Introduction

1The exploitation of animal resources (i.e. hunting and/or livestock farming) in ancient societies is often only considered in terms of subsistence strategies in relation with the consumption of edible animal products. The way herds are managed is however closely related to social organisation, landscape exploitation and the definition of a group’s territory. Herding can be associated with the polymorphic mobility of human groups, allowing for the exploitation of distant biotopes that cannot be occupied all the year round. This system may lead to the formation of territories that encompass non-adjacent areas (i.e. lowlands and highlands separated by territories under the control of other groups). However, demonstrating the existence of such herding strategies in ancient times and defining their scale, seasonality, and the nature of mobility patterns requires detailed archaeozoological or textual evidence that is often not available. This paper aims at reconstructing the herding strategies of four Caucasian Late Chalcolithic and Kura-Araxes settlements located respectively in the Kura (Dzedzvebi in Georgia) and Araxes (Ovçular Tepesi, Zirinçlik and Duzdağı in Nakhchivan, Azerbaijan) basins, by using a series of new data provided by archaeozoological and isotopic analyses.

2The South Caucasus is characterised by highlands rich in pasturelands, obsidian and ore, and rather arid lowlands where salt deposits can be found in a few locations (notably in the Araxes river basin). These topographic and environmental settings seem favourable to a vertical mobility of the herds. Although the mobility of the herd is an asset for coping with natural constraints (climate, paucity of the vegetation), environmental determinism cannot in any case explain the appearance or spatial distribution of mobile pastoralism in the Caucasus or anywhere else (Ferret 2012; de Planhol 1977). Various forms of mobile pastoralism are well documented in the ethnographic records from the Caucasus and adjacent areas such as Iran, eastern Anatolia and northern Iraq (Digard, Papoli-Yazdi 2008; Elliott et al. 2015; Hütteroth 1959; Karça, Koşay 1954; de Planhol 1958; Tapper 1997; Thevenin 2011). It is tempting to assume the antiquity of mobile pastoralism in the Caucasus by analogy with these traditional systems, with the risk of falling into the “tyranny of the ethnohistoric record” (to borrow the catchphrase of Wobst 1978). This approach, which assumes the existence of highly mobile and specialised pastoral practices in Late Prehistory from the activities of mobile groups that were studied by ethnographers in the late 19th and 20th centuries may well be a methodological pitfall: to quote Hammer and Arbuckle, one should remember that “these forms of pastoralism (in the 19th and 20th centuries) were the result of distinctive and unique historical trajectories and represent a narrow cross section of variation in pastoral strategies” (Hammer, Arbuckle 2017).

3This paper will discuss the evidence relating to mobile pastoralism in the South Caucasus during the 5th to 3rd millennium BC time span, with a view to elucidating the circumstances of its development.

Definition of pastoralism and preconceptions about mobility

4Pastoralism means the use of the extensive grazing of herbivores (ruminants) on rangelands. Herds are usually moved over distances that may sometimes be very short. The word “nomad” refers to residential mobility. Communities whose subsistence strategies rely on pastoralism are not necessarily mobile, and not all nomadic groups practise pastoralism (Ferret 2012). The word “nomad” has been regularly misused in archaeological literature when referring to “portable” material culture or “light” architecture but without empirical evidence of people’s mobility (among other exemples, see Shimelmitz 2003). The archaeozoological evidence presented in this paper only sheds light on the mobility of the herds, and cannot be used to infer that the whole community was moving with them all the year round. Any demonstration of residential mobility, duration or seasonality of settlement occupation cannot directly rely on archaeozoological data. Therefore, the concept of nomadism will not be discussed in this paper.

5Preconceptions about the behaviour of various animals are also commonly invoked as an indicator of mobile herding systems. Pigs are usually not associated with mobility. As they require regular watering and are rather slow walkers (Redding 2015), they are not fit for long-distance mobility in arid areas of the Near East. However, wild boars living in mountain areas in Iraq do seasonally move from the river valleys to the hill slopes to winter in oak woods (Hatt 1959). Pig-herding strategies associated with vertical mobility are known from mountainous areas such as the Pyrenees, where animals were driven to the montane mixed forest during the autumn to feed on fallen acorns and beechmast (Aynaud, Barreneche 2009). Therefore, the presence of pigs in an archaeozoological assemblage does not rule out the possibility of a seasonal, short-distance vertical mobility in the mountainous areas of the Near East (Zeder 1996). The seasonality and ecological setting (i.e. montane mixed forests) of this mobility are however not compatible with that characterising the transhumance of ruminant species such as goats, sheep and cattle. In the Near East, cattle are also usually considered an indicator of sedentariness (Howell-Meurs 2001). This might be due to the fact that most of the ethnographic works cited above focused on tribes highly specialised in sheep and goat herding. As they need regular watering, cattle are not adapted to long-distance mobility in arid landscapes, but they are well suited for shorter vertical journeys. In the Alps, they are frequently encountered in high altitude pastures around 1,900 m asl (Gorlier et al. 2012; Tato et al. 2011). Cattle are also present in the herds of the Bakhtiari tribes, who travel across the Zagros Mountains for 45 days and cross a mountain pass at approximately 2,900 m asl (as in the movie called Grass by Merian C. Cooper, filmed in 1924). The antiquity of vertical mobility involving cattle is attested by the presence of their bones in the 4th millennium BC highland campsite of Godedzor in the Lesser Caucasus (Palumbi et al., this volume).

