Excavations at the Mount Roman Villa, Maidstone, 1994

EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 MARK HOULISTON with contributions from E. Allison, R. Bendrey, G. M. Cruise, J. Davies, L. Harrison, D. Mackreth, R. I. Macphail, I. Riddler, M. Robinson and A. Savage. THE EXCAVATION The Mount Roman Villa at Maidstone is one of the most substantial Romano-British buildings to have been found in the Lower Medway valley, an area of Kent already noted for the density of its Roman remains. Situated just 15m from the east bank of the Medway (Fig. 1), the villa was first exposed in 1843 when large masonry projections were seen protruding from the river bank. The resulting excavation revealed what appeared to be a double range of rooms, aligned roughly east-west, and perpendicular to the river (Charles, 1847). The remains continued to the north but their excavation was prevented by the presence of an orchard. Later, perhaps in 1884 when a railway cutting and tunnel were constructed nearby, a large mound of spoil 3-4m thick was deposited over both excavated and unexcavated areas preventing all further examination of the main villa ranges south of the Cavalry Barracks and east of the tow-path. In 1970 and 1971 the Maidstone Area Archaeological Group under the direction of A. Miles carried out excavations along the line of the tow-path. This work exposed further walls, interpreted as the remains of a western range. The following year a proposal was made to re-route the Maidstone to Chatham road across the area and the Group, now under the direction of David Kelly, undertook a series of excavations within the southern perimeter of the barracks. This work was funded by the Department of the Environment and continued intermittently until 1980, although by that time the plan to build the road had been dropped. The excavations uncovered a courtyard and 71 MARK HOULISTON Rochester Erto or; ;| ROMAN VILLA MAIDSTONE Canterbury « Dover • Maraswno CAVALRY BARRACKS Fig. 1. Site Location 72 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 associated rooms in the area north of the main villa ranges (Kelly, 1992). The rooms on the west side of this courtyard, which were in line with those of the postulated western range, appeared to have formed a bath suite. Excavations were also undertaken in a limited area south of the barracks wall during 1978; evidence for two further rooms was found. The site of the villa occupies a prominent position on the eastern bank of the river Medway. As its name implies it is situated on higher ground, set some 6m above the level of the adjacent river. Although the natural form of the area has been masked to some extent by previous and current construction works, the land surface clearly slopes away both upstream and downstream as well as sharply towards the river. Towards the east the land appears to rise more gradually; presumably it was on this side that a track or driveway would have connected the villa to the main Rochester road. The river Medway in the immediate area of the site has cut down through the surrounding Lower Greensand deposits of the Hythe Beds and Atherfield Clay into the underlying Weald Clay (Worssam, 1963). Above these strata on the eastern side of the river cutting outcrops of river terrace gravels and landslip deposits have been noted. The landslip deposits predominate in areas to the east and south of the site. Capping the sequence in all places is a layer of alluvial brickearth. Building stone is in plentiful supply in the Maidstone area. The Hythe Beds contain layers of good quality ragstone while a short distance from the site, to the north and east, there are extensive outcrops of Folkestone Beds, from which ferruginous sandstone, or 'carstone', derives. The nearest source of flint is in the Middle and Upper Chalk deposits of the North Downs. A line of springs, roughly parallel to the river, runs northwards from the site. The first is situated some 30m north-east of the main villa buildings. The springs mark the limit of the Hythe Beds along the eastern side of the river cutting and occur at the junction of the these beds with the underlying, relatively impermeable, Atherfield Clay. The Canterbury Archaeological Trust's involvement began in January 1992 after the revival of plans to build the road. Five evaluation trenches were cut across the road corridor, following which a small excavation was undertaken. Parts of two buttresses from the north-eastern corner of the main building were exposed. This, and the subsequent phase of archaeological work in 1994 was undertaken on behalf of Kent County Council, the specifications being prepared by the County Archaeologist. The entire project, both in the field and 73 MARK HOULISTON post-excavation work, was funded by the KCC's Highways and Transportation Department. Fieldwork was resumed in January 1994 when a watching brief was undertaken during the removal of overburden from above the main villa ranges. Roman deposits were revealed at a height above that expected and an excavation ensued. During this work all archaeological deposits threatened by the road scheme were carefully removed. These were located principally in the area to the south of the main masonry building and along a strip of land in the east. Roman deposits exposed elsewhere were cleaned and planned but not excavated. As a result certain questions concerning the development of the villa have not been answered. The strategy adopted, however, has resulted in the preservation of the unexcavated areas beneath the modern road. Intact deposits also appear to have survived to the east of the road, though not to the north or north-east of the villa building or south of the area of excavation, where considerable nineteenth-century disturbances occurred. Interim reports have been produced on the 1992 and 1994 excavations. During the excavation of the site by the C.A.T. in 1994 eight phases of Roman occupation were revealed, as follows: The Timber Buildings and Earlier Deposits Phase 1: Two V-shaped ditches, eight truncated 'drainage' ditches, and the earliest soil horizon c. AD 130-175 Phase 2: Two 'boundary' ditches and fence line c. AD 150-175 Phase 3: The southern end of an aisled building and the base of a water basin c. AD 175-225 The Masonry Villa Phase 4: A further timber building c. AD 175-225 Phase 5: The masonry building (Kelly's Phase 1) c. AD 175-225 Phase 5a: Changes to the northern courtyard (Kelly's Phase 2) c. AD 225-250 Phase 6: Additions to the masonry building (probably equivalent to Kelly's Phase 3) c. AD 275-325 Phase 7: The buttresses c. AD 275-325 Phase 8: The demolition deposits c. AD 275-325. In Phase 5 the complete plan of the main masonry villa was exposed for the first time. Excavations to the south and east of this building led to the discovery of two phases of earlier timber structures (Phases 3 and 4). The earliest of these consisted of the southernmost four bays 74 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 of an aisled building. Phases 1 and 2 consisted of Roman ditches and soil horizons, while Phases 6 and 7 represented structural alterations made to the main building. Phase 8 consisted of a small number of isolated demolition deposits. Analysis of the information gathered during the 1992 and 1994 excavations commenced with the grouping of data pertaining to the individual stratigraphic units (the Contexts: C) into Sets. These Sets were put into Groups (G), and following the incorporation of the dating information, the Groups were put into Phases. Two stratigraphic reports were produced, one for each excavation. Discussions in this report are based on the Phases and Groups discussed in the stratigraphic reports, and not on the Sets. Contexts are referred to only when the more detailed analysis of part of a Group is required. Hereafter the Group numbers quoted refer to the larger, 1994, excavation unless otherwise stated. THE TIMBER BUILDINGS AND EARLIER DEPOSITS (PHASES 1 TO 4) Early residual material Although no excavated contexts appear to predate the mid second century AD, small quantities of earlier residual material were recovered. These comprised an assemblage of 305 flint artefacts of various periods (Wilson forthcoming), a small quantity of pottery sherds, and a silver coin of the emperor Trajan. Although no associated deposits survived, the residual evidence indicates that there were earlier periods of human activity on the site. Phase 1: c.AD 130-75 (Fig. 2) At the northern limit of the excavation the southern end of a crudely cut V-shaped ditch, 2.2m wide and 1.8m deep, was excavated (G.l). It was backfilled with irregular bands of sand and clay mixed with large quantities of flint and stone fragments. The ditch was later recut by a ditch of similar proportions, and subsequently backfilled with similar material. Neither backfill contained anthropic material, and no other features of similar type were encountered. The tops of the ditches had been truncated by modern disturbances. In the main area of excavation eight truncated, V-shaped slots, no more than 0.05m deep and 0.2m wide, were discovered (G.2). They sloped from east to west, i.e. towards the river, and may represent the remains of drainage channels. The lower backfills of these features, which were silty, contained 75 MARK HOULISTON HI 'I \A I I,.-' i I PHASE 1 / \ f I / ,l> l' / Group 3: silty day / / / / © N // l If II ll & / ^"N•*-.. ^ , »r~i // Urn it - * « -^* " ^~S *T: $S# Group 2 : truncated 'i-< 10m drainage ditches Group 1 '. v shaped s—lots / Fig. 2. Phase 1 features the remains of small invertebrates characteristic of damp, often shaded places. Although the ditches would have provided the required dampness, and scrub or other lush vegetation the required shade, the habitat of the invertebrates may have been located a short distance to the north of the site where the line of springs was located. The pollen recovered from the silt although poorly (and therefore differentially) preserved, is consistent with this assessment of the surrounding environment. Subsequently a site-wide truncation took place. This not only removed the upper portions of the drainage channels but cut into the surrounding deposits of natural head brickearth. The brickearth was examined by R.I. Macphail who noted the relative shallowness of the clay enriched, upper brickearth horizon ('Bt'). The thickness of the horizon was only 0.1m, which compares with thicknesses of 0.4m more normally encountered in such deposits. The truncation may have been caused by agricultural activity, though distinguishing traits such as plough marks were not noted. Above the truncation was a patchy deposit of orange-grey, silty clay 76 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 (G.3), up to 0.12m thick, which may have formed as a result of agricultural reworking of the ground. The layer contained a relatively large assemblage of moliuscan and invertebrate remains that are indicative of damp shady environments. However, if reworking of the deposit had taken place much of this material was probably residual. It may have derived from the fills of the drainage ditches (G.2). Neither of the two V-shaped ditches (G.l) contained any datable material, and only three sherds of pottery were recovered from the eight truncated drainage ditches (G.2), suggesting these ditches silted up after c. AD 70. The layer of silty-clay (G.3), however, contained a much larger assemblage of datable artefacts including sherds of BB2 and reduced sandyware pie-dishes of Monaghan class 5D, a deep-sided reduced sandyware dish of class 5F4 and a Central Gaulish samian dish which together give a date range of AD 130-75. This is similar to that of Kelly's earliest deposit, a burnt layer (Tl 8/7) 'immediately above the natural clay and below the make-up for the phase 3 floor' (Kelly, 1992, 190). A copper-alloy trumpet brooch, normally dated c. AD 100-250 formed part of the assemblage. The reasons behind the construction and subsequent backfilling of the two V-shaped ditches (G.l) are not clear. They could belong to Phase 1 or any of the earlier phases of activity suggested by the residual assemblages. The eight truncated drainage ditches (G.2) post-dated AD 70; at this time the area appears to have been damp and shaded. The ground-truncation and subsequent deposition of silty-clay (G.3) appears to relate to a period of more intensive land use, suggested by an increase in the quantity of pottery sherds recovered. These were mainly small and worn, which suggests either that they were left on the ground surface for some time, or that they became incorporated within a soil that was being agriculturally worked. The latter interpretation is most likely, although the survival of useful material evidence, particularly environmental and pollen data, was poor. Phase 2: c. AD 150-75 (Fig. 3) Two large U-shaped ditches (G.8), up to 1.2m wide and 0.85m deep, were cut east-west across the southern half of the excavated area. They were set 7m apart and parallel to each other, suggesting that they were both open at the same time. A fence line, represented by nine post pits (G.9), ran along the northern side of the northern ditch and this, too, was clearly a contemporary feature. The posts may have 77 MARK HOULISTON // -r '/ .' PHASE 2 O Group 10 /' ^ o rl . ii Group 4 0 // U£ // © // 0 Group 9 : post pits © // N // Group 8, ditches / Ip O 0 / / Group 6, posts / i // J //« =-,. JL / Group'7^, // Group 5 « // 0 1 5 10m. ^~ -~^ / / / Fig. 3. Phase 2 features been large as some were supported in pits packed with small blocks of roughly hewn ragstone. The northern ditch cut a large shallow pit of unknown function (G.4). The ditches were cut to facilitate drainage, because both had fine silts in their bases and sloped from east to west, i.e. towards the Medway. They were also probably cut to segregate different areas of land use. Evidence for land use south of the boundary represented by the northern ditch and the fence line was not recovered, but the relative absence of post-pits and stake-holes suggests that structures were not located there. This is in contrast to the area to the north where the concentration of posts (G.5 and G.6), stakes (G.49; not illustrated), and other features (G.7 and 10), was greater. One of the latter, a crudely-cut trench (G.10), was slightly different from the rest. During or after its use it appears to have been abandoned to the elements, since fine silt lenses containing molluscs with a known preference for moist, sheltered, environments built up over its base. Although for the purpose of this report all the stakes have been assigned to one stratigraphic group, their locations and relationships 78 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 to surviving layers suggests it is improbable that they were all contemporary. They can be formed into a number of irregular shapes, none of which appear convincing as structures. Both U-shaped ditches were backfilled with silty clay. The uppermost backfill of the northern ditch contained a large proportion of burnt soil and oak charcoal, sometimes with the soil and charcoal attached, and smaller quantities of slag and daub, some of which appeared to be fired clay with a dung temper. A much higher percentage of the total deposit, however, consisted of burnt cereal grain, and this suggests that the dump represents the remains of a domestic hearth or corn drying oven. The grain consisted of cleaned spikelets of spelt wheat that may have been accidentally burnt during the drying process. The backfills of the southern ditch contained an animal bone assemblage of a kind normally associated with primary animal butchery. Many of the bones had been gnawed, suggesting that the animal remains may have been dumped onto open ground rather than buried, a view reinforced by the presence of carnivorous molluscs and vertebrate remains in contemporary contexts. Neither ditch contained the kind of artefact assemblage more normally associated with domestic-type waste. Mollusc and vertebrate remains were also recovered from the backfills of the ditches, as well as from the silts in the base of the crudely cut trench (G.10). This evidence indicates the site was wellvegetated, with