Introduction

1Duzdağı is a major salt dome located some 7 km to the northwest of Nakhchivan city, along the former Silk Road linking Tabriz to Constantinople (fig. 1‑2). The salt deposits, which are still exploited today, are dotted with archaeological artefacts that are attested in large quantities over the slopes of the dome, such as ceramic potsherds and macro-lithic stone tools (Hamon et al., this volume; Marro, this volume).

Fig. 1 – Duzdağı, the main salt dome (MBA, S. Sanz).

Fig. 2 – Map of Nakhchivan (map background: O. Barge, CNRS).

2Systematic archaeological investigations at Duzdağı started in 2008 under the auspices of the Azerbaijani National Academy of Sciences in the frame of a French-Azerbaijani joint project led by V. Bakhshaliyev and C. Marro. A four-year survey was carried out between 2008 and 2011; it was followed by excavations on specific targets that aimed at investigating the Late Chalcolithic and Early Bronze Age mining systems (5th-3rd millenia BC).

3This paper will present an overview of the excavation results that were obtained over the last five years (2011-2016) in two areas, M1 (Window 2) and M9 (Window 4). M1 and M9 are located respectively on the south-western and western part of the dome (fig. 3) in two geologically distinct salt layers: excavation trench M9 was opened in Salt Layer 1 (SL1), which appears as a hard, darkish salt deposit, while M1 is located in Salt Layer 2 (SL2), which is much softer and also whiter. Excavation trenches were delimited following the archaeological remains scattered over the slopes. Since a large concentration of Kura-Araxes potsherds had been collected in Window 2, especially at the foot of the survey transect, we decided to open two excavation trenches in that area (M1 and M5), which were later grouped together under a single label, M1. Excavations were originally planned in Window 4 because an opening, which seemed to lead into a pillared tunnel (fig. 4), was visible in the slope at mid-height. This opening was labelled M9. Since a dozen Kura-Araxes macro-lithic tools had been registered at the bottom of the nearby wadi during survey operations in 2010, we wanted to check whether M9 could correspond to some early form of underground mining works.

Fig. 3 – Map of Duzdağı with salt layers and windows (CNES, Spot 5 satellite picture).

Fig. 4 – Duzdağı, M9-004, the “pillar-like” rock-salt face near the southern “entrance” (MBA, C. Hamon).

Geological study

4Archaeological investigations were completed by a geological study carried out in 2014, with a view to defining the specificities of the different salt formations attested at Duzdağı, since no other study related to the Duzdağı salt dome had so far been published.

5The formation of the salt dome of Duzdağı followed the Miocene marine transgression in the Eastern Paratethys region studied here, which ended after the Miocene. The Pliocene is represented by terrestrial deposits. The geological study has distinguished three major salt levels (SL1, SL2 and SL3, fig. 3) of marine origin, as shown by the nannoplankton species comprised in the illite-bearing clayey layers (fig. 5). Their Br composition (≥ 50 ppm) also confirms this marine origin. Between the salt levels, layers of grey-red mudstone or grey mudstone/clay were deposited during non-evaporitic stages. In Salt Level 2 (SL2), gypsum is also present, either as embedded layers in the salt or as single gypsum crystals (“selenite”) in the mudstone/clay layers: they are composed of a mixture of clay, illite and Fe-kaolinite, as shown by the SEM-EDX analysis (fig. 6).

Fig. 5 – Nannoplankton species in the illite-bearing clayey layers (B. Varol).

Fig. 6a‑d – Semi-quantitative elemental analysis “edx” of salt/halit-bearing clay layers (a) showing the presence of illite, and a mixed layer of clay and Fe-bearing kaolinite (b‑d) (B. Varol).

6According to the Strontium isotopic analysis of the gypsum contained in SL2, the salt deposits of Duzdağı can be dated to the Middle/Late Miocene (tab. 1).

7The salt levels labelled SL1, SL2 and SL3 do not have the same structural characteristics, a fact which has consequences in terms of salt exploitation techniques: the sub-vertical inclination of the salt layers in SL1 is covered by colluvium (ca 50-110 cm thick) (fig. 7); they display much higher rock faces, which are compatible with the digging of mining shafts, while the horizontal structure of SL2, combined with its lesser density, favours above-ground exploitation. It must be noted, however, that the gypsum layers visible in SL2 were subjected to soft deformations resulting from salt tectonics, as shown by their undulating structure. Similarly, folded structures are frequent in SL3 (fig. 8-9).

