Pleistocene horses ( genus Equus ) in the central Balkans

A review of the fossil horses of the genus Equus from the central Balkans, a mountainous area comprising Serbia and Montenegro, is presented in this paper. The time period covered by the finds is from the late Early to and including the Late Pleistocene, but the record is not complete: the dated finds are Late Pleistocene in age, while Early and Middle Pleistocene are poorly represented. The horses found resemble those from neighbouring countries from the same time period, probably showing the importance of river valleys as migration routes. The Morava River valley runs in a roughly south-to-north direction, connecting, via the Danube and Tisa River valleys, the Hungarian Pannonian Plain in the north with northern Greece in the south, via the Vardar River valley in Macedonia. In Pleistocene, large mammals, including horses, probably used this route for dispersal.

Accordingly, the important replacement of stenonid (zebroid) horses by caballoid or true horses, elsewhere locally documented by the sympatry of these two groups, is not known here. There is a single find of typical stenonid horses, dated to the latest Early or earliest Middle Pleistocene, but the rest are all true caballoids.
The presence of E. hydruntinus REGALIA characterizes the Late Pleistocene. The known finds of fossil horses in the Central Balkans are Late Pleistocene in age, either stray finds in loess or alluvia, or material collected during archaeological excavations in caves or rock shelters; more recently palaeontological excavations have been done. The finds have mainly been dated on the basis of archaeological or faunal evidence; we have been able to get some finds radiometrically dated, thanks to the courtesy of HÖGNE JUNGNER of the Dating Laboratory, Helsinki University.

The earliest stage: stenonid horses
Stenonid horses in the Central Balkans are first mentioned in a paper on loess plateaux and sands in Vojvodina (MILOJEVI], 1950). The paper briefly reports on a lower molar, identified by Vladimir Laskarev as Equus sp. aff. stenonis COCCHI, from the south-western Banat loess plateau. The tooth had not been relocated.

Trlica
Stenonid horses are represented from a single locality, i. e. Trlica in northern Montenegro (Fig. 1B: 11). The rarity of Lower and Middle Pleistocene deposits in the studied area makes this find unique.
Trlica is a hill composed of Triassic limestone situated north of the city of Pljevlja. A karstic cavern formed in the limestone and filled in with clastic deposits was accidently revealed during construction of the road Pljevlja-Bjelo Polje, in the early 1960's. The deposits, containing mammalian bones and teeth, were excavated during two short field campaigns in 1988 and 1990 (DIMI-TRIJEVI], 1990), and more recently, in 2001.
There are carnivore teeth marks on some of the bones, but not to such an extent that the sample could be regarded as an accumulation in a hyena den. Neither, probably, was it a natural trap, since no complete skeletons or articulated parts of skeletons have been found. The bones are mostly fragmentary, but not markedly worn by water transportation. The bones and teeth were probably scattered in the near vicinity of the locality and were brought into the cavern after a short transport by water and/or gravitation.
The large right PM3-4 (TRL 90/18/1) resembles E. major in occlusal length, but is narrower (Fig. 2a). The labial styles are massive, but not grooved; the hypocone is lingually indented. The two m1-2 (TRL 90/5/1 and 96/8) (Fig. 2b-c) differ from the other lower molars in that they are of greater length and breadth. We identify these three teeth as E. cf. major.

