The Earthquake House at Kourion

Daniel was working for the University of Pennsylvania’s mission to Kourion, paid for and run in the field by George McFadden. McFadden’s notice of January 1935 described two skeletons “of a man and a woman, entangled in such a way as to suggest a Romeo and Juliet tragedy”, one wearing “quite a lovely gold ring” (quoted in Soren 1985, p. 56). By 1940 he was calling them two women, and adding the sentence that eventually brought excavators back: “That none returned to retrieve or bury the dead testifies to the magnitude and extent of the earthquake.” The trench was abandoned for the basilica and the theatre and backfilled. Daniel died in 1948 and McFadden drowned in Episkopi Bay in 1953 (Soren 1985, p. 53).

David Soren came to Kourion in 1978 to work at the Sanctuary of Apollo Hylates, west of the city, where the temple’s upper wall seemed to have been sheared off and thrown to the north-east, and the coins in the debris ran up to the joint reign of Valentinian I and Valens and stopped (Soren 1985, pp. 54–55). To test a date he needed untouched earthquake debris in the city, and Daniel’s forgotten trench was the obvious place. Soren found it outlined and labelled “III” on a plan by McFadden’s architect Joseph Last, printed in T. B. Mitford’s The Inscriptions of Kourion (1971), and in June 1984, fifty years almost to the day after Daniel, a University of Arizona team reopened it with a sounding six metres by four (Soren 1985, pp. 52, 56).

By 1987 the sounding had grown to about 673.58 m². It exposed two buildings on either side of a narrow alley: Building A, the Earthquake House, dug completely, and Building B to the north, the so-called “Market”, dug only in part. Daniel’s “House 1” and “House 2” turned out to be one house (Costello 2014, p. 21). The Arizona mission never produced a final report. Its record is a set of preliminary reports, Soren’s 1985 article in Archaeology, and a popular book by Soren and Jamie James, Kourion: The Search for a Lost Roman City (1988), which Costello judges to lack “basic elements required for serious scholarly research” (Costello 2014, p. 24). Costello’s monograph, built from the field notebooks, locus sheets and finds, is the one scholarly study of the house, and the room numbers here are his.

Soren first read the building as an elite residence of the 1st century AD, with a tablinum (a reception room) and a possible triclinium (a dining room), which had come down in the world and been cut up into tenements by the time it fell (Soren & James 1988, pp. 119, 145). Costello looked for the evidence behind those labels and found none: “the evidence does not support the excavator’s interpretation of an upper-scale residence that was in a state of decline” (Costello 2014, p. 42). The house has no mosaics, no painted wall plaster, no apses and no peristyle, and its floors are packed earth. At its largest it covered about 397 m², against about 1,110 m² for the House of the Gladiators, one of the city’s elite houses (Costello 2014, pp. 42–43). Thomas Davis, Soren’s field director, calls it “a second order domestic structure” whose architecture “was not elaborate, but functional” (Davis 2013).

It resembles a Cypriot farmhouse set down inside a Roman city. Costello describes “a fusion of Greco-Roman and traditional Cypriot elements”: a paved court with a small Doric colonnade, and round it rooms of mudbrick on stone footings, sunken earth floors, central posts under the roof beams and a stable (Costello 2014, pp. 31, 42). Soren and James compared its standard of living to a modern farmhouse in the nearby village of Kato Kivides (Soren & James 1988, p. 143). Costello knows of no close parallel for the plan; the nearest are the Patrician House at Meiron, about half the size, with a paved court and an L-shaped stair, and the 6th- and 7th-century houses at Kalavasos-Kopetra and Agios Kononas.

The house is often called the Pompeii of Cyprus, and Soren and James drew the useful distinction themselves. Vesuvius gave warning and many people fled. At Kourion “death came without warning”, and the people found were caught doing whatever they were doing (Soren & James 1988, pp. 118–119). That is what the comparison is good for: a house whose contents were buried where they stood and never picked over. The comparison stops there. On its last day the Earthquake House was old, patched and subdivided, its colonnade already in pieces.

Nobody knows when the house was built. Digging below the late 4th-century levels was limited to small soundings, and Costello declines to give a construction date (Costello 2014, pp. 32–33). A sounding under the floor of Room 1 produced material that Soren dated “sometime between the later first and the mid-second century” (Soren 1985, p. 57), so the house can be no older than that.

