B. & Strani F. and Italian Institute of Prehistory and Protohistory (IIPP), with the support of local Salentine Speleological groups, investigated this area (de Lorentiis 1962; Cardini 1962a). The "ventarole" are generally filled with reddish sediments (called "terre rosse") in the lower part, and brownish sediments (called "terre brune") in the upper, particularly rich in vertebrate fossil remains (Bologna et al. 1994) (Fig. 1). The Equus sample was recovered from the "terre rosse" of the "ventarole" of Mirigliano, Cava Nuzzo and Cava Bianco.

san sIdero
The "ventarole" of San Sidero are located along the state road (SS16) between the villages of Corigliano d'Otranto and Maglie. The first description of the deposit and its faunal assemblage was reported by Cardini (1962a). The mammal fauna from San Sidero was also studied by other authors (De Giuli 1980, 1983Petrucci et al. 2012;Iurino et al. 2013Iurino et al. , 2015. Similarly to the Melpignano sediments, these "ventarole" include "terre rosse" layers in the lower part and "terre brune" layers in the upper part. The Equidae sample was collected from the "terre rosse" of the "ventarole" called SS6 and Cava L.

Repository and studied material
SS6 -IGF one skull ( Fig. 2A), 11 upper teeth, 10 lower teeth (Fig. 3A); Cava L -PF five upper teeth, one hemimandible, six lower teeth. grotta dI Capelvenere The site, located near the town of Santa Caterina (Lecce), occurs in a Cretaceous limestone (Calcari di Melissano Formation) at 20 m a.s.l. and about 100 m from the current seashore. The cave was discovery in 1960 and was only partially excavated in 1971, 1974and 1975(Borzatti von Löwenstern 1961Giusti 1979Giusti , 1980. Outside the cave, a nearby conglomerate deposit at about 8 m a.s.l. has been referred to Tyrrhenian beach (MIS 5) (Patriarchi 1980). The stratigraphic sequence can be divided into two main complexes separated from a speleothem: in the upper part, the brownish sediment, where domestic fauna, ceramics and artefacts appeared, has been referred to Iron age. Instead, in the lower part ten levels including vertebrate fossils and artefacts have been referred to Mousterian (Borzatti von Löwenstern 1961;Giusti 1979Giusti , 1980Patriarchi 1980). The studied sample comes from the lower part of the sedimentary succession.  Zanchetta et al. 2018). Albeit the mammal remains have never been studied in detail, a preliminary mammal list was provided by Borzatti von Löwenstern (1970Löwenstern ( , 1971). Equus is found from different levels (VI, IV, III, II), where also lithic artefacts referred to Middle Palaeolithic from the complexes VI-III and Upper Palaeolithic from the complex II were found (Table 1; Appendix 1).

Repository and studied material
IGF 27 upper teeth, one hemimandible, 27 lower teeth.
grotta uluzzo C The cave is located in the Uluzzo Bay near the village of Nardò, opening into the Cretaceous limestone. The stratigraphic succession was described by Borzatti von Löwestern (1965, 1966 and Borzatti von Löwestern & Magaldi (1969). In particular, the green volcanic sand from the bottom of the complex II could be correlated with the tephra found at the top of the level F of Grotta del Cavallo, dated at 45 700 ± 1000 ka by Zanchetta et al. (2018). Instead, the complex II-I transition, consisting of a grey volcanic sand, could be correlated with the Ignimbrite Campana (CI) identified at the bottom of the level C of Grotta del Cavallo, dated at 39 850 ± 140 ka Zanchetta et al. (2018). The presence of a reworked tephra in the top of the complex IV could represent a marker for the lower deposit. Further investigations needed to confirm the age of these volcanic levels. Moreover, Borzatti von Löwestern (1965,1966) reported a preliminary list of the fossil mammals recovered from this locality. The studied sample come from the complex IV, III and II (

CardaMone
The karst infilling deposit was discovered by Cosimo De Giorgi in 1872 (Botti 1890). The site is located in a region where several quarries are opened for the extraction of a Plio-Pleistocene calcarenite, and, unfortunately, the deposit was destroyed. The mammal assemblage from Cardamone, initially described by Botti (1890), was recently revised by Rustioni et al. (2003). Based on the presence of the wholly rhino (Coelodonta antiquitatis (Blumenbach, 1799)) and the wholly mammoth (Mammuthus primigenius (Blumenbach, 1799)), the association was referred to "Mammuthus-Coelodonta Faunal Complex" and chronologically attributed to climax of the Last Glacial Maximum (22-18 kyr).

