The Pithos House at Kissonerga

Building 3 at Kissonerga-Mosphilia was excavated by the Lemba Archaeological Project, which dug the site from 1979 to 1992 under Edgar Peltenburg of the University of Edinburgh. It is a single round room 9.1 m across its walls, with one door facing south and a plastered platform hearth in the middle: in plan, a house like others of its day. Its floor held at least 37 storage vessels, “more existed originally”, the final report adds, about twenty smaller pouring and serving vessels, and some 280 to 300 other objects (LAP II.1A, pp. 37–43; LAP II.1B).

The excavators judged the stores “far in excess of the needs of a normal domestic unit and quite atypical for Kissonerga or other prehistoric sites of Cyprus” (LAP II.1B). Louise Steel called it storage on “a completely different order of scale” from anything in the periods before (Steel 2004). A Middle Chalcolithic house kept a few jars against its wall. This building kept dozens, and it is the first concentrated household store known on the island, if a household is what it was. On it rests an argument that has run for thirty years: whether one group at Kissonerga gathered wealth and power over its neighbours some four and a half thousand years ago, and whether the neighbours put an end to it.

The fire that destroyed the house is the reason it can be rebuilt. It baked the floor with everything still standing on it: the stands the jars sat in, the loose jars leaning on each other, a nest of bowls inside a barrel, the burnt timbers of the roof where they dropped. We know of no other Late Chalcolithic building on Cyprus published with its stores so completely in place.

Kissonerga-Mosphilia spreads over some 12 hectares of coastal terraces about 7 km north of Paphos, the largest Chalcolithic settlement excavated on the island. Around 3200 to 2900 BC its walled “High Sector” held the largest round houses of the Middle Chalcolithic, among them the Red Building, about 14.5 m across (the round house reconstruction shows a middling house of that world). Then the village was abandoned for about two centuries, as was Lemba nearby, though at Chlorakas-Palloures, 1.5 km away, rubbish kept accumulating from the Middle into the Late Chalcolithic (Düring et al. 2023).

The Late Chalcolithic, roughly 2800 to 2500 BC, looks different in almost everything the excavators can count. Thin, standardised Red-and-Black Stroke-Burnished ware, rich in spouted pourers and small bowls, replaced painted Red-on-White, beside big Coarse Painted storage jars up to about a metre tall. Conical stones, stamp seals, faience beads and the first cast copper tools appear, and the picrolite cruciform figures of the previous age disappear (Peltenburg 2014; 2018). Houses were generally smaller, about 6 m across, with more stone in their walls and many settings for storage jars (Thomas 2005).

Building 3 belongs to the start of this, Kissonerga Period 4a, dated by the excavators to about 2700 to 2500 BC. Its five radiocarbon dates put an end to an older idea that it stood alongside the Red Building. Period 4a was a thin scatter of buildings on the old High Sector, and the Pithos House, one of the two earliest, dominated it (LAP II.1B; Winkelmann 2020). Other big houses stood in its day, one of them 14 m across at Chlorakas-Palloures (Düring et al. 2023); what sets the Pithos House apart is what the fire kept inside it.

The house was terraced into a hillside rising to the north, with a cut about 30 cm deep, 37 cm at its edge. Many of its wall stones are calcarenite, uncommon in Period 4 and probably robbed from the Red Building: the builders quarried the great houses of the Middle Chalcolithic for stone (LAP II.1B).

The wall, numbered 46 by the excavators, survives seven courses and 0.78 m high. It is 0.50 to 0.60 m wide, two rough faces of cobbles 10 to 25 cm long with a core of mud and small stones, the largest facers, up to 45 cm, on the outside. Mud makes up 30 to 40 per cent of it: in Gordon Thomas’s words, “not a dry stone construction”. It leans inwards 3 cm in half a metre. Inside, a clay render about 20 mm thick was smoothed over the stone, “through which the underlying stonework was always evident” (LAP II.1B). The reconstruction’s wall is 0.6 m thick, render included, leaning as published, which gives an internal diameter of 7.9 m at the floor.

