The Transition to Late Antiquity on the lower Danube: Excavations and survey at Dichin, a Late Roman to early Byzantine Fort and a Roman aqueduct

By Andrew Poulter

Excavations on the site of this remarkable fort in northern Bulgaria (1996–2005) formed part of a long-term program of excavation and intensive field survey, aimed at tracing the economic as well as physical changes which mark the transition from the Roman Empire to the Middle Ages, a program that commenced with the excavation and full publication of the early Byzantine fortress/city of Nicopolis ad Istrum.

The analysis of well-dated finds and their full publication provides a unique database for the late Roman period in the Balkans; they include metal-work, pottery (local and imported fine ware), glass, copper alloy finds, inscriptions and dipinti (on amphorae), as well as quantified environmental reports on animal, birds, and fish with specialist reports on the archaeobotanical material, glass analysis, and querns.

The report also details the results of site-specific intensive survey, a new method developed for use in the rich farmland of the central Balkans. In addition, there is a detailed report on a most remarkable and well-preserved aqueduct, which employed the largest siphon ever discovered in the Roman Empire.

This publication will provide a substantial database of material and environmental finds, an invaluable resource for the region and for the Roman Empire: material invaluable for studies, which seeks to place the late Roman urban and military identity within its regional and extra-regional economic setting.

• Language: English
• Publisher: Oxbow Books
• Release date: Sep 30, 2019
• ISBN: 9781785709593

Andrew Poulter

Andrew Poulter is Emeritus Professor at the Department of Archaeology, University of Nottingham. He has led several major research projects including Nicopolis ad Istrum, the first British research excavation carried out in Eastern Europe (1985-1992); The Macedonia Research Programme; and the excavations at Dichin. He is also the author of a series of substantial research reports resulting from these projects as well as books and articles on Roman and Byzantine archaeology.

Book preview

1. Area C, the east gate

Andrew Poulter

Introduction

The location of the east gate was readily identifiable before excavations commenced. From the base of the hill a track-way, revetted into the side of the steep eastern slope, ascended to the summit where it met the west/east street (street 1). In the first year of fieldwork (1996), an area 12m north/south and 8.0m east/west was laid out and excavation began. In 1997, excavation continued through the Period 3 occupation level down to the Period 1 prehistoric soil. In 1998, the area was extended south by 4.0m to investigate the upstanding remains of the curtain-wall and the ascensus which was found to abut its northern, internal face. Work in this area was concluded after a final redrawing of the sections in 1999.

Period 1

A black/brown humic clay deposit (52097) extended across the entire area and was cut by the foundations for the curtain-wall (Fig. 1.1). Its contents included copious fragments of late Bronze Age and early Iron Age pottery, which suggests that the soil had been disturbed before the fort was built, presumably by cultivation. With the possible exception of the hand-made spindle whorl (SF720), metal finds recorded as coming from this deposit were probably intrusive and came from the dump deposits which directly overlay it.¹

Uniquely, the lowest level of the Period 3 demolition deposit (52132 and 52127) contained, not only residual prehistoric sherds, but also crushed Roman red wares and brick fragments (Fig. 1.1). Too small to be closely dated, the pottery must be earlier than 400 but is most probably 2nd/3rd century in date.² Moreover, within the core of the tower wall, broken brick fragments were reused; surely salvaged from a Roman occupation deposit (Fig. 1.2).³ Within the gateway a massive limestone block (52016) must have come from a monumental Roman structure (Fig. 1.8). It measured 1.36m in length and was square in profile (0.66m × 0.66m), cut by two Lewis holes for raising the stone into position. Together, these finds indicate that the site had been occupied in the Roman Period, before the fort was built, but that its use had been limited; no structural remains in situ, predating the late Roman fort, have been found during the excavations.⁴

Period 2

Excavation uncovered the surviving remains of a section of curtain-wall (52040), the gateway together with the upstanding portion of its northern side (52035), which were of one build with the southern side of a rectangular tower (Figs 1.3 and 1.4). To the south of the gate, the curtain-wall was abutted, on its inside face, by a staircase (ascensus) (52039), providing access to the wall-walk.

South of the gate, the upstanding portion of the curtain-wall (52040), 0.90m high and 1.80m wide, had a solid core of sandstone blocks, bonded with white mortar and stood four courses high, faced with regularly shaped sandstone blocks (each c. 0.40m long and c. 0.10m wide), above a deep foundation of mortar and roughly cut blocks of sandstone (Fig. 1.2).⁵ A rectangular extension to the foundation (1.0m × 0.80m) was bonded with the curtain-wall, projecting to the west and inside the fortified area (Fig. 1.3). The function of this addition, part of the original design for the wall, must have been to anchor the curtain more firmly into the side of the hill, above the precipitous slope which plunged sharply down to the base of the hill, immediately to the east of the gate (Fig. 0.3).⁶

Abutting the inside face of the wall (and built around the buttress), a second portion of superstructure (52039), still stood 0.80m above its foundations, its inner western side faced with regularly spaced sandstone ashlars, bonded with the same white mortar which was used in the construction of the curtain-wall (Fig. 1.5). The foundation, which was 1.70m wide, was traced for 5m before it, and the remains of the curtain-wall (52040), entered the baulk, continuing beyond the southern limit of the excavation (Fig. 1.3). Its foundation trench cut the Period 1 soil and the internal mortared face butted close up against the western side of the cut. This structure supported a staircase (ascensus), no doubt built immediately after the erection of the curtain (52040), and as part of the original design. The location of the first three steps is clearly visible at its northern end leading up from the gate entrance (Figs 1.3, 1.4 and 1.5). Each step must have been formed by placing stone blocks upon the mortared substructure.⁷ Because the full length of the stairway was not excavated, it is not possible here to determine the full height of the curtain. Where the full length of foundations for another ascensus were identified on the western side of the circuit (area F), the slabs which formed the steps were in situ and the angle of ascent could be estimated, proving that the wall-walk must have been c. 9.50m above the contemporary ground surface.⁸ The same was probably true for the curtain-wall on either side of the gate.

Fig. 1.1A. The west section, Area C.

Fig. 1.1B. Interpretation of the western section, area C. 1, Soil level, Period 1. 2, The mudbrick demolition deposit (Period 3). 3, Robber-trenches. 4, Make-up deposits for Period 3 surface. 5, Rectangular pit cut during the reconstruction of the site (Period 3). 6, Occupation level covered by robbing material (Period 4).

Fig. 1.2. Elevation across the surviving, south side of the tower, Area C.