Fondo FoCone
The site, discovered during a survey conducted by Decio de Lorentiis in the early 1960s, is located near the village of Ugento. The first excavation campaign was carried out by Luigi Cardini (Cardini 1962b (Boscato et al. 2006). The studied material was recovered from the level D, associated with lithic artefacts attributed to Middle Palaeolithic.

Repository and studied material
IsIPU seven upper teeth, seven lower teeth.

grotta laCeduzza
The cave deposit, located near the village of San Michele Salentino, was discovered by the "Gruppo Speleologico Salentino Pasquale de Lorentiis" in 1970(Coppola 2005, 2012
The protocone index of the Grotta di Capelvenere, Melpignano and San Sidero shows a trend closer to that reported for E. mosbachensis, with M 3 value higher than M 1-2 one and P 3-4 value higher or sub-equal than M 1-2 one (Fig. 6). On the contrary, the pattern of the other samples is closer to that reported for E. ferus, with M 1-2 value higher than both P 3-4 and M 3 . Finally, the sample from Tana delle Iene and Grotta delle Mura possesses a well different values compared to the others, resembling those reported for Equus hydruntinus. In fact, in E. hydruntinus the values for P 3-4 and M 1-2 are significantly lower than those of both E. mosbachensis and E. ferus, whereas they are similar in P 2 and M 3 values.
Whereas, the postflexid index has been investigated, but no trend can be detected through the time and/or differences among the considered taxa (Fig. 7).
Finally, in the standard bivariate plot of muzzle proportions (Fig. 8), two groups can be recognized. The first includes the specimen from the fossiliferous sites referred to glacial stages (Appendix 9), which display a large muzzle in relation to their length. An exception is the cranium of Equus ferus antunesi Cardoso and Eisenmann, 1989 from Fontainhas (Portugal), where the muzzle is longest. The specimen from Cardamone falls in the variability of the glacial horses, and is similar to that from Cuane de l'Arago (Fig. 8). A second group is composed by crania from deposits referred to interglacial stages, where the muzzle is narrow compared to its total length. The cranium from San Sidero falls in this variability.