How high the stone went is the first judgement a reconstruction has to make. Middle Chalcolithic houses stood on a stone socle with mud walls above, as the round house did. Thomas lists this building among the few where stone walling to full height “can be confirmed on the basis of the preserved height of the wall and the extent of the stone collapse”. The collapse also held “compacted mud from the upper walls” and a “bricky material” carrying “parallel timber impressions of a half-timbered component, perhaps where the roof joined the wall” (LAP II.1B). The reconstruction reads the two together: stone to 1.9 m above the floor, then a band of mud brick packed round the ends of the beams and joists up to a wall head 2.5 m above the floor. No wall height is published. The nearest guide is the small house later built on the ruin, whose fallen stones imply a wall of at least 1.6 m and perhaps 2.4 m, and our height sits inside that range. The 3D model offers a socle with mud above as the alternative.

The door faced south, over a threshold paved right through the wall. The west jamb is lost, so the opening measures up to 1.2 m where it can be measured at all; a stone box found fallen just inside has “the same width as the entrance” at 0.94 to 0.97 m, and the reconstruction takes 0.95 m, with a lintel at 1.55 m that is our figure. Pivot stone 1705 lay in the inner west corner, so the door swung inwards to the left as you entered, like the other pivots found in place at Kissonerga. An older pivot 7 cm below the final floor belongs to an earlier door, and a broken quern stood on edge a metre or so inside, probably a brace for barring it (LAP II.1B).

The floor was orangey clay under a thin lime-based plaster, about 2 cm thick, and the excavators found it “dilapidated” from long use, a poor foundation and the weight of the collapse. Plant silicates under the stones lying on it suggest loose matting. In the middle was hearth 828, a dished platform 1.30 m across with a firebowl 27 cm wide, surfaced with whitish plaster that still carried traces of red pigment and signs of repair (LAP II.1B).

A room 7.9 m across needs help to carry a flat roof, and the excavators thought internal supports “probably necessary”. They found three probable pairs of shallow postholes in the north and east, a single posthole in the north-west only 17 cm deep, a group of flat stones 0.80 m south-west of the hearth and more stones beside the mortar in the south-east, “about the same distance from the hearth”, which “may mean that posts were also propped on slabs, as shown in Fig. 3.9” (LAP II.1B). Under the ash lay burnt timbers running mostly north-east to south-west, some “aligned with postholes, as if fallen from those positions”. Lumps of fired mud from the roof carry the impressions of reeds.

The excavators’ own drawing, Fig. 3.9 of the final report, shows a flat earth roof on a ladder-like main beam of two parallel logs, carried on posts. Our layout is a reading of the six settings. All lie about 0.80 m from the edge of the hearth, and set on a ring 1.45 m from its centre they fall close to the corners of a regular hexagon, whose corners line up on three parallel lines. Here those lines run north-east to south-west, the direction of the fallen timbers. So the reconstruction lays three beam lines on the six supports: in the middle, the paired beam of Fig. 3.9, carried by pairs of posts at its north-eastern and south-western settings, one post under each log; to either side, a single log. Poles laid across the beams every 30 cm carry a mat of reeds, then tempered mud, earth and a skim of mud and havara, the local soft marl. The excavators read paired holes as a post and its replacement; under the main beam we read them as two posts standing at once.

Fig. 3.9 draws the main beam poking out through the wall. We know of no eaves or drip-trenches at Kissonerga, so the reconstruction keeps the beam ends flush with the outer face. The 3D model can also show the roof as Fig. 3.9 drew it, on four posts with props and joists nearly 4 m long, though that ignores four of the six settings. A conical roof of thatch is shown only as rejected: the collapse was heavy compacted mud with reed impressions, the debris of a flat earth roof, and a cone this wide would need a ring of posts the plan does not have. A loft is possible, and nothing demonstrates one.

One object in the collapse has never been explained. Unit 492, a box of stones in a burnt timber frame about 1.35 by 0.95 m, lay over charred timber and broken jars just inside the door and lined up with it; it “arrived in the position we found it after the destructive fire had commenced” (LAP II.1B). It may be part of a blocked doorway that fell in. Thomas once thought of a window, and later called the idea “a very feeble proposition” (Thomas 2005).

