Salt extraction at Provadia-Solnitsata

Provadia-Solnitsata is a Late Neolithic and Middle/Late Chalcolithic tell and salt-production complex located in north-eastern Bulgaria (Figure 1). The site also features an extensive settlement, a ‘pit sanctuary’ (consisting of numerous pits dug for ritual purposes; Nikolov Reference Nikolov2011) and two cemeteries (Nikolov Reference Nikolov, Mărgărit and Boroneant2017a, Reference Nikolov and Nikolov2017b). Settlement began here in the early Late Neolithic I (c. 5500 BC), while salt production as a major industrial activity dates to the Late Neolithic II (c. 5200–4900 BC), continuing into the Middle/Late Chalcolithic (c. 4700–4350 BC). There is no evidence for salt working during the Early Chalcolithic (c. 4900–4700 BC). Salt production at Provadia-Solnitsata is considered a key factor in the development of the complex, which exploited the natural rock-salt deposits and brine springs in the nearby Mirovo area.

The process of extracting salt at Provadia-Solnitsata uses one of the most widely documented prehistoric methods, in which ceramic vessels filled with brine were heated over a fire (open or within a kiln) in order to evaporate the water, producing dry, hardened salt ‘cakes’ (Weller Reference Weller, Nikolov and Bachvarov2012; Harding Reference Harding2013; Tencariu et al. Reference Tencariu, Alexianu, Corigă and Vasilache2015). This type of pottery can be categorised as a ‘single-use’ ware because the vessels must be broken to extract the salt cake. This explains the high fragmentation of the vessels recovered.

Ceramics for use in salt making were in active circulation at Provadia-Solnitsata during the Late Neolithic I and the Middle/Late Chalcolithic. The earlier Late Neolithic vessels consisted of mould-cast, thin-walled bowls used in dome kilns (Nikolov Reference Nikolov, Nikolov and Bachvarov2012). These biconical bowls could hold between six and 36 litres of brine. It has been estimated that the total brine capacity of each firing of one of the dome kilns was 90 litres (Nikolov Reference Nikolov, Nikolov and Bachvarov2012). The Middle/Late Chalcolithic revival of salt production at Provadia-Solnitsata featured several changes in the organisation of production, including the introduction of significantly larger pottery vessels in the form of deep, thick-walled tubs (Figure 2a–b).

Figure 2. a) Single- (a1–2) and double-lugged (a3–5) Middle/Late Chalcolithic salt-making tubs (scales in cm; illustration by M. Lyuncheva, after Nikolov Reference Nikolov, Nikolov and Bachvarov2012: 26).

The Middle/Late Chalcolithic process still followed the brine-evaporation method practised in the Late Neolithic I, albeit with improved firing techniques that made it possible to use ceramic vessels with a higher capacity. The diameters of the tub rims range between 0.6 and 0.7m, the bases between 0.3 and 0.35m, and the heights between 0.7 and 0.8m; the average volume of each vessel is estimated to be approximately 60–80l of brine, the equivalent of 114kg of hardened salt (Weller Reference Weller, Nikolov and Bachvarov2012). Often, these vessels had double pairs of vertical conical lugs, intended, given their significant weight, to facilitate handling and transport. The exterior surfaces of the vessels were intentionally rusticated, while the interiors were smoothed and often covered with a kaolin film—perhaps intended to aid impermeability and to prevent breakage during heating to evaporate the brine. The heating process took place in pits, with the vessels placed with their rims in contact with each other and the gaps between the vessels probably filled with firewood (Nikolov Reference Nikolov, Nikolov and Bachvarov2012; Figure 3). The lateral heating facilitated evaporation, aiding the salt-crystallisation process. These vessels also had an additional novel feature: mat-impressed bases—a characteristic almost entirely confined at this site to salt-making vessels. These mat impressions are an intriguing feature, since they are absent from other pottery categories at Provadia-Solnitsata and have so far not been found at other Neolithic or Chalcolithic sites in south-eastern Europe (although they have been recorded at Early Bronze Age sites in the Carpathian zone; Carvruk & Harding Reference Carvruk, Harding, Nikolov and Bachvarov2012).

