The Walandawe study area, Southeast Sulawesi

Gua Talimbue and Gua Sambagowala are located in the Walandawe area of Southeast Sulawesi, close to the Wiwirano River, which drains rolling terrain interspersed with rugged outcrops of ultramafic ophiolites (oceanic crust that has been uplifted above sea level) and limestone karsts. Both sites produced flaked stone, bone, shell and other finds, including pot sherds in their upper spits (Suryatman et al. Reference Suryatman, Bulbeck, Marwick, Oktaviana and Wibowo2016; Fakhri Reference Fakhri, O'Connor, Bulbeck and Meyer2018).

Gua Talimbue is a south-west-facing cliff-foot cave that was probably formed through ground-water processes in both the vadose (ground surface to the water table) and phreatic (saturation zone beneath the water table) zones. It includes winding chambers that extend into the limestone hill and that are accessible via at least three entrances. The main entrance is around 8m high (at the drip line), 23m wide and 13m deep. The slope of the floor is approximately 2–5°. The cave-floor deposit is 4m thick, the upper 3m predominantly resulting from wet-season fluviatile sedimentation, and includes pebbles, sand and clay. The lower metre of deposits consists of yellow-brown clay, which may indicate the presence of standing or slow-moving water at the site (O'Connor et al. Reference O'Connor2014).

Gua Sambagowala is a dolomite rockshelter located on the edge of the Wiwirano River flats. The rockshelter extends around 19m east–west, the drip line extending between 3–7m from the back of the shelter. The slope of the floor is approximately 5°. Silty material dominates the 2.40m of deposits excavated within a 1m² test pit (O'Connor et al. Reference O'Connor2014). Excavation of a 1m² test pit at Gua Mo'o hono—another cliff-face rockshelter at Walandawe—produced a larger archaeological assemblage than Gua Sambagowala, including hearth features in the pottery layers, but no clay pieces (O'Connor et al. Reference O'Connor, Bulbeck, Piper, Aziz, Marwick, Campos, Fenner, Aplin, Fakhri, Suryatman, Maloney, Hakim, Wood, O'Connor, Bulbeck and Meyer2018).

The baked-clay fragments from Gua Talimbue and Gua Sambagowala can be differentiated from the pottery excavated at the Walandawe sites (Bulbeck et al. Reference Bulbeck, Aziz, O'Connor, Calo, Fenner, Marwick, Feathers, Wood and Prastiningtyas2016; O'Connor et al. Reference O'Connor, Bulbeck, Piper, Aziz, Marwick, Campos, Fenner, Aplin, Fakhri, Suryatman, Maloney, Hakim, Wood, O'Connor, Bulbeck and Meyer2018) in their lack of identifiable rim, body or base sherds. The baked-clay surface colours generally correspond to red to reddish-brown Munsell soil colours (including light and dark shades), which contrast with the typically brown and grey hues of the Walandawe potsherds. Furthermore, although the baked clay and Walandawe pottery are similar in the presence of white grain inclusions, these grains are typically larger and more granular in the baked clay, whereas the potsherds are usually white-speckled and sometimes even free of macroscopically visible inclusions.

The pottery sherds typically also have two smoothed surfaces (interior and exterior), either or both of which may have surface patterns (decoration) and adhering residues. In contrast, the baked-clay fragments from Gua Talimbue and Gua Sambagowala show smoothing on, at most, one side, with patterns and adhering residues found only on these prepared surfaces. The Gua Talimbue and Gua Sambagowala fragments appear to have derived from clay structures, which were heated to high temperatures, effectively firing the structures’ interior clay surfaces. This process is similar to that reported for Iron Age European bloomery furnaces, where the inner 40–60mm of the clay walls is fired, but the wall core and exterior surfaces remain as unfired clay and hence are usually poorly preserved (Blair Reference Blair and Mathieu2002). This phenomenon is also similar to that recorded on intact Aurignacian hearths at Klisoura Cave 1, where there is also a contrast between a smooth interior surface and more granular exterior surface (Karkanas et al. Reference Karkanas, Koumouzelis, Kozlowski, Sitlivy, Swobczyk, Berna and Weiner2004: figs 4a & 5a), with the interior surfaces fired during the heating of the hearths’ contents to temperatures between 400 and 600°C.