some areas of shrub-like growth and others where leaf litter or thick grass-cover predominated. The presence of wet or marshy conditions was also indicated, a habitat that may have existed in areas close to the nearby springs. Few useful dating indicators were recovered from G.5, 6, 7, and 9 which contained only small, worn, pottery sherds. G.IO, 17 and 49 were aceramic. Small quantities of sherds were recovered from the silts in the bases of the two U-shaped ditches (G.8). The dates assigned to these lower fills are not distinguishable from those assigned to the upper backfills. A date of c. AD 150-175 is indicated. The backfills contained larger quantities of sherds, of which those from the southern ditch included good dating indicators. Similarities in the fills from the two ditches is consistent with the view they were backfilled at the same time. However, the absence of definitive later pottery from the northern ditch allows for the possibility that it was backfilled first. The only datable small find, a bone pin, came from the backfill of one of the stake-holes (G.49). Pins of this type are found in contexts from the middle of the second century onwards. The substantial sizes of the post-pits arranged along the northern 79 MARK HOULISTON side of the northern ditch, and the presence of structural elements to their north which were absent to their south, suggests that the fence they formed represented a significant boundary between two distinct areas of land use. The southern area was characterised by two large drainage ditches, the northernmost of which lay along the southern side of the fence line. Although it is not possible to be more specific about land use within the individual areas, an impression of the use of the area as a whole is to be gained from an examination of the backfills of the two ditches. The significance of the upper backfill of the northern ditch is that it derives from a hearth or oven that must have been in use either while the ditches were open, or earlier. The concentrated, organic nature of the remains suggests they were probably deposited directly from the site of the hearth or oven, rather than as the result of later re-deposition, and therefore belong to Phase 2. Butchery waste recovered from the southern ditch may also belong to this period. Coupled with the relative absence of evidence for domestic occupation, the material from the ditches indicates that the site lay away from areas of habitation, and was used for semi-industrial purposes, namely corn-drying and animal butchery. No areas of contemporary habitation have been located though it should be noted that the 'structural' area on the northern side of the fence line bordered a large unexcavated area - that which was later covered by the masonry villa. Evidence for early habitation may be located there. Although the environmental data recovered was limited in its scope, the image presented of a well vegetated, sometimes damp, environment with areas of shrub and leaf litter or thick grass, is consistent with the site having had a marginal status. Phase 3: c. AD 175-225 (Fig. 4) The first evidence of habitation dates to Phase 3 when an aisled building, measuring 29m north-south by 10.5m east-west, was constructed north of the northernmost of the two Phase 2 ditches; this ditch was now backfilled. The main structural element of the building consisted of two rows of substantial timber posts, approximately 0.3-0.4m in diameter (GA4). (Plate I) They were set in large pits filled with compacted clay and ragstone fragments. Set out from, and parallel to, the rows of posts were two beam slots (G.16), the probable bases for the external walls of the building. Although only the southernmost four bays of the building were excavated the structure continued northwards beyond the excavation area, perhaps for a further four bays. A short length of masonry wall was observed 80 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 // '/ / PHASE 3 im r r\jvm , / ^ v lh. fAl // Group 16 : deposits west ol 0 // aisled building N // Group/16 : robber cut for western wall J " / IW. II 1 Tl- ! Group 15 ^qp Group 16 water basin robber cut for eastern wall 10m / // / Fig. 4. Phase 3 features in a position which roughly equates with the end of the postulated eighth bay. A gap in this wall fell approximately on the line of the central axis of the building and may represent the building's northern entrance. There is some indication that the building also continued to the east, with two slots indicating the position of a possible extension on that side. To the north of the possible northern entrance of the building a sunken heptagonal feature (G.l5) was observed (Plate II). It had a smaller rectangular feature bonded to it on its eastern side. Both features were constructed out of large tiles set in pale mortar, and waterproofed, internally and externally, with opus signinum mortar. They appear to have been connected by a small duct, and may have formed part of a water basin or fountain. It seems likely that water entered the structure from the east and passed through the rectangular feature before entering the main heptagonal area via the duct. The rectangular feature may have acted as a settling tank within such an arrangement. Although no internal layers, associated with the aisled building 81 MARK HOULISTON PLATE I L • The north-east postpit of the Phase 3 aisled building survived, a pronounced darkening of the underlying deposit was noted. Chemical and micromorphological analysis confirmed the suspicion that this was caused by staining from above, and tentatively suggested that such staining may have resulted from the seepage of organic acids and liquid excreta from a stable floor associated with livestock housing. The physical incorporation of some organic material, probably from such a layer, was also identified. The aisled building may, therefore, have been a byre or a barn. Six areas of a layer of clayey-loam deposit (G.l 8) were excavated in the area to the west of the aisled building. Apart from the easternmost (C.467), which was cut by the robber trench of the external wall, all of these may also have been contemporary with the Phase 4 building. Analysis of one of the areas revealed that it consisted of biologically well-mixed topsoil and sub-soil, characteristics that are not commonly associated with in situ undisturbed soils. Molluscs retrieved from the layer consisted mainly of species with a preference for damp, sheltered environments. Generally, the data suggests the area consisted of overgrown waste ground, onto which rubbish was dumped. Evidence that two industrial processes involving cereal grain were being undertaken in the area at this time was also retrieved. In the 82 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 PLATE II --;-'. -* :« The sunken heptagonal feature area to the west of the building one of the soil 'dumps' contained a large assemblage of cleaned spikelets of spelt wheat. This material had been burnt, probably accidentally, during a process designed to dry the grain in preparation for its storage. Evidence for the second process consisted of accidentally burnt grain debris and came from the construction backfill of one of the post pits of the Phase 4 building. The debris probably derived from the rubbing of parched, malted, spelt wheat, a process undertaken to remove the husks and sprouts from the grain before grinding them for brewing. Although the material is residual, and therefore could have come from any of the earlier phases, it most probably derives from Phase 3, suggesting that brewing, using wheat, was being undertaken at this time. The pottery assemblage comprised mostly small, worn and weathered coarseware sherds. Sherds from deposits associated with the construction of the aisled building (G.14) were generally indistinguishable from those associated with its abandonment (G.M and 16). Datable material included a number of sherds from imported vessels, and others from vessels of Canterbury Sandy ware. A date of AD 175- 225 is indicated. 83 MARK HOULISTON Dating for the clayey loam deposits located to the west of the building (G.l8) came from sherds of reduced sandy and grog-tempered ware, a second-century copper alloy pin and a Trumpet brooch of the second or early third centuries AD. Although the aisled building was constructed after the northern ditch of Phase 2 had been backfilled, the boundary represented by the ditch and the associated fence line was respected by the structure's southern end. This indicates continuity in occupation from Phase 2 to 3, and is consistent with a picture of a gradual change in land use, rather than large scale redevelopment. In this regard it is significant to note that the building was constructed in an area that previously (Phase 2) contained more 'structural' activity (more post- and stake-holes) than the others. The absence of definitive later pottery from the northern ditch of Phase 2 allowed for the possibility that it was back-filled before the southern one. In this case, the southern ditch may have continued in use during the life of the aisled building. Such a possibility would also be consistent with the idea of continuity between the two phases. Analysis of the underlying deposit indicates that the excavated, southern, end of the building was probably used for livestock housing. Spelt wheat, which was being produced at this time for use in the production of beer, as well as for storage, may also have been kept in the building. There is no reason why the rest should not have been domestic. Indeed such arrangements have been detected in a number of other Romano-British aisled buildings (Smith, 1963, 1-30). A masonry wall at the possible northern end of the building, with its central entrance, may be evidence of such a use. The heptagonal water-basin or fountain was aligned on the central axis of the building just outside the entrance, and may represent an external feature deliberately constructed for ornamental effect. This supports the idea that the northern end of the building was of a higher status than that of the south. This cannot be confirmed as the north end has not been excavated. In the absence of further excavation it is not possible to tell whether the aisled building is the principal building of the area or an outbuilding situated within the grounds of a larger farm or villa. Evidence for such a structure has not been found though such remains could be preserved below the unexcavated deposits associated with the later masonry villa. More rigorous interpretation of the unexcavated areas remains for the future. Phase 4: c. AD 175-225 (Fig. 5) By the end of Phase 3 all or part of the aisled building had gone out 84 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 // '/ PHASE 4 ,/ /loi I Group 30 : a clayey/toam layer \ . // / ^ / ' I I ll ll LGro.upL 12 : a line ol post pits ll I / n G N // ^ ® POS./PIB 7— Group/19 O pOSt/pHS U. © j $ ^ 7 * / / ^ 0 --< 0~-9 ^ / 1 i * -f- n 0 1 . . . 5 10m. ~"^ \ . / , / Fig. 5. Phase 4 features and timber building of use and a structure measuring at least 7m east-west by 5m northsouth, had been built across its southern end. The structure consisted of six substantial posts (G.l9), each approximately 0.3-0.4m in diameter, set in large pits packed with redeposited brickearth and ragstone fragments. The building terminated in the west and south though may have extended further to the north and east beyond the area of excavation. It may have been of 'aisled' type, and therefore aligned east-west, though parallel beam slots, which usually define the outer walls of such structures, were not discovered. The alignments of the pairs of posts of the building were similar to that of the long axis of the Phase 3 aisled building, and roughly perpendicular to the length of masonry wall, thought to represent its northern end, observed in the unexcavated area to the north. No evidence was found of floors or occupation layers, or other features possibly associated with the building. Two posts approximately 0.15-0.20m in diameter were set in the ground to the south of the building. They lined up with the structure's two central posts, and may have marked the position of 85 MARK HOULISTON an external partition or fence. The full extent of the fence cannot be determined because further to the south the ground was heavily truncated. A post positioned in the centre of the building may indicate that the structure was partitioned internally along the same axis. Away from the building, in the south-western corner of the excavated area, two groups of features were excavated. These have been assigned to Phase 4 though neither necessarily contained features solely attributable to that period. G.34 consisted of two rectangular and two circular pits of unknown function, and two possible post pits. G.l2, which probably represents a fence, consisted of a line of four smaller pits. Also assigned to this phase is a layer of clayey loam (G.30); it overlay the backfills of the two groups of pits. The pottery found does little to date activity associated with Phase 4. The layer of clayey loam (G.30) contained two sherds dated c. AD 250+, but ground disturbance in the area means these may have been intrusive. No other useful dating indicators were retrieved. The timber building post-dates the southern end of the aisled building of Phase 3 but is not necessarily earlier than the later masonry structure. If contemporary with that later structure it may have formed part of a building, or buildings, flanking the southern side of an external courtyard similar, for example, to that at Darenth (Philp, 1973). However, the orientation of the building is more akin to that of Phase 3, and it therefore seems more probable that, even though the Phase 4 structure replaces that of Phase 3 in the excavated area, the choice of its location is a result of continuity in use between the two phases in parts of the building located outside the area of excavation. For example, the unexcavated masonry wall fragment, interpreted as the northern end of the Phase 3 aisled building, may have continued in use during Phase 4; alternatively it may represent the masonry rebuild of an earlier - Phase 3 - timber wall. Further excavation will help to test the validity of these hypotheses. The two groups of pits and the soil layer assigned to Phase 4 were stratigraphically isolated from the rest of the site and contained few datable artefacts. The soil layer and some of the pits must be later than the Phase 2 ditches, but apart from this, features in these groups may belong to any phase from 2 to 4. The reason why the pits were cut is not known, though the absence of more than a handful of animal bone fragments in their backfills suggests that whatever their functions, it was not for the disposal of domestic waste. 