Fig. 7 – SL1, diapiric salt layers covered with colluvium (MBA, B. Varol).

Fig. 8 – Salt dome (SL3) with entrance to the mine, see arrow (MBA, B. Varol).

Fig. 9 – Diapir fold of SL3 (MBA, B. Varol).

8SL1, SL2 and SL3, which may be considered as the major salt levels of Duzdağı, are separated by mudstone interlayers that are 20-110 m thick; the measurable thickness of the salt layers varies between 2 and 7 m (SL1: 5-2 m; SL2: 2-3 m; SL3: 7 m). This is of course the thickness that was visible from the slopes, that is from the dome’s edges. The layers inside the dome may in fact be much thicker, especially as the salt layers shape into a dome. It must be noted that the salt stone levels have the same chemical composition (NaCl 92-98%). The rest is made of a mudstone matrix or mud bands and interlaminae. Some layers in SL1, SL2, SL3 may contain up to 96-98% NaCl (halite); this is due to a lesser amount of mud contamination. In spite of their very similar elemental composition (Na: 18,000-18,500 ppm; Cl: 141,091-121,945 ppm), SL1, SL2 and SL3 are characterized by different depositional and structural characteristics: the salt deposits of SL1 are massive, but SL2 alternates with gypsum, while SL3 is streaked with mudstone (clay) interlayers. The streaked deposits are composed of numerous alternate layers of white- (clear) and dark-coloured halite (fluid-inclusion rich). As a result, the salt beds can easily be exploited along these interlayers/bands. On the other hand, SL3 displays muddy inter-beds (40-70 cm), which are cut by halite cemented fractures.

9The salt of Duzdağı can be correlated with the salt beds of Tuzluca (Iğdır), which are located in north-eastern Turkey, some 220 km from Duzdağı (Varol et al. 2016).

Excavations results

10According to the engineers working on the modern mine, the salt deposits of Duzdağı cover an area approximately 3 km long and 2 km wide (fig. 3, 10): archaeological artefacts were encountered in lesser or greater concentrations virtually all over the surface of these deposits. Consequently, the investigation of this huge archaeological area is only possible through a selection of excavation trenches in accordance with specific targets. Since our work programme was focused on early mining, we opened the first excavation trenches where the concentration of Late Chalcolithic and Early Bronze Age ceramic potsherds was the greatest: this was at the bottom of Window 2.

Fig. 10 – Aerial view (2016) of the trenches opened in or in the vicinity of W2 (MBA, N. Gailhard).

11A first trench (M1) was opened on what we assumed to be a Kura-Araxes work platform located at the foot of a depression that somehow looked like the entrance to a tunnel. A second trench (M5) was opened around the depression itself in order to investigate this putative tunnel. Instead of a tunnel, however, we found a series of occupation layers, characterized by the remains of animals dung in the midst of ceramic potsherds dated to the Late Iron Age. The depression turned out to be the outcome of subsurface deformation processes, probably brought about by the collapse of natural galleries resulting from underground running waters. The deceptive nature of the archaeological deposits of Duzdağı were thus apparent from the onset: ancient exploitation traces were aplenty, but it took us several years to set up the archaeological stratigraphy of the site, as well as to figure out the general organisation of salt exploitation and its evolution through time.

12A total of 16 trenches (labelled M1 to M16) were opened, seven of which are located in or around Window 2, while four were opened in Window 4, including M9 (fig. 11). Excavations proved to be extremely difficult in most places because of massive colluvium accumulation processes. Of all excavation trenches, M1 and M9 turned out to be the most fruitful since they give an overview of salt exploitation mining works from the 4th millennium BCE until the beginning of the Christian era.

Fig. 11 – M1 excavation area, state of the art in 2015 (MBA, N. Gailhard).

The earliest mining infrastructures: evidence of a cell-mining system from the Kura-Araxes period in M1

13In spite of the presence of Late Chalcolithic pottery at Duzdağı, which is altogether rare but concentrates in W2 together with Kura-Araxes potsherds (Marro, this volume), no traces of Late Chalcolithic salt extraction were found in the area we excavated in M1. Given the nature of mining activities, it is possible that the earliest traces of salt exploitation at Duzdağı, which, according to the pottery, date back to the second half of the 5th millennium, were destroyed by later miners. This is a common situation in mining archaeology: late comers tend to destroy the work traces left by their predecessors. Another explanation, however, is also possible: the earliest salt exploitation at Duzdağı may have been carried out by the surface collection of salt, not by mining. This would suggest some kind of opportunistic activity, which left very few traces and probably reflects a low intensity in salt use. We have no real evidence to favour one hypothesis over the other: all that we may say is that the surface-collection hypothesis is coherent with the relatively low number of Late Chalcolithic potsherds attested in Window 2, which stands in marked contrast with the abundance of Kura-Araxes potsherds in the same area.