The first house stood around a court, Room 15, paved with squared limestone blocks, some reused with cuttings from an earlier life. Along its south side ran a portico of six unfluted Doric columns on a low stylobate, their spacing “noticeably irregular”, and the covered walk turned south as Room 18. The main door opened from the alley on the north. Off the court lay Room 1, a single room of 3.25 by 6.25 m; beyond it Room 8, a hall of 7.65 by 6.40 m and the largest roofed space in the house; and on the west a kitchen and store (Room 14), a cistern (Room 12A) and a cupboard under the stair (Room 19), with the stair itself in the court’s north-west corner (Costello 2014, pp. 33–37).

Two additions followed, in an order no one can fix (Costello 2014, p. 37). Addition A pushed the house west to enclose Room 3 and two small rooms, 6 and 7. Addition B enclosed a yard of about 103 m² to the south-east, Room 11, with a thinner wall that abuts the old one without bonding to it (Costello 2014, p. 39).

The rest belongs to the house’s last years. A crude wall, bonded to nothing and probably never full height, cut Room 1 in two, and its western half, Room 2, became a stable with a limestone trough 2.35 m long set across it. The trough weighs “in excess of 350 kg”, has tethering holes worn right through its sides, and was probably salvaged, “possibly from a public location” (Costello 2014, p. 40 n. 242). The door from Room 1 into the hall was blocked, and so was the door from the hall into Room 7, faced smooth on the Room 7 side and rough on the hall side, so closed from the west (Costello 2014, p. 40). The house split into wings, and Rooms 6 and 7 were left with no known way in; no other doorway has been found in their walls.

The alley outside the front door was walled off to make Room 20, later partitioned into Rooms 20 and 25, and the main way in probably moved to a door on the east (Costello 2014, p. 41). The yard’s ground rose about 68 cm through dumped fill, and a lean-to and a large oven went up in it. The cistern and the cupboard under the stair filled with rubbish. By then the colonnade was down: its drums and a capital lay stacked in a corner of the court, where a shallow basin was built partly out of them (Costello 2014, p. 41). The north wall of Room 1 had been rebuilt with reused ashlars and a column drum, and the kitchen walls were patched with broken amphorae, which Costello reads as repairs after an earlier earthquake (Costello 2014, pp. 36, 40). Soren and James put that shock twenty to thirty-five years before the last, naming the Salamis earthquakes of 332 and 342 as candidates (Soren & James 1988, p. 171).

The subdivided house held at least two households and possibly three: the better-off family of Rooms 1 to 3 and the rooms above them; a poorer group in Rooms 20 and 25, perhaps tenants; and perhaps a separate occupant of the hall and yard. Rooms 6 and 7 may have belonged to a neighbour (Costello 2014, pp. 97, 106).

Every wall on the plan was found as a footing 30 to 60 cm thick, of dressed and undressed limestone in irregular courses bedded in mud mortar, the gaps packed with limestone chips and now and then potsherds (Costello 2014, p. 31). The footings survive a few courses high, and no wall stands higher.

Above them the walls were of mudbrick. “A significant amount of decayed mudbrick” filled the rooms, and neither the fallen stone nor the brick would have been enough on its own to take the walls to full height (Costello 2014, p. 31). Cypriots had built mudbrick on stone bases since the Neolithic at Khirokitia, and went on doing so into the 20th century (Costello 2014, p. 31 n. 202). The exception was the cistern, “highly likely” of stone to its full height because it was always wet; its hydraulic plaster survives 1.79 m high (Costello 2014, pp. 36–37).

Almost anything solid went into the walls: column pieces, querns, small basins and tabletops. Inside the house an inverted pilaster capital served as the step down into the hall and an upside-down column base as a grinding bench (Costello 2014, pp. 31, 34–36). The walls were finished in mud plaster with a few patches of lime, and “none of the evidence indicates that any wall of the structure was decorated with painted plaster” (Costello 2014, p. 32). Soren and James did report a painted frieze in the yard (Soren & James 1988, p. 146), but Costello records painted fragments only in debris dumped after the earthquake (Costello 2014, p. 51), and they most likely came from a grander building nearby. Floors were beaten earth, often well below the thresholds: 36 cm in Room 1, up to 43 cm in the hall and 73 cm in the kitchen, with a stone block inside the door as a step (Costello 2014, pp. 32, 34–36).