86 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 THE MASONRY VILLA (PHASES 5 TO 8) The alphabetic and numeric sequences used by David Kelly (1992,184-208) to locate individual rooms within the villa are retained below. Where previously unexcavated rooms are referred to, the sequences have been continued. Phase 5: (Kelly's Phase 1) c. AD 175-225 (Fig. 6) The villa appears to have been conceived and constructed as a single period structure. There is no evidence that the elements of any earlier buildings were incorporated within it. The main building consisted of two wings, one room wide, linked by a double range of rooms, all of an integral build. It had a portico or corridor on its eastern side joining the two wings, and a similar structure running along its western side where it overlooked the river. The building was roughly symmetric about both east-west and north-south axes, and measured 34.5m north-south by 21.5m east-west. In the north-west a bath suite was attached to the building continuing the line of the western corridor or portico. None of the recently excavated remains, and very few of those previously excavated, survived above ground level (Plate III). Floors were noted but superstructure was not. The absence of upstanding walls means that care must be taken in the identification of rooms, as some divisions noted at foundation level may not have existed above ground. A comprehensive ground truncation appears to have preceded the construction of the villa; in the area of the recent excavation the uppermost deposits associated with Phases 3 and 4 had been truncated on a horizontal plane. The trench for the foundations of the main south wall was cut from the new, reduced, ground level. The foundations filled the trench in its lower half, though in the top were built free-standing. They consisted entirely of ragstone blocks, and were set in sandy, yellow mortar except in the lowest course where they were set in clay. Despite variations in thickness and in the arrangement of the blocks the same construction method appears to have been used for the other walls of the villa, since everywhere where the foundations of the main building and bath suite were exam ined, both during the recent excavation and previously, the same arrangement of stone, mortar, and clay was seen. In the area of the recent excavation a construction backfill (G.20) consisting of fragmented mortar mixed with soil and a few ragstone and flint pieces not only abutted the southern wall, but spread across the surrounding ground surface as well. An extensive courtyard surface was laid in front of the main 87 MARK HOULISTON PLATE III *. n »-*-• ,: • t *+m „ . The excavation looking north-west towards the River Medway building on its eastern side. It consisted almost entirely of one type of stone, hassock, in contrast to subsequent surfaces which were very mixed and contained high proportions of re-used material. Although this area of the site was not excavated, and therefore exact relationships could not be determined, the absence of re-used material suggests that the surface was laid during the original phase of construction. The courtyard was laid over the remains of the postulated northern wall of the Phase 3 aisled building. There was no evidence however to establish its relationship with the water basin. The stratigraphic relationships between the walls of the main building and the recently excavated deposits were contained in a small area located at the western end, and to the outside, of the main south wall (Plate IV). Elsewhere these relationships had been removed by the 1843 excavations. In the southern corner of the excavation, two successive layers of silty loam (G.33) were excavated. These appear to have been subjected to considerable biological reworking, both in recent times and in antiquity. They were laid after the layers and features 88 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 PLATE IV • • ' - • ..,• " ^^^^^^^^sPi* :ist A construction spread relating to the Phase 5 masonry villa under excavation attributed to Phase 4, and may be associated with any phase of occupation from 5 onwards. Closer to the building, fifteen small pits of assorted sizes (G.21) cut the construction layer. Their functions are not known, though the shapes of a number of them suggest that they were cut to accommodate vertical posts. They must have been contemporary with the construction of the masonry villa or later (Phases 5 to 7). In the south-western corner of the site numerous pits of assorted sizes (G. 11, 13,31,32, and 37) were located. The shapes of many of the smaller of these suggest that they, too, were cut to accommodate vertically set posts, whereas some of the larger ones may have been constructed for the disposal of waste. There were generally few clues to their functions in their backfills, though it should be noted that artefact preservation on the site was generally very poor. Only broad phasing can be ascribed to many of the pits; some cut the backfills of the two Phase 2 ditches, and others the silty loam layers assigned to Phases 4 and 5, however, several cut only the naturally occurring brickearth deposit. 89 MARK HOULISTON In the north-east corner of the excavation the remains of a silty-clay deposit (G.39) overlay the truncation horizon associated with the construction of the villa. Groups of features also post-dated the truncation horizon (G.36, 38, 42, 43, and 44). These, too, contained few clues in their shapes or backfills as to the reasons for their original construction, though some were clearly cut to hold vertical posts. Further to the south more posts and gullies (G.40 and 41) indicated the presence of activity in that area also. It is unlikely that all of these eastern features can be solely attributed to Phase 5, though many contained small quantities of pottery sherds of approximately the correct period. The southern wing of the main building consisted of five rooms (A to D, and J). Rooms D and J probably opened on to the eastern and western corridors (or porticoes) respectively, and the central rooms, A, B, and C on to rooms E and F. The northern and eastern sides of room D were replaced during the Phase 6 rebuilding. Of the foundation walls, the two running east-west were 0.7-0.8m wide and constructed of level courses of roughly worked blocks. The north-south partition walls were narrower, c. 0.55m wide, and were apparently constructed in a herringbone pattern. These walls were not exposed in elevation during the recent excavation but details of the eastern sides of the western walls of rooms A, B, and C are shown on one of the watercolours painted by Thomas Charles of the 1843 excavation (Charles Collection (Maidstone Museum), Vol. 1. ff. 56). Areas of opus signinum flooring appear to have been encountered during the 1843 excavation of the wing, and then removed, although it is not known at what height or in what rooms they were found. The central area of the villa consisted of a block of ten rooms arranged in two, north-south aligned rows, with R, T, V, I and F in the east and L, S, U, H and E in the west. Rooms E and F and the northern halves of I and H, were excavated in 1843, and the western area of room L was first exposed during the work undertaken in 1978. The remaining areas were exposed for the first time during the recent excavation. As with the southern wing, the longer foundation walls were thicker than the cross walls, and were apparently constructed of roughly-worked blocks set in courses, rather than in a herringbone fashion as shown in a painting of the excavation by Mr Charles (Charles Collection (Maidstone Museum), Vol. 1, 56). (Plate V) The walls appear to have survived to threshold level, though no evidence for doorways was revealed. Where exposed, they were bonded together, implying one period of construction. The remains of a poured mortar floor were uncovered in room T and the remains of a tile floor during the 1843 excavation in room H (see PLATE V). The 90 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 PLATE V • Rooms H and E, looking west same excavation also uncovered the remains of a tessellated floor 'in the adjoining apartment', though it not clear whether this refers to room E or I. Other rooms excavated in 1843 appear to have had floors of opus signinum. It is not known whether floors survive in the two large central rooms, U and V, though if they do it is probably just below the level recently exposed. Several large pieces of a poured mortar surface, 0.15m. to 0.20m. thick, and a length of quarter-round moulding were exposed during the recent excavation around the sides of room U. It is less likely that floors survive north of the modern drain run, in rooms L, R, S and the north-western corner of T, because here the tops of the exposed walls were at a level lower than those located to the south. The northern wing of the main building (rooms M to Q) was similar in construction and layout to that of the southern wing, though with variations in the sizes of some of the rooms. Room Q opened onto the eastern portico, whereas room M joined the western corridor to the northern bath suite. The central rooms, N, O, and P fronted rooms L and R, which may also have comprised a corridor. With the exception 91 MARK HOULISTON of the western side of room N and the eastern side of room M, which were found in 1978, all the rooms were uncovered for the first time during the recent excavation. None of the remains appear to have survived above contemporary ground level; indeed, as with other fabric exposed north of the modern drain run, the tops of the truncated walls were below the level of the floor exposed in room T. The sides of walls were not exposed, though examination of their truncated upper limits suggests that they were constructed in a similar manner to those of the southern wing. It is assumed a corridor or portico ran along the eastern side of the building, connecting rooms D and Q. The eastern side of this room would probably have been defined by a wall similar in position and dimensions to the one built there during the next phase of construction (Phase 6). In a 'winged corridor' villa like the Mount, is seems likely that such a wall would have been relatively low, and have supported a row of columns, thereby creating a classical facade. Given the symmetry of the villa the main entrance would almost certainly have been in the middle of this wall, and again the position of such a feature is indicated by evidence from Phase 6. A corridor or portico (room K) ran from the end of the southern wing along the western side of the villa where it connected, via room M, to a bath suite (rooms 1, 2, 3 and 4). Unfortunately, the dimensions of the western wall of this corridor or portico are only approximately known. The most reasonable assumption is that it was of similar dimensions to the eastern wall of the eastern portico of Phase 6. The bath suite is discussed in detail by Kelly (1992, 188-93). Briefly, the Phase 5 structure (Kelly's first phase) consisted of a range of four rooms which continued the line of the western corridor. The presence of a bath suite was indicated, for in the central rooms (2 and 3) the surviving floors appeared to represent the lower levels of hypocaust heating systems, while the associated walls were thicker than any encountered elsewhere. The associated praefurnium would have been located either to the north, in the area later occupied by room 5, or to the south, where the slope of the ground towards the river would have made construction easier. On balance, the northern position seems more likely since if built on the southern side such a building would have detracted from the aesthetic appeal of the villa and spoiled the view of the river from it. Phase 5a: (Kelly's Phase 2) c. AD 225-250 The walls and associated deposits of Kelly's second phase of activity (Kelly, 1992, 193-202) were located to the east of the postulated bath 92 (position & size approx.) Bath suite Group 20: construction layer Group 33: £. silty loam layers 1 i i i ^ 7 ! i 9 i i a Group 21: small/pits „ , TxSS—i Courtyard Spreads > Q // Groups 11, 13, 31, 32 & 37: misc. pits © ® Lean-to '&..mmm 1993: Group 4 AREAS EXCAVATED 1994 © PHASES 5 & 5a Current Excavation : I Previous Excavations : Phase 5 ^ 3 Kelly : phase 1 Phase 5a Kelly : phase 2 Group 39: silty clay layer I I 1993:'Group 1 Groups 40 & 41 posts & gullies AREAS CLEANED & PLANNED Phase 5 "• 1994 r AREA EXCAVATED l.-.-r.'.'\ Q Groups 36, 38, 42, / 43 & 4 4 : f tf?Of misc. features Fig. 6. Phases 5 and 5a I 1 1 1 I i RI B2! ! I ' Approximate positions : | i i J the buttresses having (~~7~^ become detached from i L-f the main building _ . i ± Group 24 : robber trench for Phase 6 wall \vzm l ; S»iS PHASES 6 & 7 £^::: Current Excavation : Phase 6 Phase 7 % ^ 03 1993 : Group 3, poured mammmszi::". mm y" moriar spreads above Large sandstone block Previous Excavations : Phase 6 1 Kelly : phase 3 1993;: Group 1, silty loam deposit Pi Fie. 7. Phases 6 and 7: the masonrv buildine EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 suite (rooms 1 to 4). Although this work appears to represent a continuation of the earlier development it was not undertaken until these rooms had become dilapidated and were either abandoned or partially demolished. The deposits consisted of the western, northeastern, and eastern boundary walls of an external courtyard (the latter of which was butted by a lean-to structure) as well as courtyard spreads, and a range of four rooms located in the north-west (rooms 7 to 10). The southern end of the eastern courtyard wall, and the beginnings of its turn to the south (presumably the end wall of the lean-to), were also exposed during the 1993 excavation (1993: G.4). The eastern and western boundary walls were constructed using white mortar, though the north-eastern wall and the range of four rooms used the same yellow mortar observed elsewhere. In the north-eastern corner of the courtyard, presumably sheltered by the lean-to, a clay-lined tank was located. Associated with it was a stone drain which led north-westwards towards the river. The drain cut through the northern courtyard wall, following which the resulting gap was repaired with white mortar. The differences in mortar colour in the walls of the northern courtyard may imply different periods of construction or, as David Kelly suggests (1992, 194), they may have resulted from technical considerations. No contemporary walls appear to have survived in the area of the main villa building in the south. Phase 6: (Kelly's Phase 3) c. AD 275-325 (Fig. 7) A number of alterations were made to the main villa building during its lifetime, though these do not appear to represent major rebuilds or significant additions. In the area of the northern courtyard, however, more extensive alterations were undertaken (Kelly Phase 3). In the following descriptions all the alterations are described and discussed as one phase, though this does not necessarily imply they were formed during a single period of building construction. In the south-eastern corner of the main building the eastern and northern walls of room D were rebuilt, presumably because the originals had become damaged or fallen into disrepair. The new walls were slightly narrower (0.65m) than those of Phase 5 and were constructed using a slightly paler mortar. The eastern wall extended northwards, where it formed the new outside wall of the eastern portico (room G), presumably replacing an earlier wall of similar construction. The location of most of this wall was indicated only by the top of its later robber trench, though a large sandstone block positioned close to the middle of the trench may be part of its in situ 93 MARK HOULISTON fabric. This block, perhaps a column base, may indicate the position of the southern side of the main entrance of the villa. Courtyard surfaces containing high proportions of reused material, including broken roof tiles and mortar pieces, were laid above the original hassock surface in the area located in front of the main building, and may be contemporary with the portico rebuild. Beyond the surfaces was the water basin, though unfortunately the relationship between them could not be determined. Although the position of the water-basin indicates an association with the postulated northern wall of the aisled building of Phase 3, the date of the material in its backfill allows for the possibility that it was still in use during Phases 5, 6 and 7. Excavation of this area of the site will be required before the dating of the water basin (and its function and the mechanics of its operation) are better understood. A rectangular structure was attached to the outside of the southern wall (room W). It measured approximately 2.4m by 6.2m internally. Its foundations were c. 0.8m wide and 1.1m deep and consisted of pieces of broken tile and ragstone (G.23). Unfortunately no associated superstructure or internal floors survived. The position of the room might suggest that it comprised a small praefurnium but the absence of a corresponding opening in the outside wall precludes this. The function of the room remains unknown. Developments in the northern courtyard are discussed in detail by Kelly (1992, 202-7). On the outside of the northern wall, in a similar position to that of room W in the south, a room (11/12) of comparable shape and dimensions was constructed. Because of these similarities it is suggested that the two rooms were built at the same time. The room was divided in two with the smaller, northern, area occupied by an oven, an indication that the room functioned as a kitchen. Other developments in the northern courtyard included the rebuilding of rooms 1, 2, 3 and 4 and the construction of a new room (5) against the eastern side of room 2. Both 2 and 5 contained hypocaust heating systems which tentatively suggests that this area of the villa once again housed a bath suite. Again, the location of the associated praefurnium was not established, though at this time it was probably on the western side of room 2. A passageway (room 6) was built around the outside of room 5 and appears to have been connected with the main building. It is not clear whether this area would have been covered. It may have been built as a screen to prevent the activities being undertaken in the adjacent service courtyard from being seen by those using the bath suite. The surviving northern end of the western courtyard wall of Phase 5 may have served the same function. Most of the other courtyard walls were removed, and the yard itself was resurfaced. 