14The earliest mining installations that were brought to light in M1 date back to the Kura-Araxes period: they appear as semi-circular shaped structures that were probably designed to take advantage of the natural depressions carved by erosion in SL2, two of which were brought to light in 2012 (fig. 12). The extraction of salt in such depressions may have been carried out by one single worker who would mine the salt sideward in each depression; this activity resulted in the formation of semi-circular mining cells at a man’s height, and left a small protruding overhang (fig. 13).

Fig. 12 – Duzdağı, semi-circular mining cell, M1-004 (MBA, N. Gailhard).

Fig. 13a‑b – Duzdağı, protruding overhang, M1-004 (MBA, C. Marro).

15The salt collected from these cells was most probably extracted with the pointed stone tools found in and around M1 (fig. 14). Indeed, the technological analysis carried out by C. Hamon has shown that such cells may result from the use of pointed stone tools (Hamon 2016, pp. 510-528; Hamon et al., this volume). These cells, as well as the natural depressions, contained a number of ceramic potsherds that all date to the Kura-Araxes period.

Fig. 14 – Duzdağı: example of pointed hammer (MBA, C. Hamon).

16This cell-mining system suggests individual salt-mining strategies, where salt was mined by small groups of people, possibly family units. No considerable work investment requiring the power of an organized and specialized work force is needed to run such a system, unlike the work that would be required for digging tunnels. These groups or individuals came to collect salt for their own consumption, and possibly as well for exchange with other groups of people living in the Caucasus and beyond.

17The multiple-small-mining-groups hypothesis is supported by the variety of clays used in the making of the Kura-Araxes jars from Duzdağı (Marro, this volume): if the sizes and shapes of these jars are fairly standardized, no two jars are alike, in particular as far as the temper used for crafting the vessels is concerned, which shows a much wider range of mineral combinations than the range exemplified in the Kura-Araxes pottery from the settlement of Ovçular Tepesi, for instance. The jars from Duzdağı were seemingly brought along by villagers coming from distinct settlements, some of which may have been located quite far from the mine.

18If the salt from Duzdağı was extracted on a relatively small scale, the ubiquity of Kura-Araxes potsherds and stone tools all over the slopes of the dome testifies, however, to the importance of salt in the Kura-Araxes economy. The combination of the ceramic and macro-lithic evidence with the actual mining infrastructures and spatial distribution of the finds reflects a mining system where salt exploitation was executed on a regular basis by non-professional miners, who had nonetheless a high level of expertise and skill (Hamon 2016; Hamon et al., this volume). Salt extraction during this period may have been carried out for domestic consumption and for exchange purposes. It should be mentioned, however, that the Araxes and Urmiah basins are both rich in rock-salt deposits (Marro, this volume, fig. 1), so that salt bartering, if any, during the Kura-Araxes period would probably have had a social, more than an economic, function. But it cannot be excluded that salt from the Araxes basin also travelled far away to non-salt bearing regions such as the North Caucasus.

19The exact date and the duration of the Kura-Araxes mining activities in M1 are still a matter of debate: ceramic parallels and two ¹⁴C dates obtained from a piece of charcoal (US 67) and a sheep mandible (US 107), point to a fairly wide time span within the first half of the 3rd millennium BC (ca 3000‑2600 BCE) (tab. 2). It must be stressed, however, that the Kura-Araxes pottery, as well as its related hearths, are often mixed with Late Bronze/Early Iron Age ceramics and other hearths: it is clear that the Kura-Araxes pithead has been disturbed by later exploitation episodes. Only by the northern side of the excavation trench did we find a fairly large area with no later disturbances: a hearth was brought to light next to a scatter of Kura-Araxes ceramics (US 107), where the sheep mandible was sampled (fig. 15).

Fig. 15 – M1, Kura-Araxes ceramic scatter, US 107, located west of the mining area, US 002 and US 003 (MBA, N. Gailhard).

