A Chalcolithic Round House, Layer by Layer

Towards the end of the Middle Chalcolithic, some five thousand years ago, someone at Kissonerga in western Cyprus put a house into a pit. It was made of clay: a deep, lopsided bowl 20 cm high and about 35 cm across, with a doorway cut in its side. Figurines of stone and pottery went into the pit with it, and before the little building was buried its wall bracket, its door peg and some of the features inside were snapped off (Peltenburg 1991). The model, KM 1446, is the fullest picture of a Chalcolithic Cypriot house that has come down to us.

Outside, a painted band runs round the top and the bottom with zigzags hanging from it, and a red frame outlines the door. Inside, the walls carry festoons, hatching, zigzags and wavy lines, and the floor and everything on it are painted red. Just inside the doorway sits a small raised disc with a hollow in its top, and above the doorway a bracket: the pivot stone and the top fixing of a hinged door. A separate clay slab with a peg along one long edge, KM 1532, is the door itself. At the centre lies a square platform with a dip in the middle, and from it two low ridges run out to the wall, fencing off the quarter of the floor to the right of the door.

Full-size houses are harder to see. No wall of the period survives more than about 80 cm high anywhere on the island; excavators find a ring of stones, a dished floor, the base of a hearth and a thick layer of the mud that once stood above them. The model cannot close that gap alone. Taken at face value its proportions would give a house 6.8 m across walls nearly 4 m high, and Gordon Thomas, whose classification of the Kissonerga buildings underlies most of what follows, warned against expecting from it “a faithful rendition of all aspects of a full-scale building” (Thomas 2005). Everything above knee height here is argued from the stone stubs, the debris of burnt houses, the model, and an experimental village at Lemba where archaeologists built round houses to see what would stand up.

The house rebuilt here belongs to the Middle Chalcolithic, roughly 3500 to 2900 BC. It is a middling dwelling, 5.6 m across inside and 6.8 m over its walls, put together from excavated houses at Kissonerga-Mosphilia and Lemba-Lakkous and from the model. The largest buildings of its day were twice as wide; the smallest, little more than half its size.

Chalcolithic round houses have been dug across Cyprus, yet nearly everything known about how they were built comes from one short stretch of the west coast, north of Paphos. From the 1970s the Lemba Archaeological Project, directed by Edgar Peltenburg of the University of Edinburgh, excavated a cluster of neighbouring sites there: Lemba-Lakkous on the coast, its first dig; Kissonerga-Mosphilia, dug from 1979 to 1992; the older settlement of Kissonerga-Mylouthkia close by; and Souskiou-Laona, one of the find-places of the period’s picrolite cruciform figures. The Kissonerga final report (LAP II.1A, 1998) contains Thomas’s typology of every wall, foundation, door, floor, hearth and partition on the site, and his 2005 book sets the same buildings beside the results of his building experiments.

Elsewhere the record is older and thinner. At Erimi-Pamboula on the south coast, Porphyrios Dikaios excavated from 1933 to 1935, and Thomas counts eleven clear round houses there; Kythrea and Lapithos in the north were dug by the Swedish Cyprus Expedition. The one large body of new evidence comes from the west again. Since 2015 a Leiden team under Bleda Düring has been excavating Late Chalcolithic buildings at Chlorakas-Palloures, among them a round house reported as 14 m across. “The Chalcolithic house” is therefore a western house, and how far it holds for the rest of the island is an open question. The sequence also has a hole in it: Lemba, Kissonerga-Mosphilia and Erimi all seem to have been abandoned for about two centuries around 2900 BC, a break that looks local, since Souskiou stayed in use.

Anyone who has seen the rebuilt round houses at Khirokitia should set them aside here. Khirokitia is Aceramic Neolithic, some 3,000 years older than the houses in this report, and it was built in its own way: some of its houses had double walls, and some doorways stood as much as 1.3 m above the floor inside, with steps leading down.

A house began as a hollow. On the sloping ground at Kissonerga, builders cut a terrace and scooped out a shallow dish, deepest in the middle: 30 to 50 cm at the centre of B 1046. The dish was part of the design. Plaster curves down from the wall into it, and entering a house usually meant stepping down. In the reconstruction the floor lies 15 cm below the ground at the wall and 22 cm at the centre.