Material and methods

Settlements

6The zooarchaeological evidence presented in this paper was collected on four sites that differ in their nature, location, and date of occupation.

7Ovçular Tepesi, Duzdağı, and Zirinçlik are situated in the Autonomous Republic of Nakhchivan (Azerbaijan). Ovçular Tepesi is a small (<2 ha) village located on top of a natural hill overlooking the Arpaçay river (39°35’32’’ N 45°04’04’’ E), ca 18 km upstream from the confluence with the Araxes river (Nakhchivan, Azerbaijan). The site was first briefly excavated in 1986 by A. Seyidov and later in 2001 by S. Ashurov (Ashurov 2005). Large-scale excavations were conducted by C. Marro and V. Bakhshaliyev from 2006 to 2013 within the framework of a Franco-Azerbaijani joint-project (Marro, Bakhshaliyev, Ashurov 2009, 2011; Marro, Bakhshaliyev, Berthon 2014). The Late Chalcolithic stratigraphic sequence at Ovçular Tepesi is divided into two phases. Phase I is characterized by semi-subterranean huts, whose superstructure virtually left no traces apart from a few postholes and door sockets. Large storage pits later filled with domestic refuse (Berthon et al. 2013) were found outside and, in two cases, inside the huts. These huts might correspond to a single occupation level. Phase II is illustrated by free-standing, multicellular, mud-brick houses that are sometimes built on a stone infrastructure. Unlike Phase I, Phase II lasted much longer and takes the form of a succession of many overlapping occupation levels, which constantly reused or reorganized earlier building structures. According to a series of radiocarbon dates, Phase I took place somewhere between 4450 and 4300 cal. BC, while Phase II corresponds to the ca 4300‑3950 cal. BC range. An Early Bronze Age occupation of the Kura-Araxes culture, dated to the 3200 to 2400 cal. BC time span, is also attested but badly preserved due to the levelling of the site by bulldozers (Marro, Bakhshaliyev, Ashurov 2009, 2011; Marro, Bakhshaliyev, Berthon 2014). Preliminary zooarchaeological results were obtained from 16,047 remains, of which 4,509 (28.1%) were identified anatomically and taxonomically (tab. 1). These identified remains mostly come from the second phase of the Chalcolithic occupation with 75% of the number of identified specimens (NISP). The first phase of the Chalcolithic occupation and the Early Bronze Age Kura-Araxes occupation represent respectively 16 and 9% of the NISP. The faunal remains were retrieved from various contexts. In the first phase of the Chalcolithic occupation, the remains were found almost exclusively in pits. In the second phase of the Chalcolithic occupation, the remains mostly came from dumps in an open-air area and silos later filled with refuse. The buildings yielded few animal bones. In the Early Bronze Age Kura-Araxes occupation, the faunal remains come from filling layers.

8Duzdağı is a salt mine located close to the Araxes River valley some 7 km north-west of Nakhchivan City (39°17’23’’ N 45°18’38’’ E). Excavations were conducted from 2011 to 2016 within the framework of a Franco-Azerbaijani joint-project directed by C. Marro and V. Bakhshaliyev (Gonon et al., this volume). The salt layers were exploited by Kura-Araxes groups during the first half of the 3rd millennium BC (ca 300‑2650 BCE). The faunal remains were retrieved from an undisturbed zone where they were associated with hearths and Kura-Araxes potsherds (Gonon et al., this volume). In total, 142 remains were recorded, of which 59 (42%) were identified anatomically and taxonomically (tab. 2). The low density of faunal remains is in keeping with episodic exploitation of the salt layers (Gonon et al., this volume).

9Zirinçlik is a campsite located on the southern piedmonts of the Lesser Caucasus at an elevation of 1,290 m asl (39°19’12’’ N 45°31’58’’ E). This campsite is situated between the Araxes River valley and the highlands and has been occupied over the 4200‑3700 BC time span. Survey at the surface of the site yielded copper ore and moulds for the production of copper axes (Gailhard, this volume). Excavations were conducted in 2014 within the framework of a Franco-Azerbaijani joint-project directed by C. Marro and V. Bakhshaliyev. Very few structures were preserved beside a semi-subterranean hut and a few complete ceramic vessels. Only 116 faunal remains were recovered during the excavations. They were heavily fragmented and weathered. Hence only 8 of them (7%) were identified anatomically and taxonomically (tab. 3).