The curtain-wall foundation continued north beneath the gateway and bonded with a short section of surviving curtain (52035) which was of one build with the southern side of a rectangular tower (Fig. 1.3). Within the gateway, responds, on each side of its outer face, must have supported the springing of an arch over the entrance and would have protected the pivot sockets for two wings of a timber door, narrowing the eastern, outer entrance to 1.70m. To the rear, the full width of the gate-chamber measured 2.40m.

The back wall of the tower (52045) was only 1.20m wide, appreciably narrower than the southern wall (52035) which was 1.75m thick (Fig. 1.3). Although completely robbed out, it is likely that the eastern and northern sides of the tower matched the strength of the preserved southern tower wall which was 1.80m wide (Fig. 1.2).⁹ It presumably had at least two upper floors; its foundations were in excess of 1.70m deep.¹⁰ Although the tower walls had been completely robbed on its northern and eastern sides, their location was defined by the deep robber-trench which had removed all but the stub of its southern wall. The tower must have measured c. 5m north/south and c. 5.5m west/east. Both in detail (the narrower inside wall) and in plan, the tower mirrored the design of another tower (area E) on the western side of the defences; it also straddled the curtain-wall, projecting both beyond the defences and extending back into the fortified area. The internal clay make-up for the floor of the tower (52043) survived intact, covered by a partially preserved mortar layer. The tower internally measured 2.70m north/south and 2.0m west/east. Two flat stones lay on this compacted surface: proof that the tower originally had a slabbed floor, although most of the stones had been removed before the area was back-filled with demolition material at the beginning of Period 3 (Fig. 1.3).¹¹ The southern side of the tower stood to a height of 0.70m and was internally plastered and rendered in imitation of ashlar masonry.

Evidently, the gate, curtain-wall, the ascensus and the tower were all built as parts of the same original plan. To the south of the area, a robber-trench follows the line of the curtain-wall as far as the south-eastern corner tower (Fig. 0.3).¹² To the north, the curtain wall must have continued slightly below the modern crest of the hill, almost as far as the north-eastern corner, where a short length of robber-trench on the edge of the summit denotes its course.¹³

There was a drain (52065) beneath the gate, built from mortared blocks (0.22–0.45m in length and 0.25m thick). It stood two courses high (0.15m) and was 0.25m in width (Fig. 1.3). Immediately inside the gate, two limestone slabs (52041 and 52066) remained in situ, covering the channel. East of the gate, another section (52065) survived where it changed direction and headed north-east past the tower (Fig. 1.3). No other traces of the drain were excavated to the west, although, since it certainly functioned in Period 2, it must have continued up the slope and into the site.

Period 3

West of the ascensus, the relationship between Period 1 and the overlying Period 3 demolition deposit was particularly clear (Fig. 1.1). However, unexpectedly, there was no sign of a primary occupation surface for Period 2. Instead, the Period 1 soil formed an even but uncompacted level, immediately overlain by the demolition deposit of mudbrick (52092), dumped behind the ascensus at the beginning of Period 3. The same lack of a compacted surface was true for the northern half of the area. In the Period 3 cobbled street (52025), as well as one in situ slab (1.20 × 0.80m) covering the drain (52041), there were other flat stones (0.40 × 0.60m, 0.50 × 0.50), certainly road slabs, which were reused in the later cobbled surface.¹⁴ Within the primary demolition deposit, at the northern end of the area, a pile of broken limestone slabs had been dumped as part of the levelling process; one had a moulding which indicates that it had formerly been part of a Roman building. There were also square paving stones, up to 0.40m in size. This was followed by the dumping of more stones, each up to 1.0m in length; they did not form a surface but had been tipped into the demolition deposit, above the primary levelling, and before the subsequent layer of debris (52075) was deposited (Fig. 1.1). It is reasonable to presume that these paving stones must have been removed during the process of rebuilding at the beginning of Period 3. It seems likely that both the area immediately west of the ascensus and the roadway entering the fort had been paved originally with limestone slabs as well as the floor of the tower. Outside the gate, the road must also have been covered with road slabs; at least the mortared section of drain would have required a stone cover, even if the ascent down the hill was not flagged.

Fig. 1.3. Plan of the defences (Period 2), Area C.

Fig. 1.4. Photogrammetric photograph of the defences, Area C.

Two west/east robber-trenches (52117 and 52116), cut from the modern turf line, divided the successive levels of dump deposits into three sections, each of which reflected somewhat different processes of deposition (Figs 1.1 and 1.8). West of the ascensus, at the southern end of the area, the deposit of orange burnt mudbrick (52092), 1.30m deep, contained occasional lenses of charcoal, complete burnt mudbricks and frequently roof-tile fragments. This remarkably homogenous deposit must have been dumped and compressed as part of a single operation; it was largely uncontaminated by occupation debris and produced very few finds.¹⁵ One notable exception was a complete jar, of cylindrical form, perhaps intended for metal-working, although there was no trace of burning inside; perhaps it was in store and had never been used.¹⁶ The surface of the Period 1 soil (52097) was scorched and, since it immediately underlay the dump deposit of burnt mudbrick, it follows that the Period 3 levelling had commenced when the remains of Period 2 buildings were still hot, but, apparently, after the removal of flagstones which, as suggested above, probably covered the primary ground surface (Fig. 1.1). It would appear to represent the demolished remains of walls of a burnt Period 2 structure, probably close by, perhaps from the wall represented by the adjoining west/east robber-trench (52117).¹⁷ The flat, even top of this dumped material had been carefully levelled. Immediately above, and though deposited in a distinct operation, a thin layer of mortar was visible in both the southern and western sections, spreading north for 1.90m from the south-west corner of the area. It was also visible in the southern section, extending east from the south-west corner of the area for 1.60m. It may have formed during the mixing of lime mortar, needed for the repair the defences. Above it, there was a layer of grey soil (52079) which, though part of the same dumping processes, evidently involved the collection of soil from elsewhere on the site (Fig. 1.1).¹⁸ Immediately south of the west/east robber-trench (52117), a soil-bonded wall (52006), 0.5m wide, stood to a height of 0.75m (Figs 1.1 and 1.6). Its foundations rested firmly on top of the mudbrick demolition deposit. Its rougher western face was revetted into the grey topsoil (52079). It also bordered the south side of the street (52025) and may have been built to prevent the soft soil dump (52079) from slumping north against the wall of the Period 3 building, represented by the west/east robber-trench (52117). Only a short length (2.0m) was visible before it disappeared into the western section; at its eastern end, it had been truncated by robbing in Period 4 (Fig. 1.8).