On the edge of the cut went the socle, the stone base of the wall. In Thomas’s commonest Middle Chalcolithic type, two faces of blocks about 40 by 50 cm, calcarenite and calcareous sandstone, some of it taken from eroded outcrops on the shore, were set flat side outwards among rounder river stones, with mud and small stones packed between. These walls are 40 to 75 cm wide at the base, up to 85 cm in the largest house at Kissonerga, and the mortar was soil and water. The best preserved stand four or five courses high, 56 to 60 cm at most, and many survive to only two. The final report is of two minds about the largest house, B 206: one passage counts four or five courses in its wall, another suggests there may never have been more than the three that remain. The reconstruction’s socle has three courses above ground, 45 cm, enough to keep the mud clear of splashing rain.

Above the stone rose mud. On top of several Kissonerga socles lies uneroded mud full of the casts of straw chaff, which Thomas calls “strong evidence for a mud structure in the upper part”. His term for the technique is mudwall: wet, tempered mud heaped up by hand in horizontal lifts, without shuttering, much like cob. Pisé, earth rammed between boards, would have needed formwork for which there is no evidence. Mudbrick is unattested in any Chalcolithic house; mould-made bricks appear with the Philia horizon at the threshold of the Bronze Age. The hard lumps of “daub” from Souskiou carry no impressions of twigs or branches, so wattle and daub is ruled out too.

At Lemba the builders of the first experimental house trod soil, straw or seaweed and water together barefoot, left the mix overnight and trod it again. It went onto the wall in lifts 15 to 20 cm high, with fist-sized stones pushed into the top of each lift “as a further key” for the next. A lift needed a day to firm up in summer and four or five in wet weather. In the reconstruction the mud wall is 60 cm thick and rises 1.55 m in nine lifts, to a wall-head 2 m above the ground.

That 2 m is a judgement. Thomas estimated that the early Kissonerga houses “most likely stood a minimum of 1.80–2.0 m high”, and the collapse agrees: in B 2 at Kissonerga, fallen wall stones lie under 60 cm of compacted mud washed down from above. The Lemba experimenters chose 2 m for the headroom needed to reach jars stored against the wall, and Thomas was candid about it: “In the absence of any real evidence for wall heights the one adopted … is as valid as any other.”

Each house had a single door, hung the same way almost every time. At Kissonerga nine pivot stones have been found in place, in houses of every Chalcolithic phase, and all nine lie just inside the doorway on the left as you enter; so do two at Lemba and one at Erimi (LAP II.1A). The model shows how the arrangement worked. The door was fixed to its own slim upright post, whose foot turned in a hollow pecked into the stone and whose head was held in a bracket on the wall, so the leaf swung inwards to the left. At Souskiou one socket found in place is 8 cm across, for the post of a door about 50 cm wide, and the post stood 10 cm out from the corner of the wall, a gap the excavators also noticed in the model (Peltenburg, Bolger and Crewe 2019).

The leaf itself has survived nowhere. The door of the first Lemba house was four seasoned pine planks, doweled and glued. Chalcolithic carpenters had no saws: ground-stone axes and adzes make up 35 per cent of the finished stone tools at Kissonerga and 45 per cent at Lemba, and the first metal object of the period, a chisel tip from Erimi, dates to about 3200 BC. The reconstruction gives its door three unequal boards split from a log with wedges and dressed with an adze, pegged and lashed to two cross-bars on a heel post 8 cm thick. The 3D model offers a woven hurdle as the alternative. Both fit the model’s “solid” door.

No doorway survives to its lintel. Early doors at Lemba are 50 to 70 cm wide, and Thomas surmised a metre or more for the Middle Chalcolithic, while admitting that “complete examples of this type of doorway do not exist”. The reconstruction takes 0.9 m, with a lintel of rough poles at 1.5 m, low enough that most adults stoop, and a threshold of pebbles and lime plaster. Doors faced broadly south: south-east at early Lemba, the south-west quadrant in the big Middle Chalcolithic houses at Kissonerga (though Thomas elsewhere reads one jamb there as a door facing south-east), and almost any direction in the Late Chalcolithic. The left-hand pivot has one reported exception, a Late Chalcolithic house at Chlorakas-Palloures whose door would have swung to the right.

The window in the first Lemba house, 75 cm square and a metre above the floor, is often taken as evidence that these houses had windows. It went into the east wall because, Thomas explained, “the eastern arc of the wall was the only part still low enough to receive a window without necessitating some demolition”, and he later called his archaeological case for it “a very feeble proposition”. We know of no window in any Chalcolithic house on Cyprus.