The later stages: caballoid horses
The loess record

Te{i}a Ciglana
In an area which in former days was the periphery of the city, but now is part of the centre of modern Belgrade, a brick-works, named Te{i}a Ciglana, was opened exploiting loess "earth" (Fig. 1B: 1). Pleistocene large mammal remains were found in this earth. LASKA-REV (1926) gave a short description of the find and a list of the species identified. The bones and teeth of animals identified as Elephas primigenius BLUMENBACH, Bos sp., and Equus "woldrichi" ANTONIUS, were scattered on the surface within an area of approximately 75 square meters.
Laskarev was an experienced paleontologist, who contributed significantly to Neogene and Quaternary biostratigraphy of Serbia. His identification of the fossils should be taken seriously, with amendmants for changes in the taxonomy and nomenclature of the species. Thus in addition to the horse, there were remains of Mammuthus primigenius and of a large bovid, probably Bison priscus (BOJANUS). The generic name "Bos" was at that time applied to both aurochs and bison, but Bison priscus was much more frequent in the Pleistocene of the region than was Bos primigenius BOJANUS.
The most important information about the find is that it came from the lower part of the second loess horizon. The loess deposits in the surroundings of Belgrade constitute the southern margin of the loess cover of the Pannonian Plain. The loess reaches a thickness of over 30 m in places and covers deposits of various ages, mostly Neogene freshwater deposits and Middle Pleistocene fluvio-lacustrine sediments (the so-called "layers with Corbicula fluminalis"). For a long time it was thought that loess deposition began in the Middle Pleistocene Rissian, and was finished at the end of the Pleistocene (LASKAREV, 1922;STEVANOVI], 1977). More recent data indicate an older age. In one of the profiles of the Danube embankment, 15 km downstream from Belgrade, 12 paleosols can be distinguished. Samples of the paleosols were dated by the thermoluminiscence method, which showed that the oldest soil could be equated with the oxygen isotope stage (OIS) 15 (535 000 years BP) (BUTRYM et al., 1991). The second Belgrade loess has been correlated with the second loess in the loess profiles from Batajnica, Stari Slankamen, and Ne{tin, and dated to the second Würm stadial (MARKOVI]-MARJANOVI], 1972). The lower part of the second loess in the Mo{orin-Surduk Dukatar profile has been dated to 35 900 +/-500 BP (Lub-1905) and in the Stari Slankamen-^ot profile to 37000 +/-600 BP (Lub-1892) (BUTRYM et al., 1991). These dates compare closely with the 14C date 36680 +/-1100 BP (NLJ 306) for the uppermost part of the paleosol beneath the second Belgrade loess (MARKOVI]-MARJANOVI], 1976).
The teeth, which evidently represent a caballoid horse, although the diagnostic lower cheek teeth are lacking, are medium sized; the total tooth rows measure appr. 165-175 mm at the alveoli. In the mature PM3 (LV 21/1 and 21/25) the protocones are very short, only a little longer than in PM2. In the young adult PM 3 (LV 21/19 and 21/22), the protocones approach those of PM 4 in length. The mean protoconal length of PM3-M2 of the two individuals together is only 11.8 mm (N 8, range 8.97-13.5 mm). The hypocones in PM2-4 are lingually indented and there is a pli hypostyle overhanging the hypoconal groove; the labial styles are angled or grooved, in M1-3 they are simple, but the mesostyles may be slightly grooved lower down. The plication counts are 3-8 plications; in the molars the plis caballin tend to disappear with wear.
The size of the horse teeth from the Belgrade second loess compare best with those of late Late Pleistocene caballoids from Central Europe (FORSTEN, 1991: fig. 4), but are on the small side compared with those of the middle Late Pleistocene, corresponding to the dates of approx. 35 900-37 000 BP and 36 680 BP for the second loess and paleosol, respectively. The species E. "woldrichi", to which these specimens were originally referred, is based on a horse skull from Krems, Austria, believed to represent the Late Pleistocene "loess horse" (ANTONIUS, 1912). The species was poorly described, without photographs and measurements, and is probably not valid; in addition its age is unknown. SCHLOSSER (1916), MOTTL (1938MOTTL ( , 1941, and VERTES (1950) used the name for fossil horses from Eichstätt, Germany, Bergavölgy and Solymar, Hungary, respectively. The teeth referred to E. woldrichi by SCHLOSSER (1916) seem larger than those from the Belgrade loess, while the three specimens referred to MOTTL (1938, table 396) correspond to those of the latter.
A host of different names have been coined for Late Pleistocene medium sized to smallish Eurasian caballoid horses; more recently usually identified as E. caballus L. with "subspecies". For fossil forms, GENTRY et al. (1996) have proposed Equus ferus BODDAERT to replace E. caballus, a name originally given to the domestic horse (LINNAEUS, 1758). neath the topographic surface. By correlating the Ciglansko Brdo lithology with that of the bore-core Kovin (Bolnica), MARKOVI]-MARJANOVI] (1970) dated the equid from Smederovo to the Mindel 2.