Soren’s published reconstructions show the whole house in stone to its full height, though he acknowledged the mudbrick in his text (Costello 2014, p. 31 n. 201). The model offers that version as a rejected alternative. In its own version the stone footing rises 0.50 m above the court paving, and mudbrick of a reconstructed module, 40 by 40 by 10 cm, takes the walls to their heads; no brick size from the house has been published. Built that way the house needs about 6,600 bricks, and the finished shell weighs some 510 tonnes: roughly 950 person-days of work for a single campaign, leaving out the decades of additions and repairs the plan records.

Most of the house was roofed in earth. Costello describes one or more large beams running the length of a room with their ends set in the walls, smaller beams across them, then branches, sticks, reeds and matting, and packed mud on top, “flat or gently sloping” (Costello 2014, p. 32). A piece of fallen roof plaster still carries the impression of reeds (Soren 1985, p. 57). No thicknesses were recorded; the model’s 34 cm above the joists follows roofs built at the Lemba experimental village, a modern trial that counts only as a hypothesis. No gutters or drains were found anywhere in the house.

Room 8 was the only room with a tiled roof when the house fell, and Schiffer’s count of tile corners showed it had its full complement (Costello 2014, p. 34; Soren & James 1988, p. 141). At its centre stood a post of reused parts: six flat limestone slabs about 52 cm square, stacked to 1.14 m, and on them two column drums, 0.96 m and 1.14 m tall and 42.5 cm across, 3.24 m in all (Costello 2014, pp. 34–35). Posts like this carry the main roof beam in traditional Cypriot houses, often under the join between two lengths of timber. A squared block by the south wall may be the base of a second prop, perhaps added because insects had weakened the beams (Costello 2014, n. 214).

The roof probably had a north–south ridge and two slopes at about 22.5°, an angle inferred from the number of tiles and from impressions of overlapping tiles in the plaster that bonded them (Costello 2014, p. 35 n. 215). About 250 iron nails lay in the room, so the timbers and perhaps some flat pan tiles (tegulae) were nailed. The cover tiles (imbrices) were of the Corinthian type, with an angled section, “cemented to the underlying tegulae with lime plaster containing small river gravel aggregate” (Costello 2014, p. 35 n. 216). No tile size has been published, so the model borrows the large Corinthian pans of 56 by 48 cm found at the Late Roman village of Kalavasos-Kopetra (Rautman 2003): Schiffer’s rate of four tiles to the square metre fits that size, and rules out Kopetra’s smaller 44 cm pan. The roof then takes about 290 tiles.

The post fixes the height of the roof, and it can be read two ways. If it carried the ridge, the eaves would come down to under 2 m above the hall floor, lower than the tops of the doorways in the side walls. The model therefore puts the post under a main beam at wall-plate level, 2.85 m above the paving, with rafters rising to a ridge at about 4.25 m. That matches Costello’s description of the tiled roof on the same kind of framework as the earth roofs, a main beam carrying cross-beams, and the 2.85 m plate becomes the height of every single-storey wall in the house. The other reading is drawn among the roof alternatives, flagged as not favoured.

Rooms 6 and 7 had about 10 and 14 tiles, enough for two or three square metres each, which Costello reads as patches set into earth roofs (Costello 2014, p. 39 n. 234). The hall’s west slope shed its water towards Wall K, the wall it shared with them, so the roof over Rooms 6 and 7 must have fallen the same way, east to west, to carry that water off the house (Costello 2014, p. 39). The published reconstructions tiled these rooms too, meeting the hall roof in a valley over Wall K; the architect R. Zebrowski objected that such a valley would pour both roofs’ runoff onto a mudbrick wall (Costello 2014, p. 39 n. 235). That version appears here as rejected. The portico had no tiles: its roof was probably timber and mud, and it had come down with the columns before the end (Costello 2014, p. 34).

In the north-west corner of the court “four well-built stone steps rise westward to a nearly square landing that provides access to Room 14. From here the staircase turns roughly 90° southward with four stairs continuing up” (Costello 2014, p. 34). Beyond the eighth step the stair “continued in wood” over the cupboard, Room 19, carried on the cupboard’s east wall (Costello 2014, p. 37). The closest parallels, at Meiron, at Agios Kononas and in a Justinianic house at Pella, have “virtually identical” stairs, and all of them led to an upper storey (Costello 2014, pp. 34, 89).