The absence carries less weight than it seems to, because a window at any sensible height would have sat in the mud wall, which is gone. Older houses on the island had small openings set high in the wall, at Late Neolithic Ayios Epiktitos-Vrysi (26 by 35 cm, 1.2 m above the floor), at Kalavasos-Tenta (Todd 1982) and at Khirokitia (Dikaios 1953). The one positive argument against comes from the model, whose painted decoration runs round the whole wall. As Thomas put it, “the way in which the doorway is accommodated leaves no space for any other apertures.” The reconstruction leaves the wall closed. The 3D model can add a small opening of the Neolithic kind, 30 by 35 cm with its sill 1.2 m up, over the special floor to the right of the door, and labels it as a guess.

Floors were usually earth or clay. Thomas refused to call them “beaten” or “trampled”, since there is no sign of special preparation. Some were laid deliberately, in greenish clay at Lemba and red clay at Erimi. In the richest Middle Chalcolithic houses, lime-plaster skims up to 6 cm thick were laid together with the wall plaster, so that floor and wall formed one surface.

At the centre was the fire. Dikaios read the Erimi hearths as “central supports” for a roof post, and Bolger followed him in 1988, yet the hearths carry ash, reddening, light vitrification and a bluish tint, and they are far too fragile to bear a post (LAP I). The usual hearth is a low round platform of cobbles in mud, plastered, about 75 cm across, with a fire-bowl some 25 cm wide. Kissonerga alone has a rectangular kind, 1.5 to 2.2 m long, always with radially divided floors, and the model shows one. The best, hearth 951 in the Ridge Building (B 855), was built of stones in mud under a course of flat potsherds and a 4 cm skin of clay, fired maroon round its 30 cm bowl and orange-red beyond. In a small hollow beneath it lay a fragment of a figurine. The reconstruction’s hearth follows this one: 1.6 by 1.3 m and 8 cm high.

From two corners of such hearths, in three Kissonerga houses, low ridges of small cobbles rendered with havara or reddish soil ran out to the wall. They are about 20 cm wide and 10 cm high, and they mark off the same part of the house each time: the quarter on the right as you come in. The model shows it too. Later floors build up on one side of the ridges only, so the two zones were resurfaced separately.

In the great houses of the Middle Chalcolithic climax, that quarter received what Thomas identifies as the most labour-intensive element of any Chalcolithic house: a bed of fist-sized cobbles under 5 to 10 cm of burnt lime mixed with unburnt limestone, floated so that the finest fraction rose into a smooth, hard top, about 20 cm deep in all and curving up into the wall plaster. Peltenburg remarked that such floors “required even more labor than was expended on roofing beams” (Peltenburg 1991). In the Red Building, B 206, the quarter became a room of 40.5 m², almost a third of the house, and its first floor was painted red; the excavators think it was probably “vivid red” all over, as in the model. Such floors come mostly from great houses; the reconstruction gives its middling house one to show the arrangement whole.

What went on there is unknown; nothing found on it says why the quarter was fenced off, floored with lime and coloured red. The seams of these houses seem to have mattered. At Lemba a limestone figurine, the “Lemba Lady”, lay where a partition channel met the wall, and the Kissonerga model and its figurines were buried beneath a house.

Stores were kept in the room, in plastered basins sunk into the floor and storage jars held upright in rings of stones; the reconstruction sets one of each along the wall to the left of the door.

No Chalcolithic roof survives, and it is the part of these houses drawn most freely. Dikaios postulated a cone of wattle and daub, “although no evidence for this proposition is presented”, as Thomas observed; for Neolithic Khirokitia he proposed corbelled domes. Popular pictures still give these houses thatched cones. The evidence points to something flatter and much heavier.

The best witness is the Pithos House at Kissonerga, which burned. On its floor lay the roof: charred timbers, daub with reed impressions on one face, and compacted mud. The fire-blackened stones of the wall lay on top. The excavators concluded that “the roof collapsed and stones and compacted mud from the upper walls followed subsequently”, the way a heavy roof of timber and earth falls. At Kissonerga-Mylouthkia Thomas read reed- and timber-impressed daub as “the first mud layer over reed matting of a flat timber-and-earth roof”, and his study of the Kissonerga building materials found that flat roofs on reeds or regular long branches “were commonplace”.