Gua Talimbue: excavation, sequence and baked-clay distribution

The Gua Talimbue excavation focused on two areas of the cave. The adjoining A, C and D squares to the west were excavated to a depth of approximately 0.55m, allowing the recovery of an extended human skeleton buried with glass beads and an iron ornament (see Bulbeck et al. Reference Bulbeck, Aziz, O'Connor, Calo, Fenner, Marwick, Feathers, Wood and Prastiningtyas2016). The small quantity of baked clay recovered from these squares was not recorded in detail and so is not reported here. Near the centre of the cave, the adjoining squares B and E were excavated in 50mm spits to a depth of just over 4m (Figure 2). Larger finds were recorded in situ, while smaller finds were recovered by wet sieving through 5mm and 1.5mm mesh sieves. We separated the baked clay into plain and patterned fragments, recorded the weight of the fragments to the nearest 0.5g (Table S1 in the online supplementary material (OSM)), and assigned each fragment to the relevant 50mm spit recorded during the excavation (Table S2).

Figure 2. Stratigraphic section of Gua Talimbue squares B and E, with calibrated dates and baked-clay distribution (sources: Table S1; Suryatman et al. (Reference Suryatman, Bulbeck, Marwick, Oktaviana and Wibowo2016) incorporated under a CC BY-NC-ND 4.0 licence) (figure by D. Bulbeck).

Eighteen AMS determinations on unidentified plant charcoal samples collected in situ during the excavation and from the sections of square B are presented in Table 1. The calibrated dates are presented in Figure 2, with the exception of one out-of-sequence date of 3726±33 BP (D-AMS 004041). A Holocene midden deposit containing discrete features of red clay, charcoal concentrations and lenses of white ash extended down to a layer of imbricated freshwater shellfish at approximately 3.20m beneath the surface (around spit 65). The underlying deposit beneath the shellfish layers consists of yellow-brown clay with sparse evidence for habitation, and is featureless in section, apart from rock inclusions. Although stratigraphically inverted, the two available radiocarbon dates, c. 19 000 cal BP (spit 71, 3.67m) and 13 500 cal BP (spit 73, 3.74m), confirm a Pleistocene date. Excavation was discontinued at a depth of 4.20m, without reaching bedrock, due to the difficulty of working within the 2 × 1m pit.

Table 1. AMS determinations from Gua Talimbue square B, calibrated with IntCal13 (Bronk Ramsey Reference Bronk Ramsey2013)

The charcoal samples immediately overlying the imbricated shellfish layer are spatially associated with red-clay features, while the charcoal samples from higher in the stratigraphy are not (Figure 2). Potential bioturbation of the upper deposit is indicated by the presence of insect casts and roots recorded during the excavation from spits 1–10 (0.50m depth) and sparse rootlets from spits 11–22 (1.16m depth). Based on the stratigraphic position of the dated samples, the Holocene midden deposit appears to have mostly accumulated in three pulses, rather than evenly over time. These three pulses of deposition, c. 4100–4500 cal BP (spits 10–21, 0.50–1m depth), 6500–7500 cal BP (spits 25–45, 1.30–2.30m depth) and 9500–9900 cal BP (spits 56–64, 2.90–3.20m), are associated with 95.4 per cent calibrated confidence intervals that nestle together (Table 1), and a steep gradient of accumulated deposit over a short period of time (Figure 3).

Figure 3. Median calibrated Holocene dates from Gua Talimbue square B (95 per cent confidence—see the OSM), plotted against the stratigraphic depth of the dated samples (source: OSM computer code output) (redrawn by D. Bulbeck).