Ciglansko Brdo, Smederovo
The material cannot be located, but the figures (MARKOVI]-MARJANOVI], 1970) show right pm2-pm4 and left pm2-m2 of a caballoid horse, originally identified as the large stenonid E. cf. suessenbornensis WÜST, probably mainly because of the presumed Middle Pleistocene age of the find. Neither the size of the specimens nor the scale of the figures is given. In the left jaw, m1 has a deep ectoflexid reaching to the double knot. Morphologically these teeth do not differ from those of other caballoid forms.
The Natural History Museum and the Faculty of Mining and Geology have in their collections remains of Pleistocene large mammals found in alluvial deposits, chiefly in deposits of the greatest rivers in the region, the Danube, Sava, Tisa, and Morava. Most numerous are the massive skeletal parts of the largest animals, e.g. the teeth and bones of mammoth and crania of bison, reflecting the selection of the collectors'; horse remains are scarce. Some specimens have exact data about the circumstance of their finding, while others only have a note on their alluvial origin. Several specimens are considered alluvial finds solely on the basis of the sediment preserved in the bone crevices.
The horse metatarsus (No. 603), in the collections of the Natural History Museum, was found from Ostru`nica, a small town west of Belgrade ( Fig. 1B: 2). In 1949, when the supporting columns of a railway bridge crossing the Sava River were built here, large mammal fossils were found in deposits known as the «Sava layers» (VESELINOVI]-^I^ULI], 1952). Originally dated to the Holocene (LASKAREV, 1938;STEVANOVI], 1977), the «Sava layers» were shown to contain Late Pleistocene mammals identified as Bison priscus, Mammuthus primigenius, and Cervus sp. (amended) (VESELINOVI]-^I^ULI], 1952). According to the museum's inventory book, the metatarsus was found together with M. primigenius, Megaloceros sp., and Bison sp., indicating a stratigraphic age similar to the find described by VESELINOVI]-^I^ULI], (1952).
We compared the bone (No. 603) with metapodials from the Late Pleistocene of the Risova~a cave (FORSTEN & DIMITRIJEVI], 1995) and with bones of similar age from Hungary in a Simpson's ratio diagram (Fig. 7). The bone is as large and massive as those compared with it.  The upper and lower jaws (Nos. 1591 and 1397) (Figs. 8,9b) in the Natural History Museum are said to come from alluvial deposits, but lack locality data. There are no data on the mandible (RGF 87/01) (Fig.  9a) held in the Faculty of Mining and Geology, but river pebbles in the canine alveolus and dark sand with muscovite particles in the incisor alveoli and in cracks in the bone, clearly show its alluvial origin.
The mandibles (No. 1397 and RGF 87/01) come from medium-sized caballoid horses, with the premolar rows measuring alveolarly 86.7 and 90.5 mm, respectively (Table 3). The total pm2-m3 row of RGF 87/01 measures 177.6 mm. The pm2 lack "protostylids" and m2-3 have shallow ectoflexids. These jaws and teeth correspond in size and morphology to those of late Late Pleistocene caballoid horses of Eurasia, identified under several local names but probably synonymous with E. ferus. In terms of size, they also correspond to the uppers from the Belgrade second loess. Left mandible (RGF 87/01) has been δ13C dated to > 40 000 BP (Hela-506 & Hela-507) (JUNGNER, pers. comm.).
No. 1591, δ13C dated to 36300+/-1700 BP (Hela-505) (JUNGNER, pers. comm.), represents a larger horse, the to-tal teeth row PM2-M3 measures alveolarly 194.3 mm, but these teeth are less worn than those of the two previous specimens, as shown by the still deep post-protoconal groove in PM3, reaching towards the prefossette (Fig. 8, Table 3). The protocones of PM3-M2 vary between 16.85-18.4 mm in length, the plication counts between 11-13 plications, including well-developed plis caballin. The hypocones are lingually indented and plis hypostyle are indicated; in M3 the pli hypostyle closes the hypoconal groove as a lake. Premolar labial styles are well grooved, also molar styles may be grooved at some point along the crown. A single PM 3-4 (ZP 5) from the Sava River is also large, but has a short protocone, measuring 11.65 mm. These specimens, with PM3-4 occlusal lengths of 32-35 mm, correspond in size to large horses from the Late Pleistocene caves of Hungary, from caves and rock shelters in Serbia and Montenegro, identified as E. sp. and E. mosbachensis-abeli REICHENAU-ANTONIUS (RAKOVEC, 1965;MALEZ, 1975;, 1995), and to large Middle Pleistocene horses from neighbouring Greece and Rumania, identified as E. abeli ANTONIUS and E. insulidens SAMSON (MELENTIS, 1966, SAMSON, 1975.