Fig. 1.5. Elevation of the inner face of the ascensus and curtain wall, Area C.

Fig. 1.6. Elevation of the earth-bonded stone wall, Period 3, area C.

Fig. 1.7. Plan of the pit, underlying the final levelling dump (Period 3), Area C.

Immediately to the west of the revetment wall, and just below the modern turf line, a structure (52071) was found, cut through the underlying make-up deposit (52079) and into the underlying demolition level (52092). It comprised two parallel soil-bonded stone walls, both of which were reinforced with clay; it was 0.60m wide and stood 0.40m high. At its eastern end, the channel had been removed by Period 4 robbing within the central part of the area (Fig. 1.8). The upper surface of the stones was burnt red and the space between the two walls was filled with ash which also spread out around the structure. It was evidently the flue of an oven. The absence of any distinctive finds leaves its purpose uncertain. Unless those who used the structure were very poor cooks, it is unlikely to have been used for baking bread. Although no traces of slag were found in or close to the flue, it seems most likely that the oven was used for an industrial purpose; the recycling of copperalloy material is a possible function and this would have left fewer traces than if it was used for iron-working or the recycling of glass.¹⁹ Above, a silty grey soil (52001) formed a build-up, which post-dated the abandonment of the site and merged with the modern turf line. It was largely devoid of small-finds, except for some small fragments of burnt mudbrick (Fig. 1.1).

Fig. 1.8. Occupation level (Period 3), Area C.

Between the two deep robber-trenches (52117, 52116), the sequence was somewhat more complex (Fig. 1.1). Layers of silty clay (52127, 52093, 52132, 52125), containing small burnt mudbrick fragments, bone, slag, glass and tile fragments, angular stones, sherds of pottery, charcoal flecks and river-worn cobbles which formed successive dump deposits, notable, not only for the general absence of mudbrick, but also for the large number of small-finds. The lowest deposit contained flint blades, probably coming from the underlying Period 1 soil.²⁰ Cutting through these layers was a vertical-sided, rectangular pit (52103), with shallow postholes at both its southern and northern corners (Fig. 1.7). It was 0.90m deep and 2.40m north/south. Its width could not be ascertained as it continued into the western section (Fig. 1.1).

Since, as described below, it was filled with a similar dump material to that through which it was cut and was itself sealed by a subsequent levelling deposit, its use must have been short-lived, occurring during the process of dumping and levelling up for the Period 3 occupation surface. It was filled with successive layers, containing fragments of mudbrick, pottery, stones, ashy silt, charcoal, angular stone fragments and river-worn cobbles, as well as a range of small-finds, no different in composition from those retrieved from the deposits which had been dumped before it had been cut.²¹ A final layer of sandy silt (52063) sealed the top of this feature (Fig. 1.1). North of the robber-trench (52116), the sequence was similar to those already described. The lowest visible context (52095) was a level layer of clay (probably identical to 52132, south of the robber trench (52116), producing pottery sherds, bone, tile and slag, followed by another dump of clay (52089) with limestone blocks, burnt mudbrick and a variety of finds.²² Unusually, the next level (52139) was represented by a short stretch of small, compacted stones and tile fragments, all c. 80mm thick; it resembled a metalled surface. However, though it probably continued west of the area, it was not found east of the section and it was immediately covered by a black sandy silt, containing ash, mixed with bone and pottery sherds, then by another layer of sandy silt (52075), containing lumps of burnt mudbrick and slag (Fig. 1.1).²³ It therefore seems that the layer of stones was part of the dumping in this part of the area and was not an occupation surface. The final layer of infill (52059) raised the surface to the same level as street 1; it contained mostly clay with some silt, as well as more fragments of burnt mudbrick.

Towards the northern end of the area, the eastern side of the Period 3 occupation surface had slumped down slope and robbing of the curtain-wall had removed most of the remains of Period 3 buildings. Just some sections survived, immediately east of the cobbled street (Fig. 1.8). Of particular interest was the pair of robber-trenches, entering the area from the west.²⁴ The southern robber-trench (52117) continued down slope, bordering the cobbled street, to abut the western side of the ascensus. The staircase could still have been accessible from the roadway and probably continued to be used in Period 3.²⁵ The northern robber-trench (52116) was situated immediately north of street 1. It descended the slope to abut the west side of the tower, just south of a presumed central doorway. Both robber-trenches must have contained walls built upon substantial foundations.

Portions of a building were identified to the west of the defences (Fig. 1.8). Traces of the occupation surface (52033) lay directly above the final layer of make-up (52093). Associated with it, a short section of wall (52031) 0.60m wide, stood only one or two courses high (c. 0.30m), and continued as a line of rubble (52076) before entering the northern section; parts of its superstructure had also collapsed to the west where a jumble of stones remained close to the building’s surviving foundations.²⁶ It is notable that the wall runs up to the edge of the robber-trench (52116) which it meets at right angles. Since this cut (52116) must denote the course of a wall and, with its deeply robbed foundations, it can be assumed that the robber-trench took out the exterior wall of a building and that the slighter walls must have served as internal partitions.²⁷ Subsequent to its demolition, another short section of a west/east wall (52009) overlay the southern end of the primary partition. This was 0.70m wide and stood 0.33m high, plastered with clay on its western face. The relationship with the deeply robbed outer wall (52116) is unclear although the remains of a scorched clay floor containing burnt mudbrick and charcoal lumps (52033) lapped up against its northern face. To the west, and continuing into the western section, another compacted surface (52018) was found, its existence supported by the discovery that it was cut by a solitary posthole. It extended as far south as the robber-trench (52116) and was no doubt contemporary with the external wall, the foundations of which the trench must have originally contained. A tumble of stones and mudbrick appears to respect the edge of this floor surface; it runs north/south and parallel with the partition (52031). Some finds came from the debris.²⁸ Its remains are too insubstantial to have been part of an external supporting wall; it was only 0.40m wide: certainly therefore another internal partition. The line of the outer wall, as defined by the deep robber-trench (52116), extended west beyond the section and the southern side of the building must have used the outer face of the tower’s rear wall (52045). In this period, any entrance into the tower could just have been from the interior of the building although that would seem most unlikely because the Period 3 occupation level was c. 1.5m above the Period 2 ground level. It is most improbable that there could have been access into the tower at ground level (Figs 1.1 and 1.8). After the remodelling of the site, the entrance into the tower must have been blocked; a similar situation affected the western wall where the ground-floor entrance into the medial tower was no longer accessible in Period 3.²⁹ The floor (52018) was overlain by an ash layer, representing the final destruction level.³⁰