10Dzedzvebi is an archaeological settlement situated in the Mashavera River valley in Georgia, some 50 km south-west of Tbilisi (41°22’08’’ N 44°23’13’’ E). This settlement is closely related to the gold mine of Sakdrisi (Stöllner 2016; Stöllner et al. 2014). Excavations were conducted between 2007 and 2015 within the framework of a German-Georgian joint-project directed by T. Stöllner and I. Gambashidze. The settlement was occupied during the Late Chalcolithic period in the first half of the 4th millenium BC and then by Kura-Araxes communities during the last quarter of the 4th millenium BC and the first half of the 3rd millenium BC. A later Iron Age occupation greatly disturbed the previous layers and produced a large archaeozoological assemblage that is beyond the scope of this paper. Preliminary zooarchaeological results were obtained from 2,788 remains, of which 581 (21%) were identified anatomically and taxonomically (tab. 4). The Late Chalcolithic and Kura-Araxes assemblages represent respectively 60% and 40% of the NISP.

Zooarchaeological methods

11Most of the faunal remains were collected by hand and eye at the four sites. Water sieving was performed on selected contexts such as floors, refuse dumps, pits or ashy layers from Ovçular Tepesi, Zirinçlik and Duzdağı. For instance, more than 7,000 litres of soils were water-sieved at the site of Ovçular Tepesi. We therefore consider that our sampling and collecting methods are adapted to the collection of even small-size species or juvenile individuals. All fragments were counted and weighed, including unidentified remains. Most of the specimens were analysed at the excavations houses in Nakhchivan and Georgia without the help of reference collections. Anatomical and taxonomical identification was performed with the help of atlases (Barone 1999; Pales, Garcia 1981a, 1981b; Pales, Lambert 1971a, 1971b; Schmid 1972) and additional specific criteria (see Berthon 2011 for the main methods used by the author and their references). Selected specimens were exported to the Muséum national d’histoire naturelle (Paris, France) for further identification with the comparative anatomy collection. The analysis of the age at slaughtering for sheep and goats is based on use-wear pattern and crown height (Ducos 1968; Payne 1973; Vigne, Helmer 2007). Harvest profiles have been built by using a correction of each age group to figure in the histogram. This is linked to the various lengths of the age classes (Ducos 1968; Vigne, Helmer 2007).

Isotopic methods

12Since enamel is not remodelled after the completion of mineralization, a sequence of isotopic variations preserved along the tooth crown height constitutes a record of individual history over tooth growth (Bocherens et al. 2001; Sharp, Cerling 1998). Stable isotopes of oxygen, carbon and strontium will be considered in this study. The oxygen isotopic composition in tooth enamel (δ¹⁸O) derives from that of meteoric water that is linked to local temperature, causing a seasonal variation in the oxygen isotope ratios of precipitation, with the highest values occurring in the warmest months and the lowest values in the coldest months (Gat 1980). Therefore, it becomes possible to attribute a sample taken from a specific place along the tooth crown to a specific period of the year. The carbon isotopic composition (δ¹³C) of the tooth enamel is related to that of the plants eaten. The carbon isotopes ratios in the plants differ according to the metabolic pathways for carbon fixation during photosynthesis (for instance C₃ and C₄; see Vogel, Fuls, Ellis 1978), together with other factors such as altitude (Körner, Farquhar, Wong 1991). The strontium isotope ratio (⁸⁷Sr/⁸⁶Sr) in tooth enamel is derived from that of the eaten plants, which in turn reflects the strontium absorbed by the plant through water and soil ions – itself a reflection of bedrock geology (Ericson 1985).

13As a means to provide evidence of seasonal changes in the diet of herbivores, as well as the geological background of the pasture, the sequential analysis of stable oxygen, carbon and strontium isotopes is a valid approach for characterizing herding strategies in general and herd mobility in particular (Balasse et al. 2002).

14Eleven individuals (nine caprines and two cattle), all from Ovçular Tepesi, have so far been analysed. Since this study is still in progress, the following results should be considered with caution. A sequential sampling was performed along the crown of the sheep, goats and cattle lower molars by drilling the enamel on the buccal side of the teeth. The number of samples taken from one tooth varies from 14 to 20 for cattle teeth and from 12 to 16 for sheep and goat teeth. Enamel samples were treated in 0.1 M acetic acid (CH₃COOH) – 0.1 ml solution/0.1 mg sample – for 4 hours (Tornero et al. 2013).

15Pretreated enamel samples weighing ~600 μg were analysed for δ¹⁸O and for δ¹³C on a Kiel IV device interfaced to a Delta V Advantage IRMS at the Service de spectrométrie de masse isotopique du Muséum national d’histoire naturelle in Paris. Samples were tested for reaction under vacuum with 100% phosphoric acid (H₃PO₄) at 70°C in individual vessels and the produced CO₂ was purified in an automated cryogenic distillation system. The accuracy of the data was checked through the analysis of our internal laboratory carbonate standard (Marbre LM), whose theoretical values are the following: –1.83 ‰ for δ¹⁸O and 2.13 ‰ for δ¹³C (normalized to the international standard NBS 19). Results are expressed in V‑PDB.