Two concentrations of red-clay features appear in the stratigraphic section (Figure 2). The upper concentration, positioned between approximately 0.60m and 1m below the surface (4100–4500 cal BP), includes a large discrete feature of red clay (not collected) on the south-west wall of square B at around 0.70m depth, as well as other red-clay features associated with lenses of white ash; nodules of red clay were observed during excavation in spits 6–19 (0.30–0.95m). The deeper concentration, between 2.70–3.20m below the surface (9100–9900 cal BP), comprises lenses of red clay, along with red-clay clasts. Both the upper and lower concentrations of red-clay features correspond to peaks in the quantities of baked clay recovered during excavation, even though the uppermost peak of recovered baked clay (spits 5–9, 0.25–0.45m) stratigraphically overlies the upper concentration of red-clay features in the stratigraphic section (Figure 2). Between spits 30 and 54 (1.60–2.70m), there were no patches of red-clay deposit in the sections, very little baked clay was recovered and red-clay mottling was observed only rarely. In summary, it is possible to identify a general association between the presence of red-clay features in the section and the recovery of baked clay from the deposit.

The excavated cultural remains are presented in Table S3 and Figures S1–2. Freshwater shellfish was usually the dominant find by weight in the midden deposit in spits 1–51 (0–2.60m), dating back to c. 7500 cal BP. Pottery was recovered down to spit 3 in square B and spit 6 in square E. In both cases, the pottery overlies the stratigraphically uppermost concentration of baked clay. The interpolated date for the initial appearance of pottery lies between 2000 cal BP (with reference to square B) and 3000 cal BP (with reference to square E). Between spits 52 and 64 (2.70–3.20m)—dating back to around 9900 cal BP—lithics are generally the dominant find by weight, accompanied by moderate quantities of vertebrate and freshwater shellfish remains. The underlying Pleistocene deposit produced few finds, apart from modest quantities of lithics.

The baked clay is assigned to plain and patterned categories. Square B yielded 634.5g of plain baked clay between spits 2 and 80 (0.07–3.98m) and 1188g of patterned baked clay between spits 5 and 64 (0.23–3.19m). Square E yielded 1542.5g of plain clay between spits 1 and 66 (0.07–3.24m), and 463g of patterned baked clay between spits 1 and 63 (0.07–3.10m) (Table S2). Both the plain and patterned categories were discontinuous in their vertical distribution (Figure 2), and tailed off in the Late Pleistocene layers, which yielded only plain, not patterned, baked clay.

Gua Talimbue: analysis of the baked-clay patterns

A classification of the 299 patterns found on the interior surfaces of the 249 patterned fragments from Gua Talimbue is presented in Table S4. The most frequent class comprises shallow linear furrows labelled ‘striations’ (n = 156, 52.2 per cent), which includes white-infilled, irregular parallel (Figure 4), regular parallel and cross-hatched (Figure 5) striations. Striations sometimes occur in parallel sets (n = 22), including cross-hatching, but are also found in more loosely configured sets (Figure 4). Deeper linear depressions are labelled ‘corrugations’ (n = 73, 24.4 per cent) when in parallel formation (Figure 6), and ‘incisions’ (n = 13, 4.3 per cent) when appearing in isolation (Figure 4). Two fragments with surface ridges complete the total number of 244 linear patterns. The remaining 55 patterns include 27 irregular impressions (Figure 6) and 28 curvilinear patterns, including 11 with semi-circles, nine with dashes (Figure 7), six with lunate patterns (Figure 8) and two with quadrangular stamps. The curvilinear patterns in particular are reminiscent of motifs found on decorated pottery, even though they do not resemble the particular motifs on the decorated Walandawe pottery (Bulbeck et al. Reference Bulbeck, Aziz, O'Connor, Calo, Fenner, Marwick, Feathers, Wood and Prastiningtyas2016; O'Connor et al. Reference O'Connor, Bulbeck, Piper, Aziz, Marwick, Campos, Fenner, Aplin, Fakhri, Suryatman, Maloney, Hakim, Wood, O'Connor, Bulbeck and Meyer2018).

Figure 4. Photograph (above) and illustrations (below) of baked-clay fragments from Gua Talimbue square B spit 15, showing incisions (far left and far right), sets of parallel white-infilled striations (second and fourth from left) and irregularly parallel striations (centre) (photograph by Fakhri; illustration by D. Bulbeck).