Despotovac
A L PM3-4 (LV 20), in the Faculty of Mining and Geology, University of Belgrade, was collected by Laskarev, most probably when surveying the Despotovac area ( Fig. 1B: 4), where important Neogene mammalian fossils were found (LASKAREV, 1949). The tooth is appr. 90 mm high, large, with a long protocone and grooved labial styles, resembling those of No. 1591.

Stubal
Three large upper horse cheek-teeth, R and L PM3-4 (ZP 6/1 and 6/3) and L M1-2 (ZP 6/2), probably from a single young adult individual, were found by @IVADIN PETRONIJEVI], a Serbian Neogene mammal specialist, in Stubal, a village near Blace in Central Serbia (Fig. 1B,9). Little is known about the circumstances of the finding, except that the teeth come from loose sediment, most probably alluvial sands. The crowns are high, from >84 to 98 mm, and the protocones long, range 15-16.4 mm.

Cave deposits of Serbia
The Late Pleistocene fossil record is best known from caves, both because the cave environment furnished favourable conditions for fossilization and because archaeological excavations have focussed on caves. The cave faunas are mostly rich and diverse, and Palaeolithic artefacts, where found, make dating and age comparisons feasible.
Except for the second phalanx, the horse bones show severe damage attributed to Crocuta gnawing. The epistropheus is gnawed along the neural spine, and the caudal articular surface and processes show similar damages. The lateral part of the trochlea tali and the plantar surface of the astragalus show grooves made by Crocuta teeth. The lateral side of the proximal end of the radius is chewed off and there are two pits on the proximal articular surface, probably made by Crocuta canines. The radius also shows scratches, but they are not so clear nor in a characteristic position as to be classified with certainty as butchering marks, and they could have been made by stone ertefacts.
The horse teeth are large, corresponding in size to those from the Risova~a, Baranica, and Ravani~ka caves ( Fig. 10) (Table 4). Two R lowers (Nos. bb1 and bb2) (Fig. 10n-o), although worn, are particularly large; they may be m1 and m2 (alternatively, mp4 and m1) of one individual. In this sample, the protocones are medium long (mean 14.8 mm), the plication counts slightly high (mean appr. 9.8 plications). The ectoflexids tend to be shallow in the lower molars.
The astragalus (No. 98) is medium sized, but the medial phalanx (No. 284), with a total length of 58.4 mm, is as large as the bones from Risova~a.

Lazareva Cave
The Lazareva Cave, sometimes referred to in the literature as Zlotska Cave, is located at the foot of the Ku~aj mountain, 3 km northwest of the village Zlot, on the left side of the Lazareva River, a tributary of the Zlotska River (Fig. 1B: 10). The entrance is at an altitude of 291 m or 6.7 m above the Lazareva riverbed (LAZAREVI], 1978 1963, 1964, and 1968, revealing Iron and Bronze Age remains (TASI], 1971).
The Pleistocene layers of the Lazareva Cave have not been systematicaly excaved, but scattered remains of several Pleistocene mammalian species have been reported, such as Ursus spelaeus, Panthera spelaea, and Crocuta spelaea (CVIJI], 1893, 1895; @UJOVI], 1929). One cave visitor happened to be @IVADIN PETRONIJEVI], who collected an equid tooth and four canines of Ursus spelaeus, which are now in the collections of the Faculty of Mining and Geology, University of Belgrade. Two equid teeth from the same locality, found by an unknown collector, are held in the Natural History Museum, Belgrade.