20It results from this situation that the duration of the Kura-Araxes salt-exploitation episode in Window 2 is difficult to assess with precision: are we faced with a series of small-scale, but regular, mining activities over several centuries or with a relatively short (a few years or decades) but more intensive mining episode? In this particular case of horizontal stratigraphy, all the organic matter left across the pithead should be dated through radiocarbon analyses in order to map Early Bronze Age activities both spatially and chronologically. The ceramic evidence, which is fairly homogenous typology-wise, certainly does not support the first hypothesis (salt mining over several centuries), but it seems difficult to say more for the time being. It must also be stressed here that the duration of Kura-Araxes salt extraction in M1 does not necessarily reflect the overall duration of Kura-Araxes activities in other parts of the dome, whether it be in SL2, SL1 or SL3, which may have been much longer: the ceramic evidence actually suggests that salt extraction may have started well back into the 4th millennium BC, since a number of potsherds at Duzdağı display close similarities with the Kura-Araxes pottery from Şorsu, a mobile camp located at the foot of the Sirab piedmonts. This camp has been dated by radiocarbon analyses to the first half of the 4th millennium BCE.

M1 during the Late Bronze/Early Iron Age (LBA/EIA): evidence of a slab-extraction system

21Another salt-extraction phase documented in M1 belongs to the Late Bronze/Early Iron Age, two periods that are still ill-defined in terms of ceramic repertoire, which explains why a broad LBA/EIA label is generally used to refer to the 1500‑900 BCE time span.

22The LBA/EIA mining zone is located only about 2 to 5 m from the Kura-Araxes exploitation area (fig. 16). The salt deposits exploited by the LBA/EIA miners lie closer to the core of Salt Level 2, while earlier mining infrastructures are located along its periphery. The LBA/EIA miners had to cross both the pithead and the mining cells used by Kura-Araxes workers in order to reach their work area, while still using the pithead themselves: all this explains why the latter is in such a bad state of preservation. The location of the LBA/EIA mining infrastructures further inside the dome also meant that these infrastructures were later covered by a much thicker layer of colluvium than those located along the dome’s edge, where the Kura-Araxes mining cells were found. While the layer of colluvium was about one meter thick around the slope’s edge, it reached a thickness of 5 to 8 m inside the slope, where the LBA/EIA miners extracted their salt. This situation probably results from the cliff created by the Iron Age exploitation system, but the erosive process was also sped up by the geological layout on top of the salt bed, which consists in alternate layers of salt and clay. Earthquakes, which affect the area from time to time should also be taken into consideration to account for these massive erosion processes.

Fig. 16 – Plan of mining area M1, which includes ex-area M5 (T. Gonon).

23The work area opened by LBA/EIA miners was an opencast, just as during the Kura-Araxes period, but the exploitation methods they used were totally different. The mining area we brought to light measures 3x3 m; it rests directly upon the salt bed. This area was probably already covered by colluvium by the time it was re-opened for salt exploitation at the end of the 2nd millennium BC; its thickness at the beginning of LBA/EIA mining activities, however, remains unknown. A sophisticated technique of extraction was used: salt was exploited by cutting large slabs into the salt bed (fig. 17). Seven steps, each corresponding to the cutting of one slab about 5‑10 cm thick, were brought to light. Thanks to the orientation of these steps, it appears that the LBA/EIA miners started to exploit the salt deposits from the north-west, since the resulting flight of stairs goes down from east to west. Judging by the orientation of the step trench, extraction probably started from the wadi located north of M1.

Fig. 17 – F2 (“Window 2”)-M1, slab-extraction area (MBA, C. Marro).

24The steps are 30 to 50 cm large and the longest is 4 m long. We do not know the precise measurements of the extracted slabs, as none were found during the excavations. But judging by their width, they may have been as long as 50 cm and more or less rectangular in shape.

25Ethnographic examples give an idea of the possible techniques employed for processing the salt slabs of Duzdağı. At Taoudenni in West Africa (present-day Mali), the extraction of rock-salt slabs was described by Odette du Puigaudeau, a French ethnographer, who visited the mine in the 1930s (du Puigaudeau 2001 [1940]). She describes how the slabs were trimmed after extraction into a standardized shape, by retouching the edges with a pick to prepare them for trade. If the slabs happened to break, the miners would cut them into smaller blocks, so that nothing would remain unsold, or unused. According to O. du Puigaudeau (2001 [1940], pp. 153‑155), the slabs from Taoudenni weighted between 25 and 40 kg; up to 10 slabs were transported on camel backs, together with other load packages. From the pictures taken by this ethnographer, it seems that these slabs, which were meant for trade, measured ca 1.20 m per 0.50 m.1 These slabs were and are still transported by camel caravans from Taoudenni along a 1000 km route through the Sahara; this caravan, and its bi-annual journey, is known as the Azalaï.