Figure 5. Photograph (above) and illustrations (below) of baked-clay fragments from Gua Talimbue square B spit 19, showing regularly parallel striations (far left), irregular impression (centre) and cross-hatched striations (far right) (photograph by Fakhri; illustration by D. Bulbeck).

Figure 6. Photograph (above) and illustrations (below) of baked-clay fragments with corrugations from Gua Talimbue square B spit 5, showing regularly parallel striations (first, second and fourth from left) and irregular parallel striations (third from left) (photograph by Fakhri; illustration by D. Bulbeck).

Figure 7. Photograph (above) and illustrations (below) of baked-clay fragments from Gua Talimbue square B spit 58, showing dashes arranged perpendicularly (first and third from left), fields of semi-circular impressions (second from left) and semi-circular incisions (far right) (photograph by Fakhri; illustration by D. Bulbeck).

Figure 8. Photograph (above) and illustrations (below) of baked-clay fragments from Gua Talimbue square E spit 63 with lunates—lengthwise repetition producing rounded zigzags (left) and repetition perpendicular to length (right) (photograph by Fakhri; illustration by D. Bulbeck).

Both the quantity and character of the patterns vary with stratigraphic depth (Table 2 & Figure S3). Together, the patterns encountered between 0.03–0.48m and between 0.53–1.16m—which cover the Mid/Late Holocene—account for 236 (79 per cent) of the total patterns, with approximately equal numbers of patterns in the spits that correspond to the 4100–4500 cal BP pulse of deposition (0.53–1.16m) and in the overlying spits. Linear patterns dominate (91 per cent), followed by irregular impressions (8.5 per cent). The patterns encountered between 1.35 and 2.26m, which includes the 6500–7500 pulse of deposition, account for a mere 5 per cent of the total, and comprise mostly linear but also some curvilinear patterns. The most notable contrast with the Mid/Late Holocene patterns is provided by the Early Holocene patterns encountered between 2.69–3.20m, which covers the 9500–9900 cal BP pulse of deposition. Although comprising only 16 per cent of the total patterns, they account for 89 per cent of the curvilinear patterns, including semi-circles, dashes, quadrangles and lunates.

Table 2. Vertical distribution of Gua Talimbue patterned baked-clay fragments (summarised from Table S6 in the OSM). Linear patterns include striations, corrugations, incisions and ridges. Irregular patterns are impressions not elsewhere included. Curvilinear patterns include semi-circular, dash, lunate and quadrangular patterns. Percentages may not sum exactly to 100.0% due to rounding.

Based on the similarity of patterns, surface colour and relief, and excavated depth, we grouped the patterned fragments into 46 sets that potentially correspond to distinct baked-clay structures (Tables S5–6). This analysis shows that convex surfaces were frequent on the patterned fragments encountered between spits 26 and 64 (1.25–3.19m depth), but not on those from the spits above spit 26, which mainly have a concave or unevenly smoothed surface. To explain this difference, we can use the basin-shaped, baked-clay structures from Klisoura Cave 1, Greece, as a comparative model, which are significant both for their completeness and their construction by pre-pottery foragers. These structures exhibit a thickened ‘rim’ with a coarse, convex surface along the top of the basin-shaped structure, and a concave surface on the inside of the structure beneath the rim (Karkanas et al. Reference Karkanas, Koumouzelis, Kozlowski, Sitlivy, Swobczyk, Berna and Weiner2004). The combined information on dating, patterns and surface shape suggests the following chronology and interpretation for the decorated baked clay at Gua Talimbue:

1. 1. C. 9900–8800 cal BP: frequently, convex surfaces derived from the ‘rim’ of clay structures, with predominantly curvilinear patterns, consistent with the production of motifs designed to enhance the visual appearance of the structures.

2. 2. C. 7000–6500 cal BP: based on a small sample of patterned baked clay, mainly linear patterns on a variety of convex, concave and undulating surfaces.