Figure 3. Pit filled with salt-evaporation pottery (illustration by J. Hristov, after Nikolov Reference Nikolov, Nikolov and Bachvarov2012: 27).

Vessel shape, size and features

The mat-impressed vessels from Provadia-Solnitsata are classified as mass-produced, medium- to large-sized, conical or oval brine evaporation vessels with two handles (Weller Reference Weller, Nikolov and Bachvarov2012; Nikolov Reference Nikolov, Nikolov and Bachvarov2012, Reference Nikolov, Mărgărit and Boroneant2017a). The analysed bases with adjoining walls show similar ranges of base thickness (mean = 14.98mm; standard deviation (hereafter SD) = 7.93; range = 9–33mm) and wall thickness (mean = 14.82mm; SD = 3.99; range = 9–32mm; Figure 4; see the online supplementary material (OSM)). These values support the hypothesis that the containers had a standardised shape and size, and specific features (i.e. mat impressions)—all indicative of mass-production.

Figure 4. Base/wall thicknesses of 50 fragments from Provadia-Solnitsata, where both bases and walls were present (graph by M. Andonova).

The high concentration of fragmented pottery vessels associated with brine evaporation is a key characteristic of salt-production sites (Weller Reference Weller, Nikolov and Bachvarov2012). At Provadia-Solnitsata, these single-use vessels are of a plain, undecorated form and produced from local clays. They contain a high concentration of organic/inorganic inclusions in the temper, a roughened exterior and a smoothed interior, suggesting an intentionally rapid production process involving the assembly of separate base and wall components. A coiling technique was probably used to produce the vessel bases, with clay coils flattened on a mat to form a disc (Weller Reference Weller, Nikolov and Bachvarov2012). This coiling technique may have served to increase the impermeability of the bases. Many of the ceramic tubs at Provadia-Solnitsata also had a clay (kaolin) coating on their interior surfaces. Weller (Reference Weller, Nikolov and Bachvarov2012) offers two hypotheses concerning this coating: either it was added at the end of the salt-crystallisation process, when plastering the internal surface could have caused a thermal shock that facilitated the separation of the salt cake from its container, or the clay was applied before heating in order to thermo-regulate the vessel's surface and prevent thermal shock during the crystallisation process.

Weaving techniques

Most of the ceramic fragments from Provadia-Solnitsata bear the impressions of woven mats that are either diagonally twilled (64 per cent) or, more rarely (12 per cent), made by simple plaiting (Adovasio Reference Adovasio1977; Wendrich Reference Wendrich1990; Figure 5; see also the OSM). For unequivocal identification of either technique, however, the presence of a mat-edge (selvedge) is required, but no such evidence was identified among the 100 mat-impressed sherds analysed.

Figure 5. Weaving techniques at Provadia-Solnitsata (graph by M. Andonova).

Diagonal twilling consists of multiple strands plaited into separate plaits, or one large plait with numerous strands (Figure 6a–b). Simple plaiting also contains multiple strands, usually forming one plait (Figure 6c–d). Twilling elements are always an odd number, while simple plaiting could comprise an even number of strands. In both techniques, a large surface (fabric) could be created by continuous plating with multiple strands, or by stitching together several plaits (Figure 7). A compact weave could be achieved by the application of force by the weaver (e.g. pulling with fingers or sitting on top of the mat), or by a loom-like device fitted with weights (Adovasio Reference Adovasio1977; Wendrich Reference Wendrich1990).

Figure 6. Diagonal twill in (a) 1:1:1 and (b) 2:2:2 ratios; simple plaiting in (c) 1:1 and (d) 2:2 ratios (illustration by V. Galabova).

Figure 7. Stitching of two plaits with rear (‘hidden’) stitch (illustration by M. Andonova, after Wendrich Reference Wendrich1990: 61).