Two absences point the same way. A cone needs “a greater number of internal posts than is warranted by the evidence”, in Thomas’s judgement. A roof with eaves drips: the 40 cm eaves of the first Lemba house cut a drip-trench round its wall, and birds burrowed into them. No drip-trench has been found round any excavated Chalcolithic house, so the reconstruction has no eaves. Le Brun found evidence for flat roofs at Khirokitia in 1989, a precedent some 3,000 years older and no more, and the excavators of Chlorakas-Palloures give their Late Chalcolithic houses a “probable flat roof” of loam.

The case has weak points, and Thomas named them. A flat roof on a round building is “arguably less common, if not unusual” in the ethnographic record; no ancient roof has been measured in section; and the great Middle Chalcolithic houses at Kissonerga have no internal postholes at all, although an earth roof spanning 10 to 14.5 m must have needed support. Postholes, 5 to 25 cm across, form no consistent pattern. Thomas’s explanation is practical: a heavy earth roof holds its posts down on flat stones, as in Cypriot vernacular building, without deep holes, and in the Pithos House flat stones near the hearth suggest that “posts were also propped on slabs”. At Lemba, soil added over the years to stop a roof leaking split a beam, which then had to be propped. “A continuation of this process through time,” Thomas wrote, “would lead to a proliferation of posts within the building in an apparently random pattern.”

The reconstruction follows the system Thomas judged “probably more accurate” for smaller houses: a spine beam of pine or mulberry, the woods found in structural contexts at Kissonerga, 24 cm thick and 6.8 m long, laid at right angles to the door, with 40 rafters 12 cm thick every 33 cm. Over the rafters lie a reed mat 5 cm thick, a bedding of brush, 6 cm of tempered mud, earth rising from 10 cm at the edge to 20 cm at the centre so that rain runs off, and a cap of havara, the white mix of decayed limestone and clay, 8 cm thick. The havara cap is the least certain layer. It follows Cypriot tradition and the Lemba experiments, where dry havara, spread and then soaked, set into a waterproof skin and havara mixed wet cracked. The roof stops flush with the wall; the gutters and stone spouts tried at Lemba washed away in their first winter.

The 3D model can swap in two other roofs. The radial roof is the one built on the first Lemba house: four posts carrying a ring beam, with rafters radiating to the wall-head. Thomas called that house “a hybrid that never existed in prehistory”, and its posts would interfere with the special floor. The cone shows what the houses were not: a thatched cone some 3 m high needs a ring of posts and throws its rain beyond the wall, cutting the drip-trenches that excavators never find. Nor is there a case for domes: the inward curve Dikaios saw in the Erimi walls is, Thomas argues, an illusion made by plaster curving down into dished floors.

Above the model’s door stand nine small projections in two rows. Thomas read them as the ends of roof timbers; Peltenburg and Bolger see “protomes”, ritual or status symbols, or a porch. Nothing else supports projecting rafter ends, so the reconstruction trims them flush.

Thomas insisted on a distinction: plaster means true lime plaster, made with burnt limestone, and every other coating of mud, clay or havara is render. The commonest finish at Kissonerga, in every period, is clay or havara render smoothed straight onto the wall. In the Pithos House it lies about 2 cm thick over rough stone, “through which the underlying stonework was always evident”. A Late Chalcolithic render from Kissonerga, up to 28 mm thick, is “a dense, very fine white-grey clay”, and one fragment of it carries painted bands of red and black.

The Red Building shows how far this could go. Its walls were replastered many times in lime coats about 2 mm thick, to a total of 20 to 35 mm, in layers of “a distinct pink and a white colour”. The excavators first called it “Painted Unit 206”; what they found in its wall plaster were small irregular inlays of pink, never more than 10 cm across, set into the white 10 to 20 cm above the floor. Painted fragments occur elsewhere, red on lime plaster at Lemba and red traces on a white slip at Souskiou. Whether full-size walls carried the model’s festoons and zigzags is unknown; the 3D model offers them as an option, alongside the Red Building’s inlays and a plain wall.

Outside, the reconstruction departs from a familiar image. The houses rebuilt at Lemba were rendered in red soil and straw, following the recent Cypriot custom of red soil for walls and white havara for roofs, and it is easy to picture these houses as red. Thomas himself warned that the custom was recent and might not reflect prehistoric choices. The excavated renders at Kissonerga are mostly clay and havara, several of them white or grey, so the reconstruction is rendered in pale havara, with red as an alternative. Red belongs on the floors: in the Red Building, on stained earth floors elsewhere at Kissonerga, and across the floor of the model.