94 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 During the excavation undertaken in 1993 the remains of a silty loam deposit (1993: G.l) were discovered immediately above natural brickearth. This deposit appears to represent the lowest surviving portion of what was originally a much thicker layer. The layer has been tentatively identified as a cultivated soil as it was well-mixed. It contained a number of fragments of tile and seventy-three sherds of abraded pottery, the bulk of which were second century in date. The layer may have been agriculturally worked right through to the onset of Phase 7, at which time it was sealed by spreads of poured mortar. It is not known whether similar deposits extended into areas located away from the main villa buildings. In the future, a closer examination of surviving soils located in the unexcavated areas to the north and east of the villa, utilising techniques such as soil micromorphology, may shed more light on this and other 'cultivated' deposits. Phase 7: c. AD 275-325 (Fig. 7) At the onset of Phase 7 the room located against the outside of the southern wall of the main building (room W) was demolished. Even its foundations were substantially removed (G.24) which suggests that material was being acquired for re-use. A buttress (B3), attached to the southern wall, was located approximately in the centre of the remains of room W. Although the stratigraphic links between it and the surrounding deposits had been previously removed by the cut of the 1843 excavation, its position indicates that it was built after the removal of that room. In total, eight buttresses were exposed, six during the recent excavations (B3 to B8), and two in 1843 (BI and B2). Variations in size have been noted, which are principally in their lengths, although all apparently used the same materials and were built using the same construction techniques. Because of this, it is assumed that they were all built at approximately the same time. Six of the eight buttresses were located around the south wing, implying that this was the part of the building most in need of strengthening. The two at the west end were discovered in 1843 along with part of the associated end wall (Charles, 1847, 86-8), apparently having become detached from the remains of the villa (Plate VI). The more northerly of two (BI) was about 50 per cent longer than the other (B2) and appears to have included fabric located on the inside of the main wall, suggesting that some of this wall may have been replaced at the same time. A third buttress may have been built in the same area, extending southwards from the 95 MARK HOULISTON PLATE VI : I? I [ * i I s •AL * -A * -- *• *> -4, * j£- X i # "V v aft Ji *•/. r " - * i - - i i i t f r r l C < f - ' H i i ' f i - ' in (t*M4* J^ * ^ A site plan of the 1843 excavation from amongst the paintings and sketches by Mr Charles (Charles Collection, Maidstone Museum) 96 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 PLATE VII The southern wing, looking west (Charles Collection, Maidstone Museum) west end of the south wall, mirroring the one at the east end, though no evidence of this has been found. Only the foundations of the buttresses exposed during the recent excavations survived. B3, which was positioned halfway along the south wall, was exposed on three sides. It measured 0.85m deep, making it shallower than the wall against which it butted. It appears to have been built by placing courses of ragstone blocks, alternating with mortar, directly into the construction trench. Three sides of B4 were also exposed, but only the south side of the adjacent buttress B5, and the top of the foundations of B6. Buttresses B5 and B6, however, were exposed during the 1843 excavation (Plate VII). During the excavation undertaken in 1993 the remains of two buttresses (1993: G.2) were exposed (B7 and B8) at the north-eastern corner of the northern wing, mirroring those exposed in the south-eastern corner. These buttresses, previously discovered by Kelly (1992, 187), were of similar construction to the others. B8 was 0.9m deep (Plate VIII). Dumps and spreads of poured mortar and other materials were also 97 MARK HOULISTON PLATE VIII Buttress B8 revealed during the 1992 evaluation recovered during the 1993 excavation (1993: G.3). The lowest consisted of a fusion of poured mortar patches which thickened with its proximity to the intersection of B7 and B8. The spread abutted B7 and was probably formed during the construction of the buttresses. Phase 8: c. AD 275-325 In only two areas were deposits recovered that might date to the end of the life of the building. G.26 consisted of a layer of clay-loam mixed with large amounts of building demolition rubble, and it was located in the north-eastern corner of the area excavated in 1994, whereas G.27 comprised the backfill of the heptagonal feature and was located in the unexcavated area east of the main building. It, too, consisted of loamy soil mixed with large proportions of building demolition rubble. Care should be taken with the interpretation and dating of both these Groups, since either may relate to constructional activities associated with earlier phases of activity. 98 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Post-Roman activity [Phase 9] G.25, 28, 45, 46 and 47 consisted of assortments of post-Roman features of unknown function. G.29 corresponded to the southern limit of 1843 excavation and its backfill, whereas G.48 consisted of materials backfilled into rooms 'A' and 'D' at the end of that excavation. Dating of the Masonry Villa In contrast to Phases 1 to 4, where the assemblages of pottery recovered were far larger and more diagnostic than those of the equivalent phases from former excavations in the area (Pollard, 1992, 214-26), the assemblages from the masonry villa phases were considerably smaller. Because of this, the dates already established have been retained as a framework for the current discussions, modifications being made only where the new information extends or refines the old. Few datable artefacts were retrieved during the recent excavation from deposits associated with the construction and occupation of the first masonry villa (Phase 5). Those which belong to this Phase did not extend the date range beyond that already established for Phases 3 and 4 (c. AD 175-225), though the emphasis of the range is slightly later than that previously suggested by Kelly (Kelly Phase 1). The silty loam layers (G.33) incorporated a few later sherds, and one of the pits located in the south-western corner (G.l 1) contained a fragment of a shale bracelet of a type normally found in third- or fourthcentury contexts, but either of these may have been contemporary with one of the later phases of occupation. It is probable that a number of the pits assigned to the first period of villa occupation (Phase 5) were contemporary with the second phase of activity previously identified in the northern courtyard (Kelly Phase 2). However, as none of the recently excavated contexts could be tied in with these developments, the date range assigned to them (c. AD 225-50) has been retained. Slight adjustments have been made to the dates suggested by Kelly to accommodate recent changes in ceramic dating, and to keep the ranges expressed in this report internally consistent. The potential of the artefact assemblages to date any of the later additions to the masonry villa (Phases 6 and 7) is minimal, though the deposits (G.24) backfilling the trenches created by the removal of room W prior to the construction of the Phase 7 buttress (B3) contained a small group of artefacts datable to the late third or fourth centuries. These included New Forest ware, 'ceramique a l'Eponge\ 99 MARK HOULISTON and a fragment of a shale bracelet. Generally the assemblages from Phases 6 and 7 were indistinguishable, and their dating is consistent with the latest range suggested by Kelly (c. AD 275-325), though this had to be slightly extended to include the possibility of fourthcentury occupation. Deposits assigned to the possible demolition of the villa contained only small quantities of ceramics and no other datable artefacts. Some late Roman pottery, including sherds of an Oxford Parchment mortarium dated c. AD 240-300, was recovered, though this came from the backfills of the heptagonal feature (G.27) and should be treated with caution since the feature may have been filled in many years before the demolition or abandonment of the main villa buildings. Small quantities of residual and unstratified early fourth-century pottery sherds were recovered from both the recent and previous excavations (Lyne, 1994). This is consistent with the view that occupation at the site continued into the early years of the fourth century. Discussion of the Masonry Villa The villa is not of the 'courtyard' type as previously supposed (Kelly, 1992, 184-188) but conforms to the more common 'winged corridor' design. It appears to have been built as a single period structure, for there were no butt joints or changes in fabric to indicate that earlier walls were incorporated within it. This view is reinforced by the surprising degree of symmetry observed about both the east-west and the north-south aligned central axes. A comprehensive ground clearance preceded the construction of the masonry villa, and the identification of a single truncation horizon over a large area of the site is a further indication that development occurred during a single phase, rather than during a succession of lesser phases. Although all the wall foundations were constructed using the same basic arrangement of clay, mortar and ragstone, differences were observed. The four longer walls in the two wings and the three in the main range were wider and constructed using level courses of ragstone blocks, while the cross walls were narrower and constructed in a herringbone fashion. It is assumed the longer walls were the main load-bearing walls of the villa. A courtyard surface appears to have been laid across the front of the building at this time. The surface consisted of hassock, a by-product of ragstone quarrying, which suggests that the stone used in the construction of the villa came from a local quarry. The identification of a horizon contemporary with the construction 100 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 of the Phase 5 villa during the recent excavation was important because it established a direct link between the main villa ranges and the excavated deposits to the south. This helped to create a broad phasing for many of the layers and features investigated, though most could not be tied down to individual periods of development. The situation was slightly worse in the south-western corner where there was no stratigraphic link with the construction horizon and the masonry ranges. Some features in this area could be assigned to almost any of the site phases. Generally there were few clues to indicate the original functions of the features assigned to Phases 5-7, though the shapes of many of them suggested that they were constructed to hold vertical posts. A few of the larger features may have been cut for the disposal of domestic waste. The remains of a two silty loam layers (G.33) were identified. However, these could not be assigned to any but the broadest phase bands. There has been no evidence from the recent or previous excavations to indicate what the functions of any of the rooms may have been, and no suggestions as to which may have been heated, though floor surfaces including opus signinum, tiles and tesserae were observed. The functions of the long rooms located along the eastern and western sides of the main range which joined the projecting wings at each end, are less conjectural, since such arrangements are a common feature of 'winged corridor', villas in Britain. Generally, these rooms consisted of colonnaded corridors or porticoes, and were built against the front wails of villa buildings. One of the unusual features of the arrangement at the Mount is the existence of such porticoes on both sides of the main range. The eastern room (G) faced the direction of the Roman road to Rochester and, as such, probably contained the main entrance to the villa. The Phase 6 rebuild of the outside wall of this room contained what appears to be the base for a column marking the southern wall of such an entrance. The western portico overlooked the river and may have been built with this vista in mind. The functions of the porticoes would have been to link up with the internal corridors (rooms E and F, and L and R), and the linking room in the north-west (room M) to connect all the various internal rooms of the building. The porticoes would also have contributed to the overall appearance of the villa, creating an impressive sight to all passing travellers, whether they were viewing it from the main road or from the river. The four rooms of the bath suite were attached to the main building, where they extended the line of the western corridor, thus demonstrating that they were conceived as an integrated part of the original structure. Associated contemporary features such as service 101 MARK HOULISTON buildings and courtyard spreads may have existed in the area to the east of the bath suite, though the only features like these excavated belonged to Phase 5a. By the second quarter of the third century the bath suite had fallen out of use and had become partially demolished. The reasons for this are unclear. Although new rooms (7-10) were constructed at this time on the northern side of a new courtyard there is no evidence that they were constructed as a replacement bath suite. The new suite may have been contained in a detached block located away from the main building. Further developments during this period included a lean-to built along the eastern side of the northern courtyard, which had a clay lined tank at its northern end. By contrast no contemporary developments appear to have taken place in the main building. Later, during the end of the third century or the beginning of the fourth (Phase 6) a number of alterations were made to the main villa building. A new room was added to the southern side of the southern wing, and the eastern portico and parts of room D were rebuilt. New yard surfaces were laid in front of the building on its eastern side. Changes in the northern courtyard were more substantial; they included the rebuilding of the original bath suite with the addition of an extra room, and the construction of a possible kitchen against the northern side of the northern wing. Later on, but still during the same period of time (Phase 7: c. AD 275-325), large buttresses were added to support the southern wing and the eastern corner of the northern wing. This necessitated the removal of the room attached to the southern side of the building. It is interesting to note that despite the fact that this room had its foundations removed almost in their entirety, presumably for reuse, no evidence of reused fabric was observed in any of the exposed buttresses. This may indicate the existence of additional contemporary building activity, work that was not detected during the recent,' or previous, excavations. Variations