A Stone House at Sotira-Kaminoudhia

Kaminoudhia means “small lime kilns”. The name belongs to a flight of south-facing terraces on the northern edge of the modern village of Sotira, 288 m above sea level and about 400 m north-east of the Neolithic hill of Sotira-Teppes (Swiny 2003, pp. 3, 10). Under the terraces lies a village of the Early Bronze Age covering at least a hectare. Its houses were clusters of rooms, roughly rectangular, trapezoidal or D-shaped, packed so close that neighbours shared walls, and reached along narrow passages. Every wall was stone.

The house rebuilt here stood in the south-west corner of the excavated northern area: a vestibule with its own hearth, a main room with a double hearth, a bench and a stone set in the floor to take a post, and a back room where eight cooking pots were found. Next door, behind a shared wall, is a room entered over a threshold cut from a single block, with a stone mortar sunk in its floor.

Almost nothing above knee height survives. The tallest wall of the block, the far side of the passage, stands 87 cm high in eight courses; the rest are lower. Everything above that line in the model is argued: the height of the walls from other walls that fell flat elsewhere on the site, the roof from twentieth-century village houses, the doors from a threshold and a pivot stone found in neighbouring rooms. The model marks the difference on every layer.

Stuart Swiny found the site on survey in 1978 and at first assigned it to the Philia culture, the horizon that opens the Cypriot Bronze Age (Swiny 1985, p. 44). He excavated it for the Cyprus American Archaeological Research Institute and the University of Minnesota: he directed the season of 1981, the geologist George Rapp the second main season in 1983, and a smaller season followed in 1986. In Area A, at the northern edge of the settlement, 375 m² were cleared on bedrock never more than a metre down; Areas B and C, further downslope, lay under deeper hillwash that had kept their layers better (Swiny 2003, pp. 3, 5, 10). The final report, edited by Swiny, Rapp and Ellen Herscher, appeared in 2003, and almost everything here comes from Swiny’s chapter in it on the settlement.

The Philia label did not outlast the digging. The houses of Area A belong to two phases, and the whole of the excavated village is Early Cypriot III, the last part of the Early Bronze Age (Swiny 2003, p. 7). Philia pottery was found only in four of the tombs. Jennifer Webb and David Frankel, both excavators of the Early Bronze Age village of Marki-Alonia in the centre of the island, suggest in their reviews that the settlement began as a smaller Philia foundation whose earliest levels lie outside the excavated areas (Webb 2004, p. 374). The houses that were dug are later, and the pottery on their floors is Red Polished, the burnished red ware of the period.

The date rests on ten radiocarbon measurements made at Oxford on charcoal, two of them from the main room of this house. The later phase, to which the house belongs, has a median date of 2262 BC, or 2236 BC after an allowance for the age of the wood burnt, and Swiny puts the latest occupation of the settlement at about 2200 BC (Swiny, Rapp and Herscher 2003, pp. 504–505; Swiny 2003, p. 66). These are old single measurements with errors of 75 to 100 years, each calibrating to a span of several centuries, and none of them fixes the end of the village. “About 2200 BC, on the excavators’ dates” is as far as the evidence goes.

Swiny does not describe houses. He describes units, the numbered rooms and open spaces of the plan, and he is careful about what joins them. The house of this report is our grouping of three of them, Units 40, 7 and 18, which he describes together “because they form a single architectural unit” (Swiny 2003, p. 23): a vestibule, a main room and a back room, linked by doorways, matching the “two to three intercommunicating rooms” he gives as usual in the village (Swiny 2003, p. 5). The fourth room of the model, Unit 6, lies east of the back room across the party wall WW. No doorway joins them; Unit 6 opens onto the passage. Whether one family used all four rooms, or two lived back to back, cannot be told, and Swiny identifies no households anywhere on the site (Webb 2004, p. 374).

The block was chosen because its way in can be followed from the passage to the back room, because Unit 7 holds the one piece of excavated roof evidence in Area A, and because Unit 6 is, in Swiny’s words, the best-preserved room there (Swiny 2003, p. 21). Its weaknesses belong on the page too. Unit 18 was only a third excavated, so a hearth and a bench suspected there remain unproven and the model leaves them out (Swiny 2003, p. 28). The height of the walls comes from walls some distance away, the doors borrow hardware from neighbouring rooms, and the roof is analogy from beams to skin.