Nobody found that storey, because it was built of timber and mudbrick and fell into the rooms below. Costello’s case rests on where things lay (Costello 2014, p. 89). In Rooms 1 to 3 the coins and the best objects were found high in the collapse, above the floors: a bronze lampstand whose shaft is a spiral-fluted column on three dolphin heads, a marble tabletop, a silver-sheathed bone hairpin. In Room 2, 170 coins lay with a glass jar in a spread in front of the door to the court, beginning some 85 cm above the bottom of the deposit (Costello 2014, pp. 58–59). The girl found in the stable lay on top of the mule. If the upper floor’s beams were as insect-eaten as the hall’s, it would have given way towards its middle and dropped its load, and its people, onto whatever stood below.

The stable points the same way. In traditional Cypriot houses the living rooms lie “close to, and normally above, those sheltering animals so that their radiant heat can provide warmth” (Costello 2014, p. 89 n. 383), and the excavators heard of the same arrangement at the old village of Paramali near Akrotiri (Soren 1985, p. 57). Costello concludes that the evidence “does not absolutely prove the existence of an upper floor”, but together makes “a compelling circumstantial case for one above Room 1, Room 2, and Room 3” (Costello 2014, p. 89).

The model follows him and offers two other readings: Soren’s early view of Room 2 as a small courtyard open to the sky or half roofed, because few tiles lay in it (Soren 1985, p. 58; Soren & James 1988, p. 114), and a plain single-storey house. No storey or riser height has been published, so the model derives them from the stair: eight stone risers of 20 cm reach 1.60 m, seven timber risers of about 21 cm reach an upper floor at 3.10 m, and the upper walls stop at 5.50 m under a flat earth roof. No windows are modelled. With the walls surviving only as footings we know of none, and the evidence is silent either way.

Costello names rooms by what was found in them, and the model does the same.

The hall, Room 8, was where work and meals happened. West of the post stood a stone table 86 by 80 cm, weighing about 110 kg, on three limestone legs and probably a fourth of wood (Costello 2014, p. 35). In the north-east corner an upturned column base carried a dark gabbro grinder, stone from the Troodos mountains, with a low seat beside it (Costello 2014, p. 36). There were copper-alloy pitchers and a basin, probably fallen from shelves, a spearhead, a four-tined iron fork, a lead cup, a hanging lamp with seven wicks (Soren & James 1988, pp. 141–142), a pair of proportional dividers lying on a patch of laid sand, and the jawbone of a cat.

The kitchen, Room 14, was also the main store: a domed mudbrick oven took up about 15 per cent of its floor, beside a pithos of about 151 litres and amphorae. Room 6 was a decanting room, with a funnel, a deliberately holed amphora and ten vessels holding some 370 litres between them. Room 7 held 19 amphorae, many cut down to necks and bases; the household reused broken amphorae, and their sherds turn up as chinking in the kitchen walls (Costello 2014, pp. 91–92). Room 1 had a small pithos, a copper-alloy jug and a lidded cooking pot on the floor. Room 2 held the trough, the mule’s iron chain and its tack, fallen from the wall where it had hung (Costello 2014, p. 58). Room 20, where three people slept, held a glass juglet, an amphora and nothing of value.

Some things were missing. There were no loom weights or other textile tools, although at least three of the dead were female (Costello 2014, p. 103). There was no household shrine; a ring and a bell are the only religious objects (Costello 2014, p. 104). Querns appear only as building stone, so bulk grain was ground somewhere else (Costello 2014, p. 94). And no wall was painted.

Soren’s team read the destruction from the layers in the hall, and their account has three waves (Soren & James 1988, pp. 143–144). The first brought the tiles down, “sliding them down into the room as the central column buckled”, and toppled the top courses of mudbrick. After a pause of three to five seconds a second wave did most of the damage, and a weaker third followed. The seismologist Terry Wallace put figures on it: about four seconds at 0.1 g for the first wave, which people could survive; about ten seconds at 0.35 g for the second, “the killer”; five seconds for the third; about 7.25 in magnitude overall. Those figures are Wallace’s estimates, modelled from an assumed epicentre, and they rest on one team’s reading of one room. The geologist Reuben Bullard believed the third wave had rippled the stable’s wall and floor and cracked the trough. “Wallace, however, is not convinced by this model of the tertiary wave’s effects” (Soren & James 1988, p. 144). The model shows the first two waves as inferred and the third as disputed.