The lime came from kafkalla, the hard limestone crust of the local soils. At Lemba the experimenters layered kafkalla and charcoal into a dome-shaped clamp, 100 kg of stone to 17 kg of charcoal, and burned it from nine in the morning until about three in the afternoon. The result was a “primitive but very effective cement” of the kind found in the floor plaster of Kissonerga (Thomas 1999).

The Lemba Experimental Village was begun in 1988 beside the ruins of Lemba, run by Gordon Thomas within Peltenburg’s project. Its first house, built by Paul Croft, was 10 m across and took 140 person-days of building alone: 29 for the terrace and a stone plinth of 12 to 20 tonnes, 57 for a mud wall of about 35 m³ that used 926 barrow-loads of soil and some 11,000 litres of water, and 43 for the roof. In 1990 two people built a second, stone-walled house 5 m across in about seven weeks, without trowels, levels or metal tools, and found stone hammers as effective as steel.

The 3D model’s Build mode scales these figures to the house reconstructed here, step by step, from its own geometry and ordinary engineering densities. Digging the scoop and the socle trench moves about 14 m³ of earth. The socle takes some 450 facing blocks and 1,100 cobbles and weighs around 18 tonnes. The mud wall needs about 17 m³ of mud, 30 tonnes once dry: 460 of the Lemba builders’ barrow-loads (the Chalcolithic builders had no barrows) and 5,500 litres of water. The roof asks for a spine log, 40 rafters with a combined length of about 110 m, and some 2,200 canes for the mat; with its mud, earth and havara it weighs about 23 tonnes, not far short of the wall beneath it. Burning lime for the wall coats and the platform floor uses about a tonne of kafkalla and 170 kg of charcoal. All told, about 114 tonnes of material are dug, carried and placed, water included.

Scaled part by part from the first Lemba house, the building work comes to about 97 person-days: 31 for levelling, the scoop and the socle, 28 for the mud wall, 20 for the roof, 6 for the door, render and plaster, and 11 for the floor, hearth and lime platform. The figure leaves out what the Lemba builders left out too: quarrying and carrying stone, digging soil, fetching water, felling timber, cutting reeds and burning lime.

Then the house had to be kept up. At Lemba the first roof leaked after almost every big storm for a year or two, because springy myrtle under the soil stopped it compacting, and the render weathered worst on the west and south-west, the storm sides, and had to be replaced. A Late Chalcolithic house rebuilt at Kissonerga-Mosphilia in 2010 told the same story. By 2021 winter rain had brought down part of its wall near the door, and Paul Croft, who had built the first Lemba house thirty-three years before, was rebuilding it.

The Chalcolithic round house grew out of timber. At Kissonerga-Mylouthkia the excavators traced an “uninterrupted development from a pit, to semi-subterranean post-frame structures to a mud-built Chalcolithic round house with stone foundations” (LAP III.1). The earliest houses at Lemba were small, reportedly down to about 3 m across, with mud walls set straight on the ground, greenish clay floors and narrow doors facing south-east. Their builders pressed broken artefacts and near-complete pots into the walls. The habit lasted: in a Late Chalcolithic hut at Kissonerga, a Red-on-White spouted bowl was smashed into the wall beside the door.

Stone footings arrived with the early Middle Chalcolithic, and houses grew to 5 to 8 m. Then, in the climax at Kissonerga, a few became very large. The Red Building measured about 14.5 m across and enclosed 132.7 m² (LAP II.1A; earlier accounts give other figures), and its red east room alone was bigger than most later houses; the Ridge Building reached 10.5 m. These great houses had calcarenite socles, lime floors, pink-and-white walls, rectangular hearths and wide stone-jambed doors, with a paved track 2.6 to 3.2 m wide running between them. The first pithoi appear. Small rectilinear houses with rounded corners stood alongside, so square buildings were no Late Chalcolithic novelty.

After the break around 2900 BC, the houses at Kissonerga and Lemba were different in kind. Most were 4 to 6 m across, with 7 to 20 m² of floor. Their walls held far more stone, laid in thick mud, perhaps to half height, perhaps all the way up; the Pithos House wall still stands 78 cm. Floors were plain earth. Ridges and platforms vanished, and storage crowded in: B 204, a hut 3.9 m across, held at least five storage vessels. Doors faced any direction, houses were grouped round shared yards, and at Kissonerga a frame of posts in two rough rings, B 375, was raised over graves.