were observed in the lengths of the buttresses, though it should be noted that the excavated remains consisted of foundation material and not superstructure. It is significant that although the main villa was constructed without buttresses the need for such supports became necessary later on. The evidence suggests that the buttresses were added because problems occurred with the stability of the original structure. This is surprising given the depth of the foundations, although the problems may have been caused by the effects of the spring line which crossed the area. Alternatively their incorporation may have been made necessary by the addition of an extra floor. Although it would have been easier to extend the building on virgin ground than upwards 102 PLATE IX A reconstruction by Ivan Lapper, showing the villa in its fully developed state (Phase 7) EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 such an alteration may have been undertaken if the motivation was status. The villa appears to have gone out of use during the same period (Phase 8: c. AD 275-325), though dating evidence was poor. This may imply that the changes made necessary by the addition of the buttresses were ultimately unsuccessful, though other explanations can be sought. The reconstruction drawing (Plate IX) represents an attempt to show the villa in its fully developed state (Phase 7). Evidence came primarily from the results of the excavations at the Mount, though this was interpreted using a variety of external sources. Observations of the fabric of the walls revealed that the four long walls in the two wings, and the three in the main range, formed the principal load-bearing walls of the building, an important factor when considering the arrangement of the superstructure. Three separate reconstructions were considered and drawn up, based on three distinct roof and wall arrangements. The version presented in this report is the simplest; it consisted of a building comprising three large pitched roofs, one above each of the two wings, and one over the central range. Each was supported on joists which lay across the lines of the load-bearing walls. The central roof, although of a much wider span, was lower, which allowed it to butt against the walls of the two wings. When considering the heights of the various parts of the villa it was interesting to note that although the structure was built without buttresses the need for such supports became necessary later on. Whatever the reasons behind their construction their presence indicates that in its final development the wings of the building were at least two floors high. The evidence from the villa site at Meonstoke (de la Bedoyere, 1991, 134) shows that two storeyed structures were known in Roman Britain. It also shows that such structures could be supported on foundations that were less substantial than those at Maidstone. In the reconstruction painting the wings have been interpreted as having two floors, but the main range is only three quarters of the height of the wings. In such an arrangement there would have been space for upper dormitory levels in rooms H, I, S and T, while the central rooms, U and V, could have had the height of one and a half rooms. The pairs of rooms located along the sides of the wings (E and F in the south, and L and R in the north) were narrow, and may have been corridors, perhaps with stair access to the upper levels. On the eastern and western sides of the building the colonnaded corridors or porticoes must have been built in such a way as to allow 103 MARK HOULISTON light into the main rooms beyond. The porticoes are one storey high in the reconstruction. This would have allowed them to butt against the sides of the main range (one and a half storeys high) but still allow light to pass through windows located in the wall above them into the rooms beyond. For the purposes of the reconstruction the northern service courtyard (Kelly Phase 2) has been retained, as has the water basin located in the area east of the eastern portico, though this may have gone out of use as early as Phase 5. The gardens to the east and south are purely hypothetical, as is the area to the north. There are reasons for proposing the existence of the wharf shown in the background, though again this is not based on any excavated evidence. THE FINDS The objects recovered from these excavations largely amplify those discussed previously by Kelly (1992, 209-35). At the same time, there is evidence, principally from the ceramics, both for activity prior to the masonry villa and for a continuation of occupation there into the fourth century. THE SMALL FINDS (Ian Riddler) The small sequence of 46 objects and 49 fragments of waste material of Roman date amplify earlier discoveries and, taken together, they provide an indication of the material culture of the villa, as well as providing some information about the range of activities which were carried out there (Kelly, 1992, 209-14). Few can be closely dated, but most undoubtedly lie within the chronological range established previously and there are no finds, other than ceramics, of an earlier date (Kelly, 1992, 181-2). The most significant discoveries may well be those which indicate the range of handicrafts carried out at the villa. The production of textiles is attested indirectly by the presence of spindle-whorls and needles, and the smithing of iron (if not its smelting) is borne out by the small assemblage of metallurgical residues. Kelly had previously drawn attention to a quantity of antler waste, most of which came from contexts of third-century date (Kelly, 1992, 213-4). Sections of red deer antler tines and beams had been smoothed in part and prepared for splitting into quadrants, using a technique which is well-attested in post-Roman antler assemblages (Ulbricht, 1978, 25-32). The waste gathered from the Trust's excavations comes from two related contexts of late second- or early thirdcentury date and adds further detail to the earlier discoveries. The objects are described here by functional category. Small find number (SFN), Group (G.) and phase information is given for each object. 104 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Dress Accessories - Pins Two pins were recovered from the villa, one of which is made of copper alloy, whilst the other has been produced from bone. The copper alloy pin (SFN 271: G.l8, Phase 3, Fig. 8, 1) expands evenly from the point to the head, which is lightly tapered. It belongs to Cool's type 24, the dating of which is centred on the second century (Cool, 1990, 170 and fig. 12.7). The corresponding version of this pin, produced in bone, belongs to Greep's type A2.1, whose floruit has also been given as the second century; the type went out of use around the middle of the third century (Greep, 1995, 1117). At both Canterbury and Colchester copper alloy pins of this type have come from fourth-century contexts, but these examples are thought to be residual (Blockley et al., 1995, 1013 and fig. 424.278; Crummy, 1983, 30 and fig. 31.507). The bone pin (SFN 14: G.49, Phase 2, Fig. 8, 2) has a globular head and lightly swollen shaft and it can be ascribed to Crummy's type 3 and Greep's type B 1.1 (Crummy 1983, 20-2; Greep, 1995, 1117). This is the most common type of Roman bone pin, both in Kent and elsewhere (Philp, 1981, 156-9; Greep, 1995, table 28; Crummy, 1983, 21-2). One example from previous excavations at the Mount has also been published (Kelly, 1992, 210 and fig. 11.10). Pins of this type occur in contexts from the middle of the second century onwards, although they are most familiar in deposits of third- and fourthcentury date. This particular example comes from a context of late secondcentury date and is therefore reasonably early within the sequence of pins of this type. It is inherently likely that the bone pins from the Mount were made I » @ 0 1 Fig. 8. Copper alloy pin (no. 1); bone pin (no. 2). Scale 1:1 105 MARK HOULISTON at the villa, as was the case also both at Chalk and Springhead (Johnston, 1972, 137-9 and fig. 17; Penn, 1957, 74; MacGregor, 1985, 44). Dress Accessories - Brooches (Don Mackreth) All of the brooches are made from a copper alloy. They are described here by type. Late La Tene: (SFN 712: G.18, Phase 3: Fig. 9, 3) Forged. The integral spring has four coils and an internal chord. The bow is relatively broad at the top, tapers to a pointed foot and has a rounded front. Although classifiable as a Nauheim Derivative, this brooch is at some remove from the earlier stages which represent in form and decoration the original, only lacking the necessary framed catch-plate; the front of the bow is rounded and the basically triangular appearance of the bow has given way to something much weaker. However, as the bow is still in essence much broader than deep, it just falls within the general class. The dating of similarly weak and plain designs is: Baldock, 25-80 (Stead and Rigby, 1986, 109, fig. 41,28); Hod Hill before 50 (Brailsford, 1962, 7, fig. 7, CI8: Richmond, 1968, 117-9); Verulamium, Claudian sherds (Stead and Rigby, 1989, 17, fig. 10, 7); Chichester, Claudian (Down, 1978,280, not illus.), 43-60 (Down, 1974,144, fig. 8.15,12) and 44-Flavian (Down, 1989, 186, not illus.); Gorhambury, 43-62 (Neal et al., 1990, 115, fig. 121, 1); Colchester, 49-61, two examples (Hawkes and Hull, 1947, 312, pi. 92, 55-6); Chilham, Kent, mid first century (Jessup, 1939, 270, fig. 3); Baldock, 50-70 (Stead and Rigby, 1986, 109, fig. 41, 26); Dorchester, Oxon., 50-70 (Frere, 1962, 137, fig. 27,3); Colchester, 61-65 (Hawkes and Hull, 1947, 312, pi. 92, 59); Fishbourne, 43-75, three examples (Cunliffe, 1971, 100, fig. 37,14,19,20) and before 75/80, three examples (ibid., fig. 16-8); Baldock, 70-90, two examples (Stead and Rigby, 1986, 109, fig. 41,33 and not illus.); Gorhambury, 75-100? (Neal et al, 1990, 115, fig. 121,4); Richborough, before 80/90 (Bushe-Fox, 1949, 108, pi. 25,3); Chichester, late first to mid/late second century, three examples (Down, 1978, 280, fig. 10.27,29; Down, 1989, 186, not illus.) Trumpets: (SFN 227, G.18, Phase 3: Fig. 9, 4) The spring is mounted between two pierced lugs behind the head which has a cast-on loop raised on a pedestal with two cross-grooves, each with a line of scorper-graver ornament. The trumpet head is plain and slightly shallow with an arris on each side. The knop is a single prominent and thick moulding with diagonal cuts, their direction alternating and the face between being slightly rounded. There is a pair of cross-mouldings above and a single one beneath at the top of the lower bow which is rounded between bordering grooves. The foot runs all the way round and is made up of three mouldings. The Trumpet has an overall range beginning before 75 and ending in the main 175-200. However, it may be worth looking at the dating of those which have the same spring-fixing arrangement as that used here: Harlow, Essex, P,1 80-100 (France and Gobel, 1985, 79, figs. 41,74); Derby, 80-140 (Dool et al., 1985, 291, fig. 128,29); Shakenoak Farm, P, first half second century (Brod- 106 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 o —ucaasnasb 1 o o ^ © i Fig. 9. Copper alloy brooches (nos. 3-6). Scale 1:1 ribb et al., 1971, 117, figs. 47, 66); Wroxeter, 110-20 (Bushe-Fox, 1913, 26, figs. 10,8); Silchester, 120-160/70 (Cotton, 1947, 145, fig. 8,2); Chelmsford, 120/5 - 160/75 (Drury, 1988, 94, fig. 62,10); Dover, P, 130-160/70 (Philp, 1981, 151, figs. 33,77); Nettleton, 140-150 (Wedlake, 1982, 127, figs. 53,51); Derby 150-175 (Dool etal., 1985,291-3, fig. 128,31); Canterbury, P, 175-300/ 20 (Mackreth, 1995, 963, fig. 404, 26). 107 MARK HOULISTON (SFN 276, G.3, Phase 1, Fig. 9, 5) The spring is mounted as that in Brooch 4. There are the stubs of a very small loop on the head which is properly trumpet-shaped. The knob has been replaced by a peltate plate, the arms ending apparently in knobs, although they join the main body of the brooch. The lower bow is broad, with a chamfer down each side, and tapers slightly towards the pseudo-penannular form at the foot. The only real evidence for the applied white metal trim which is invariable on the type is the trace here and there of the solder and the more definite mark for one eye and the strip down the centre of the trumpet. The group of brooches to which this example belongs has a plate of some form mounted as a replacement for the more usual knop. Whether the large size of this type (cf. Richardson, 1960) or the smaller one like the example here, both were tricked out with white metal trim, and it is the use of that which provides the main dating evidence. A review of its use suggests that it hardly dates before 125 and runs on into the early third century (Mackreth, 1996, 304-5). Zoomorphic: (SFN 49, Unstratified, Fig. 9, 6) The spring is mounted like those of the previous two brooches. The plate is in the form of a horse standing right. The ears are well pricked, and marked by grooves as is the mouth. The eye is a recessed annulus with mid-blue enamel. The head is set back on a long neck and the very long tail has a coil, to form a ring, and rises well above the back to end in two lobes. The body of the beast has two cells for enamel, the larger runs from the haunch to the shoulder and lower neck and is filled with discoloured material, the smaller one lies over the front leg has mid-blue enamel. The brooch had once had a plate of white metal trim covering the lower half of the neck, all the body and the tail, its presence being marked by a coating of solder. Brooches in the form of animals, as opposed to birds, are not as common as might be thought, the few which are seen make a disproportionate impression on the casual observer. This item is of British manufacture, not only because of the use of a sprung pin, but also because of the applied white metal sheet, a technique which is practically unknown on the continent. The date-range of this style of decoration was given under the last brooch. That would normally be enough comment, but there is a strong suspicion that brooches representing horses, especially a horse and rider, may have had an extended life in a religious context: brooches of this pattern were found in far greater numbers than could be accounted for by chance on the Late Roman site of Lamyatt Beacon, Somerset (Leech, 1986). There should be no doubt that they were made when the evidence suggests, in the second century, but could obviously have a greatly enhanced life in the right context. Dress Accessories - Shale bracelets (SFN 777, G.24, Phase 7: Fig. 10, 7) (SFN 782, G.ll, Phase 5: Fig. 10, 8) Fragments of two shale bracelets were recovered from contexts of late secondor third-century date. Both are undecorated and sufficient survives in each case 108 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 / !' \ «« 5' 8 i » \ Fig. 10. Shale bracelets (nos. 7,8). Scale 1:1 for their internal diameters to be measured, which extend to 50mm and 65mm. This places them within the range established for similar objects from Canterbury (Leveson-Gower, 1995, 1188-90). Both at Canterbury and at Dorchester two groups of bracelets have been identified, based on a consideration of their internal diameters (Calkin, 1972; Leveson-Gower, 1995, 1190). The smaller group, with diameters of 40-50mm., has been associated with children, and the larger group, with diameters of 65-70mm, with adults. On that basis, No. 7 would have been worn by a child and No. 8 by an adult. The majority of shale bracelets are undecorated and come from late Roman contexts. At Caerleon, Canterbury, Colchester, Portchester, Verulamium and Winchester, they have principally been found in fourth-century deposits, whether from settlement contexts or from graves (Evans and Metcalf, 1992, 187; Leveson- Gower, 1995, 1190-3; Crummy, 1983, 36-7; Webster, J, 1975, 226-8; Frere, 1972, 152-4; Goodburn and Grew, 1984, 77; Clarke, 1979, 301 and 312). As grave-goods, they are found more often in burials of the second half of the fourth century (Crummy et al., 1993, 136-7). Equally, however, undecorated shale bracelets are attested from third-century contexts on a variety of sites and those from Kent villas tend to be relatively early in date (Lawson, 1976, 250 and fig. 4; Allason-Jones, 1991, 273; Ocock and Syddell, 1967, 214; Penn, 1957, 105 and fig. 18.3). 