Immediately above the demolition deposit, a closely packed spread of water-worn cobbles (52056 and 52025) formed the surface of street 1, 3.10m wide, descending to the east towards the gate (Figs 1.1 and 1.8). Finds from its surface included a mid-6th century coin: confirmation that the street was used in Period 3.³¹

The road surface was covered with a layer of ash and charcoal: the destruction level dating to the end of Period 3.³² It contained rubble and a wide variety of finds, including two 6th century coins.³³

Along the north side of the street, a gulley (52030) had been cut and its sides reinforced with stone slabs (Fig. 1.8). It descended towards the gate and, to the west, disappeared into the section. Most of the paving slabs, apart from two close to the entrance into the gate (52041), had been removed, no doubt during the demolition process at the beginning of the period. However, the portion of the drain which underlay the gate may have continued to function; its side walls were raised by an extra course but it was bonded, not with mortar, but with soil. West of the gate, the original drain was buried (without cover slabs) within the cobbled street so it could not have helped to drain away surface water. It is no wonder that a new drainage channel was necessary, particularly since, in Period 3, the slope down to the gate had been substantially steepened with the dumping and raising of the occupation surface. The new channel remained open until the final destruction when it was filled with burnt debris.³⁴ Outside the gate, a lightly cobbled spread (52068) indicates that the track leading down to the bottom of the hill had been cobbled, almost certainly in Period 3 but not in Period 2 when at least the drain passing out of the gate must have been covered by paving slabs (Fig. 1.8).³⁵

The large stone (52016), found in the gateway, was not in situ: it overlay robbing holes dug into the threshold of the gate. Even so, it could not have been moved very far from its original position. Either it had been part of the gate structure or it had been used to block the entrance during the final period of occupation.³⁶

Period 4

Robbing of the curtain-wall had taken out most of the defensive walls’ superstructure. The robber-trenches produced a few items, including two 6th century coins.³⁷ In this case the primary trench (52117) was too deep to bottom safely; it was in excess of 1.75m deep and was 0.90m wide. The second robber-trench (52116) was also too deep to fully excavate; it was 1.75m wide and also at least 1.75m deep (Fig. 1.8). These are extraordinarily deep trenches to have only taken out the robbed foundations for just Period 3 buildings. On the surface, the robber-trenches continued west for a few metres before apparently joining with others, visible on the surface, which appear to belong to two separate buildings, one to the north and the other to the south of street 1. Since the robber-trenches were deeper than the Period 2 level, it seems likely that the original structures were first built at some point during the 5th century and that the walls were then reused as foundations for a building in Period 3.

Above the northern part of the area, a conspicuously loose mound of earth with some mudbrick and stones (52008) followed by a clay and silt deposit with stone fragments (52136) deeply buried the Period 3 occupation level and both must have formed a spoil heap created during post-medieval robbing.³⁸ The same may partly be true of the later deposit (52001) which merged into the turf line and which contained smaller quantities of stone and burnt mudbrick fragments (Fig. 1.1).

Discussion

The fortifications were solidly and expertly constructed. Unlike the western defences, no evidence survived for tile courses; they may have been reserved for the superstructure, starting at a height above the surviving sections of walling. The gate could have formed a separate part of the building operations and was erected first along with the tower, reusing large limestone blocks to build the superstructure; one candidate is a reused Roman block (52016) in the gateway which had been abandoned during post-medieval robbing.³⁹

There was only one large tower protecting the north side of the entrance; another tower to the south, on the other side of the gate, would have been superfluous; a direct approach up the precipitous, almost vertical slope from flat ground would have been difficult, if not impossible, to scale. Alternatively, anyone coming up the side of the hill on the steep path which led to the gate, was exposed to fire from the tower where an enemy had to skirt around it before reaching the entrance. Tactically, the defensive position was well-chosen. Since the wall-walk was c. 9.50m above the Period 2 ground level, the tower may have had at least three stories and could have been as high as c. 20m.⁴⁰ The gate would have had a two-leafed door, its pivot holes protected by the projecting ends of the curtain-wall which would also have supported the springing of an arch over the entrance. Inside the tower and across the full extent of the area west of the stairway and probably east of the entrance, the surface area was probably paved with flagstones, laid as one with the limestone slabs which must have paved street 1 (under which a drain was constructed) as it ascended to the top of the plateau.

At the gate, excepting two floor slabs remaining in the tower, the rest of the flagstones must have been removed during the reconstruction of the site at the beginning of Period 3; some of the slabs were simply dumped within the backfill used to raise the level within the gate. Two others were incorporated within the cobbled Period 3 roadway. There must have been a pressing reason why so many of the stones had been taken up as part of the primary stages in reconstructing the fort in Period 3; perhaps they were needed to repair the defences or to rebuild Period 2 structures.

The demolition level, at the beginning of Period 3 – and the reoccupation of the fort after its destruction – is of particular interest. Immediately north of the ascensus, the deep deposit of almost pure mudbrick suggests that it came from the demolition of buildings, their robber-trenches visible on the surface within the defences, either side of street 1 (Fig. 0.3). This was followed by the dumping of looser, more mixed black soil which must have come from further afield. Paving slabs from the street by the gate were then removed, leaving only one in situ although other, smaller road slabs were incorporated within the cobbled surface which formed the roadway in Period 3. An attempt to raise the level of the drain underneath the gate chamber would have been scarcely a solution to the steep slope and the need to divert surface water away from the entrance. An ad hoc solution was to cut a drainage channel down the north side of the new Period 3 cobbled street. To the north of street 1, a series of dumps, rich in small-finds, raised the occupation level quite dramatically. However, after the first levels of soil and destruction material had been deposited, there was a temporary halt to the process when a square or rectangular pit with flat bottom and vertical sides was cut through the already deposited debris. It could not have functioned for any length of time; it was itself backfilled with destruction debris and clay before the make-up was laid for the occupation surface and the contemporary cobbled street; whatever its function, it belongs to the remodelling of the interior at the beginning of Period 3. The cut extended west beyond the baulk but two shallow postholes were visible in its south-eastern and north eastern corners (Fig. 1.7). Given its position in the sequence, it clearly belongs to the primary remodelling of the interior. There was no sign that it had been used for habitation. A possible interpretation is that it was connected with the repair or remodelling of the gate structure. When using a crane, great care was needed in ensuring the legs of the apparatus were firmly imbedded in the soil (Vitruvius, De Architectura X 2.3). However, where heavy weights were involved, a polyspaston was best employed since it could raise a stone block weighing as much as 3,000kg. To operate effectively and safely, the windlass was replaced by a treadwheel, operated by one or two people. Given that it would have to be placed over the already dumped, still un-compacted demolition soil, there would have been a danger that the treadwheel might topple over. Digging a trench 1m deep to support the wooden base timbers of a rectangular frame would seem a reasonable precaution. Although it appears from the best contemporary illustration of the machine that exists (the tombstone of the Haterii in Rome), the treadwheel is set close to the crane, this need not have been always the case; they could be set back, on a rectangular base and not right up against the crane itself.⁴¹ The cut was 3.50m from the curtain-wall, which, if it did contain a treadwheel, would leave space for the arm of the crane to be raised and lowered inside the curtain-wall and behind the tower. An additional advantage of the polyspaston was that, since it had a single pole, it could be moved from side to side before the lift began, so as to place the stone in the correct position (Vitruvius, De Architectura X.2.9). In the constricted space available, this would have been a practical solution if the gate or tower required reconstruction. Whatever the explanation for this feature, the raising of the occupation level had immediate consequences. It appears that the ground floor of the tower was backfilled and the entrance blocked so that, if used at all, the ground floor could only have been reached by ladder from the first storey.⁴²