DISCUSSION
The taxonomy of Middle to Late Pleistocene European Equus remains controversial (Forstén 1991; van Asperen 2012). The large variability of the morphological features and biometric traits of caballoid horses has been the subject of controversy amongst many authors. No consensus exists on how to define this variability, as it is either treated as being intra-specific or inter-specific events (Azzaroli 1983;Forstén 1988;Cramer 2002;van Asperen 2012). This unresolved taxonomic issue has led a proliferation of taxa, identified as either species or subspecies (E. mosbachensis, E. steinheimensis Von Reichenau, 1915, E. achenheimensis Nobis, 1971, E. taubachensis Freudenberg, 1911) (see van Asperen 2012 for discussion). In the Italian fossil record, the specific attribution of the fossil samples from Middle Pleistocene sites reflects this and the comparison of the protocone index carried out in this work allow to refer the material from San Sidero to E. mosbachensis.
In this scenario, the analysis of a relatively large sample of Equus fossils from late Middle to Late Pleistocene localities of AP allows us to reassess the taxonomy and the evolutionary trend of local horse species. Based on the results of the statistical analyses and the comparison of the protocone indexes, E. mosbachensis is identified for the first time from few Apulian fossiliferous sites, among which are included San Sidero, Melpignano and Grotta di Capelvenere. Equus mosbachensis possesses larger upper and lower teeth than those of E. ferus (Figs 4,5;Tables 2,3) and different values of the P 3-4 , M 1-2 and M 3 protocone index (Fig. 6). The Grotta di Capelvenere, Melpignano and San Sidero samples displays a M 1-2 protocone index value lower than those both P 3-4 and M 3 (Fig. 6) (Sarti et al. 1998(Sarti et al. , 2002, Grotta Paglicci (Boscato 1994), Melpignano (Rustioni 1998  fig. 4. -Boxplot of the length of the lower teeth considered for chronology: A, second premolar (P 2 L); B, third-fourth premolar (P 3-4 L); C, first-second molar (M 1-2 L); D, third molar (M 3 L). For the groups, see Table 1 Tables 2, 3), the protocone index in the materials of E. ferus displays M 1-2 value higher than those both P 3-4 and M 3 (Fig. 6). Furthermore, for the samples from Tana delle Iene and Grotta delle Mura, the protocone index differs. The values of P 3-4 and M 1-2 are much lower than those of caballoid horses. This atypical profile of IP index in the fossil materials of Tana delle Iene and Grotta della Mura could be due to the small size of the available samples.
Finally, following the literature, the postflexid index fails to discriminate Equus species from Middle to Late Pleistocene and no trend can be observed (Fig. 7). In accordance with Boulbes (2010), the significant variation of the postflexid index could be related to tooth ontogeny (relative wear).
Whereas E. mosbachensis is widespread in Europe during the Middle Pleistocene, its presence in Italian Peninsula was quite scarce (Gliozzi et al. 1997), and documented only from few sites: Cesi (Ficcarelli et al. 1997), Venosa-Notarchirico (Palombo & Alberdi 2017), Fontana Ranuccio (Biddittu et al. 1979. In AP, the Mosbach horse is identified for the first time in a few localities, which unfortunately lack of absolute radiometric determinations. As for the European material, in the AP the Mosbach horse was well-distinct for its large teeth sized, which is larger to the wild horse (E. ferus) (Figs 4, 5; Tables 2, 3) and display different values of the protocone index (Fig. 6). According to several authors (Guadelli 2007; Uzunidis 2017), fig. 5. -Boxplot of the length of the upper teeth considered for chronology: A, second premolar (P 2 L); B, third-fourth premolar (P 3-4 L); C, first-second molar; (M 1-2 L); D, third molar (M 3 L). For the groups see Table 1. GEODIVERSITAS • 2022 • 44 (2) record prevents an in-depth reconstruction of the caballoid horse lineage (Berzi 1972;Caloi & Palombo 1987;Strani et al. 2018, 2019, Strani 2020. The last occurrence of E. ferus in the Italian Peninsula took place during the end of Late Pleistocene to Early Holocene (18-9.1 ka BP). During the end of Late Pleistocene (16-12 ka BP), E. ferus was well diffused across the Italian Peninsula, as documented by Leonardi et al. (2018) at Grotta delle Mura, Grotta Paglicci, Palidoro, Romito and Vado Arancio. During the Early Holocene however, its presence was exclusively reported from Grotta delle Mura (Bon & Boscato 1993) (Leonardi et al. 2018). The radiometric dating of level 3 indicates an age ranging from 17913-1738 to 13009-12688 cal BP, whereas that of level 2 varies between 9451-9125 to 9527-8982 cal the last occurrence of E. mosbachensis took place during the late Middle Pleistocene, probably during the MIS 6, although no general consensus was reached (Boulbes & van Asperen 2019). However, the first historical appearance of E. ferus in AP is from Grotta del Cavallo during the early Late Pleistocene (< 109 ka) (Zanchetta et al. 2018). Therefore, a new dispersal of Equus species could have taken place during the late Middle Pleistocene. This possible scenario is consistent with the results of the aDNA analysis performed on E. ferus, which revealed that the wild horse originated at about 240.000 years ago (late Middle Pleistocene), differing therefore from the earlier form of Equus (George & Rider 1986). Unfortunately, the taxonomic uncertainty for the Middle Pleistocene sample from Italian Peninsula  Table 1.
The wild horse is a common element of the mammal assemblages from AP during the late Aurelian, showing a homogenous body size through the time. According to van Asperen (2012), size oscillations of E. ferus can occur in response to climatic change, with the specimens from glacial stages being smaller and more robust, possibly as an adaptation to colder environmental conditions (Mayr 1956;James 1970;Lindstedt & Boyce 1985;Blackburn et al. 1999). On the contrary, interglacial horses could be larger with more slender limb proportions. Some populations of interglacial E. ferus can also be characterized by small and robust individuals (always less robust than glacial ones). The body-size of E. ferus of the Apulia region during the MIS 7-5 to MIS 2 is quite constant and no changes can be detected. According to van Asperen (2010), this stasis could suggest that the wild horse BP (Leonardi et al. 2018;recalibrated using Oxcal v. 4.4, IntCal20 curve). Recently, new radiometric analysis has been performed on Grotta dei Cervi, which records the presence of E. ferus at 10175-9701 cal BP (De Grossi Mazzorin & Montefinese 2017) (recalibrated using Oxcal v. 4.4, IntCal20 curve). A long gap was detected in the horse fossil record, between the occurrences from Grotta delle Mura and Grotta dei Cervi and those from the bronze age localities of Santa Rosa di Roviglio (4149-4112 to 3492-3355 cal BP) and Montale (4089-3057 to 3370-3219 cal BP) (recalibrated using Oxcal v. 4.4, IntCal20 curve). This gap has been interpreted as the local excinction of wild E. ferus, which was later reintroduced in the Italian Peninsula by recent human populations. This disappearance during the Early Holocene could be linked to a marked reduction of steppe-and tundra-like landscapes (Leonardi et al. 2018). Whereas, the crania from interglacial stages (n = 2) possess an elongated and narrow muzzle, as the specimens from Middle Pleistocene sites of Lunel-Viel (MIS 11) (France) (Bonifay 1980;Eisenmann et al. 1985) and Mosbach (MIS 13) (Germany) (Maul et al. 2000) (Fig. 8; Appendix 9). An exception is represented by the cranium from Last Glacial (MIS 2; 22, 730 ± 835 ka) of Fontainhas (Portugal), since its proportions fall outside the variability of the glacial caballoid horses ( Fig. 8; Appendix 9). The proportions of the two studied skulls, San Sidero and Cardamone specimens, corroborated the glacial/ interglacial separation. Indeed, the skull from Cardamone biochronologically referred to Last Glacial falls in the variability of the glacial group, representing one of the largest specimens ( Fig. 8; Appendix 9). Contrary, the proportions of the skull from San Sidero (MIS 9-8) differ from those of the glacial group, and are similar to those of Mosbach and Lunel-Viel ones ( Fig. 8; Appendix 9). Finally, considering the large variability of the skull size of caballoid horses during the Middle and Late Pleistocene, no evolutionary trend can be recognized. Nevertheless, based on the proportions of the muzzle, two groups are identified from the Middle to Late Pleistocene of Europe, which reflect an adaptation to climate (glacial and interglacial stages). Therefore, the proportions of the muzzle of caballoid horse clearly reveal important information on climatic conditions and palaeoenvironment.