Ravani~ka Cave
The cave is located in the immediate vicinity of the Ravanica monastery (Fig. 1B: 7). The entrance is at an altitude of 235 m, only 6 m above the riverbed of Rava-nica, a tributary of the Morava River. The total length of the cave is 1049 m (PETROVI], 1976).
]IRI] (1952) described cave bear remains from the Ravani~ka Cave. Two equid teeth from this locality, found by @IVADIN PETRONIJEVI}, are in the collections of the Faculty of Mining and Geology, University of Belgrade. The teeth are L PM3-4 (ZP 2/2) of a caballoid horse and R M3 (ZP 2/1) of E. hydruntinus. In the Natural History Museum, Belgrade, there is a total of five teeth, L and R PM3-4, L M1-2, and L pm3-4 (Nos. 377, 375, 376, and 380) of a horse and R PM3-4 (No. 378) of hydruntinus. The uppers of the horse are large, with long protocones and 5-11 plications, including marked plis caballin.

Baranica
The cave is situated in the wider surroundings of the town Knja`evac, on the right bank of the Trgovi{ki Timok River, at an altitude of 260 m (Fig. 1B: 6). The entrance is in the form of a low rock-shelter, while behind it a network of cave channels is developed.
Archaeological  to a depth of 2.2 m. Flint artefacts were found, as well as an extremely rich and diverse fauna of both large and small mammals, birds, frogs, fish, and gastropods. Among the flint artefacts from layer 2, characteristic early Upper Palaeolithic types can be distinguished, indicating an age < 41 000 BP (MIHAILOVI] et al., 1997).
Some of the large mammal bones were water worn, probably from having been transported down from the upper levels of the cave by underground water currents. Tooth marks and gnawing damage could be seen on other bones, mostly caused by cave hyena. Some ten metres above the cave entrance in the same carbonate rock, a road cuts through a karst cavern, which was excavated in 1997. Animal remains recovered showed strong predator selection and were damaged by gnawing to such an extent that it was concluded that the place was a cave hyena den and raptor haunt. It is supposed that it was connected with the lower cave and it has accordingly been named Baranica II. The palaeontological finds from the lower cave and the cavern are marked "BAR" and "BAR II", respectively.

Vrelska Cave
The cave is in the town Bela Palanka, at an altitude of 545 m (PETROVI], 1976) ( Fig. 1B: 8). Since 1986 it has been repeatedly explored by hydrogeologists, who found the remains of Pleistocene mammals. In 1990 paleontological excavations were done in the cave, revealing a fauna of mammals, birds, reptiles, amphibians, and fish (MARKOVI] & PAVLOVI], 1991; DIMITRIJEVI], 1997a). The material collected during the hydrogeological works (mainly Ursus spelaeus) is kept in the Bela Palanka Regional Museum, while the spoils of the excavations are kept in the Natural History Museum, Belgrade.
There are only two tooth fragments of horses: an upper PM/M and a lower pm/m. Although broken, the protocone and the metaconid-metastylid double knot, respectively, clearly show caballoid features.

Rock-shelters of Montenegro
Pleistocene vertebrate remains are known from only a few cave localities in Montenegro (DIMITRIJEVI], 1997c), which mainly comprise the remains of the cave bear, but another type of karst localitiy, rock-shelters, have yielded rich and diverse Pleistocene faunas.
Excavations have been done in the Crvena Stijena rock shelter in the Trebi{njica River valley (BENAC, 1975) (DJURI^I], 1996). These sites are situated in mountainous areas, at an altitude of more than 500 m. They are multilayered and contain numerous flint artefacts and faunal remains (MALEZ, 1975;MALEZ et al., 1988;DIMITRIJEVI], 1996MIHAILOVI] & DIMITRIJEVI], 1999). The composition of the fauna, especially the frequency of the various mammal species, is largely influenced by the hunting and subsistence strategies of Palaeolithic and Mesolithic human societies, which enjoyed the advantages of these natural shelters.