Although no buildings close to the defences would seem to have been envisaged at the beginning of Period 2, a stone and earth bonded structure abutted the south side of the gate and was part of a large building visible within the interior; its outline indicated by the surface depressions left by stone-robbing and, within the area, by a notably deep west/east robber-trench (52117). This formed part of the northern side of the structure, bordering the cobbled street leading down to the gate; it was probably built before the destruction at the end of Period 2. Certainly, it must have existed in Period 3, otherwise the position of the remains of the wall (and its foundations) would not have been so precisely located by the post-medieval stone robbers. Similarly, another west/east robber-trench (52116) to the north must also represent the deep foundations of a Period 3 building, quite probably already constructed in Period 2. Like its neighbour, this wall belonged to a building, also visible on the surface thanks to the post-medieval robber-trench which extended north from street 1. The remains of slighter walls (without foundations) north of the larger wall (52116) must have supported internal partitions. To the east, this building abutted the internal face of the tower. As a general rule, it seems that slight stone and earth walls supported internal divisions between different rooms, whereas the deeply robbed, wider robber-trenches must have taken out exterior walls along with their foundations.⁴³

Scant traces of the Period 3 floor levels survived post-medieval robbing. However, the end of occupation was signalled by a level of burnt debris over the cobbled road surface and within the remains of the building north of the street, from which a complete amphora was retrieved. It had been reused: it was full of carbonised grain.⁴⁴ Of the coins from a dispersed hoard, the latest issue provides a terminus post quem of 579 for the destruction and end of occupation.⁴⁵

Bibliography

Adam, J.-P. 1994. Roman Building Materials and techniques, London.

Notes

152052 : copper alloy fragment (SF410). 52097 : hand-made ceramic spindle whorl (SF720), iron fragment (SF718), dog (SF715), flint flake (SF11367), part of a knife blade? (SF716). 52099 : fragment, part of a crucible? (SF11102). 52121 : flint blade (SF13007). Although these finds are all attributed to the period 1 soil, this level lies immediately below the cobbled track (period 3), south of the gate, and the fragments listed may well derive from the period 2 surface or the period 3 demolition deposit. The finds (with the possible exception of the flint blade from 52097) probably came from the overlying period 3 demolition deposit, remnants of which were recorded as contaminating the prehistoric soil level. No metal finds have been recorded from period 1 deposits elsewhere on site.

2No red wares of early Roman type came from occupation deposits at Dichin. Whether the well-dated 2nd to 3rd century local pottery continued to be produced into the 4th century remains uncertain; Falkner (1999), 94–96. See Swan, pp. 457–458.

3The Roman material must have come from the site itself; the lowland around the fort was probably flooded during the Roman and late Roman periods. Apart from the causeway leading south from the site, there would have been no other place free from flooding except on the hill itself. For the topography of the site, see above, pp. 6–8.

4If the stone block was not from a building on site, it may have been moved from Nicopolis for the reconstruction of the aqueduct, then salvaged for use in the fort, just as the stone channels were clearly taken from the aqueduct for reuse; see area E, pp. 67–70. However, the evidence rather suggests the existence of a major building (probably a temple) at Dichin in the Roman period, see above, pp. 9–10.

5This width is similar to the curtain wall on the west side of the circuit where (area E) it varied slightly between 1.55 and 1.90m.

6The building of such projections was no doubt a precaution taken along the length of the circuit; for another interpretation see area F, pp. 184–185.

7See the lower section of the staircase in area F, p. 187.

8See area F, p. 187.

9Such was the case with the medial tower on the west side of the defences, see p. 120.

10 For the number of floors and the heights of towers, regularly underestimated in books on Roman military architecture, see Poulter (1995), 39.

11 One was 0.45m by 0.30m, the other 0.40 by 0.35m.

12 See area J, below, p. 186.

13 It may well be that the absence of a robber-trench for most of the northern part of the east wall is because subsequent robbing and the dumping of spoil down slope has largely obscured it.

14 The drain must have been covered by more than just the single large slab left in situ . Presumably, the street was surfaced with limestone slabs in Period 2 and the paving would have extended north past the entrance into the tower and south along the back of the ascensus .

15 52092 : clay fragments, part of a loom weight (SF707), a complete loom weight (SF706). Iron nail (SF495).

16 See Swan, p. 541.

17 See below, p. 18.

18 52079 : copper alloy coin; 445–50 (Cat. No. 123).

19 For the cutting up of copper alloy objects, surely for recycling in Period 2 and Period 3, see Ghey, p. 291.

20 52127 : a copper alloy coin 395–401 (Cat. No: 32). 52088 : iron fragment (SF474), nail (SF487). 52091 : fragment of an unknown object (SF714). 52124 : glass bead (SF11277). 52126 : glass bead (SF11294). 52127 : flint blade (SF13006), green glass handle, part of vessel (SF11296). 52132 : flint blades (SF13003, SF13003, SF13004, SF13004, SF13005, SF13005), flint scrapers (SF13002, SF13002). Lead fragments (SF730, SF730).