Frankel doubts that a standard house can be defined at Kaminoudhia at all, because its rooms were renovated, subdivided and rebuilt many times over. The vestibule is itself a late change: its threshold and the wall WAQ beside it were added to a passage that had been wider (Swiny 2003, pp. 24–25). The model shows the block as it stood in its last phase.

The way in is corridor 32/33, a passage 11.3 m long and on average 1.2 m wide that runs along the south front of the block. Its floor climbs eastwards with the hill, and it served, among other things, as a dump for household rubbish (Swiny 2003, p. 31). Six slabs stand upright against the foot of its north wall and two against the south. Orthostats like these turn up singly or in short rows across the site, mostly in the passages; Swiny suggests that some protected mud plaster from rising damp and that some were simply slabs in store (Swiny 2003, pp. 55–56). They never formed a continuous base course, and the model places only those that were found.

At the west end of the passage a threshold of two flat fieldstones, laid end to end, leads into Unit 40, a vestibule about 2.0 by 1.6 m whose floor stands 20 cm above that of the room beyond. A hearth, Ft 9, fills its corner: a box of mud plaster with walls 4 cm thick and 18 cm high round a firebox 32 by 48 cm, braced on one side by a stone. Swiny cannot say whether this space had a roof and gives both readings (Swiny 2003, pp. 24–25); the model roofs it by default and lets the reader take the roof off. A doorway 1.1 m wide, with no threshold recorded, opens north into the main room.

The best doorway on the block belongs to the neighbour. Unit 6 is entered from the passage over a threshold cut from one block of limestone, 1.32 m long, 23 cm wide and standing 27 cm above the floor, with two pecked patches on top, probably for the door jambs, that would have framed an entrance 87 cm wide (Swiny 2003, p. 21). Swiny’s text puts this door in the south-west corner of the room; his plan, a photograph captioned with the threshold on the east, and a stone mortar that sat “1 m northwest of the door” all put it in the south-east, and the model follows them. Forty centimetres in front of it the passage floor holds Ft 11, a carefully shaped hollow 50 by 64 cm and 10 cm deep, whose purpose is unknown; it would have hindered access to Units 6 and 16, Swiny observes, “unless traffic was meant to pass over or through it” (Swiny 2003, p. 31).

No door leaf survives anywhere on the site. The pivots do. In the doorway of Unit 16, east of Unit 6, an oval stone 30 cm long with a pecked cup 5 cm across lay at one end of the threshold, where the heel post of a swinging door would have turned in it (Swiny 2003, p. 26), and Swiny’s summary speaks of thresholds “complete with well-used pivot holes” (Swiny 2003, p. 59). The model hangs each door on a heel post turning in such a stone, with a leaf of split and adzed boards: the village had copper axes and chisels, and we know of no saw. The height of the doors is a guess, since every lintel sat above every wall top that survives.

Unit 7, the main room, is “slightly apsidal or D-shaped”, about 3.3 m across at the widest and some 17 m² by Swiny’s figure (Swiny 2003, p. 23); the published plan gives a little less. At the foot of its curving north wall the builders had cut into the soft bedrock to level the floor, and against the middle of that wall stands the double hearth Ft 8: two fireboxes side by side, each 18 cm wide and 32 cm deep, edged with thin limestone slabs set on edge and packed in front with buff mud plaster 20 cm thick. Along the south wall runs a bench of up to three courses of stone, about 2 m long, 50 cm wide and 42 cm high (Swiny 2003, pp. 23, 61). The floor held pounders, querns and a mortar, three spindle whorls, some 750 sherds, and a flat axe of arsenical copper about 5.7 cm long with copper chisels (Swiny 2003, p. 24; Swiny, Rapp and Herscher 2003, p. 374). Swiny calls the room multi-functional.