CONCLUSIONS
Our results highlight as the application of statistical analysis and the reconstruction of the protocone index in a large dataset represents a potential tool to redefine taxonomical attribution of equid fossil material in order to improve current biochronological information on key localities and areas of the Italian Peninsula. Most of the examined material originates from archeological sites where human exploitation produced a strong impact on fossil remains, which are often highly fragmented. Furthermore, Equus material from these localities was not studied from a paleontological perspective and most of the samples was taxonomical ascribed to Equus ferus according to the chronology of the deposit. Therefore, our results allow us to redefine the taxonomical attribution of the material from Grotta di Capelvenere, Melpignano and San Sidero, which is referred to Equus mosbachensis. For the fossil material from the deposits attributed from the late Middle Pleistocene (MIS 7-6), Late Pleistocene (MIS 5-2) and Early Holocene the attribution to E. ferus is confirmed.
The presence of E. mosbachensis is reported from few AP localities, which lack of absolute radiometric dating, whereas the first historical appearance of E. ferus is from the early Late Pleistocene of Grotta del Cavallo (MIS 5). Therefore, the Mosbach horse could disappear during the late Middle Pleistocene. In addition, in accordance with Leonardi et al. (2018), the last occurrence of E. ferus in Italian fossil record is from Early Holocene sites of Grotta delle Mura and Grotta dei Cervi. Finally, the proportions of the muzzle of caballoid horses fluctuate as response to climate, where wide and short muzzle was found in specimens from deposits attributed to glacial stages and viceversa.   Table 1.