109 MARK HOULISTON Household items (SFN 280, G.14, Phase 3: Fig. 11, 9) A complete iron slide-key of Manning type 2 was recovered from one of the post-holes of the aisled building. The straight end bit of the key includes three lightly-raised teeth. This is the most common form of slide-key to be found in Roman Britain (Manning, 1985, 93). A similar example came from Springhead, where a variety of keys were recovered (Penn, 1957, fig. 16). (SFN 212, G.40, Phase 5: Fig. 11, 10) The only spindle-whorl to come from the site has been fashioned from the base of a cream-slipped Upchurch type flagon. A similar spindle-whorl, cut from the base of a black burnished vessel, came from earlier excavations at the Mount (Kelly, 1992, 210 and fig. 11.12). Spindle-whorls produced from pot bases tend to be the most prolific to be encountered on Roman sites (Blockley et al, 1995, 1170 and fig. 517; Bishop, 1995, 32-3; Crummy, 1983, 67). (SFN 211, Unstratified: Fig. 11, 11) A fragmentary hone, which is sub-rectangular in shape and is relatively broad in proportion to its length, has been made from a fine-grained grey micaceous siltstone, possibly derived from the Hythe Beds. Transverse sharpening marks are visible on both faces. Similar hones are known from Roman contexts in Canterbury and they may have been widely distributed across Kent during this period (Blockley et al., 1995, 1210). Two hones found at Springhead may possibly have come from a similar source (Penn, 1957, 104-5). (SFN 559, Quern fragments, G.25, Phase 9: Not illus.) (SFN 693, Quern fragments, G.l 1, Phase 5: Not illus.) Twenty-six small fragments from basalt lava querns, weighing 539g, were retrieved from two separate contexts. All of the fragments are relatively small and it is not possible to distinguish between pieces of upper and lower stones, or to retrieve any metrical data. It has been suggested that the trade in basalt lava querns was most pronounced during the early Roman period, and that it may have been connected with military activity (Peacock, 1980, 50; Parkhouse and Evans, 1992, 191; Mcllwain, 1980). Equally, however, basalt lava quernstones are well represented in late-Roman contexts at Canterbury, Colchester and elsewhere, and it is likely that they remained in use across the entire Roman period (Blockley et al, 1995, 1206; Buckley and Major, 1983, 73-6). (SFN 355, G.16, Phase 3: Fig. 11, 12) A fragmentary pestle, produced from a piece of Lower Greensand, came from a context of late second- or early third-century date. The lower surface, which has a diameter of 50 mm, is largely intact, but the shaft has fractured. It may possibly have been of elbow-shape, in common with Purbeck marble pestles from Colchester and Richborough, but this appears to be unlikely, given the length of the surviving fragment (Crummy, 1983, 77; Dunning, 1968, 112). As with several examples from Richborough, the pestle may have been formed from the re-use of a section of stone. 110 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 13 Fig. 11. Iron key (no. 9), spindlewhorl (10), honestone (11), stone pestle (12), antler cylinder fragment (13). Scale, nos. 10,11,13 at 1:1; nos. 9 and 12 at 1:2 111 MARK HOULISTON (SFN 273, Ceramic gaming piece, G.l9, Phase 4: Not illus.) A disc, produced from a greyware sherd and 60mm in diameter, came from a context of late second or early third-century date. Discs of this type are known from a variety of sites of both Roman and post-Roman date but they have rarely been considered in any detail, until recently. Amongst the uses suggested for them are household weights, vessel lids or mats, reckoning counters and gaming pieces (Crummy, 1983, 94-5; MacGregor, 1978, 33; Addyman and Priestley, 1977, 139). In her detailed analysis of the series from Colchester, Nina Crummy concluded that it was most likely that they served as gaming pieces. With this in mind, it is worth noting that Roman board games were pursued both indoors and out, and it is possible that counters of this comparative crudeness and large size were better suited to outdoor games. Certainly, some of the boards incised on stone at Rome and Ostia are of a commensurate size for these pieces (Schadler, 1995, 89-95). Discs of this form could easily be stacked and it is conceivable that this took place with the race-games of XII Scripta and Alea, although it remains difficult to reconcile Roman gaming pieces and board games (MacGregor, 1976, 4). Industrial residues Twenty-seven fragments of ironworking residues, weighing a little under 1kg, were recovered from fifteen separate contexts. A further fourteen fragments came from layer 1, or were unstratified. There were no concentrations of material. All of the fragments have a porous, vesicular structure and they can be identified as residues from smithing, rather than smelting. Ironworking residues have attracted comparatively little attention in Kent villa studies. None are mentioned in accounts of previous excavations at the Mount, and they are absent from reports of excavations at Chalk, Darenth, Keston, LuUingstone, Orpington and Snodland, although they have been mentioned in accounts of other settlements in the area (Johnston, 1972; Philp, 1973, 59; 1991; 1996; Meates, 1987; Birbeck, 1995). Evidence for both smelting and smithing was revealed at Springhead, however (Penn, 1957, 70). The sparse quantity of material from the Mount does at least indicate that iron smithing took place there and it is a reasonable assumption that many of the pieces of structural ironwork found at the site were locally produced, using smelted material gathered either from the Weald, or more locally. The small quantity of residues recovered from these excavations is insufficient to allow for any broader economic considerations to be made (cf. Branigan, 1989, 47-8; Fulford and Allen, 1992, 204-5). One of the most interesting aspects of the results of earlier excavations at the villa was the evidence which emerged for antler working (Kelly, 1992, 210-4 and fig. 12). Ten fragments of worked antler came from a make-up layer of the second half of the third century. Other fragments of antler appeared to belong, in contrast, to Phase 1 of the activity within the villa, which was dated to the second half of the second century (Kelly, 1992, 181-2). A further eight fragments of antler, weighing 126g, were retrieved from two separate contexts of late second- or early third-century date. Several horn cores were also found in the same deposit. Two fragments of antler beam or tine have 112 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 been hollowed, but show no other traces of working. A further six fragments include two pieces which again show this characteristic hollowing of the interior of the beam, and two sections of junctions of tine and beam are also present. There is clear evidence, across the largest fragment, for the use of the saw as well as the knife in working the material. These fragments of antler are clearly related to the earlier assemblage, both for their dating and for the tools used in their working. At the same time, however, they differ for the practice of hollowing the antler beam before it has been cut into segments. The antler waste illustrated previously by David Kelly is particularly important for the evidence presented of the methods used in its working. It is clear that antler beams were separated laterally into sections, which have then been scored with the aid of a saw. These sections contain deep grooves and if the antler is tapped with a wedge, requisite sections of the beam will be removed. This method of antler working is well attested in the post-Roman period, both from fragments of waste and from the presence of antler wedges, one of which remains in situ in a section of antler from Haithabu (Ulbricht, 1978, 25-32 and taf. 26-7; MacGregor, 1985, 55-8 and fig. 34). The hollowing of sections of antler can be related to the production of lathe-turned antler cylinders, a fragment of which also came from the same group within the villa (SFN 282, B, Phase 4: Fig. 11, 13). It has been sawn laterally at both ends and the outer surface is scored by five unevenly-spaced grooves, which were added during the turning process. The fragment may have come from a cylindrical furniture mount, or possibly from a hinge. Objects of this general form are familiar from a number of sites and examples have been published recently from Caerleon, Exeter and London (Zenkiewicz, 1986, 207 and figs. 73.1 and 76.34-5; Evans and Metcalf, 1992, 190; Holbrook and Bidwell, 1991, 267 and fig. 122.6; Jones, 1987). This fragment is of the length and diameter of a hinge. It is unusual, however, for hinges to be produced in antler, rather than bone, and the majority were manufactured from cattle metapodia (MacGregor, 1985, 203-4; S. Deschler-Erb, pers. comm.). The closest parallel for this object is provided by a fragmentary bone cylinder from Caerleon (Zenkiewicz, 1986, fig. 76.34). Lathe-turned cylinders were produced from both bone and antler in a reasonably wide variety of shapes and sizes. Although they are generally thought to be furniture mounts, it can be difficult to establish the function of fragmentary examples. Latheturned cylinders were also used as dice boxes and unguent containers during the Roman period, although these objects, which were also produced in both bone and antler, are generally of a broader diameter than the fragment seen here (Biro, 1994,41 and 61-2). CERAMIC (Andrew Savage) A total of 4,067 'Belgic' and Roman sherds, weighing just over 35.5kg, was recovered from the excavations. Most of the pottery is of second- to early third-century date. Although a few sherds of pre-Conquest and Flavian-Trajanic date were recorded, these are all residual and no deposits could be dated, on the basis of their ceramic content, earlier than the mid second century AD. 113 MARK HOULISTON At the other end of the scale, several sherds of probable fourth-century date were recovered from Phase 7 deposits associated with the final period of occupation of the villa buildings. The quality of the assemblage as a whole is poor. Most of the sherds are small and worn, or are heavily weathered, resulting in some cases in a complete loss of the original surface. This has led to some difficulties of identification. It is almost certain, for example, that quantities of black-burnished ware (hereafter abbreviated to BB2) have, in the absence of a visible slip or burnish, been subsumed within the broader fabric category of North Kent Reduced Sandy ware. No distinction was made between these fabrics by Kelly in his earlier account of Roman ceramics from the villa (Kelly, 1992, 214), apparently on the basis of reservations expressed by Monaghan and Pollard regarding the use of the term 'BB2' (Monaghan, 1987, 171-2; Pollard, 1987, 198-9; 1988, 88). These reservations concern the variations in fabric and form within BB2 and associated fabrics which are encountered close to their production sites in the south-east; some of these variants do not meet the generally accepted defining criteria for BB2 proposed by Farrar (1973, 84) which were based on examination of the narrower range of BB2 types found in northern Britain. Nonetheless, the term BB2 is still both widely used and broadly understood and Pollard himself used it in his discussion of pottery from the villa and in his quantification of forms in Kelly's Phase 2 construction deposits Q21-X22/3-4 and B/l/21-22/3 (Pollard 1992, 232-35). For these reasons the term is retained in this report. In general, vessel profiles are severely truncated, largely due to breakage at the junction of rim and neck or shoulder. In addition, much of the pottery was residual in the contexts in which it was found, and this is most obvious with the material from phases 4-8. As a consequence, the dating of the masonry villa sequence, as suggested by the associated pottery and the identification of other salient characteristics of the villa assemblage, must continue to rely principally on the evidence published by Kelly (1992, 214-26). Material associated with Phases 7 and 8 does, however, indicate the possibility of abandonment in the early fourth century rather than the late third century, as he has suggested. This report concentrates on the discussion of material from selected groups in phases 1-3, supported by the quantification of fabrics and forms and by the illustration of selected representative pieces. These three phases are considered to pre-date Kelly's Phase 1, the earliest phase of the masonry villa, which was dated by him to the second half of the second century (Kelly, 1992,190-3). Pottery recovered from phases 2 and 3 of the recent excavations suggests that the construction of the masonry villa might now be placed slightly later, in the late second century. The assemblage is dominated by both coarse and fine wares of north-west Kent/Thameside origin. The most significant additions to the range of fabrics seen previously comprise pre-conquest flint-tempered wares and glauconitic sandy wares. Among the coarse wares, reduced sandy wares and blackburnished types predominate, in addition to a smaller grog-tempered component. Finewares, which are mostly Upchurch-type fabrics Nl-4 (Monaghan, 1987, 252-3), comprise c. 12 per cent by sherd count of the total assemblage. Pollard's quantified late second to mid third-century group from the Mount 114 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 (Pollard, 1992, 232-3) included just 3 per cent fine wares, assessed by Estimated Vessel Equivalents (EVEs), of which 2.3 per cent were fine reduced and oxidised wares (Upchurch-type wares) of a possible local origin. The proportion of fine wares is described there as proportionately small when compared with assemblages from other similarly dated sites in the area, including Rochester, Springhead, Chalk and Cobham Park (Pollard, 1992, 233). In contrast, the quantified groups from phases 1-3 of the excavations described here include 15 per cent of fine wares, when quantified in the same manner. Unfortunately, however, the total EVEs value for all forms in these groups amounts to only 468 (equivalent to 4.68 vessels), compared with a figure of 4214 (42.14 vessels) in Pollard's group, making detailed comparison difficult. All the pottery has been examined and identified by fabric type and the fabrics in each context have been quantified by both sherd count and weight. A complete list of the fabrics present and a discussion of the criteria used to determine the fabric groupings is held in the archive. The pattern of incidence of the ceramics broadly follows that which has already been observed on the site and previously discussed (Pollard, 1992, 232-5). Identifiable 'Belgic' and Roman forms in selected groups within Phases 1-3 have been quantified by EVEs using rim sherds, and the results are presented in Table 1. This table also records all the fabrics identified in these groups with the exception of the Iron Age flint-tempered wares. Within quantified groups, representative examples of all form types have been illustrated (Figs. 12-14). In describing vessel forms reference has been made, when possible, to the typology established for Upchurch and Thameside Roman pottery (Monaghan, 1987). When describing BB2 forms it has been considered appropriate to use this locally-derived typology in preference to Gillam's work (Gillam, 1970) which, although frequently used, is based on material from northern frontier sites. Wherever appropriate, descriptions have been formulated to be compatible with the nomenclature used previously by Kelly and Pollard. In the following discussion the ceramics are considered by phase. In general, this report is restricted to an assessment of the dating evidence provided by the pottery, rather than any broader considerations of ceramic trends within this area and period. The quantity of material and its condition preclude any attempts to examine functional trends in pottery supply to the site, as has been undertaken recently for Beadlam Roman villa, for example (Evans, 1996,85-7). Early