26Judging by the width of the steps left by salt extraction, the slabs extracted from Duzdağı were certainly smaller; this specificity might be explained by the means of transportation available to LBA/EIA miners, which, according to the faunal remains found in M1, were not camels. The hoof of an equid, without the phalanx bone, was found in Locus M1.016; its relative good state of preservation must be accounted for by the salt contained in the sediment. Considering its size, this hoof probably belonged to a domestic donkey (fig. 18). This is the only piece of evidence that would point to the use of pack animals on the mine during the LBA/EIA period, but it should be kept in mind that such remains are less likely to be found on a mining site than food leftovers, since this is an unlikely place for pack animals to be slaughtered.2 Evidence of possible food is provided by the other bones collected during the excavations, which reveal the presence of caprines, including juveniles, during the LBA/EIA period3: these are probably the remains of the meals taken by miners on the mine. The age of the juveniles suggests these activities took place in early spring.

Fig. 18a‑b – Left: hoof of a LBA/EIA domestic donkey from Locus M1-016; right: example of modern hoof for comparison (MBA, N. Gailhard).

27The tools related to the slab-extraction area in Duzdağı are different from the ones used for the Kura-Araxes cell-mining system (Hamon, this volume): they are much less sophisticated. No clear tool traces could be seen from the salt surface, probably because they remained in the open air and suffered from erosion. A number of wooden wedges were also found in the salt debris, immediately to the east of the slab-extraction area. These wedges may have been used in the mining process to detach the slabs; a tricky operation that probably required a whole set of different tools.

28Another important point about salt mining during the LBA/EIA period is the change in work organization and scale: in contrast to the Kura-Araxes system, the extraction of slabs, which probably had standardized dimensions, suggests that salt mining had developed into a specialized task carried on by one or several groups of trained miners. These slabs were then probably sold to merchants, who may have traded them over long distances. No archaeological evidence, however, is available as regards the existence, the scope and organization of such a trade.

29After the abandonment of the LBA/EIA mining area, the erosion process started again, and the slab-extraction zone was covered by a thick layer of colluvium. Some form of salt exploitation was later carried on in the Late Iron Age or in Antiquity, as suggested by the pottery and the remains of animal dung left further up the colluvium deposits. These animals may have been used for carrying salt, but the botanical analysis carried out on the dung by N. Boenke gave no further clues concerning their identification. Moreover, the extraction area related to these putative pack animals is still unknown. One radiocarbon reading (M1-027) dates this phase to the 200 BC‑30 AD time span (tab. 2).

30As for the slab-extraction zone, its dating is provided by the pottery and four ¹⁴C dates, one of which (1200‑1000 BCE cal., tab. 2) comes from a hearth (M1-045) located to the west of the extraction zone. This time span is coherent with the handful of LBA/EIA potsherds that were found in this area.

A case of semi-underground salt exploitation: the overhang mining works of M9

31Lastly, a third system of salt extraction was brought to light in M9 (fig. 3), where an overhang quarry was found in the earliest salt deposits, SL1, which are characterized by their hardness. M9 is located above a narrow wadi, close to a pass leading to the northern stretches of the salt dome, some 700 m to the north of M1. A similar overhang quarry is visible just after the pass, but it has not been studied.

32The salt beds from SL1 are not horizontal: they follow a sub-vertical inclination that probably results from tectonic salt processes: the salt layers present a 30 to 40°C angle, from west to east. Consequently, other methods were developed to mine salt in that area, which was extracted by cutting vertically into the rock face: this resulted in the formation of “chambers” aligned in a row under the overhang. These “chambers” actually result from erosion processes, which progressively sealed the mining works by filling the gap between the work floor and the overhang: there should not be considered as original mining chambers. Thanks to this process, which partly shielded the rock face from erosion, ancient tool traces were preserved in different parts of the rock face.