21 52063 : ceramic fragment, part of lamp (SF11362). Copper alloy fragment, part of a sheet (SF433). Iron nail (SF432). 52069 : lead fragment, part of folded sheet (SF441). 52073 : bone ‘amulet’ (SF448) Iron fragment, part of nail (SF442), glass bead (SF496). 52096 : iron fragment, part of a strip (SF719). 52102 : copper alloy, pierced sheet fragment (SF724).

22 Copper alloy coins; 52089 : 388–92 (Cat. No. 29), 404–08 (Cat. No. 41), 406–08 (Cat. No. 43), 423–25 (Cat. No. 85), late 4th–5th c. (Cat. Nos. 159, 172, 202, 203, 204). 52089: copper alloy object (SF497). Iron fragments (SF477, SF488, SF489, SF491, SF493, SF702, SF704, SF705), fragments (SF494), part of objects (SF468), nail (SF484), nail stem, part of nail (SF499), object (SF490), objects (SF476), tang, part of trowel (SF478), iron / copper alloy object (SF479).

23 Copper alloy coins; 52075 : 388–408 (Cat. Nos. 47, 48, 67, 71), 408–23 (Cat. No. 81), 425–35 (Cat. Nos 89), late 4th–5th c. (Cat. Nos. 196, 197, 198, 199, 200, 201). 52075: copper alloy ear-ring (SF450), fragment (SF470), unknown objects (SF467, SF483), part of sheet with rim (SF459). Iron blade, part of knife (SF472), fragments (SF471, SF481), part of a nail? (SF486), hobnail (SF466), nail (SF482), objects (SF456, SF458, SF480), bolt-head (SF449). Glass object (SF461). Lead fragment (SF465).

24 For these robber-trenches, see below, Period 4.

25 This presumes that the lower step of the ascensus was not incorporated into the robbed outer wall of the building. Note that, in area F, the lower steps of the ascensus were covered by the mudbrick demolition deposit in Period 3; see above, p. 38.

26 52031 : part of ceramic aqueduct pipe (SF443).

27 Compare this with the internal wall in building 1, area K or the internal walls of the north-west room in building 2, area D.

28 52010 : iron dog (SF242), fixtures (SF253), part of a pin? (SF244). Glass sherds, part of vessels (SF257, SF258). 52057 : iron fragment, part of a nail? (SF429), lump (SF430), object (SF425). Glass fragment (SF428).

29 See area E, below p. 123.

30 52018 : iron dog (SF241), glass fragment (SF434).

31 A copper-alloy coin: 52025 : 556/7 (Cat. No. 393). That this is part of a dispersed coin-hoard, see Ghest, p. 278. 52025 : iron fragment (SF287), nail (SF288), shaft of a nail (SF290), lead folded sheet (SF299). 52056 : amulet (SF417). Iron fragments (SF420, SF422), unidentified object (SF421).

32 See Ghest, hoard 2, pp. 241–242.

33 Copper alloy coins; 52001 : 348–64 (Cat. No. 16), 579 (Cat. No. 402), 52002 : 348–50 (Cat. No. 14) (evidently residual), 518–27 (Cat. No. 384). 52017 : 539/40 (Cat. No. 387). Part of a dispersed coin hoard.

52001: iron fragment (SF275), lump of corroded iron (SF271), nail (SF227), shaft of a nail (SF211), fibula (SF260). Lump of lead (SF225). Stone rim, part of a basin? (SF274). 52002: ceramic fragments, parts of floor tile (SF201), tiles with paw print (SF11168). Iron dog (SF203), nail (SF202), glass fragments, parts of vessels (SF228). Stone rim, part of basin (SF277), worked stone (SF276). 52004: iron nails (SF229, SF230), lead lump (SF210), weights (SF208, SF209). 52007: iron fibula (SF426) and cylindrical fragment, nails (SF218, SF220, SF231, SF232, SF233, SF278), barb-spring padlock case (SF212).

52013: copper alloy fragments, parts of sheet /vessel (SF215), fragments (SF245). Iron fragment (SF404), rectangular ferrule or binding (SF401), metal object (SF408), nails (SF216, SF217, SF234, SF235, SF291), t-staple (SF246), fragment (SF296), iron and lead object, function unknown (SF264), glass sherds, parts of vessels (SF255), glass slag? (SF406), slag melted around a tile, from a furnace (SF263), whetstones (SF13016, SF13017). 52017: antler (red deer?), worked antler (SF 222). Iron blade, part of a knife (SF243), link (SF224, SF282), nails (SF285, SF286, SF284, SF283), glass sherd, part of a vessel (SF236). 52019: ceramic, Laconian tiles (SF11386, SF11385). Antler fragment, part of a comb (SF403). Iron corner, fragment (SF252), rectangular ferrule or binding (SF267), nails (SF237, SF238, SF240), nail shafts (SF226, SF249), a barb-spring padlock (SF239), strip of iron (SF247), T-staple (SF251), corroded lump of iron (SF250). Glass, part of a vessel (SF261), lead fragments (SF266, SF248).

From the general destruction deposit a number of additional finds were recovered; 52050: Copper alloy coins; half follis, 527–38 (Cat. No. 391), follis, 569/70 (Cat. No. 393). 54001: Follis, 569/70 (Cat. No. 394).

52050: Part of comb made from antler (SF409). Two iron nails (SF415). Glass fragment, part of vessel (SF427). Pewter, part of vessel (SF11390). Although this material belongs to the 6th century, the coins are not necessarily part of the coin-hoard found on the road surface.

34 52022 : glass, part of vessel (SF300).

35 See above, p. 13.

36 The north gate was certainly blocked, presumably late in Period 3. If the east gate was also blocked, there must have been another entrance which has not been identified during the excavations. Certainly, in area C, the coin finds from the roadway suggest that the road functioned during the later 6th century so any blocking of the gateway must have occurred only towards the end of Period 3.

37 Copper-alloy coins; 52026 : 425–35? (Cat. No. 120), 574/5 (Cat. No. 397). 52054 : 541/2 (Cat. No. 392). It is not possible to assume that these finds were part of the hoard found on the street; all came from robber-trench fills.

52023: antler? (Red deer?), worked antler (SF293). Ceramic loom-weight (SF268). Iron dog (SF292), padlock case? part of a lock? (SF270). 52026: iron nail (SF297), piece of lead (SF402). 52060: iron fragment (SF418). 52100: copper alloy fragment (SF726).

38 52008: clay daub (SF407). Iron nails (SF223, SF279, SF280). Part of a quernstone (SF221).