One new fitting remains a puzzle. In seven Late Chalcolithic houses at Kissonerga, a broken saddle quern stands upright on its edge, 1 to 1.5 m inside the door, directly between the door and the hearth. It has been read as a door brace, yet the stones often sit nearer the hearth than the door, one is set lower than any door would reach, and Thomas asked why anyone would use broken querns for the job. What they were for is unknown.

A few Late Chalcolithic buildings were large. The Pithos House at Kissonerga measured 9.1 m across, according to the final report; the 10 m of earlier accounts came from preliminary publications. At Chlorakas-Palloures, the large round house has a charred roof post reportedly dated to between 2875 and 2665 cal BC. With the Philia horizon, from around 2500 or 2400 BC, the island’s builders turned to rectilinear houses of several rooms, and the Early Bronze Age began.

The Pithos House, B 3 at Kissonerga, is the best recorded destruction of a Chalcolithic building on the island. It had seen long use: it has five radiocarbon dates, a patched hearth, a door pivot that may have been replaced, and a cache of axes sealed beneath its jar stands. At the end its storage jars were massed against the back wall in purpose-built stands, with more standing loose, propped against one another, to within 2 m of the door. The final report counts 37 storage vessels and notes that there were once more; the figure of over 40 pithoi still in circulation comes from preliminary accounts.

The fire brought the roof down first. Charred timbers, reed-marked daub and compacted mud lie on the floor, and the fire-blackened stones of the upper wall lie over them. A child’s body was found trapped in the western collapse of the wall and roof. Above the roof timbers lay something that remains unexplained: a neat box of stones, 1.40 by 0.97 m, bedded two deep in hard mortar inside a frame of burnt timber, the same width as the entrance 1.45 m away. It fell after the fire had started. A blocked doorway, a window, a loft and a fitting on the roof have all been suggested.

Why it burned is open. Peltenburg read the building as a central store, and the final report describes stores “far in excess of the needs of a normal domestic unit”. Work on the Palloures houses points to a practical difficulty: with floors, walls and roof all made of earth, the only parts of a Chalcolithic house that burn are its posts, beams and door, and “burning Chalcolithic buildings is difficult”. A house burnt as thoroughly as this one may have been fired on purpose, an idea that has yet to be tested.

Most houses ended more quietly. B 2 at Kissonerga was left with few finds and its partition walls carefully taken down before it was demolished, and a buttress of the Red Building was later built over its foundation. At Souskiou the mud upper walls seem to have been dismantled at abandonment, perhaps for re-use, in what the excavators call “a formal occasion”. Houses left to the weather filled with fine silt, washed and blown into the scoop, and then grew a laminated crust of clay riddled with insect burrows: the roof had gone and rain was falling on the floor.

What survives is what the reconstruction’s Tour mode shows at its last stop: a ring of stones two to five courses high, a dished floor with its skims of plaster, the base of a hearth, the stumps of ridges and a basin, all under a spread of fallen mud.

The roof is the largest gap. A flat earth roof is well supported; its timber layout is a matter of argument, and no one has shown how the great houses of the Middle Chalcolithic, up to 14.5 m across and without a single internal posthole, were roofed at all. Walls of 1.8 to 2 m are an inference, and the thickness of the roof and its havara cap come from what worked at Lemba. We know of no smoke-hole, wall vent or sooted ceiling from any of these houses; the 3D model can add a smoke-hole, and it is a guess. The reeds of the roof mat have never been identified to species.

The painted walls, the projections over the model’s door, the upright querns and the burning of the Pithos House all remain open. We know of no published estimate of how many people lived in a house. Floor areas run from under 7 m² for the smallest huts to about 133 m² for the Red Building; turning them into households needs modern assumptions about space per person, and about whether a given building was a dwelling at all.

For centuries, in house after house at Kissonerga and Lemba, the same part of the floor was fenced off by ridges, floored with lime in the grandest houses, coloured red and resurfaced apart from the rest. Nothing found there says what it was for. Before the clay model went into its pit, someone snapped off its wall bracket, its door peg and some of the fittings inside. The two ridges on its floor survive, still marking off the quarter on the right.