33M9 consists in three adjacent “chambers” or sections (fig. 19), which are aligned from north to south over a total width of about 30 m. The central section could not be studied in detail: during the 19th or 20th century AD, people tried to exploit the remaining salt deposits with explosives. Extraction tests destroyed most of the previous tool traces that had been left in this area and also destabilized the roof, making it unsafe. Safety measures had to be taken before we could work in M9. It is interesting to note that such measures had also been taken by the miners themselves, as shown by the remains of wooden posts and poles that served to prop the overhang and prevent it from collapsing (fig. 20).

Fig. 19 – Plan of the excavation area in M9, showing the three sections of the overhang area (MBA, T. Gonon, G. Gadebois, 2018).

Fig. 20 – Duzdağı, M9, remains of a wooden propping system (MBA, C. Marro).

34The south section was also difficult to study. We met with an important amount of debris that had accumulated over the work floor, which could only be reached over one square meter: the overhang had also been destabilized in that area by the use of explosives in the central section. The other problem regarding the south zone is that it has been mined from three sides (west, south and east); which, from the outside, gives the impression that the rock face is a pillar located at the entrance of a tunnel (fig. 4). This was precisely why we had opened a trench in M9 to start with. Since the overhang seemed to be dangerous, only a limited excavation trench was opened on its western side, in front of this apparent pillar.

35The northern section is the most interesting, as neither its shape, nor the tool traces were damaged by later mining works. A test trench was opened, which yielded organic material through which the overhang quarry system could be dated: a hearth (M9‑102) was discovered immediately on top of the salt bed and dated between 346 and 49 BCE cal. (2 sigmas). Another radiocarbon date (400‑200 BC cal.) was provided by a seed stuck under the sole of a rope sandal that was found in the central section (M9‑101). This sandal, probably some kind of an overshoe used to protect the miners’ shoes, was found in the central part during our first visit. It was lying over the surface of the mining debris, which suggests that the colluvial process had been fairly limited under the overhang itself after M9 was abandoned. Altogether, four ¹⁴C readings concur to date the overhang mining works to the 400‑30 BC time span, that is to the Late Iron Age (beginning of the Parthian period) (tab. 2).

36As with the LBA/EIA slab-extraction system in M1, the shapes of the salt chunks extracted in M9 are visible in the negative over the rock face. These traces, however, show that salt in the Late Iron Age was no longer extracted by slabs, but in small blocks (fig. 21), which measured about 32x22 cm (thickness unknown). These blocks were thus much smaller in size than the slabs extracted during the LBA/EIA period, which implies major differences in salt-extraction methods, but probably also in distribution processes. Salt-extraction methods are partly linked to the specific density and structure of the salt deposits in SL1, which are sub-vertical and much harder than in SL2: they are not suitable for extraction in slabs. Conversely, the production of relatively small salt blocks probably also had consequences in terms of trading value: could the salt of Duzdağı be as valuable in small blocks during the Late Iron Age as it was in large slabs during the Late Bronze/Early Iron Age? If not, what determined the Late Iron Age miners to choose the hard salt deposits of SL1 in W4, next to the softer and more easily accessible deposits of SL2 in W2? It is impossible to answer these questions for the time being, but it is probable that the small Late Iron Age blocks of M9 were much easier to carry and distribute than the large slabs produced earlier: they could easily be piled in a bag and transported by small-sized pack animals, such as donkeys.

Fig. 21 – Duzdağı, M9-003, salt-block rough-out (MBA, C. Marro).

37Another specificity of the Late Iron Age salt extraction in M9 is that the traces of extraction are often grouped together in shallow “niches”, which cover a surface of about one square meter (fig. 22). These niches seemingly correspond to the individual work place of one miner, possibly the result of a single day’s work.

Fig. 22 – Duzdağı, M9, north section, “niches” (MBA, T. Gonon).

A few words about Late Prehistoric salt extraction in Salt Level 3 (SL3)

38Work in Salt Level 3 (SL3) was carried out in 2015-2016 in Window 6, where the 2008-2011 survey had also registered a large concentration of Kura-Araxes pottery, together with a few Late Chalcolithic potsherds. But as in M1 during the first two years of excavation, only Late Iron Age pottery and a few hearths could be brought to light in that area, where two trenches were opened (M15 and M16). The thickness of the colluvium layer, which in places seems to have accumulated over 50 m, is clearly a serious obstacle to the study of early salt mining in many places of the dome, in particular in W6.4 At any rate, two seasons of excavation in that area were not sufficient for us to bring to light the Kura-Araxes salt-extraction mining works corresponding to the large concentration of potsherds collected in 2011. The geological study of SL3 shows that these salt deposits compare in hardness with those of SL1, but the structure of the SL3 salt deposits is more homogenous: they are also certainly suitable for extracting blocks.