The debris supports the outline. The fallen tiles in the hall formed a single layer, with gaps “especially around the perimeter” (Costello 2014, p. 47), as they would if the roof sagged at its centre and the tiles slid inwards. Walls tended to fall to the north-east, though “not every piece fell to the northeast”, and part of the hall’s south-east wall “just gave way” (Soren & James 1988, p. 144). At some point the upper floor over Rooms 1 to 3 came down and spilled its coins, lamp and tabletop into the rooms below.

Two of the best-known details have soberer readings. Soren described the stable trough as hurled against the wall and snapped “like a candy cane” (Soren & James 1988, p. 114). The field report Costello cites says it stood on two rough ashlar blocks, “one of which cracked during the earthquake” (Costello 2014, p. 40 n. 242). Soren also described a threshold block of more than 100 pounds “ripped loose” and turned through ninety degrees (Soren & James 1988, p. 113). Costello found 10 to 15 cm of fill between that block and the threshold beneath it, so it is more likely a door jamb that fell after the fill had built up (Costello 2014, pp. 37–38).

The posture of the three people in Room 20 “may provide credence” to a quake at night or early in the morning (Costello 2014, n. 399). Afterwards no one dug; Costello found no sign that rescuers or looters ever disturbed the deposits (Costello 2014, p. 44). Wind laid yellow soil over the ruins, and when Kourion was reoccupied in the early 5th century, as a smaller Christian town, its people used the house to dump rubbish (Costello 2014, pp. 15, 51).

Costello counts seven people in the house: two adults in Room 1, a child in Room 2, a man in the yard and three people in Room 20 (Costello 2014, pp. 97, 106). The model marks where each was found with a small numbered pin and shows no bodies. The Room 20 group is on display in the Kourion museum at Episkopi; the child and the mule are in store.

The two adults in Room 1 are Daniel’s. His photograph shows one complete skeleton crouched on its side, arms across the chest and hands over the face, and beside it only the legs and feet of a second, whose upper body ran into the unexcavated baulk (Soren 1985, p. 56). McFadden’s “Romeo and Juliet” stuck as the excavators’ nickname for them, until Soren’s verdict on the photograph: “there never was any Romeo and Juliet death embrace” (Soren & James 1988; Eyewitness to Discovery, p. 217). The bones are lost and their sexes unknown. One of them wore a gold ring, also now lost, on the same finger as a copper-alloy ring (Costello 2014, p. 110).

The child in Room 2 was about thirteen by her teeth, which also suggest she was a girl, though her bones were the size of a modern ten- or eleven-year-old’s. She lay face down over the hindquarters of the mule, her head turned to the south-west, with the joining piece of a bone hairpin just east of her (Soren 1985, p. 58). The team called her “Camelia”. Soren and James told her last minutes as a story, woken by the mule, going out to calm it, the jar of coins falling beside her, and framed it themselves with “If our conclusions are right” (Soren & James 1988, pp. 114–115). They thought she was climbing out through the rubble of the first wave when the second struck. Costello points out that a few seconds leaves little time for that, and that a body lying on top of the mule “strongly suggests” she came down with the upper floor (Costello 2014, p. 89 n. 387). The mule was still tethered to the trough by its iron chain.

The man in the yard was in his fifties, heavily built, with the arthritic back of someone who did hard physical work. He died in a doorway, caught between fallen ashlars, and scavengers, the excavators thought, later carried off his lower body (Soren & James 1988, p. 145). Which doorway has not been published, so his pin is approximate.

In Room 20 a young woman of about nineteen, 1.44 m tall, lay in the north-west corner holding an infant of about eighteen months to her breast and face. A man about 1.60 m tall lay behind her, his head at her shoulder and his left hand at the child’s back. Soren and James describe three painted arched blocks lying on her (Soren & James 1988, pp. 165–167), but Costello places the carved blocks and painted fragments in a dump laid down after the earthquake, separated from the bodies by a sterile layer (Costello 2014, p. 51). By the man’s hand lay an iron ring and a copper-alloy ring engraved with the chi-rho monogram of Christ and an alpha; Soren and James also read an omega, which Costello could not see (Costello 2014, pp. 80–81). These three became the excavators’ “Christian family”. The ring is a Christian object. That they were a family is an inference from their posture, and Soren and James allowed that it is “theoretically possible that they are all complete strangers” (Soren & James 1988, p. 166). In the same room lay a tintinnabulum, a hanging bell meant to ward off harm, perhaps hung over the child (Costello 2014, p. 104).