residual material Although no features could be dated earlier than the mid second century AD on the basis of associated finds, small quantities of residual pottery in later contexts attest to later Iron Age and Flavian-Trajanic activity in the area. There are twenty sherds of flint-tempered pottery of first millennium BC to first-century AD date (Macpherson-Grant, 1996). Most of these have been identified and dated on the basis of their tempering characteristics. Seven, including the rim of a simple closed form and a comb-finished body sherd containing a moderate to profuse quantity of glauconitic sand, might be of later Iron Age to 'Belgic' transitional type (c. 100/50 BC-AD 50). Two others, one of which is also combfinished, can probably be dated to c. AD 25-75/100. The remaining sherds can 115 MARK HOULISTON only be assigned a broad Iron Age date. They are probably no earlier than c. 600 BC and might even be contemporary with the Late Iron Age/'Belgic' material. At least one sherd of grog-tempered ware, representing ajar with an internally thickened bead-rim comparable with Thompson type C3, is probably pre- Conquest in date (Thompson, 1982, 234-7). A number of other grog-tempered and shell-tempered sherds, although they lack diagnostically early typological traits, might also date to that period, but they could equally well be later in date. A small number of residual sherds of glauconitic sandy ware were also noted. These wares are particularly abundant in the area of the Medway valley (Pollard, 1988, 31). They are considered to be characteristic of the first millennium BC, their manufacture perhaps declining in the early first century AD (Pollard, 1988, 32). The glauconitic wares identified here, which include a few TABLE I. QUANTIFICATION BY ESTIMATED VESSEL EQUIVALENTS (EVES) using rim sherds of forms in selected groups in Phases 1-3. Fabric and Form North Kent Reduced Sandyware: Closed forms Pie-dish with roll-rim of rounded profile Pie-dish with roll-rim of triangular profile Dish Dog-dish Jar Everted-rim jar Everted-rim jar of 'cooking pot' type ?Neckedjar Necked, roll-rim jar Bead-rim jar Everted rim jar or bowl Bead-rim bowl Lid-seated jar 'BB2' Dog-dish Pie dish with roll-rim of triangular profile (decorated) Pie dish with roll-rim of triangular profile Pie dish with roll-rim of rounded profile Dish (decorated) Dish Everted-rim jar of 'cooking pot' type Phase 1 Total * 9.0 * * 5.0 2.8 3.9 4.3 * * % 22.9 12.7 7.1 9.9 10.9 Phase 2 Total * 31.4 3.0 * 40.7 * 39.82 23.5 21.5 6.1 8.0 5.3 40.0 * % 10.1 1.0 14.0 13.7 8.1 7.4 2.1 2.75 1.8 13.8 Phase 3 Total * * 3.0 * 11.8 25.4 * 6.2 9.8 8.0 4.2 5.6 * * % 2.2 8.8 18.9 4.6 7.3 5.9 3.1 4.2 116 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Fabric and Form 'Belgic' Grogged Ware/'Patch Grove' Closed forms Jar Everted-rim jar Probable jar Everted-rim bowl Unclassified 'Patch Grove' Ware (grog-tempered) Closed forms Jar Everted-rim jar Everted-rim jar or bowl 'Belgic' Shelly Ware Dish Jar Glauconitic sandyware Closed forms Jar Jar or bowl Unclassified 'Belgic' Grogged Ware Closed forms ?Hand-made Grog-tempered Ware Closed forms Chaff-tempered Ware Closed forms Canterbury Fabric LR2 Jar Jar or bowl Fine White-Cream Ware Flagon or beaker Misc. Coarse Buff Sandyware Closed forms Other Coarse Ware Closed forms ?Gaul or South-east England fabric Mortarium ?Kent fabric Mortarium South Spanish Amphorae Dressel 20 Phase 1 Total % 2.2 5.6 * 4.8 12.2 * * * * * * Phase 2 Total % * * 7.60 2.6 * 4.50 1.6 * 2.80 0.1 12.90 4.4 * 4.50 1.6 * * * * * Phase 3 Total % * 18.3 13.6 * * * 12.9 9.6 * * * * * * * * * * 117 MARK HOULISTON Fabric and Form Fine reduced Upchurch-type ware Closed forms Beaker Neckless, globular beaker Poppyhead beaker Beaker/flask ?Carinated bowl Everted-rim jar Carinated jar Flanged dish Unclassified ?Flask Fine oxidised Upchurch-type Ware Closed forms Cupped-mouth flagon ?Flagon Carinated bowl Colchester Colour-coated Ware Beaker Simple-rim beaker Comice-rim beaker Fine Buffware Comice-rim beaker Phase 1 Total * * 8.9 2.7 Central Gaulish 'Rhenish' Colour-coated Ware Beaker ?Early Nene Valley Colour-Coated Ware Bag-shaped beaker Lower Rhineland Fabric 1 ('Cologne' Ware) Beaker Totals 39.3 % 22.6 6.9 100 Phase 2 Total * * 9.5 5.4 * * * 24.3 2.8 * 293.6 % 3.3 1.9 8.4 0.1 100 Phase 3 Total * * 13.7 * 6.2 * * * * * * * 134.6 % 10.2 4.6 100 All the fabrics present are listed, with the exception of Iron Age flint-tempered wares. Quantified groups: Phase 1, Groups 2 and 3; Phase 2, Groups 5-9; Phase 3, Groups 14 and 16. For ease of comparison this table is closely modelled on that used by Pollard in his quantification of pottery from the villa (Pollard 1992, 234-5). Within each phase the total given for an individual form refers to its EVEs value; the percentage describes that figure as a proportion of the aggregated EVEs values of all quantified forms from a phase. The presence of forms which are not represented by rim sherds is indicated thus *. drastically truncated simple squared rims, demonstrate a variety of temper mixes, including pure sand, sand-and-grog and sand-and-flint. At least one sherd displays 'Belgic' style comb decoration. In general terms, this residual Iron Age pottery means that an early date cannot be entirely precluded for the Phase 1, G.l ditches which, unfortunately, were aceramic. 118 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 A possible spatial differentiation in the distribution of flint- and glauconitetempered wares is suggested by their absence from deposits excavated by Kelly. None were identified by Pollard (pers. comm.) and none were seen by the present writer in the course of his own examination of the relevant material. This pattern of incidence may, however, reflect the differing nature of the two excavations. Most of the early pottery came from features which lay outside the boundaries of the villa structure, and Kelly excavated few deposits of this type. Other early pottery types include a few sherds of fine grey Upchurch-type ware of late first- to early second-century date. Most are carinated beakers and bead-rim platters, cf. Monaghan classes 2G and 7A (1987, 68-71, 158-9). There are also several examples of wheel-thrown North Kent reduced sandy ware bead-rim jars. These, whilst possibly dating to the second half of the first century, could be as late as the mid second. The scarcity of pottery of mid firstto early second-century date was also noted by Kelly, who indicated a 'general absence of pre mid second-century pottery on the site' (Kelly, 1992, 193). Phase 1: c. AD 130-75: G.2 and 3: Table 1; Fig. 12, Nos. 1-9 Only three sherds of pottery were recovered from G.2 contexts. All represent closed forms, two in hand-made grog-tempered ware and one in wheel-thrown North Kent Reduced sandy ware. The latter sherd is likely to be Flavian in date but may possibly be much later. Given the shallowness of these features and their truncation, an intrusive origin for this material cannot be precluded. In G.3 contexts the coarse wares comprise mainly North Kent reduced sandy wares (43 per cent by sherd count). Less abundant fabrics include grogtempered ware (10 per cent), shell-tempered ware (5 per cent) and Thameside BB2 (9 per cent). Finewares mainly comprised 18 per cent fine reduced and oxidised Upchurch-type ware (Fig. 12, No. 1), in addition to two sherds of central Gaulish samian and a single sherd of a cornice-rim beaker (Fig. 12, No. 2) which is of probable north Gaulish or south-east English origin. Useful dating evidence for activity in this phase is provided by BB2 and reduced sandy ware triangular and rounded roll-rim pie-dishes of Monaghan classes 5D and possibly 5C (Fig. 12, Nos. 6 and 7). Decorated class 5D vessels were made from c. AD 120. Undecorated class 5C vessels, although they may also have been made from around the same date, are considered to mostly post-date c. AD 150 and had largely supplanted class 5D by the late second century (Monaghan, 1987, 140). The poor condition of these sherds means that of four dishes in Phase 1 contexts only one, a decorated example, can confidently be allocated to either category (Fig. 12, No. 6). Other useful evidence includes two sherds of central Gaulish samian representing Drag 18/31 dishes dating to the mid second century, and a deep-sided reduced sandy ware dish cf. Monaghan class 5F4, dated AD 130/140-230/300 (Fig. 12, No. 9). None of the pottery is necessarily of late second-century or later date. Whereas Phase 1 produced only a single example of a necked, roll-rim jar, cf. Monaghan class 3H2, dated AD 150-250/300, similarly dated jars and bowls comprised 37 per cent of the pottery associated with Kelly's Phase 2 construction levels. 119 MARK HOULISTON The dating of the pottery associated with Phase 1 contexts is similar to that of Kelly's earliest deposit: a burnt layer (T18/7) 'immediately above the natural clay and below the make-up for the phase 3 floor' (Kelly, 1992,190). This provided a mid second-century terminus ante quem for his Phase 1 construction. The assemblage consists of small, worn sherds. The small size of the sample precludes any useful statistical analysis, and no significant chronological distinction could be made by other means between groups of material from different sets. Both the nature and condition of the pottery is therefore consistent with the suggestion that these deposits represent areas of a single, contiguous 'soil' layer. Early activity is represented by two residual sherds of glauconitic sandy ware. The presence of a few sherds of coarse, hand-made grog-tempered ware is of particular interest. Although these might date to the first century AD, none display characteristic 'Belgic' traits of either form or decoration. AU the sherds are plain and represent closed forms, which are probably jars. Hand specimens cannot be distinguished by fabric from a ware previously identified on the Mount site in late second to mid third-century contexts by Pollard (Kelly 1992, 233). If they are the same and are not residual 'Belgic' pieces his interpretation of the type as representing a localised, short-term shortfall in the supply of more typical north-west Kent sandy wares in the early to mid third century may need to be reconsidered. Hand-made grog-tempered wares were also identified at LuUingstone, where 'their production may have died out during the [second] century, or continued as an 'undercurrent' to the predominantly sandy coarse wares, as in East Kent' (Pollard, 1987, 21). Unfortunately, the evidence here is again inconclusive, and merely hints at the possibility of longer-term supply. Phase 2: c. AD 150-75: G.5-9:Table 1, Fig. 12, Nos. 10-21 and Fig. 13, Nos. 22-31 Material associated with the two parallel east-west ditches (G.8) is sealed by deposits associated with the aisled building (Phase 3, G.H and 16). This is particularly important, insofar as it suggests a terminus ante quem for the construction of the villa. The pottery was examined to see how long the ditches might have remained open, and whether they were contemporary. The shallow, silty lower fills of both ditches contained little pottery. The lowest fill of the north ditch yielded only two sherds of glauconitic sandy ware, both of which are of probable pre-conquest date. The material was sealed by a context which contained a late first- to early second-century carinated bowl of Monaghan type 4G (Fig. 13, No. 23). In the south ditch, the lowest fill yielded seventeen sherds, representing a single undecorated, coarse grey sandy ware roll-rim pie-dish of Monaghan type 5C4, dated AD 150/170-250 (Fig. 13, No. 27). Another context produced one sherd of a similar, probably undecorated BB2 pie-dish of Monaghan type 5C1, dated AD 120/150-230/250 (Fig. 13, No. 26). The upper fills of the two ditches represents material dumped into them during their backfilling. The south ditch yielded sandy wares of Hadrianic to Antonine date, two sherds of apparently undecorated BB2 roll-rim pie dishes of Monaghan type 5C4, dated AD 150/170-250 and one of type 5C1, dated AD 120 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 120/150-230/250 (Fig. 13, Nos. 25, 26 and 27) and two sherds of central Gaulish samian, one of which has been dated to the second half of the second century, whilst the other is considered to be of Antonine date. There were also a few sherds of a very hard-fired coarse grey sandy ware with purplish-brown external surfaces. Further sherds were recovered from Phase 3, 014. The ware has previously been recognised in West Kent, sherds occurring 'not infrequently, but ... in a small minority' (Pollard, 1988, 127). This fabric corresponds with Canterbury fabric LR2 common in East Kent in the late second to third centuries (Pollard, 1988, 99; 1995, 702-3). Pottery associated with the upper fills of the north ditch included a sherd representing a fine oxidised Upchurch-type cupped-mouth flagon (cf. Pollard, 1988, no. 159) which can be dated to mid-second to early third century. Nothing in the south ditch is necessarily later than this, although there was a sherd of Antonine central Gaulish samian and two BB2 dog-dishes of Monaghan type 5E2, dated AD 110/120-210/300 (Fig. 13, No. 24). Dating the ditches on the basis of their ceramic content is made difficult not only by the small sample - particularly from the lower (accumulative) fills - but also by the possibility that the lower and upper fills were mixed by various biomechanical processes, as described above. Nonetheless, pottery from the upper dump fills of the south ditch clearly suggests a termination date for this feature in the later second century, after c. AD 150. It might tentatively be suggested that it did not, on the basis of the small quantities of late second- to third-century material present, remain open much later than c. AD 175. Although the relative absence of late second-century material in the northern ditch might suggest an earlier termination date for this feature, the small sample sizes involved again make comparison difficult. Neither ditch contained any pottery that was necessarily of third-century or later date. The other quantified Phase 2 groups (G.5, 6, 7 and 9) together, contained only thirty-six small, worn sherds, which include few useful dating indicators. The material in G.5 can only be dated to the late first century or later. The other groups all contained pottery of AD 110 or later. G.6 yielded a coarse grey sandy ware roll-rim pie-dish, probably undecorated, of Monaghan type 5C1, dated AD 120/150-230/250 (Fig. 13, No. 26). G.9 yielded a similar dish, but with a rim of triangular profile, cf. Monaghan type 5C2, dated AD 110/150-210 (Fig. 13, No. 31). Although none of these groups yielded pottery of necessarily late second- to third-century type, which suggests that deposition may have ceased in the mid to late second century, the size and condition of the sample and the low proportions of contemporary pottery in later phases mean that a later date cannot be precluded. The other groups of this phase (10,17 and 49) yielded a little pottery of mid to late second-century date at the latest. Phase 3: c. AD 175-225: G.l4, 16 and 18: Table 1, Fig. 13, Nos. 32-37. These deposits, representing the earliest villa construction phase, are of considerable significance as they provide a terminus post quem for Kelly Phase 1 masonry building construction levels, which were dated to the second half of the second century. The construction of the masonry villa was 121 MARK HOULISTON placed in the second half of the second century on the basis of a very small sample of material, consisting of sherds from five pots, one of which, a BB2 dog-dish (Kelly, 1992, fig. 13, no. 1), could be dated to c. AD 130/140+. In the northern part of the villa, a terminus ante quem (sic) was provided by sherds of 'half-a-dozen' fabrics which included a fine reduced Upchurch ware copy of a samian Dr 38 bowl, dated to the second half of the second century. Although the pottery assemblage associated with the Trust's Phase 3 deposits comprises mostly small, worn or weathered coarse ware sherds, it still suggests that Kelly's Phase 1 construction can be placed in the late second century, at the earliest. G.H (construction backfills