Beyond a doorway 1.2 m wide, over a mud sill about 8 cm high that Swiny thought “most likely pisé” (Swiny 2003, pp. 23, 60), lies Unit 18, 5.5 m from east to west and up to 3.8 m wide, about 17.5 m². In the third of it that was fully dug lay eight Red Polished cooking pots, “an unusually high number”, with up to thirty-one small bowls, a stone that may have been a weight or a tethering stone, and a gaming stone (Swiny 2003, p. 28). The pots point to cooking or storage.

Unit 6 has the fullest floor. Its hearth, Ft 10, against the east wall, is an L-shaped kerb 14 cm high round a firebox about 33 by 46 cm lined with buff clay, and the wall face above it carries reddish-yellow mud plaster up to 46 cm, the top of the surviving wall. A stone mortar 36 by 38 cm is sunk with its rim flush with the floor, and white lime plaster patches the floor in the north-east corner. On the floor lay some twenty grinding and pounding tools, four whole or restorable pots, fragments of a copper dagger, two spindle whorls and 1,176 sherds, the most from any room in Area A. Swiny reads the room as given over to food, “from pounding and grinding to cooking and consumption” (Swiny 2003, pp. 21–23).

He finds no specialisation between rooms anywhere in the village (Swiny 2003, p. 5): grinding, cooking and spinning turn up in room after room, as in other villages of the Early Bronze Age. Frankel adds a warning that applies to every such reading: without a clear idea of how things came to be thrown away or left behind when the village emptied, the finds on a floor are a weak guide to what a room was for. The lime-plastered bins found elsewhere in Area A are absent from these four rooms, and the model has none.

“All the walls at Kaminoudhia were built of angular fragments of the locally abundant tabular limestone” (Swiny 2003, p. 59). The rock of the Sotira hills is the Pakhna Formation, marly chalks and limestones, reddish where the iron in them has weathered (Rapp 2003, p. 463), and it splits along its beds into slabs often 10 to 15 cm thick. The builders used them as they came: the stones are untrimmed, and their flat faces are the bedding planes. Counting stones on the tops of walls, Swiny found two sizes, a block about 30 by 22 by 10 cm that makes up 86 per cent and a larger one about 55 by 30 by 15 cm (Swiny 2003, p. 59).

Each wall has two faces of these slabs with a core of rubble between, “roughly coursed” and bedded in “a liberal application of mud mortar”, normally 3 to 4 cm thick in both the bed and the upright joints, with no small stones wedged in to tighten them. Partitions are only a double line of facing stones (Swiny 2003, p. 59). Most party walls are about 50 cm thick, and the east wall of Unit 6 is 70. There are no footings: the first course sits on the bedrock as it slopes. The corners of the original rooms are bonded, and later cross-walls butt against them with a straight joint, which is how the order of building can be read from the plan (Swiny 2003, pp. 21, 55–56).

“No sun-dried mudbrick was used in wall construction” (Swiny 2003, p. 59), though the villagers knew how to make it: three chaff-tempered bricks, 5 cm thick and about 40 cm long, stand on edge in a feature in Unit 5, the room north of the house (Swiny 2003, pp. 20, 59–60). This is the sharpest difference between Kaminoudhia and Marki-Alonia, where the houses were built of mould-made mudbrick on stone footings (Swiny 2003, p. 65). The simplest explanation is the rock underfoot, which hands the builder ready-made slabs. At Alambra-Mouttes, though, builders put mudbrick on stone footings “although excellent building stone was available on site” (Swiny 2003, p. 66), so a supply of stone did not settle the matter everywhere. The model can swap its upper walls for the Marki method, a metre of stone socle with brick above, as a regional contrast; for this site it is rejected.

Nor were the walls plastered all over. “No evidence for the consistent rendering of either interior or exterior walls with mud plaster was recorded at Kaminoudhia” (Swiny 2003, p. 60). Mud plaster clings round hearths and bins, as on the wall behind Ft 10, and white lime plaster was found on walls only in two open courtyards, Units 2 and 12 (Swiny 2003, pp. 54, 61). The whitewashed village of some modern pictures has nothing to stand on. A mud render would weather off low, exposed wall stubs, so its absence does not prove the walls were bare, and the model offers a rendered interior as a speculative option. Its default is what was found: pale stone with mud in the joints, and plaster where the fires were.