The jars were kept in two ways. Along the back wall, opposite the door, ran a band about 1.5 m deep “set aside for permanent storage”. “Eight stands stood against the back wall, 6 in front of them … and one more had been placed between the second row and the hearth”: fifteen found, “a minimal number”. A stand was a shallow hollow in the floor, 0.5 to 1.0 m across and 9 to 31 cm deep, ringed with stones, set into a low mud bank a few centimetres high along the wall, and wedged tight with worn-out grinding tools: querns, rubbers, pestles, pounders and hammerstones. One more setting stood against the west wall and held the tallest jar in the house, KM 3300, 95 cm high, one of the best preserved because the wall shielded it from the worst of the fire (LAP II.1B).

The jars needed the stands. The common storage jar of the period is tall and ovoid, with an inturned “hole” mouth and a dimpled base 5 to 18 cm across that cannot stand on its own. The temporary store stood straight on the floor, “the unstable ones propped up by adjacent vessels”, filling “much of the area from the fixed pithoi to within 2 m of the doorway”: between the back rows and the hearth, between the door pivot and the hearth, and to the right of the entrance. The excavators reckoned 17 permanent and 18 temporary jars. Nearest the hearth stood smaller vessels: bowls, spouted pourers and slender collared jars (LAP II.1B).

None of them was stacked. The preliminary reports had jars piled on jars, and the final report corrects them: “Contrary to assessments in preliminary reports, maximum diameters allow the full complement to rest on the floor: there was no need for stacking” (LAP II.1B). The model holds 36: 18 fixed, counting a coarse barrel on its own stand against the south-west wall, and 18 loose, all on the floor in a single tier. Of the 37 storage vessels, one, known only from its base, is left out of the model.

Most of the big jars are Coarse Painted holemouths, thin, chaff-tempered and porous, with streaky orange-red to red-brown paint; one carries a criss-cross “tartan” pattern. Those with complete profiles stand 62 to 95 cm tall and held between about 58 and 200 litres. With them were thicker-walled jars, wide-mouthed barrels with relief ribs, and smaller jars and flasks. Diane Bolger, who studied the pottery, read the porous holemouths as stores for liquids, cooled by evaporation through their walls, and the barrels as stores for dry goods (LAP II.1B). The lids found were a small chalk disc and plaster stoppers, all too small to close a pithos mouth: the jars stood open or under covers of some perishable material.

The figure of about 4,000 litres is the excavators’. Bolger’s team calculated each jar’s volume from its profile restored on paper, and Peltenburg’s text gives the total (LAP II.1A; LAP II.1B). The jars with volumes published in the catalogue come to about 2,100 litres between them. Many of the rest survive only as rims and body sherds, and for those the reconstruction has to assume a size. We gave them capacities in the range of the measured jars, so that the whole set comes to 3,983 litres: the model agrees with the excavators’ total because it was built to, and only 53 per cent of that total rests on measured jars. In the model the 18 fixed vessels, 17 jars in stands and barrel KM 5581, hold 2,487 litres and the 18 loose jars 1,497.

What was in them, we do not know. Burnt seeds of wheat, barley, grape, lentil, pistachio, fig and olive were found with the jars, but “none … in sufficient quantity to conclude that it was being stored”, and the fire mixed what there was. Chemical tests on 24 sherds found no lipids in any of the storage fabrics, pithos fragments included, so the excavators thought it “unlikely that the pithoi were used for oil, fat or wax storage” (LAP II.1B). The tests have limits: the jars beside the stone blocks were not sampled. Grain, pulses, water or wine would leave no fatty trace, and a white powdery deposit inside three jars has never been analysed.

For scale, and only for scale: if every jar had held grain, the store would have weighed about 3 tonnes, enough to feed a household of five for more than three years at half a kilo a head a day. The sum is ours; we know of no published one. Peltenburg estimated that the building housed six or seven people (Peltenburg 1993).

In the middle of the back rows, with open floor in front, stood two big stones: a drum 54.5 by 47.5 cm with a central cavity 23 cm across, and a thicker block 68.5 by 52 cm with a long depression. Peltenburg compared them with the Iron Age olive presses at Tel Miqne-Ekron in Israel and suggested that olive oil “may have comprised a significant proportion of bulk storage”. The final report itself sets out the case against. Seymour Gitin, who dug at Miqne, judged the arrangement about half the size of the presses there; the basin has no spout or raised rim; the ground-stone catalogue classes one stone as a pot stand and the other as a mortar; and the residues and seeds give “little … positive support” (LAP II.1B). The archaeobotanist Murray found that “the botanical evidence does not support the substance being olive oil” (Lucas 2014). The model shows both stones where they lay and offers the press only as a speculative overlay.