Conclusions

39The excavations carried out at Duzdağı between 2011 and 2016 have yielded an unprecedented wealth of data that suggest widely different salt-exploitation practices over the 4500 BC‑30 BC time span. This variety in practices certainly reflects changing technologies that are themselves linked to specific economic systems: from the probable opportunistic hand gathering of salt during the Late Chalcolithic period to the propping of overhang mining facilities during the Late Iron Age, the goals and thus the work investment of Nakhchivani communities at Duzdağı through time are clearly not the same. The three geological salt layers from which the salt dome is made have different plastic and mineral specificities, a fact that must also be taken into account when reflecting on the evolution of salt-extraction methods, aims and location.

40Unsurprisingly, it seems that the most accessible salt layer (SL2), which at the same time is the whitest and the softest, was first exploited during the Late Chalcolithic and the Early Bronze Age. Even if related extraction facilities have not been found in W6 yet, Salt Layer 3 was also certainly exploited during the Kura-Araxes period, as suggested by the massive presence of potsherds and stone tools. But this was seemingly not the case of Salt Layer 1 in W4: judging by the available evidence, these dark and hard salt deposits only started to be mined during the Late Iron Age. This kind of salt is mainly used nowadays in the pastoral economy: blocks of grey salt are distributed to flocks of caprines and cattle grazing in summer in the uplands, for instance, but some of them are also sold in the valley village shops for the animals living or grazing nearby (fig. 23).

Fig. 23 – Salt blocks for sale at the village shop of Kültepe, see background, upper left side (MBA, C. Marro).

41No real clues have so far been found that would help to reconstruct the use of salt during the Late Chalcolithic or the Early Bronze Age, whether it be from SL2 or SL3. Two stone tools were collected from Uçan Ağıl in the Sirab region, a mobile camp that yielded a Kura-Araxes occupation layer (Gailhard et al., this volume). One of these tools resembles the pointed Kura-Araxes hammers from Duzdağı, but, as a stray find, its stratigraphic position is unclear. If the presence of Duzdağı stone tools was confirmed in the Kura-Araxes occupation layer of Uçan Ağıl, this would testify to the use of salt for herding purposes in this period, since Uçan Ağıl is clearly a pastoral campsite. This would also support the hypothesis according to which individual Kura-Araxes groups mined salt for themselves to meet their own needs. At any rate, it would show that mining at Duzdağı was carried out by part-time miners, who carried their own mining tools around, and acted as shepherds, and possibly metallurgists, the rest of the time.

42The relative sophistication of salt-slab extraction during the Late Bronze/Early Iron Age in SL2 suggests on the other hand that salt exploitation was the work of professional miners in the frame of an entirely different economic system: this system was connected to commercial purposes, with a whole range of possible final uses, from herding to food preservation or leather making. Work was probably divided between miners and traders and possibly other groups of middlemen; but no archives, written or otherwise, are attested that may document the existence of such a system at Duzdağı.

43It is clear that not only the methods and the scale, but also the goals of salt extraction have changed through time at Duzdağı. However, the elusive nature of salt prevents further analysis of its use and distribution scope, since salt remains are unlikely to be found on ancient sites (settlements or graves), unlike artefacts or other minerals. An interesting case, however, must here be mentioned, with the discovery at Ovçular Tepesi of a jar that has probably been used in relationship with salt: the surface of this jar (Jar 19045) has bubbled and in places flaked off, as if it had been covered with salt (fig. 24). Interestingly enough, patches of swollen and flaked clay are only attested over the upper half of this jar, as if the lower part, which was buried, had had no contact with the salt. This jar is one of the rare clues that suggest the use of salt on a Late Chalcolithic settlement. Other cases of ceramic containers with a corroded surface, usually visible at the bottom of the vessels, are also attested at Ovçular, but Jar 19045 is the only example where corrosion has actually evolved into some kind of a swollen mineral crust.

Fig. 24 – Storage jar from Ovçular Tepesi (Locus 19047), with a mineralized clay crust, Late Chalcolithic (MBA, C. Marro).

44Otherwise, clues testifying to the use of salt in Late Prehistory are conspicuous by their absence: it seems so far nearly impossible to trace back the circulation routes of the salt extracted from Duzdağı before the advent of written archives.