3. 3. C. 4500–2500 cal BP: predominantly linear patterns on mostly concave and unevenly smoothed surfaces. The range of linear patterns may reflect the application of surface-finish techniques, impressions left by plant materials lining the interior surfaces, and designs to enhance visual appearance prior to use and any re-use.

Two different scenarios may explain the concentrations of baked-clay fragments associated with phases of rapid sediment accumulation between 4100–4500 cal BP and 9500–9900 cal BP (Figure 9). In scenario A, which best fits the Mid/Late Holocene evidence, the cave occupants produced an inset hearth by scooping surface sediment into a ring around a shallow pit, which was then lined with clay. Patterns were applied to the internal surface of the structure, which was hardened during heating. In scenario B, which best fits the Early Holocene evidence, the structure may have comprised a brazier-like hearth constructed partly above the ground surface of the cave, as also described for the Aurignacian clay hearths at Klisoura Cave 1 (Karkanas et al. Reference Karkanas, Koumouzelis, Kozlowski, Sitlivy, Swobczyk, Berna and Weiner2004). Rapid sedimentation helped to keep the structure intact, while heating during use hardened its interior.

Figure 9. Two scenarios to explain the association between baked clay and intervals of rapid sediment accumulation at Gua Talimbue (illustration by D. Bulbeck).

Hearth construction and use in the vicinity of the excavated squares may have been discontinuous between the Early and Late Holocene. This would explain the absence of Mid-Holocene baked-clay fragments, apart from the small quantity associated with the pulse of sedimentation between spits 26 and 45 (1.38–2.36m, 6500–7500 cal BP). As for the upper stratigraphy, spits 1–9 (0.07–0.48m) produced approximately the same quantity of baked clay as spits 10–22 (0.53–1.17m), which implies that hearth construction continued in the vicinity of the excavated squares after 4000 cal BP. Furthermore, due to a slower rate of deposition, spits 1–9 correspond to a notably longer time interval than spits 10–22, resulting in numerous baked-clay fragments, despite lacking the favourable preservation conditions provided by the rapid accumulation of deposit in spits 10–22.

Gua Sambagowala

A single test pit, pit A, was excavated at Gua Sambagowala (Figure 10). The finds comprised a shell midden associated with potsherds in spits 1–2, and a pre-pottery shell midden in the 36 lower spits (Tables S7–8). Spits 7–38 yielded 293g of baked clay, including two small fragments of baked clay with roughly parallel striations from spit 11 (c. 4400 cal BP) and spit 30 (c. 5500 cal BP). Freshwater shellfish was the dominant cultural material between spits 3 and 38 (0.31–2.40m depth), at a density between 24.5 and 74.2g/kg of excavated sediment.

Figure 10. Stratigraphic section of Gua Sambagowala test pit A with calibrated dates and baked-clay distribution (sources: Table S7; Fakhri (Reference Fakhri, O'Connor, Bulbeck and Meyer2018) incorporated under a CC BY-NC-ND 4.0 licence) (illustration by D. Bulbeck).

Traces of hearths, such as burnt stone and white ash, were recorded in 12 spits between spits 4 and 32 (Table S8). Although these traces were recorded in only two of the 11 spits with baked clay, the latter should also be interpreted as representing the remains of hearths, particularly as burnt animal bone was found embedded in one fragment. While burnt nut remains and wood charcoal were consistently present from spits 4–21 (0.38–1.53m depth), only wood charcoal was present between spits 22 and 38 (1.60–2.40m depth). This disparity, due either to dietary change or preservation conditions, suggests that spits 21 and 22 mark a chronological divide between c. 3500–4500 cal BP (spits 4–21) and c. 5500–6000 cal BP (spits 22–38).

In summary, the Gua Sambagowala excavation produced a small assemblage of baked clay dating to c. 5500–6000 cal BP, and a larger assemblage dating to c. 4000–4500 cal BP. The smaller, earlier assemblage falls within a gap in the chronology of the Gua Talimbue baked clay, whereas the larger, later assemblage corresponds to a peak in the presence of baked clay at Gua Talimbue.