South-east of the hearth, on a flat slab, stood a stone mortar of dense calcarenite, 38 cm high and 45 cm across; “no pestle was recovered in its vicinity”. Against the west wall, behind the jars, lay a heap of about twenty stone tools and two intact triton shells (LAP II.1B). The thirty axes and seventeen adzes Peltenburg counts from the building (Peltenburg 2014) include nineteen axes and six adzes from an older cache beneath the stands, which belongs to an earlier phase of the house.

With fragments of the burnished bowls, the most common things on the floor were conical stones: small chalk cones about 3.5 cm high, which first appear at Kissonerga in Period 4. Peltenburg and Bernard Knapp read them as tokens for recording goods in and out; Steel sets them in Paul Halstead’s idea of “social storage”, households storing against bad years (Peltenburg 1993; Knapp 2008; Steel 2004). They might equally be stoppers, for which there is “no direct evidence” (LAP II.1B). The model scatters 30 of them, a number of our choosing.

The bowls may be measures. The small hemispherical bowls average 1.3 litres and the ovoid bowls 1.39, “much more standardised than in previous periods”, and the excavators wondered whether they were used to dole out goods “beyond the Pithos House”. A nest of five bowls and a triangular bowl lay inside a barrel on its stand (LAP II.1B). Cones for counting and bowls of a set size are the props of every argument that goods were handed out from here, and both rest on interpretation.

No one has built a Pithos House, but in 1988 archaeologists at Lemba began building round houses to see what the work involved. The first, 10 m across with a stone plinth and a mud wall, took 140 person-days of building: 29 for the terrace and plinth, 57 for the mud wall and 43 for the roof (Thomas 2005). The 3D model’s Build mode scales those rates to this house, step by step, from its own geometry and ordinary densities for stone, earth and timber.

Cutting the terrace moves about 14 m³ of hillside, some 930 basket-loads. The wall is the big job. Stone to the wall head makes it about 51 tonnes of stone, 9,000 or so stones as an order of magnitude and 2,100 head-loads, laid in mud that takes some 4,300 litres of water together with the roof. On top go about 900 mud bricks for the band round the timber ends. The roof asks for six post seats, 8 posts, beam logs about 19 m and 18 m long for the main and side lines, the heaviest single log some 190 kg, 26 joist poles with a combined length of about 190 m, and around 10,000 reeds. Finished with its mud, earth and skin, it weighs about 25 tonnes, 540 kg/m² over the room below. All told about 150 tonnes of material are dug, carried and placed, water included.

The building work comes to about 260 person-days, 150 of them for the stone wall, where every load above head height has to be handed up. Like Lemba’s, the figure leaves out gathering the stone, fetching the water, felling the timber, cutting reeds and burning lime, and it leaves out the potters’ work of making the jars. Ten people working every day would need about four weeks; a crew of that size, and building in the dry months so that the mud could set, are our suggestions.

The excavators took the burnt house apart from the top down, and the order in which things lay tells the order in which they fell. On the floor, with the jars and everything else, lay ash and burnt timbers from the roof, some still aligned with the postholes they had stood in, and a dense spread of fine ash 5 m long and a metre wide against the south-east wall. Above that came the roof itself: charred timber, lumps of mud with reed impressions and compacted earth, and the stone box that fell in once the fire was under way. Over it all lay the upper walls, “wall and roof? stones in disarray, together with bricky material, plaster patches, ash and burnt, compacted mud. They had crashed down onto the pithoi in such a way that sherds and stones became inextricably mixed.” The excavators’ reading is that “the roof collapsed and stones and compacted mud from the upper walls followed subsequently” (LAP II.1B). The whole deposit was about half a metre deep, with objects caught in it up to 0.57 m above the floor.

The jars were “severely burnt”, many of their bases burnt where they stood; only those shielded by the wall came through “tolerably well”. The body of an infant lay in the western collapse of the upper wall and roof: limb bones and fragments of a thin skull, with no teeth or jaw, and no closer age is published (LAP II.1B). The excavators took it as evidence of how suddenly the house came down. The 3D model marks the find-spot in its excavated state with a plain label and shows no figure.