The number of strands impressed on the bases from Provadia-Solnitsata was limited by the size of the ceramic surfaces that survived and especially by the state of preservation. Nevertheless, the number of recorded strands averages 14 (mean = 14.2; SD = 6.91; range = 2–39mm; see Table S1 in the OSM). Amongst both the fragments exhibiting twilling and plaiting, the mean angle between the strands is approximately 90° (mean = 90.27°; SD = 0.87; range = 88–92°; Table S1). Hence, application of medium force while weaving should be assumed, as the angle would otherwise be acute. The generally right-angled strands also suggest a somewhat flexible mat.

The dominant weaving pattern of the diagonal twill has a 1:1 ratio (1 strand under, 1 over, 1 shift; 58 per cent; Table S1), with a few examples of a 2:2 ratio (2 strands under, 2 over, 1 shift; 7 per cent). The examples of simple plaiting also show two ratio variations: 1:1 (8 per cent; Figures 7c & 8) and 2:2 with doubled strands (5 per cent; Figure 6d). In some cases, the patterns display ‘errors’ or ‘misses’, such as the doubling of strands in a diagonally twilled weave by passing twice under or over the same element (Figure 9f), along with irregularities due to stitching to create a larger fabric (Figures 8 & 9f).

Figure 8. Mat impressions from Provadia-Solnitsata: a–d) diagonal twilling; e–h) simple plaiting (figure by M. Andonova).

Figure 9. Types of leaf/stem venation recorded at Provadia-Solnitsata: a & d) keel absent; (b & e) keel present; (c & f) presence of conspicuous lateral veins (figure by M. Andonova).

The mat-weaving raw material

The clearest imprints on the Provadia-Solnitsata mat-impressed bases provide information on the types of plants selected for weaving. Stereoscopic analysis excludes the presence of dicotyledonous plant parts, such as wood splints, bark, bast or branches as possible materials for mat weaving because of the presence of prominent parallel venation of their strands and the type of weaving techniques, which require long and flexible wefts. Instead, the stereoscopic analysis shows flat-leaf lamina/stem fragments with parallel venation and pronounced main and secondary veins, which were attributed to monocotyledonous species (Metcalfe Reference Metcalfe1960: xvii). We assume that there are no circular or triangular cross-sections in the analysed material, given that no evidence for folding or compressing of the impressed plant parts was identified. The determination of a cross-section of a monocotyledonous leaf lamina through its tangential plane should be made with caution. Longitudinally, three types of leaf laminae/stem fragments are visible in the mat impressions, according to their venation: 1) with no conspicuous main vein (also called the midrib or keel (Figure 9a & d)); 2) with a clear main vein (keel) (Figure 9b & e); and 3) with two conspicuous lateral veins (Figure 9c & f). The average lowest strand width is 5mm (mean = 5.14mm; SD=1.8; range = 3–14mm), while the highest is 7mm (mean = 7.55mm; SD = 2.36; range = 4–16mm; Table S1).

These descriptions could apply to several monocotyledonous species recorded ethnobotanically as plants used in basketry and weaving in the Balkans (Bichard Reference Bichard2008; Nedelcheva et al. Reference Nedelcheva, Dogan, Obratov-Petkovic and Padure2011; Table 1). Those considered native to Bulgaria (Euro+Med 2006–2011) and present in the study area today (Jordanov et al. Reference Jordanov, Kitanov and Valev1964; Stoyanov et al. Reference Stoyanov, Stefanov and Kitanov1966) include sedges (Cyperaceae), rushes (Juncaceae), grasses (Poaceae) and cattails (Typhaceae). Based on leaf-blade width and venation features, some species can be excluded from the Provadia-Solnitsata sample, as their specific features are unrecorded in the analysed impressions. These include wide-leaf species, such as cattails, species with cylindrical stems and/or leaves, such as rushes, and species with a triangular leaf cross-section, such as sedges. Hence, flat-leaved grasses, including reeds and wild and cultivated grasses (Metcalfe Reference Metcalfe1960, Reference Metcalfe1971; Cutler Reference Cutler1969) are more likely to have been used. To reduce the range of possible choice of species, a brief overview of the palynological and archaeobotanical data related to Provadia-Solnitsata is necessary.