The walls of the block survive to between 35 and 87 cm, and the tallest wall standing anywhere on the site reaches 1.2 m (Swiny 2003, pp. 29, 31). The height of the rest comes from walls that fell flat. When a long wall topples in one piece, its courses land in order on the ground beside it, and the length of that spread, added to the stub still standing, gives the least height the wall can have had. Swiny found three such falls, each a section of wall that “had toppled southwards as a unit … with each course of stones remaining in situ” (Swiny 2003, p. 56).

The clearest is WI, a wall 60 cm thick in Unit 1, some 8 m north-east of the house. Up to ten of its courses lay “stacked like fallen dominos at 45°” to the south (Swiny 2003, p. 11). Nine of the fallen stones were measured: 12, 12, 13, 12, 15, 13.5, 12, 12.5 and 15 cm thick, 117 cm in all. Allowing at least 3.5 cm of mortar for each course adds 31.5 cm, and the stub still standing adds 45 cm. The sum, 1.93 m, is a minimum.

In Area B the deeper soil kept more. There the wall WH had fallen into Unit 13 as a spread of courses 2.5 m wide, which with its standing stub gives a minimum of 3.04 m. In the courtyard, Unit 12, a section of wall at least 2.1 m high and 1.8 m wide lay on the floor, beginning 2 m out from the wall it came from; counting the gap as well would make that wall 4 m high, which Swiny calls unlikely (Swiny 2003, p. 56). His preliminary report of 1985 had given “at least 3.7 m” for what is probably the same fall (Swiny 1985, p. 44); the final report drops the figure. His conclusion is that some walls stood “to at least 2 m and probably closer to 3 m in height, without any use of mudbrick or pisé” (Swiny 2003, p. 56).

The model takes 3 m from the floor of the back room to the top of the roof at the north edge of the block, with the wall tops falling gently south under the roof. That leaves about 2.1 to 2.5 m beneath the beams in the rooms of the house, close to the 2.5 m inside the burnt village house at Old Paramali whose roof is described below. An option cuts the walls to 1.93 m, the WI minimum alone; under a roof some 38 cm thick, beams included, that leaves about 1.0 to 1.4 m under the beams, below head height.

One part of this argument stands apart from everything else in the report. WI, Swiny notes, toppled downhill “after the structure was abandoned” (Swiny 2003, p. 11). Whether it came down in an earthquake or slowly gave way to the slope, its courses measure the same. The height follows from the geometry of the fall, and the cause of the fall is a separate question, taken up below.

“No direct evidence for the manner in which buildings were roofed was discovered at Kaminoudhia” (Swiny 2003, p. 57). Several rooms burned, yet no charred beams were found, and no fire-hardened clay carrying the prints of reeds or brushwood. Swiny gives three reasons. The roofs were probably mostly dry soil over brush, which leaves no impressions even when it burns. Beams that could still be used were taken away when houses were abandoned, as he saw in the Kyrenia district after 1974 (Swiny 2003, p. 53). And in a village house at Old Paramali that burned in 1985, the charred beams under some 20 cm of earth left no baked impressions at all (Swiny 2003, p. 57). Across Cyprus such impressions are rare, known from Neolithic Khirokitia to Late Bronze Age Maa-Palaeokastro (Swiny 2003, p. 57).

What Swiny offers instead is the roof of the traditional Cypriot village house, framed as a condition: if the roofs of Kaminoudhia resembled those built up to the 1950s, these are the methods that would have been used (Swiny 2003, p. 57). Round beams span the room, 16 to 30 cm apart and usually about 20. Densely packed reeds or branches go across them at right angles, then a well-trampled bed of brush, thyme or spiny burnet, then 10 to 13 cm of dry earth, sealed with 5 to 7 cm of mud plaster tempered with chaff and finished with a thin skin of tamped clay renewed every year. The whole is 20 to 40 cm thick. Such roofs are called flat, but they fall gently towards one of the long walls so that rain runs off, and Swiny gives no angle. The model’s roofs fall south, down the hill, at 1 in 30, a figure of our own. The Chalcolithic round house on this site carries a roof on the same principle under a cap of havara, the local chalky marl; this one ends in mud and clay of a pale greyish buff. The pitched thatch of some modern pictures of prehistoric villages has no support: we know of no roof tile or thatch from prehistoric Cyprus.