16The remaining pretreated enamel samples were processed for ⁸⁷Sr/⁸⁶Sr isotopes in a Picotrace class ten clean room at the Yale Metal Geochemistry Center (Department of Geology and Geophysics at Yale University) by using in-house distilled ultra-pure acids. Approximately 5 mg of enamel were dissolved in 1 ml 6.2 M hydrochloric acid (HCl) in 5 ml acid-cleaned Teflon beakers and put on a hot plate at 100°C overnight to digest. Two drops of pure hydrogen peroxide (H₂O₂) were added to dissolve potential organic matter and samples were then evaporated dry. Dried samples were then dissolved in 1 ml 2 M nitric acid (HNO₃). A third of the sample (0.333 ml) was dried down and raised in 1 ml weak nitric acid (5% HNO₃ v/v) for measuring the quantity of major and trace elements on a Thermo Element XR ICP‑MS. Major and trace element abundances were measured using an in-house analytical standard, which was made using certified single-element standards to resemble the sample geochemistry as much as possible and to avoid potential matrix effects. A split of each remaining sample solution was purified for ⁸⁷Sr/⁸⁶Sr analysis using an ESI PrepFast‑MC‑Sr system (Romaniello et al. 2015). The resulting elution of strontium was evaporated on a hot plate and then raised in 1 ml weak nitric acid (5% HNO₃ v/v) for analysis on a Thermo Neptune MC‑ICP‑MS using NIST SRM 987 as a bracketing standard (average ⁸⁷Sr/⁸⁶Sr ratio of 0.71034+/-0.00001, 2SD). Analytical accuracy was assessed with the USGS carbonatite standard COQ‑1. We measured values of 0.70335±0.00005, consistent with the value of 0.70331±0.00002 (2SD) reported by Grünenfelder et al. (1986) for rocks from the same geologic unit.

Results

Pastoral signature of the faunal spectra

17Before examining the faunal spectra, it should be noted that the sites that appear to have been occupied for shorter, maybe seasonal, stays yielded very small faunal assemblages. The faunal spectra of all the sites presented here are dominated by domestic ruminants (i.e. sheep, goats and cattle). They represent more than 80% of the NISP in each single assemblage. Hunting is attested, but the remains of wild species represent only a small part of the assemblages. They are more frequent at Dzedzvebi (5.4% of the NISP in the Chalcolithic assemblage, 6.6% of the NISP in the Kura-Araxes assemblage) than at Ovçular Tepesi (5.2% of the NISP in the Phase I Chalcolithic assemblage, 1.9% of the NISP in the Phase II Chalcolithic assemblage, 0.9% of the NISP in the Kura-Araxes assemblage) (fig. 1). No remains were attributed to wild mammals in the assemblages from Duzdağı and Zirinçlik. Hunting occurred at Dzedzvebi and Ovçular Tepesi, but it did not play a significant role in the subsistence strategies of these communities. The people who settled at Ovçular Tepesi also fished in the nearby river (Berthon et al. 2013).

Fig. 1 – Relative representation of domestic caprine, cattle, domestic pig, other domestic mammals and wild mammals remains in percent of the NISP; OT = Ovçular Tepesi, DZ = Dzedzvebi, Chalc = Chalcolithic, KA = Kura-Araxes.

18Considering the importance of domestic ruminants in the assemblages presented here, we may conclude that the Chalcolithic and Kura-Araxes communities of the Kura and Araxes basins focussed on pastoralism, as opposed to a mixed farming strategy including a significant exploitation of domestic pigs. The assemblages, however, differ from each other in the relative proportions of the different species. In all assemblages Caprinae outnumber cattle. The highest frequency of cattle remains is found at Dzedvebi, where it amounts to respectively 36.40% and 49% of the total bulk in the Chalcolithic and Kura-Araxes levels. Cattle are less frequent on the sites located in Nakhchivan. At Ovçular Tepesi cattle remains represent only 6% of the number of sheep, goat and cattle bones in the Chalcolithic Phase I assemblage. Then it increases up to 19.1% in the Chalcolithic Phase II and 21.2% in the Kura-Araxes assemblage. The assemblage from Zirinçlik contains one cattle bone but the NISP is too low to calculate a percentage. No remains from Duzdağı were strictly identified as cattle but three bones belong to the “large-sized mammal” category (i.e. the size of cattle, equid or red deer). The possibility that some cattle were present in the herds of the Kura-Araxes groups that exploited the salt mine cannot be ruled out. Caprinae herds were composed of both sheep and goats in various proportions. Sheep represent from 60 to 70% of the number of Caprinae remains identified to the genus level. Such sheep/goat ratios are compatible with a goal of herd security (Redding 1984).

19Although Caprinae remains outnumber cattle in all the assemblages, the economic importance of the latter should not be underestimated. The weight of identified specimens (WISP) is a rough indicator of the contribution of each species in terms of quantity of animal products. Considering the WISP, cattle was the main species for the subsistence economy at Dzedzvebi during both the Chalcolithic and Early Bronze Age occupations. Conversely, during Phase I of the Chalcolithic occupation at Ovçular Tepesi sheep and goats represent 65% of the WISP. Caprinae and cattle contributed equally to the subsistence economy during the Phase II of the Chalcolithic occupation and the Kura-Araxes occupation at Ovçular Tepesi.