Table 1. Monocotyledonous basketry species recorded ethnobotanically (A = Bichard (Reference Bichard2008); B = Nedelcheva et al. (Reference Nedelcheva, Dogan, Obratov-Petkovic and Padure2011)) and archaeobotanically (C = Marinova (Reference Marinova and Nikolov2008); Hristova et al. (Reference Hristova2019)).

Available pollen data from the sites closest to Provadia-Solnitsata—Varna Lake, Lake Durankulak, Shabla Ezerets and Bolata (Marinova & Atanasova Reference Marinova and Atanasova2006; Filipova-Marinova et al. Reference Filipova-Marinova2013; Tonkov et al. Reference Tonkov, Marinova, Filipova-Marinova and Bozilova2014)—suggest that the local palaeo-habitats during the Late Neolithic, Middle Chalcolithic and Early Bronze Age consisted of a relatively open forest-steppe (Figure 1; Table 1). The surrounding forests were dominated by oak (Quercus sp.) and hornbeam (Carpinus sp.), along with abundant light-demanding species, forest understorey and grassland vegetation. The humid zones in close proximity to Provadia-Solnitsata were covered with riparian forests of willow (Salix sp.), alder (Alnus sp.) and European ash (Fraxinus excelsior), along with communities dominated by reeds, cattails, sedges and rushes.

Increased local agricultural activity is documented during the Late Neolithic and Chalcolithic periods, as attested by frequent Hordeum-type pollen, scarce Triticum-type pollen, and a continuous presence of Cerealia-type pollen (Filipova et al. Reference Filipova-Marinova2013). In addition, the dominant cereal crops recorded archaeobotanically at Provadia-Solnitsata (Marinova Reference Marinova and Nikolov2008; Hristova et al. Reference Hristova2019; E. Marinova pers. comm.) are einkorn (Triticum monococcum), emmer (T. dicoccum), cf. new type wheat (cf. T. thimopheevi), free-threshing wheat (T. aestivum/durum) and barley (Hordeum vulgare) (Table 1). Alongside the species recorded in the Late Neolithic layers, wild grass species (needlegrass (Stipa sp.) and Timothy grass (Phleum sp.)) and fragmented reed stems (Phragmites sp.) from Middle Chalcolithic architectural features are also relevant to mat production.

Of the monocotyledonous species potentially used to make mats at Provadia-Solnitsata (Table 1), needlegrass can be excluded, as its narrow leaf-blade width does not match any mat impressions from Provadia-Solnitsata (Stoyanov et al. Reference Stoyanov, Stefanov and Kitanov1966). Hence, amongst the archaeobotanically recorded species, the grain cultivars, wild Timothy grass and reeds remain valid choices. Of the ethnobotanically recorded species, reed canary grass (Phalaris arundinacea) can be excluded, because the length of its leaf blade is too short, reaching a maximum length of 0.35m (Stoyanov et al. Reference Stoyanov, Stefanov and Kitanov1966). Ecologically, the species listed above are found in two habitat types: reeds on marshy lakeshores, and grain cultivars and Timothy grass in open meadows and cultivated fields. Mature Timothy grass has the longest leaf blade (1.2m), while barley reaches 1m, einkorn 0.9m, and emmer 0.8m (Stoyanov et al. Reference Stoyanov, Stefanov and Kitanov1966; Zorovski et al. Reference Zorovski, Popov and Georgieva2018).

The leaf length of a plant is important because the plait must be continuous and homogeneous, and the weaver must be efficient, for example, when making multiple adjustments (e.g. stitching, extending or repairing). Long leaves enable more elaborate weaving, featuring, for example, rear (hidden) stitching and invisible extensions. Nevertheless, the archaeobotanically identified species that would be suitable for mat weaving in Chalcolithic Provadia-Solnitsata all have similar characteristics. Hence, at the current stage of research, the discussion cannot be extended to species level.