One excavated clue fits this kind of roof, and it is in the house. In the middle of Unit 7, 1.7 m from its east wall, lies a stone, IS10, still in place, with a shallow round hollow pecked into its top; a second stone, IS11, lay nearby, and the two hollows are 12 and 17 cm across (Swiny 2003, p. 23). Swiny reads IS10 as the footing of a post under a crossbeam, which in turn carried shorter timbers resting on the east wall. The model follows him in principle: a heavier girder runs north to south over the post, with short joists on either side. The stone is excavated, the post is inferred, and the layout of the timbers is ours. Other clues are thinner: in Unit 6 a pale-brown soil under the fallen stones “could be roof material”, and a quern found high in the tumble “may have been on the roof” (Swiny 2003, p. 22).

Swiny puts the weight of such a roof at about 500 kg per square metre, “a serious consideration when determining the width of a room and the size and availability of beams” (Swiny 2003, p. 57). He shows no working. A check from his own material: the burnt Paramali house, 3.20 by 9.50 m inside, lost a roof of “some thirteen tons”, about 430 kg on each of its 30.4 m². Adding up the model’s roof layer by layer, with densities we have assumed, gives 36 kg/m² for the beams alone, 48 with the reeds, 54 with the brush, 246 once the earth is on, 348 with the mud plaster and 375 kg/m² for the finished roof, dry. The earth and the plaster make up 78 per cent. A roof soaked by winter rain, or built up towards Swiny’s 40 cm, would approach or pass his figure.

The beams set the size of the rooms. Swiny’s average room is 3.5 m wide; with 40 cm bedded into each wall a beam is about 4.5 m long and, at 8 kg a metre for seasoned cypress, weighs 36 to 40 kg, a load two people can manoeuvre into place and a donkey can carry two of (Swiny 2003, p. 58). The beams he recorded in the old houses of Paramali are round cypress poles 10 to 14 cm across, adzed and slightly pointed at both ends (Swiny 2003, pp. 75–76). The model’s poles are 12 cm across at 22 cm centres across the short way of each room; the longest is 4.4 m. Which wood the villagers used is not known: the one conifer fragment among the charcoal could be pine, cypress or juniper (Hansen 2003, pp. 449–451), and all three grow round the site today.

No estimate of the labour in an Early Cypriot house has been published, so the figures here are our own, worked out by the model from its plan, its heights and ordinary rates for hand work. They cover the four roofed rooms, Unit 6 included, with 3 m walls, and they are orders of magnitude.

The walls come to 71 m³ of masonry, of which 16 m³ survives in the courses found standing and 55 m³ is reconstructed above them. Taken as three parts stone to one of mud, that is about 150 tonnes: some 120 tonnes of limestone in roughly 5,800 slabs of the two published sizes, bedded in 18 m³ of mud mortar that needed something like 5,300 litres of water to mix. None of the stone needed dressing; it only had to be prised from its beds and carried.

The roofs cover 48 m². They take 89 poles, 280 m of timber in all, weighing about 2.4 tonnes with the girder and post of Unit 7; about 580 kg of reeds or thin branches; 5.8 m³ of dry earth, some 9.3 tonnes carried up onto the roof; and 2.9 m³ of chaff-tempered mud. The finished roofs weigh about 19 tonnes dry, on walls nowhere thicker than 70 cm.

Quarrying, carrying and laying the stone, fetching the water and building the roof come to about 460 person-days, of which the walls take about 290 and the beams, felled, trimmed, adzed, carried and set, about 100. A team of ten could do it in about 45 working days. With walls only 1.93 m high the total falls to about 340. Felling and seasoning the timber, and carrying it from wherever it grew, could add a good deal. The one seasonal remark in the excavation report is Swiny’s: a roof of dry earth implies building between spring and autumn (Swiny 2003, p. 57).