Medena Stijena
The Medena Stijena rock shelter is situated about 20 km south-east of Pljevlja, in the ]ehotina canyon, at an altitude of 780 m (Fig. 1B: 12). Upper Palaeolithic chipped stone assemblages were found in several layers of the lower stratigraphic complex, and Mesolithic and Eneolithic/Bronze Age archaeological material in the upper stratigraphic complex of the rock shelter (MIHAILOVI], 1996).
The poor preservation of the bones, due both to predepositional fragmentation and depositional conditions, has meant in that less than 5% of the 1747 pieces of bone and teeth collected could be assigned to species. The following species were found in the lower stratigraphic complex: Ursus arctos LINNAEUS, Equus sp. (caballoid), Sus scrofa, Cervus elaphus, Bison priscus, Bos/Bison, Capra ibex, and Rupicapra rupicapra (LIN-NAEUS) (DIMITRIJEVI], 1996).
A single equid R pm2 (MS 88/95/1) and a proximal MC III (MS 86/109/1) were found in the lowermost layer, together with artefacts of Early Epigravettian type (mainly dated 25 000-13/12 000 BP) (MIHAILOVI], 1996). The fossils belong to the City Museum of Pljevlja, Montenegro. Although fragmentary, the pm2 appears large to medium in size; the proximal breadth of the metacarpal is 50 mm, indicating medium size.

Crvena Stijena
The Crvena Stijena rock shelter is in the Trebi{njica River valley, near the village of Petrovi}i, western Montenegro, at an altitude of approximately 700 m (Fig. 1B:  13). Archaeological excavations there were done in 1955-1964(BENAC, 1975. The locality comprises one of the most complete Middle to Late Pleistocene sequences in Europe, with more than 20 m of deposits and 31 distinguished layers, without as yet having reach- The equid material is kept in the Institute of Quaternary Geology and Palaeontology, Zagreb, Croatia, and lacks collection numbers. It consists of: three R and L PM2; six R and L PM3-4; six R and five L M1-2; four R and six L M3; L and two R pd 3-4; R and two L pm2; six R and eight L pm3-4; ten R and twelve L m1-2; four R and three L m3. There are no limb bones. The horse material comes from layers XX-XXV of Crvena Stijena, believed to span the late Riss Glacial -?Amersfoort Interstadial, i.e. mainly the last Interglacial (BRUNNACKER, 1975) or the marine oxygen isotope sta-ge (OIS) 5e. Maximal frequency of the horse is in layer XXIV, believed to correspond to the Interglacial, with a high frequency also in layers XXI-XXII, believed to correspond to the Würm I (OIS 5). According to BRUN-NACKER (1975), the climate of layer XXII was still relatively warm.
The horse is dentally as large as those from the Ri-sova~a, Jerinina, and Baranica caves (Table 4). In a phenogram, constructed on the basis of ten upper cheek tooth measurements (Fig. 12), these samples cluster together. The protocones are medium long, the plication counts slightly high. Although the coefficients of variation (100 × s/Mean) for M3 and m3 occlusal lengths are high, probably because specimens in different stages of wear were included in the calculations, there are no morphological indications of two different caballoid species in the sample, nor has E. hydruntinus been found here.