Outside the house, in 1987, the excavators found a man in his early twenties killed by a fall of roof tiles (Soren & James 1988, p. 170). Soren’s figure of 500 dead for the city is a guess scaled up from the bodies found (Soren & James 1988, p. 171). We know of no ancient DNA or isotope study of any of these people.

Soren’s date rests on coins. The 1984 season alone produced 365 copper-alloy coins, none later than the early reign of Valens, with no coins of Gratian, struck from 367, and none of the usurper Procopius, struck from September 365 (Soren 1985, pp. 55–56, 59). The numismatist Eugene Lane counted 40 coins of Valens, 33 of an early type with the emperor’s name split across the portrait and 7 unsplit, and concluded that “a date in mid-365 seems dictated by the numismatic evidence alone” (Soren & James 1988, p. 112). Soren’s first range, in 1981, was 365 to 372. He narrowed it to the great earthquake of 21 July 365, described by Ammianus Marcellinus, after the seismologist Nicholas Ambraseys suggested the link (Soren 1985, pp. 54–55).

The coins fix one thing: the house fell after about 364. The single day is much less secure. Costello did not re-examine the coins; Lane’s identifications were never published, and his notes could not be found (Costello 2014, p. 27). The one late coin in Costello’s catalogue with a secure findspot is Daniel’s, “5 cm above one of the skeletons” in Room 1, an early issue of Valens (Costello 2014, p. 111). A coin gives only the earliest date its context can have, and issues from distant mints could take years to reach Cyprus, which weakens the ratio of split to unsplit legends as a clock (Costello 2014, p. 15 n. 104).

Geography weighs against 365 as well. That earthquake ruptured off western Crete, throwing down Gortyn, Eleutherna and Kisamos, about 860 km from Kourion. At that distance the shaking would have done little damage, if any; what Ammianus describes further east was the tsunami, and Kourion stood high on its acropolis with no commercial harbour (Costello 2014, p. 15 nn. 105–107). Emanuela Guidoboni’s catalogue of ancient earthquakes puts the Cypriot destruction at about 370, from Jerome’s Life of Hilarion; Ambraseys later preferred about 375. Costello’s verdict is that no “smoking gun” exists and that the event happened “sometime during the later 4th century AD, possibly ca. AD 370–380” (Costello 2014, p. 15). The later 4th century is secure. The year is not, and the model’s collapse carries no date. The island’s wider run of earthquakes is set out under earthquakes in ancient Cyprus, and the period under Roman Cyprus.

The plan comes from Costello’s schematic phase plan, whose scale bar proved about 5 per cent short when checked against the excavators’ state plan; the model corrects for that, and wall positions are good to perhaps 20 or 30 cm. Where plan and text disagree the model follows the plan: the court measures 4.1 m north to south in the text and about 5.3 m on the plan, and the kitchen 3.14 or 3.16 m against about 2.8 m. The published plans put north at the top, but their own arrow shows that this “north” points roughly north-east, about 40° off, and whether the arrow marks true or magnetic north is not stated. The model keeps the plan’s axes so that its sections cut square across the rooms, and the drawings carry a rotated north arrow.

No wall height, brick size, storey height or riser height has been published. All of them here are derived: single-storey walls from the hall roof, the upper floor from the stone stair, the brick from a reconstructed module. In the house as first built the portico columns stand at an irregular spacing of the model’s choosing, since the positions were never published, and no taller than 2.76 m, the most the surviving drums allow (Costello 2014, n. 207). The mud plaster is a pale yellow-buff, the colour of the local soil, without the terra-rossa red of modern replicas; the tiles are a pale pink-buff borrowed from Kopetra; the timber is sawn and nailed. Building B across the alley is a plain block, because only part of it has been dug.

Some questions stay open in the model because they are open on the ground: how anyone reached Rooms 6 and 7 after their door was blocked, whether a gap in the north wall of Room 20 was a doorway, and in which doorway the man in the yard died. The tour’s second-to-last stop shows the house as the excavators left it: footings, floors, the trough, the post of drums and slabs in the hall, and seven small pins.