The village was abandoned around 2200 BC and never lived in again (Swiny 2003, pp. 53, 66). How it ended is disputed, and the dispute is on record. Swiny argued for a catastrophe: an earthquake, a conclusion Rapp reached independently in his chapter on the geology, and a fire that ran through the settlement (Swiny 2003, pp. 53–54). Walls with coursed tumble fell south, which he took for the direction of the shock, since most earthquakes felt in southern Cyprus start offshore to the south. He added that the village was never reoccupied, that human bones lay in three rooms, that Unit 6 “collapsed with artifacts still in place”, and that in some deposits, in all three areas, up to three-quarters of the chipped stone was burnt, which he thought “hardly … accidental”.

Rapp’s own words are measured: the tumble with its blocks still in their arrangement is “evidence of a possible earthquake” (Rapp 2003, p. 466). Swiny’s chapter itself supplies the complications. WI, the best height evidence in Area A, fell “after the structure was abandoned” (Swiny 2003, p. 11). Several walls, those of Unit 6 among them, had tilted south under the pressure of the slope (Swiny 2003, p. 21), and a slope that pushes walls south will in time bring unroofed ones down in the same direction as a north–south shock; the direction of fall cannot tell the two apart. The south wall of Unit 6 fell north, into the room. Area C, with the most burning, has no coursed tumble at all. Fire is better attested than the earthquake, and a single event across the whole site has not been shown.

The two reviewers of the final report doubted the catastrophe. Webb pointed out that a sudden end should leave a household’s durable equipment in place, yet 80 per cent of the 74 registered vessels are incomplete, which makes “the overall case for catastrophic abandonment difficult to sustain” (Webb 2004, p. 374). Frankel found Swiny’s presentation “designed to suggest a drama involving fire, destruction, and ‘projectile points’”, and offered a plainer reading of the human remains.

Those remains are few. In Unit 6, the room next to the house, a human thigh bone lay on the floor 50 cm from the north wall, beside the base of a Red Polished pot, with half a pelvis 40 cm east of it (Swiny 2003, p. 22). In Unit 16, next door again, three human long bones lay scattered on the rock, and Swiny asked whether they belonged to the same skeleton (Swiny 2003, p. 27). The only complete skeleton is in Area C, in Unit 22: a young woman lying on her right side with her legs drawn up and her hands raised to her face, in the same ashy debris as the floor, one leg over a large bowl that had been mended in antiquity and was found smashed (Swiny 2003, p. 51; Schulte Campbell 2003, p. 433). Swiny wrote that “this does not appear to be a burial”, and that since she did not lie under coursed tumble “seismic activity is not an obvious cause” (Swiny 2003, p. 51); the volume’s chronology chapter nonetheless calls her probably an earthquake victim (Swiny, Rapp and Herscher 2003, p. 505). Webb and Frankel read her as a burial made in an abandoned building, for which Marki offers parallels: seven burials there were made inside the settlement (Webb 2004, p. 375). The count is two people, perhaps three, and none of them is shown to have been killed by falling stone.

The model offers all three endings and takes no side. Its as-found state shows the block as excavated, roofs and upper walls gone and the rooms filled with 40 to 60 cm of jumbled stone, with the south wall of Units 18 and 6 lying where it fell north into them. Only the pattern of the fall is excavated; the cause is a switch. The “Pompeii of the Bronze Age” and the earthquake village with three victims go further than any of this evidence.

The roof is borrowed from houses of the twentieth century AD, some four thousand years later, and its pitch, its edge and any parapet or spout are unrecorded; the vestibule may have had no roof at all. The doors are the other large gap. Their thresholds, their widths and one pivot stone survive, while their leaves, jambs, lintels and heights are reconstruction, and the Unit 6 doorway stands in the south-east corner against the text’s south-west. Windows: we know of none. No wall of the block survives above 90 cm, and the fallen walls that give the heights are described as coursed stone with no opening noted; the model can add one small vent high in the wall of the main room, labelled a guess.

Two-thirds of the back room were never dug, and Swiny’s published areas for Units 7 and 18 are larger than his plans allow; the model follows the plans. Unit 6 has no door into the house, and whether the same family used it is unknown. Nothing in the block itself records how tall it stood.

The one door that can be measured belongs to the neighbour: a single block of limestone 1.32 m long, lying across the doorway of Unit 6 on the floor where it was set, with two pecked patches on its top where the jambs stood, 87 cm apart.