Equus hydruntinus REGALIA
The presence of the small Middle to Late Pleistocene stenonid, Equus hydruntinus, was previously established in the Risova~a Cave, near Arandjelovac in Central Serbia (RAKOVEC, 1965;FORSTEN & DIMITRIJEVI], 1995). The coexistence of this species with a much larger, caballoid horse also characterizes the caves Baranica, Jerinina, and Ravani~ka. A further similarity lies in the low number of hydruntinus remains found in relation to those of caballoid horse remains found, as is the case with most European finds of hydruntinus. In the Ravani~ka Cave, there were two hydruntinus teeth compared with only five caballoid specimens, but the sample in its entirety is too small to be conclusive regarding their original frequency. Whenever caballoid horse remains are found in large numbers in a cave, hydruntinus is generally also present, Crvena Stijena being an exception.
All over its range in Europe, the Middle East, and the Caucasus, Equus hydruntinus is identifiable by its dental morphology, although a single tooth may resemble small/much worn specimens of the caballoid horse. The protocones, although short, may have an anterior extension, a "heel", particularly in the molars. The protoconal length, i.e. the development of the heel, is variable within samples. A separate "subspecies", E. hydruntinus davidi Alimen was established, believed to differ in having relatively long protocones (ALIMEN, 1946), but significant differences in the protoconal length between local samples have not been demonstrated. The lower cheek teeth are typically stenonid with V-shaped lingual grooves, protostylid plications indicated, and mostly deep molar ectoflexids. Occasionally, the molar ectoflexids may be shallow, resembling those of the Asiatic Wild Ass, E. hemionus PALLAS. Generally small, the limb bones vary locally in size, but the slender metapodials and proximal phalanges are characteristic, an additional resemblance with E. hemionus.
The few hydruntinus teeth listed above do not differ from the teeth described from elsewhere of the species' range, either in their size or morphology (Fig. 13). The mean protoconal lengths are: PM3-4 7.98 mm, M1-2 9.25 mm; there may be a short heel. The plication counts vary between 6-11 plications; the plis caballin were obliterated early by wear.
The Equus hydruntinus is believed to have preferred dry and temperate climatic conditions, but the species has been found in France under glacial conditions also (PRAT, 1968). In the Bacho Kiro Cave, Bulgaria, hydruntinus occurs from layer 7 to and including layer 13, dated from 29 150 +/-950 BP to >47 000 BP (MOOK, 1982;GINTER & KOZLOWSKI, 1982), and under climate conditions varying from cool-dry or cold-humid to warm-dry and warm-humid (KOWALSKI, 1982). The maximum frequency of hydruntinus is in layer 11, dated > 40 000 BP (MOOK, 1982), when the climate was becoming warmer and increasingly humid, preceding the maximum cold of layer 12 (KOWALSKI, 1982;GIN-TER & KOZLOWSKI, 1982). It is not known from which layers of the Risova~a, Baranica, Ravani~ka, and Jerinina caves hydruntinus derives, neither the age nor the climatic conditions are known.

Pleistocene/Holocene discontinuity in the horse record
In the Central Balkans, as well as elsewhere in Europe, the records indicate that the caballoid horses disappeared at the end of the Pleistocene. There are no finds of caballoid horse from the Early Holocene faunas of Serbia, although due to extensive archaeological excavations these faunas are much better known than those from the Late Pleistocene. The Equus hydruntinus survived and locally in the Neolithic even formed an important part of the fauna and of man's subsistence (BÖKÖNYI, 1984).
The further history shows, once again, the caballoid/stenonid turnover. The latest representative of the stenonid horses, E. hydruntinus, became extinct in the Middle Neolithic (BÖKÖNYI, 1974) or as late as in the Copper Age in Spain (BOESSNECK, 1967), while the caballoid horse reappeared in the Eneolithic, this time as domesticated animals.