20Absolutely no domestic pig remains were identified in the faunal assemblages from Nakhchivan (Ovçular Tepesi, Zirinçlik and Duzdağı) and in the Chalcolithic assemblage from Dzedzvebi. On the contrary, a limited, but significant, number of domestic pig bones were identified in the Kura-Araxes assemblage from Dzedzvebi (n=11, 4.8% of the NISP). Domestic pigs are usually not found in Kura-Araxes faunal assemblages (Davoudi et al. 2018; Decaix et al. 2019). They are however present, albeit in small quantities, at the sites of Köhné Pasgah Tepesi (Davoudi et al. 2018) and Gegharot (Badalyan et al. 2014).

21The faunal spectra suggest that the Chalcolithic and the Kura-Araxes communities of the Kura and Araxes basins were highly specialised in domestic medium and large-size ruminant herding. Such a pastoral strategy is compatible with an extensive system aiming at exploiting grazing areas outside the vicinity of the settlement. However, the faunal spectrum is not in itself a proof of herd mobility. It is thus necessary to investigate the demographic composition of the assemblage in closer detail.

Demographic evidence of mobility

22The age at slaughtering can be a good indicator of the mobility of the herd (Arnold, Greenfield 2006). This indicator is based on the assumption that the absence of individuals from a given age group on the site may be linked to the fact that these animals were away during a particular period. For instance, if a herd is moved to spend the summer in the highlands two months after the birth season, it is likely that no animal aged between two and six months will be found in the lowland settlement’s assemblage. This method, however, implies (1) that the age range attributed to the various tooth-wear stages (Payne 1973; Vigne, Helmer 2007) are correct, (2) that the season of births is known, and (3) that the births season is short. Tooth-wear speed can vary depending on the abrasiveness of the food. However, a deviation from a standard due to abrasive feeding conditions would be certainly proportional to the use-life of the tooth and therefore mostly visible in older animals. Since the age groups of interest for evidencing mobility are the youngest ones, we consider that the age estimates provided by Payne (1973) are reliable. Concerning the birth season, it is known that sheep have a seasonal reproductive behaviour in temperate latitudes (Balasse et al. 2017). Nowadays in Nakhchivan, the introduction of the ram into the herd is well controlled and births occur in February and early March (personal observations by R. Berthon). During the European Neolithic period, the birth season of domestic sheep also started in late winter to early spring (Balasse et al. 2017). More problematic is the length of the birth season. Early domestic near-eastern sheep had a short birth season of two and a half months (Tornero et al. 2016b), while this season extends to three to four months for 6th and 5th millennium BC European Neolithic sheep (Balasse et al. 2017). A longer birth season would certainly blur the relevance of the age-at-slaughtering criterion for evidencing herd mobility. Birth seasonality is not yet known for Chalcolithic and Kura-Araxes sheep in the southern Caucasus, but we assume that births occurred from late winter onwards for an average length of three months.

23The sheep kill-off patterns of Phases I and II at Ovçular Tepesi (Chalcolithic), as well as the Chalcolithic and Kura-Araxes assemblages of Dzedzvebi, lack younger individuals – Payne classes A and B, from birth to six months old (fig. 2‑5). Admittedly kill-off patterns from Dzedzvebi are small and group both sheep and goats. Nevertheless, considering that most of the births occurred in March and April, no sheep were slaughtered on these sites before September. We consider this as an important piece of evidence for demonstrating the mobility of the herds and their absence from the surroundings of the settlements during the spring and summer seasons. One can argue that the youngest animals, those younger than two months old, were probably not slaughtered for their meat and that the exploitation of milk was achieved without killing them (Helmer, Vigne 2004). Individuals assigned to the B class, however, are frequently found in archaeozoological assemblages (Helmer, Gourichon, Vila 2007) and their absence here is striking. Individuals assigned to the B class are for instance present in the kill-off patterns of Godedzor, a 4th millennium BC high-altitude summer settlement located in Armenia (Palumbi et al., this volume).

24Only a few sheep and goat teeth were available for analysis from the Kura-Araxes assemblage of Ovçular Tepesi (n=6). They show that at least some animals were present on the site during the spring and summer seasons (fig. 6).

Fig. 2 – Kill-off pattern of sheep (Ovis aries) from Ovçular Tepesi, Chalcolithic Phase I, in corrected percent of the number of identified dental specimens (Payne 1973).

Fig. 3 – Kill-off pattern of sheep (Ovis aries) from Ovçular Tepesi, Chalcolithic Phase II, in corrected percent of the number of identified dental specimens (Payne 1973).

Fig. 4 – Kill-off pattern of sheep and goats (Caprinae) from Dzedzvebi, Chalcolithic, in corrected percent of the number of identified dental specimens (Payne 1973).

Fig. 5 – Kill-off pattern of sheep and goats (Caprinae) from Dzedzvebi, Kura-Araxes, in corrected percent of the number of identified dental specimens (Payne 1973).

Fig. 6 – Kill-off pattern of sheep and goats (Caprinae) from Ovçular Tepesi, Early Bronze Age Kura-Araxes, in corrected percent of the number of identified dental specimens (Payne 1973).