San Sidero
The IGF16329 is a splancnocranium in poor state of preservation. The sutures are completely fused. In dorsal view, the nasal is narrow with the anterior part getting thinner, whereas the muzzle is long and broad. In lateral view, the frontal and nasal bones are flat, where the nasal incisive fossa opens at the level of the posterior border of P 2 . In ventral view, the palatal is large at the level of M 2 -M 3 . The specimen IGF16329 has permanent teeth and the right and left cheek toothrows are complete. The presence of the canines allows to refer the cranium to male. The incisors are disposed in semicircular, where the I 1 and I 2 are buccolingually elongated whereas the I 3 is mesiodistally elongated. The canine is mesiodistally elongated, with an evident crest along the margin of the teeth. Generally, the sketch of the caballine and protoconule folds is low complicated. The hypoconal grove is more pronounced in the premolars (P 3 and P 4 ) than that of the molars (M 1 and M 2 ), instead the protoconal groove is pronounced.

Cardamone
The CC467 is a well-preserved cranium, except for parietal, frontal, nasal, zygomatic and vomer bones, which are incomplete. The cranium is elongated rostrocaudally with the nasal and frontonasal sutures not completely fused. In dorsal view, the nasal bones are narrow and the frontal bone became wider at the zygomatic processes. In lateral view, the frontal is flat and the zygomatic process is robust. The nasal incisive fossa opens at the level of the middle part of P 2 , whereas the infraorbital foramen is large and opens at around the middle part of P 4 . In ventral view, the incisive bone is narrow, the palatine fissurae is long and the interincisive canal is large. The palatal is wide at the level of M 2 -M 3 , the basisphenoid is robust as well as the basal part of occipital bone. The retroarticular process is robust, the mandibular fossa is marked and hypoglossal foramen is large and mesiodistally elongated. The specimen CC468 has permanent teeth and the canine is not present, therefore the cranium can be attributed to female. The incisors are not preserved, whereas the left and right cheek toothrow are complete. Generally, the sketch of the caballine and protoconule folds is low complicated. The hypoconal grove is more pronounced in the premolars (P 3 and P 4 ) than that of molars (M 1 and M 2 ), whereas the protoconal groove is low pronounced. The CC468 is very well-preserved cranium, only lacking the anterior part of incisive bone. The cranium is elongated rostrocaudally with the sutures completely fused. In dorsal view, the nasal is narrow with the anterior part getting thinner, whereas the zygomatic processes of the frontal bone are robust from which diverged the two temporal lines ending posteriorly to a short sagittal crest. In posterior view, nuchal crest is welldeveloped and the braincase has a rough surface. In lateral view, the frontal bone is flat, whereas the nasal and parietal bones are slightly convex. The infraorbital foramen is large and open around the posterior border of P 4 . The zygomatic process is robust and the occipital bone is posteriorly directed.
In ventral view, the palatal bone is larger at the level of M 2 -M 3 , the basisphenoid bone and the basal part of occipital bone are robust, whereas the vomer bone is thin. The mandibular fossa is marked, the retroarticular process is robust and hypoglossal foramen is large and mesiodistally elongated. The specimen CC468 has permanent teeth, which includes the presence of left canine. Therefore, the cranium can be referred to male. The canine is mesiodistally elongated, with an evident crest along the margin of the teeth. The left and right cheek toothrow are complete, excepted for incisors. Generally, the sketch of the caballine and protoconule folds is low complicated. The hypoconal grove is more pronounced in the premolars (P 3 and P 4 ) than that of molars (M 1 and M 2 ). whereas the protoconal groove is low pronounced.