Discussion
The taxonomy of the caballoid horses from the European Pleistocene is still in disarray, due to the morpho-logical uniformity of these animals. There are size differences, probably partly clinal, partly temporal. A number of species names have been coined, mainly on the basis of the size of the horse and its stratigraphic age, but specimens (mainly isolated teeth and limb bones) from the type localities or referred to various species cannot be morphologically clearly differentiated from one another and are connected by forms intermediate in size, thus depriving the names of meaning and usefulness. However, at a number of localities, two caballoid forms, differing in size, seem to have occured in sympatry (see discussion in FORSTEN, 1993), indicating distinct species.
On the basis of limb massivity and protoconal lengths, SPASSOV & ILJEV (1997, 1998 tried to differentiate Pleistocene species of Equus in an Equus germanicus/latipes Nehring-Gromova Group, massively built with long protocones, and an E. gmelini Antonius Group, more gracile in build and with short protocones. They showed (SPASSOV & ILJEV, 1998) that the protoconal length can be roughly correlated with the dental occlusal length (their Fig. 3), as pointed out earlier (FORSTEN, 1982), and that a positive correlation between the MC III diaphysal breadth and length (their Fig. 4), in both cases with only small differences in the proportions between representatives assigned to the two groups. They concluded that the domestic horse is polyphyletic, deriving from both groups. Polyphyly of the domestic horse was suggested earlier (LUNDHOLM, 1949;GROVES, 1974), but considered unlikely on genetic and molecular grounds (FORSTEN, 1988).
In the present material, short protocones occur in both maxillas (LV 21/) from the Belgrade loess, the mean protoconal length in PM3-M2 is 11.8 mm, range 8.97-13.5 mm, in the PM 3-4 (ZP 5) from the Sava River alluvia, the protocone length is 11.65 mm, in the two PM 3-4 (No. bb and ZP 7/1) from the Lazareva Cave, the lengths are 11.35 and 14.5 mm, respectively, and in the PM 3-4 and M1-2 (BAR II 97/7/24 and 97/27/1) from the Baranica Cave, the lengths are 10.9 and 10.8 mm, respectively. Except for the mature maxilla from the Belgrade loess and the No. bb from the Lazareva Cave, with worn teeth measuring 22.3-31.8 mm and 35.1 mm in crown height, respectively, the teeth are only moderately worn, with crown heights of 51.7-75.2 mm, indicating that in these specimens the short protocones are not due to wear. The teeth from the Sava River, Lazareva and Baranica caves are large, the PM3-4 measuring 31.1-34 mm occlusally. Where limb bones are known, e.g. from Risova~a, Jerinina, and Baranica II, they are also large and massive (FOR-STEN & DIMITRIJEVI], 1995).
We do not believe that rhe European Pleistocene horses of the genus Equus can be differentiated into the two groups of SPASSOV & ILJEV'S (1997, 1998, simply based on the basis of the protoconal length. In addition, the type specimen of E. germanicus, a skull (in the Berlin Humboldt Museum MB 1972.31.86; NEH-RING, 1884: pl. 5) from Unkelstein near Remagen, Germany, has short protocones. Especially within local samples, we believe that more evidence is required for species differentiation. To avoid competition for resources, sympatric species ought to occupy different ecological niches, particularly among caballoid horses, with wide ecological requirements; this should be re-flected in the morphology. Where seemingly two taxa of caballoid Equus occur, they usually differ noticeably in size (e.g. MUSIL, 1962). In some of the cave samples, e. g. Jerinina and Baranica, single teeth do differ in that they are larger than the rest. We consider here these specimens large variants within taxonomically homogeneous samples, but acknowledge their difference in size.
The dated specimens do not support a decrease in size with time, as observed in the caballoid horses in other areas (FORSTEN, 1991(FORSTEN, , 1993, rather an oscillation in size is indicated. The earliest find described here, the horse from Crvena Stijena, dated to the Last Interglacial or approx. 120 000 BP, is large. It is matched in size by stratigraphically younger horses from the Riso-va~a (> 36 400 BP, Malez in litt.) and Baranica caves (< 41 000 BP), and by No. 1591 from alluvium (36 300 BP). The two medium-sized jaws (No. 1397 and RGF 87/01) from alluvia, dated > 40 000 BP, correspond in size to the uppers (LV 21) from the Belgrade second loess, dated appr. 36-37 000 BP. Taking into account the slight uncertainty with radiometric dates, all these rather differently sized finds could be roughly contemporaneous. The majority of the dated finds from the Central Balkans thus fall in the middle Late Pleistocene or OIS 3, approx. 30-50 000 BP, comprising the interstadials Denekamp, Hengelo, and Moershoofd with colder stadials in between (ZAGWIJN, 1989).

Conclusions
The earlist stage in the records of the Pleistocene horses in the Central Balkans is represented by the late Early Pleistocene stenonid horses. Remains were found of two species, Equus major and Equus stenonis and from a single locality Trlica in northern Montenegro.
The next stage is represented by the existence of solely caballoid horses throughout The middle and beginning of the Upper Pleistocene. The remains were found mostly in alluvial and cave deposits. There are two size groups: large forms mainly from caves and rock shelters and medium-sized forms from the alluvial and loess deposits. A size decrease with time is not confirmed in the Central Balkan caballoids.
In the Upper Pleistocene, the stenonid line is again present, this time with the species Equus hydruntinus.
At the end of the Pleistocene, the caballoid horses disappeared, while the Equus hydruntinus continued to exist into the Neolithic.