25Thus, it appears from the sheep and goat kill-off patterns of the valley sites, such as Ovçular Tepesi and Dzedzvebi, that the flocks were away in summer, which suggests that the pastoral practices of the Chalcolithic and Kura-Araxes communities of the Kura and Araxes basins included seasonal mobility. Isotopic analyses have been further carried out to test this interpretation, as well as to gather information on this particular form of mobility (vertical or horizontal).

Isotopic evidence of mobility

26Eight of the eleven individuals under study display lower δ¹³C values during the summer (where δ¹⁸O values are the highest) compared to the winter (where δ¹⁸O values are the lowest) (fig. 7). This seasonal variation of the δ¹³C values supports the hypothesis of a seasonal summer mobility of the herd towards high elevation pastures with moist ¹³C-depleted vegetation (Makarewicz, Arbuckle, Öztan 2017; Tornero et al. 2016a, 2017). Few of these individuals (n=2) had a diet with higher δ¹³C values (>–7.7‰) in winter (fig. 8). This type of fodder likely contained a significant proportion of C₄ plants. It is to be noted that none of the individuals had access to this ¹³C-enriched diet during the summer. This pattern of seasonal variation of the δ¹³C values does not correspond to the natural availability cycle of the plants, since C₄ species are more abundant during the warmest months of the year (Yamori, Hikosaka, Way 2014). Hence this pattern would rather suggest a change in pasturing places: the animals were probably taken to the uplands, since the quantity of C₄ plant greatly decreases in high altitude (Hatami, Khosravi 2013; Li et al. 2009). The question is whether this winter ¹³C-enriched signal corresponds to winter foddering with C₄ plant-enriched hay or rather to winter grazing in pasturelands with higher C₄ plant abundance. No definitive answer may be given to this question yet, but it should be noted that the enamel samples with the highest δ¹³C values each yielded a very distinct ⁸⁷Sr/⁸⁶Sr value (fig. 8). This means that in the case of foddering, hay would have been collected in at least three different geological units, whose distance from Ovçular Tepesi is yet unknown. C₄ plants are more abundant in arid but also salty environments (Hatami, Khosravi 2013; Pyankov et al. 2010). This is an important point, since we cannot exclude winter horizontal mobility towards salty pasturelands in the lowlands: the surroundings of the salt mine of Duzdağı may have been a possible destination, particularly in relationship with the seasonal exploitation of the salt deposits.

Fig. 7 – Results from stable carbon (δ¹³C; black symbols) and oxygen (δ¹⁸O; open symbols) isotope analysis (V‑PDB), and strontium isotopic ratio (⁸⁷Sr/⁸⁶Sr; red symbols) of the tooth enamel bioapatite of a sheep’s lower second molar from Ovçular Tepesi, Chalcolithic Phase I (erj = enamel‑root junction).

Fig. 8 – Results from stable carbon (δ¹³C) and strontium (⁸⁷Sr/⁸⁶Sr) isotope analysis of the tooth enamel bioapatite of sheep, goats and cattle from Ovçular Tepesi, Chalcolithic Phase I and Phase II; each sample corresponds to the maximal or minimal δ¹³C value in a given tooth; the season (winter or summer) is attributed in accordance with the δ¹⁸O value of the sample.

27The variation of ⁸⁷Sr/⁸⁶Sr values within a single tooth also testifies to the seasonal mobility of the animals. Nine of the eleven individuals display a significant seasonal variation of these ⁸⁷Sr/⁸⁶Sr values. We are not able to map the routes followed by these individuals yet, since the making of a map with biologically available ⁸⁷Sr/⁸⁶Sr values in Nakhchivan is still in progress. However, we may note that these patterns of mobility are quite heterogeneous. The graphs obtained from the ⁸⁷Sr/⁸⁶Sr values of some individuals do not even overlap with other graphs (fig. 9). This could be explained by the diachronic evolution of the mobility patterns, which is not necessarily visible in the stratigraphy. Herders may also change their mobility patterns and routes every year in order to cope with possible inter-annual variations in the availability of grazing resources. Since our isotopic approach covers one year of an individual’s life, this heterogeneous picture could result from inter-annual change in the location of winter and summer pastures. Lastly, a third hypothesis is also possible. Ovçular Tepesi may have been settled by different groups, contemporaneously or not, each having a different seasonal mobility pattern.

Fig. 9 – Results from stable (⁸⁷Sr/⁸⁶Sr) isotope analysis of the tooth enamel bioapatite of sheep, goats and cattle from Ovçular Tepesi, Chalcolithic Phase I and Phase II; the maximum and minimum values are plotted for each tooth; open symbols indicate that the maximal or minimal portion of the curve was missing.

28The preliminary isotopic results obtained from the sheep, goat and cattle teeth of Ovçular Tepesi (Chalcolithic Phases I and II) evidence various forms of herd mobility, which include seasonal vertical mobility: tooth OT Cap3 M2 clearly belonged to a sheep involved in pastoral transhumance (fig. 7). This specimen displays a pattern of inverse cyclical isotopic variation characterized by high summer season δ¹⁸O values coincident with low δ¹³C values, which suggests that this animal moved to moist, high elevation pastures during the summer months. The variation of the ⁸⁷Sr/⁸⁶Sr values throughout the year also shows that vertical mobility was correlated with a trajectory towards an area characterised by a different geological background.