39 For the significance of this stone, see p. 22.

40 Heights of walls and towers are regularly underestimated in the literature, see Poulter (1995), 39.

41 See Adam 1994, 46–7.

42 In all probability, the ground floor was no longer used. Although here the evidence has been removed by robbing, it is quite clear that the ground floor of the tower on the western defences (area E) was backfilled with demolition material at the beginning of Period 3.

43 For similar internal partitions see area D, the north-western room and the southern dividing wall (Period 2), as well as the internal wall in building in K1 (Period 2) and, in the same area, building K2 (Period 3).

44 See Swan, A23b, p. 541.

45 See Ghest, p. 257.

2. Area D, two storeyed houses/barracks south of street 1

Andrew Poulter

Introduction

In 1996, the first year of the excavations, a small area (4.4 × 2m) was dug to a depth of 2.0m in the north-eastern corner of building 1, a small, apparently square building set back behind a larger rectangular structure (building 2), fronting on to street 1 (Fig. 2.1). Both buildings were visible on the surface before excavation commenced; robber-trenches, dug to remove the walls and foundations, defined the outline of three buildings (Fig. 0.5). Apart from building debris immediately below the turf line, no discernable occupation level was found. However, at a depth of 1.70m below topsoil, a compacted layer of earth-bonded stones represented the surviving fragment of building 1’s lower foundations. Consequently, at the beginning of the 1997 season, it seemed acceptable to remove the topsoil across the full extent of the building, down to a depth of 1.50m before resuming excavation. This proved to have been a mistake. When excavations commenced in area E in the same year, it was recognised that, where it survived, the Period 3 occupation level was immediately below topsoil.¹ In area D, the use of a JCB had removed any traces of the admittedly scant remains of 6th century occupation within the primary area. In 1997, the full extent of building 1 was uncovered and the area was enlarged by 1.20m to the west, 1.0m to the north, and 2.0m to the south of building 1. To the east, although the area extended c. 2.30m beyond the eastern robber-trench, no attempt was made to continue excavation in this direction.² The total area of excavation in 1997, which included building 1, was 12 × 12m. In 1998, exploration was confined to the interior of the building and the western side of the area, beyond the building’s west wall. In the same season, the area was extended to the west by 1.5m, then by 5m to the north to investigate the southern end of building 2. It was during excavation of the northern extension that the robber-trench for the west wall of an adjacent structure (building 3) was identified (Fig. 2.1). The following year no excavations were carried out in area D, but in 2001 the area was further extended to the north by 12m and as far as street 1, uncovering the full length of building 2, including both the remains of the Period 3 floor level, then the full extent of the building in Period 2.

Building 1

Period 1

As elsewhere, the floor of the building and the exterior surface, west of D 1, comprised a layer of brown/black sandy clay with silt, containing prehistoric pottery but no sign of structures. No attempt was made to excavate this deposit which formed a compacted surface contemporary with the Period 2 building.

Period 2

In plan, building 1 was roughly square, its north wall 7.60m in length and its south wall slightly longer at 8.50m. The east wall must have been 7.90m in length, its western wall 7.60m. In width all surviving sections of the foundations (53023, 53024 and 53027) were 0.70–0.80m in width and were made of limestone blocks, tightly packed, with small amounts of broken tile. All had been deeply robbed, surviving somewhat better for the east wall than the west. Only in the centre of the robber-trench taking out the south wall (53027) did the wall still stand up to the modern ground level (Fig. 2.2).³ Elsewhere, the robber-trenches closely following the foundations, and were all c. 0.87m deep.

The western robber-trench (53015) differed from the others in that a series of semicircular postholes protruded from its western side, sealed beneath the destruction deposit which ended Period 2 (53013).⁴ They were spaced at roughly 1.5m intervals and must have contained timbers which would have flanked the outer face of Building 1 (Fig. 2.2).

Fig. 2.1. Plan of buildings 1, 2 and 3 and location of sections, area D.

The cut at the south-western corner (53018) penetrated 1.28m below the Period 2 ground level, immediately west of the building, and 0.18m deeper than the bottom of the robber-trench for the west wall (53015). Its flat base, 0.35m in diameter, contained tile fragments laid flat to support the vertical post it must have originally contained. The second hole (53019) was similar in shape and dimensions; it was 0.30m in diameter and 1.28m deep, 0.18m deeper than the bottom of the robber-trench (and consequently deeper than the bottom of building 1’s foundations). Its base contained flat limestone blocks to support another timber upright. The hole to the north (53020), possibly recut on its south side, was also circular in profile although, half way down, the width increased from 0.34m to 0.60m where it reached its flat base; the total height was in excess of 1.18m. The adjacent hole (53021) was 0.32m in diameter, and was more than 1.25m deep, penetrating below the bottom of the robber-trench. At the junction between the western and northern walls, a final hole (53022), 0.26m in diameter and at least 1.29m deep, continued down to at least 0.25m below the bottom of the foundations which, like the others, it must originally have abutted.⁵

The circular shape of these holes and their depth suggests that they contained substantial vertical timbers and were never filled with stone and silty clay as was the robber-trench which had taken out the west wall of the building. Also, where bottomed, the discovery of tile or stone bases makes sense only if they were intended to spread the weight of a vertical post. These timbers must have been used to brace the outer face of the wall which, on this side of the building, inclined southwards at a gentle but perceptible angle. The only other hole which occupied a similar position on the north wall (53029) was only 0.16m in diameter and 0.34m deep. If it did perform the same function, it evidently did not represent such an effective support as the much larger posts on the west side of the structure. Even so, a post here could have helped to brace the northern face of the building where the wall ran across the slope and did not need such a substantial support as did the west wall.

Curiously, there were a series of much smaller postholes along the inner face of the building’s walls. Two shallow postholes were identified along the inner face of the robber-trench which took out the eastern wall. One (53028), at the southern end, appears to have been only 0.21m deep and 0.16m wide but its position is mirrored by another posthole (53035) for a timber which must have been originally flush with the western face of the eastern wall; it was only 0.29m in depth and 0.28m wide Similarly, the location of other postholes along the inside walls of the building may also have contained posts, serving a similar function. Two holes were identified on the inside of the north wall. One (53025) appears to have been only 0.18m deep but 0.25m wide, the other (53026) was 0.68m in diameter but only 0.25m deep. Close to the foundation trench for the south wall was another (53097), distorted in profile, perhaps because the timber it contained had been recovered during the salvage operations before the construction of the Period 3 building. On the inside face of the south wall, another badly distorted hole (53052), had a flat base and was 0.15m wide and 0.24m deep. To the west, a shallow hole (53095) 0.2m wide and 0.04m deep (and similarly disturbed?) may belong to this group. Though much less impressive than those along the outside of the western wall, these internal posts may have been used to give support to the internal faces of the structure. It seems certain that these measures, particularly in the case of the west wall, were designed as bracing for the earth and stone superstructure. This explanation is supported by the discovery of similar bracing in buildings of 5th and 6th century date in the fort of Iatrus where earth and stone walls were often strengthened with regularly spaced vertical timbers, both along the outer and inner faces of buildings.⁶

Fig. 2.2. Plan of building 1, area D.