29Thus, the available evidence testifies to the existence of seasonal vertical mobility from at least the middle of the 5th millennium onwards, that is roughly when extractive copper metallurgy started to develop in the South Caucasus (Gailhard et al. 2017; Gailhard et al., this volume): one wonders whether pastoral strategies had a role in the exploitation of highland territories and their mineral resources by Chalcolithic communities.

Discussion

The antiquity of the pastoral system and vertical mobility in the Araxes valley

30The almost exclusive exploitation of domestic sheep and goats characterizes subsistence economies in the Araxes valley since the arrival of domestic mammals in this area towards the end of the 7th millennium BC (Berthon 2014). For instance, sheep and goats represent up to 90% of the mammalian NISP in the Neolithic occupation levels at Aratashen and Khaturnarkh-Aknashen (Bălăşescu et al. 2010; Vila et al. 2017). This specialised form of animal exploitation seems to derive from a pastoral tradition that appeared during the Neolithic; it is not a specificity of Chalcolithic or Kura-Araxes herding strategies. But the case with herd mobility is not so clear since the isotopic data on the Neolithic cattle and caprines of the Araxes valley is not available yet. Judging by the existing demographic and isotopic data, the practice of mobility in herding strategies can only be demonstrated from the end of the 5th millennium BC onwards.

31If we turn to kill-off patterns during the Neolithic, the information retrieved from Aknashen indicate that the herd was present at this lowland settlement during the spring and summer seasons (Bălăşescu et al. 2010; Vila et al. 2017). The demographic evidence is less clear at Khaturnarkh-Aknashen since the younger age groups are represented only by a few individuals in some of the occupation levels (Bălăşescu et al. 2010; Vila et al. 2017). Furthermore, to the best of our knowledge, no Neolithic settlement with domestic fauna has been yet identified in the piedmont or the highlands surrounding the Araxes valley. The intensification of archaeological investigations in the highlands might however change this view in a near future.

32Judging by the available data, it is possible that herd mobility in the Araxes valley only appeared with Chalcolithic pastoral practices some time during the 5th millennium BC. The preliminary results from stable isotopes analyses point towards prevalent seasonal vertical mobility (eight out of eleven analysed teeth). The herds were moved to the moist highlands during the warmest part of the year. The emphasis put on sheep and goats is reflected in the symbolic world of these human communities: the only animals illustrated at Ovçular Tepesi during the Late Chalcolithic are caprines (fig. 10), which are depicted in relief on large vessels (see Gailhard et al., this volume). From an iconographical point of view, these animals closely resemble the caprines incised over the rocks of the highland pastures at Gəmiqaya in Nakhchivan (Gailhard et al., this volume). Although there is no absolute dating for the petroglyphs of Gəmiqaya, the caprines depicted over these rocks clearly link the valley-settlements of the Chalcolithic with highland pastures.

Fig. 10 – Relief-decorated jar with caprines from Ovçular Tepesi, Chalcolithic Phase I, Locus 12128 (MBA, N. Gailhard).

The appearance of specialised pastoralism in the Kura Valley

33Contrary to the Araxes valley, subsistence strategies in the Kura valley during the Neolithic were not completely focused on sheep and goat herding. Cattle and pigs also played a minor role (Berthon 2014; Benecke 2017). The only exception to date is Hacı Elamxanlı Tepe, one of the earliest Neolithic settlements in the area, where sheep and goats represent 90% of the NISP. Kill-off patterns at the Neolithic sites of Aruchlo, Kamil Tepe and Hacı Elamxanlı Tepe show that at least part of the herd stayed in the vicinity of these lowlands settlements during the spring and summer seasons (Benecke 2017; Nishiaki et al. 2015). The only piece of evidence suggesting vertical mobility during the Neolithic period in this area could be the Bavra Ablari rock shelter situated at 1,660 m asl in Georgia. Although the faunal assemblage is composed mainly of sheep and goat remains, however, their domestic or wild status is still under investigation (Varoutsikos et al. 2017).

34Beside Dzedzvebi, very little information concerning the exploitation of animal resources in the Kura basin is available for the Chalcolithic period. The only published zooarchaeological analyses come from the sites of Tsopi (Nebieridze 2010) and Damtsvari Gora (Varazashvili 1992). We have unfortunately very few details about the chronological sequence, or the collection and identification methods. The information provided by these two assemblages is very different from the picture that emerges from Dzedzvebi. They are characterized by the presence of domestic pigs, a significant proportion of wild game at Tsopi (18% of NISP), and a high proportion of cattle (72% of NISP) at Damtsvari Gora. The Chalcolithic faunal assemblage from Dzedzvebi is so far the only one in the Kura area that displays a specialised pastoral-fauna spectrum and a kill-off pattern compatible with seasonal mobility. In that sense, there is a clear break between the strategies of animal exploitation on the Neolithic sites of the Shomu-Shulaveri tradition and that of the Chalcolithic community at Dzedzvebi.

Did mobile pastoralism play a role in the emergence of mining activities in the Caucasus?