The floor of the building (53110) was the same compacted clay surface found in other Period 2 buildings.⁷ Only one find came from the overlying occupation deposit, and that may well have been residual.⁸ The surface was cut by a few pits. Apart from the postholes described above and set around the inner edges of the floor, only three certain cuts were found. Two on the western side (53103 and 53101), both shallow and with no packing, are unlikely to have contained posts.⁹ A rectangular feature (53051), 0.32m wide but only 0.08m deep, was located in the southern part of the floor; its function remains uncertain; it was too far north of the south wall and too shallow to have served as bracing. But one somewhat larger hole (53034), 29m wide and 0.35m deep, more closely corresponds to the dimensions of the pits found within D 2 (and in K 1).

Slight traces of burning covered the floor of the building: the remains of a destruction deposit, although the absence of large timbers may be significant; either the fire was so intense that it turned all the roof beams into ash or, more probably, the roof timbers had been removed before the rest of the building burnt down.

Fig. 2.3A. Section, west side of area D.

Fig. 2.3B. Interpretation 1 mudbrick demolition (Period 3), 2, 4 and 5 robber-trenches (Period 4), 3 grey ash destruction deposit (Period 3).

Period 3

The destruction level was directly overlain by a deep deposit of burnt and unburnt mudbrick (53115) which, notably in the higher deposits, included blocks of limestone, probably from the building’s superstructure, a deposit which also included fragments of tile (Figs 2.3–2.5). This demolition deposit contained a remarkable number of small-finds but, conspicuously, no coins.¹⁰

The Period 3 occupation level was immediately below topsoil. Pottery recovered from the deposits excavated in the trial pit dug in the first year of excavation included an assemblage of sherds datable to the 6th century.¹¹

Period 4

Inevitably, interesting finds came from the robber-trenches (53015 and 53012).¹² Of immediate interest is a clay pipe which confirms the post-medieval date for the robbing of the foundations of Period 1.

Deposits immediately west and north of building 1

Period 2

Excavation continued down to the black humic soil which formed a ground surface predating the construction of the defences; it produced only prehistoric pottery. Above, there was an irregular spread of mortar, petering out to the south but extending north and then east along the southern edge of the robber-trench which took out D2’s south wall; this, as elsewhere on site, can be ascribed to the construction level when the fortifications were being built. The compacted top of this prehistoric soil formed the occupation surface.

Fig. 2.4. The west section 2, area D. Note the mudbrick demolition deposit and the ash destruction level which denotes the end of Period 3 (compare with Fig. 2.3).

Fig. 2.5A. Section across Building 2, area D.

Fig. 2.5B. Section 1 across building 2, area D. 1, foundation trench for east wall of north-western room. 2, mortar spread (Period 2). 3, soil (Period 1). 4, destruction level (Period 2). 5, mudbrick demolition (Period 3). 6, floor make-up (Period 3). 7, robber-trench (Period 4).

Outside the building to the west, there were seven cuts in the occupation surface, probably all postholes, mostly deeper than those within the building.¹³ Three were haphazardly arranged (53044, 53043, 53042) and their function can not be determined (Fig. 2.2). However, two adjacent holes (53099 and 53108) were deeper than the others, both c. 0.50m in depth and opposing another pair of holes to the north (53106 and 53045), the second of which was just as deep. It is possible that these four postholes contained vertical timbers which may have been part of the same or similar structures.¹⁴ Whatever the function of these postholes, it seems that at least one (53106) had its timber removed before building 1 was demolished; its fill contained a cow bell (SF 1131), surely part of a cache of cow bells, found c. 1.0m to the east, on the lip of the robber-trench (53015). Beyond the northern wall and truncated by the robber-trench (53082) taking out the south wall of building 2, was the only cut (53037) found between the two buildings; it was of unusual size (0.68m. wide although apparently only 0.25m deep), probably part of the period 4 robbing (Fig. 2.4).¹⁵

A destruction level (33080 and 58591) overlay the occupation surface, immeditely west of the building, thinning out towards the south but deeper towards the north end of building 1 (Figs 2.3, 2.4 and 2.5). It was essentially black ash, admixed with limestone blocks and fragments of tile, and included a significant number of small-finds, notably a deposit of cow-bells.¹⁶

Period 3

Above the destruction level, there immediately followed a deposit largely comprising burnt and unburnt mudbrick (53115/53116), deeper towards the southern baulk, evidently intended to level up the slope which had existed in Period 2 (Figs 2.3 and 2.4). The levelling would have required the importation of more mudbrick, some of which came from area E.¹⁷ The deposit immediately west of the building contained a large quantity of coins.¹⁸ This is in marked contrast with the mudbrick dump within the building which produced no small-finds and no coins at all. Since, in several cases, coins were corroded together, it is likely that many of them had been in close association at the time of the destruction, perhaps in leather purses.¹⁹

Though not recorded during the machining of the deposits above the demolition level, the top of the mudbrick dump (53115) was clearly visible in the western section and, above it, a layer of fine white powder (53118), most probably the remains of wind-blown thatch, which was formed during the final destruction of the site at the end of the Period (Figs 2.3 and 2.4).²⁰

Period 4

The robbing of the building’s walls and foundations demonstrates a direct relationship between building 1 and the reoccupation of the site in the 6th century. It is certain that building 1 must have been rebuilt on exactly the same plan as its Period 2 predecessor, reusing the earlier stone and earth-bonded walls as foundations for the Period 3 building; the Period 2 ground level was c. 2m below topsoil and it would have been impossible for the early foundations and walls to have been so precisely followed by the robber-trenches unless their presence was readily detected on the surface. This could only have happened if the Period 3 building followed the same dimensions and location as the Period 2 building. Consequently, the robber-trenches not only removed the Period 3 walls but continued deeper to remove all but the lowest portions of the foundations of the