35The data related to the herds’ demography strongly suggest that Chalcolithic, and to some extent Kura-Araxes, communities in the Kura and Araxes basins moved their animals on a seasonal basis. The mobility of the herds, however, encompasses a large variety of practices that do not necessarily include transhumance towards the highlands. The results from the stable isotopes analyses evidence a seasonal vertical mobility at Ovçular Tepesi. The Chalcolithic groups who occupied this settlement at the end of the 5th millennium would move their animals to highland pastures, where the ore resources are located. Hence, it seems relevant to discuss the role of mobile pastoralism in the emergence of mining activities.

36As a matter of fact, the faunal assemblages presented here do not, with the exception of Duzdağı, come from actual mining sites but rather from settlements occupied by communities mastering copper smelting and, for the Kura-Araxes occupation at Dzedzvebi, gold processing. Drawing on taxonomic, demographic and isotopic data (for Ovçular Tepesi), we suggest that Chalcolithic and Kura-Araxes communities who settled at Ovçular Tepesi, Dzedzvebi, Zirinçlik, and Duzdağı, shared similar herding strategies based on the almost exclusive exploitation of sheep, goats and cattle, which included some kind of seasonal mobility.

37The focus on sheep and goats is not in itself closely related to the intensive exploitation of mineral resources. In the Araxes valley, this type of animal-exploitation strategy appears together with other characteristics of the Neolithic way of life, much earlier than the beginning of ore-smelting or mining activities. Information on herd mobility during the Neolithic period in the Caucasus is not available yet, since isotopic analyses on the fauna from Kültepe I in Nakhchivan are still in progress. So far, our work has provided evidence for the seasonal vertical mobility of the herds from the 5th millennium BC onwards, which is concomitant with the first evidence of copper smelting at Ovçular Tepesi and salt collection at Duzdağı (see Gonon et al., this volume). In the Kura area, Neolithic economies were not so much focused on sheep and goat exploitation, but a specialised pastoral herding strategy with probable seasonal mobility is evidenced at Dzedvzebi during the Chalcolithic, where groups practising this type of animal exploitation strategy also mastered copper smelting.

38The evidence testifying to the role of mobile herding strategies in the development of mining is still scanty. But we may note that metallurgical activities were performed on pastoral mobile sites, whatever their size, as shown by the examples of Zirinçlik and Uçan Ağıl. The piedmont site of Zirinçlik for instance yielded mould fragments and copper ore, indicating that the people who occupied this site were clearly involved in copper metallurgy (Gailhard et al., this volume). Although the faunal assemblage is very small, the fact that only domestic sheep, goat and cattle remains were identified, together with the location of the site, suggests that Zirinçlik was indeed occupied by groups practising mobile pastoralism. This is also the case for Uçan Agil, where copper ore and a copper nugget have been found in the Kura-Araxes occupation levels (Gailhard et al., this volume). We can therefore surmise that herding strategies that included vertical mobility contributed to the exploitation of copper ores situated in the piedmonts and to the circulation of ores and copper items.

Conclusions

39The beginnings of herd mobility in the Caucasus are still unknown, but this paper tried to elucidate the circumstances in which mobile pastoralism developed in the South Caucasus during the 5th to 3rd millennium BC, as well as its relation with the concomitant rise of the exploitation of mineral resources. We wanted to test the hypothesis according to which the vertical mobility of the herds contributed to the exploitation of the minerals (ores) located in the piedmonts and their distribution across a larger area. The faunal spectra of four sites located in Nakhchivan (Azerbaijan) and Georgia was thus studied through zooarchaeological and isotopic approaches.

40Available results suggest that the Chalcolithic and the Kura-Araxes communities of the Kura and Araxes basins were highly specialised in domestic sheep, goat and cattle herding. Such a pastoral strategy is compatible with an extensive system aiming at exploiting grazing areas outside the vicinity of the settlement. However, the faunal spectrum in itself is not a proof of herd mobility. We therefore investigated the demographic composition of the assemblage in greater detail: it appears from the sheep and goat kill-off patterns of the valley sites, such as Ovçular Tepesi and Dzedzvebi, that the flocks were away in summer, which suggests that the pastoral practices of the Chalcolithic and Kura-Araxes communities of the Kura and Araxes basins included seasonal mobility. Isotopic analyses have been further carried out to test this interpretation, as well as to gather information on this particular form of mobility (vertical or horizontal). Most of the analysed specimens (i.e. sheep, goat and cattle lower molars) display a pattern of inverse cyclical isotopic variation characterized by high summer season δ¹⁸O values coincident with low δ¹³C values, which suggests that these animals moved to moist, high elevation pastures during the summer months. The variation of the ⁸⁷Sr/⁸⁶Sr values along the year also shows that the areas targeted by vertical mobility are characterised by a different geological background. Although it is premature to discuss the origins of vertical transhumance in the Caucasus, we can now prove that this peculiar form of herd mobility was effective at least by the last quarter of the 5th millenium BC. As a working hypothesis, we surmise that herding strategies that included vertical mobility have contributed to the exploitation of mineral resources and to their distribution. This hypothesis seems so far to be confirmed by the evidence produced by paleometallurgical studies (Gailhard et al., this volume).