of the post-pits of the aisled building) yielded trace quantities of several fabrics which are largely absent in the Phase 2 deposits beneath them. They include a single sherd of an eastern Gaulish Curie 79 samian dish of late second- to early third-century date, a sherd of central Gaulish 'Rhenish' colour coated ware, which is most common in late secondto early third-century contexts in Britain (Pollard, 1988, 99), and eighteen sherds of a hard-fired sandy ware, cf. Canterbury fabric LR2. One of the latter sherds exhibits a burnished band on the shoulder, a trait which is typical of the ware in late second- to mid third-century contexts in East Kent, where it is most commonly found (Pollard, 1988, 99; 1995, 702-3). Smaller quantities of the latter fabric were recovered from Phase 2, G.8 contexts. Although some of the pottery may thus belong to the third century, one cannot be certain that it is not confined to the second century. Packing deposits contained sherds of coarse reduced sandy ware latticedecorated jars of probable 'cooking-pot' type which are likely to date to the Hadrianic period or later. The sample sizes are too small to allow for any meaningful comparisons between packing and backfill deposits. The pottery associated with the G.l6 beam-slots is broadly similar to that associated with the G.H post-holes. Although the assemblage lacked eastern Gaulish samian, context 250 contained a sherd of a rouletted globular or bag-shaped beaker in Lower Rhineland fabric 1, possibly similar to forms which are dated c. AD 180-250 (Anderson, 1980, fig. 8, no. 6 or 7) and examples of reduced sandy ware everted-rim jars of Monaghan types 3J9 and 3H7, both dated AD 170-250+. There is no reason to suppose, on the basis of the pottery, that G.H and 16 were not contemporary. A preponderance of decorated pieces among a number of BB2 and reduced sandy ware pie-dishes suggests that the pottery from G.18 soil layers is mostly of mid to late second-century date, although the assemblage also contains undecorated BB2 pie-dishes and reduced sandy and grog-tempered wares of broad late second- to mid third-century date. The material cannot be chronologically distinguished from that associated either with Phase 2 deposits or with the stratigraphically later groups in Phase 5 to which it can be related. Table 1 quantifies by EVEs the forms present in Phases 2 and 3. The data can be compared with a published late second- to mid third-century quantified group from the Mount (Pollard, 1992, 234-35) which was associated with Kelly's Phase 2 construction, dated to the first half of the third century, 'perhaps about AD 225-230' (Kelly, 1992, 202). 122 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Phases 4-6: c. AD 175-225: G. 11-13, 19, 20-22 and 30-44: Fig. 13, No. 38 The potential of the pottery to date activity associated with Phases 4-6 is minimal. The range of pottery forms and fabrics present is very similar to that associated with Phase 3 and the latest material is probably of broad late second- to early third-century date. There are also a few later pieces. Phase 4, G.30 contained two sherds representing a bead-and-flange reduced sandy ware dish (cf. Monaghan type 5A) and a Nene-valley colour-coated ware jug or flagon, both of which are likely to be mid third century or later. Unfortunately, however, these sherds may have been intrusive, as the deposit was much disturbed both by ancient and by recent activity. Material of similar date was recovered from G.33 in Phase 5: a sherd of another bead-and-flange dish and one of Oxford colour-coated ware. There are nineteen sherds of samian, representing fifteen vessels, none of which have been dated later than the Antonine period, and this suggests that much of the intrinsically undatable pottery dates to the second century. All of the groups contained generally small, worn or weathered sherds, with just one exception. C.204 in Phase 5 (G.32) contained a single, substantially complete but fragmented roll-rim necked jar in a fine oxidised fabric of probable Upchurch ware (Fig. 13, No. 38). The pot is a little unusual in exhibiting three 'pressed-out' bosses on the shoulder, a feature not previously noted by the present writer on vessels of Upchurch-type. Phases 7 and 8: c. AD 275-325: G.24, 26 and 27: Fig. 13, No. 39 and Fig. 14, Nos. 40-48 The Phase 7 (G.24) trench segments yielded a small group of pottery, which included several sherds of probable late Roman grog-tempered ware. This ware is generally assigned a fourth-century date in West Kent. Pollard suggests that it did not perhaps become 'a significant element in [West Kent] pottery assemblages until the second or third quarter of the century' (Pollard, 1988, 149). There was also a sherd of New Forest colour-coated ware and one of a probable ceramique a I 'Eponge bowl. Neither of these finewares are common in Kent and both are also usually assigned a fourth-century date there. A New Forest greyware bowl was found by Kelly associated with late third-century construction deposits (Kelly, 1992,224). A shale bracelet (SFN 777) of middle or late Roman date was also recovered from this phase. Ceramique a I 'Eponge is distinguished by a generally hard oxidised paste and a glossy, often yellow-orange slip on which may be superimposed darker blurred or mottled decorative motifs. Finds are concentrated in south-western France, between the valleys of the Loire and the Gironde (Richardson, 1986, 130). Useful descriptions and discussions of the ware have been published by Richardson (1986) and Fulford (1977, 45-6). It should also be noted that a quantity of material associated with Kelly's Phase 3 deposits, but which was not seen either by him or by Pollard included a sherd of Rettendon Ware, which is similarly of fourth-century date (M. Lyne, pers. comm.). Other deposits (Phase 8, G.26 and 27) probably associated with the demolition of the villa, contained a small quantity of pottery in poor condition. Only 123 MARK HOULISTON a single vessel in either group can be conclusively dated to the mid-third century or later, namely an Oxford Parchment mortarium of Young's type M17 (G.27), dated c. AD 240-300 (Young, 1977). A vessel of similar form and date was recovered from Kelly Phase 3 demolition layers. The present example was represented by several large, fresh sherds and it might tentatively be supposed that a utilitarian coarse ware vessel in such fine condition would not, if it were in regular use, have been discarded much after AD 300. Although the sample of material from Phases 7 and 8 is very small and the possibility of intrusion cannot be precluded, the presence of both pottery and small finds of probable fourth-century date in layers associated with the occupation of the villa clearly challenges Kelly's assertion that demolition occurred before the end of the third century. An absence of wares characteristic of the mid to late fourth century and common then, such as Portchester 'D' sand-tempered ware which was recovered from topsoil and unstratified deposits, suggests that demolition may have occurred in the early fourth century. The topsoil and unstratified deposits With the possible exception of layers in Phases 7 and 8 (above), no stratified fourth-century deposits were excavated. Activity of this period on or in the vicinity of the Mount site is certainly attested, however, by the presence of late pottery in the topsoil and other unstratified deposits. On the basis of a subjective analysis there were far higher proportions of late Roman grog-tempered wares, colour-coated wares, hook-rim sandy ware jars and bead-and-flange dishes in these layers than in stratified deposits (a finding duplicated by Kelly) and the corpus of wares is extended by the presence of Alice Holt ware, Rettendon ware and Portchester 'D'/Tilford greyware. All of these except the last were also recovered by Kelly in equivalent deposits. The illustrated pottery (Figs. 12-14) Phase 1. G.3: a silty clay layer, possibly an agricultural soil, 1. Fine reduced Upchurch-type ware. Neckless globular beaker with short everted rim, cf. Monaghan 2H, dated late first to early second century. C.428. 2. Fine buff ware. Cornice-rim beaker. Original surface missing. C.237. 3. Grog-tempered ware of 'Belgic' or 'Patch Grove' type. Everted-rim jar. Bodysherds of the same vessel exhibit a poorly applied patchy external burnish. C.896. 4. North Kent Reduced Sandyware. Necked roll-rim jar cf. Monaghan 3H2, dated AD 150-250/300+. C.896. 5. Shell-tempered ware. Dish with lid-seated 'bead'-rim cf. Pollard 1987, no. 287, dated first century AD and perhaps also first century BC. C.428. 6. BB2. Roll-rim pie-dish cf. Monaghan 5D1.8, dated AD 110/20- 180/200, Burnished lattice decoration. C.896. 7. North Kent Reduced Sandyware. Roll-rim pie-dish cf. Monaghan 5C1, dated AD 120/150-230/250. Original surface missing. C.237. 8. BB2. Roll-rim pie-dish cf. Monaghan 5C1, dated AD 120/150-230/250. 124 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 •v T5 -Js 10 11 / r y 13 14 15 17 r 18 ? \ 19 E -? ; 21 Fig. 12 Roman Pottery (nos. 1-21). Scale 1:4 125 MARK HOULISTON Most of the original surface is missing; it is not possible to say whether it was decorated. C.237. 9. BB2. Roll-rim pie-dish cf. Monaghan 5C2, dated AD 120/150-210. Most of the original surface is missing; it is not possible to say whether it was decorated. C.237. Phase 2. G.6: post-pits 10. North Kent Reduced Sandyware. ?Everted rim jar or bowl. C.772. 11. BB2. Roll-rim pie-dish cf. Monaghan 5C1, dated AD 120/150-230/250. Probably undecorated. C.772. Phase 2. G.8: fills of the U-shaped boundary ditches 12. North Kent Reduced Sandyware. Bead-rim jar cf. Monaghan 3E3, dated AD 10/40-150/170. C.194. 13. North Kent Reduced Sandyware. Bead-rim jar, cf. Monaghan 3E/F. C.415. 14. North Kent Reduced Sandyware. 'Hooked' bead-rim jar cf. Monaghan 3F1, dated AD 40-150/170 or 3F4, dated AD 70-150. Horizontal groove on shoulder. C.415. 15. North Kent Reduced Sandyware. Necked everted-rim jar, cf. Monaghan 3H1, dated AD 100-250/300+ or 3H2, dated AD 150-250/300+. C.415. 16. North Kent Reduced Sandyware. Very hard-fired. Lid-seated jar cf. Monaghan 3L1, dated AD 70-150; class 3L was popular in Kent from the late first to mid-second centuries AD. C.415. 17. North Kent Reduced Sandyware. Everted-rim jar cf. Monaghan 3H. C.415. 18. North Kent Reduced Sandyware. Everted-rim jar, cf. Monaghan 3J1, dated AD 110/120-150/190. C.194. 19. North Kent Reduced Sandyware. Everted-rim jar of 'cooking-pot' type with tooled lattice decoration, cf. Monaghan 3J1, dated AD 110/120- 150/190. C.415. 20. Grog-tempered ware of 'Patch Grove' type. Everted-rim jar with corrugated shoulder and neck, cf. Pollard 1988, no. 19 (Tester and Caiger, 1954, no. 6), dated second century. C.415. 21. Grog-tempered Ware of 'Belgic' or 'Patch Grove' type. Everted-rim jar or bowl with neck cordon. C.194. 22. Grog-tempered ware of 'Belgic' or 'Patch Grove' type. Everted-rim bowl. C.194. 23. Fine reduced Upchurch-type ware. ?Carinated bowl cf. Monaghan 4G, dated late first to early second century AD. C.669. 24. North Kent Reduced Sandyware. Simple-rim dog-dish cf. Monaghan 5E2, dated AD 110/120-210/300. C.668. 25. North Kent Reduced Sandyware. Roll-rim pie dish cf. Monaghan 5C1, dated AD 120/150-230/250. C.461. 26. North Kent Reduced Sandyware. Roll-rim pie-dish cf. Monaghan 5C1, dated AD 120/150-230/250. Undecorated. C.194. 27. North Kent Reduced Sandyware. Roll-rim pie-dish cf. Monaghan 5C4, 126 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 < 22 V 23 26 W \ 24 T * 25 \ 27 __^ V 29 r 30 31 32 / 33 \ 34 28 X 35 36 37 ^ 39 ^ Fig. 13 Roman Pottery (nos. 22-39). Scale 1:4 127 MARK HOULISTON dated AD 150/170-250. Most of the original surface is missing; it is not possible to say whether it was decorated. C.415. 28. Fabric and form as No. 27. Probably undecorated. C.415. 29. Fabric and form as No. 27. Undecorated. C.121. Phase 2. G.9: post-pits, fence line 30. North Kent Reduced Sandyware. Hooked bead-rim jar cf. Monaghan 3F6.1. Vessels in class 3F are generally dated within the range AD 40-150/170. C.689. 31. North Kent Reduced Sandyware. Roll-rim pie-dish cf. Monaghan 5C2, dated AD 120/150-210. Original surface missing. C.713. Phase 3. G.H: construction backfills, post-pits, aisled building 32. North Kent Reduced Sandyware. Lid-seated ?jar cf. Monaghan 3L1, dated AD 70-200/250. C.450. 33. North Kent Reduced Sandyware. Everted-rim jar of'cooking-pot' type, cf. Monaghan 3J3, dated AD 150-220/240. Possibly burnished on shoulder. C.411. 34. Grog-tempered ware of 'Belgic' or 'Patch Grove' type. Everted-rim jar. C.411. Phase 3. G.16: robber backfills, external walls, aisled building 35. North Kent Reduced Sandyware. Hooked bead-rim jar cf. Monaghan 3F3, dated AD 40-150/170. C.207. 36. Grog-tempered ware of 'Patch Grove' type. Everted-rim jar. C.250. 37. Fine reduced Upchurch-type ware. Flanged dish cf. Monaghan 5B5, dated AD 60/70-130. C.207. Phase 5. G.32: pit backfill 38. Fine oxidised Upchurch-type ware. Traces of a white or cream slip. Roll-rim necked jar with three 'pressed-out' bosses on the shoulder. C.204. Phase 7. G.24: backfilled robber trench representing the final phase of villa construction 39. BB2. Everted-rim jar of'cooking-pot' type cf. Monaghan 3J3, dated AD 150-220/240. C.729. 40. Fabric and form as No. 39. C.715. 41. Fabric and form as no. 39. C.586. 42. North Kent Reduced Sandyware. Everted-rim of a jar of possible 'cooking-pot' type cf. Monaghan 3J9, dated AD 170/190-250+. Traces of burnish or smoothing on top of rim and exterior. C.518. 43. North Kent Reduced Sandyware. Necked roll-rim jar. Smoothed/wiped surfaces. C.716. 44. North Kent Reduced Sandyware. Lid-seated jar with heavily moulded rim. C.716. 45. Grog-tempered ware of 'Patch Grove' type. Everted-rim jar. C.729. 46. Grog-tempered ware of possible late Roman type. Everted-rim jar. Faint vertical wipe marks beneath smoothed rim and shoulder. C.715. 128 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 y 40 41 " ^ 42 44 45 T y* ^~ ••.".-•V^y^ T 46 ) » « Fig. 14 Roman Pottery (nos. 40-48). Scale 1:4 47. 48. Grog-tempered ware of probable late Roman type. Everted-rim jar with shallow tooled 'chevron' decoration on shoulder. Patchily fired. Lightly and roughly burnished on exterior. C.715. Grog-tempered ware of probable late Roman type. Everted-rim jar. C.715. 129 MARK HOULISTON BUILDING MATERIALS (Louise Harrison) The combined assemblage from the 1992 and 1994 excavations consists of 1.780kg of brick, 9.815kg of flue tile, 17.490kg of imbrex, 40.635kg of tegulae and 2.975kg of miscellaneous, unidentifiable fragments. The most significant findings are included in this report. Records of the material derived from the earlier excavations (Kelly, 1992) are included in the archive. Tile The tile has been sorted by fabric and type, the former identified using a 10x/ 20x binocular microscope. A number of the fabrics have been classified only by their sandiness (i.e. the relative number of quartz grains) though these differences may not indicate a different tile kiln merely a different clay source. Fabric types and numbers of fragments are quantified in Table 2. Descriptions of Fabrics 1, 3, 7, 8, 10 and 11 have recently been published (Harrison, 1998, 150-151). The remaining fabrics identified at Maidstone are described below. The tiles are red in colour and of a hard, well-fired consistency unless otherwise stated. Fabric 2: very fine and neither sandy or micaceous. There are little or no quartz grains and few other inclusions. Fabric 4: usually dark red to purple in colour. It is characterised by a large quantity of inclusions, comprising abundant large white and clear coloured quartz grains up to 1 mm diameter and a large quantity of burnt flint and ironstone nodules also measuring up to 1mm across. Additionally the tile fabric has a distinctive very coarse sanding applied to the back of the tile, also consisting of the above inclusions. A number of examples of this fabric have TABLE 2. QUANTITIES OF BRICK AND TILE TYPES BY FABRIC Fabric Tegulae Imbrex Brick Flue tile Misc (unident- _ _ _ ifiable) 1 4 8 10 3 2 13 11 7 14 28 3 16 20 16 8 2 - 6 - 16 5 6 15 10 5 _ 1 2 5 14 2 „ _ 2 - m .. „ „ 44 „ _ . 9 3 2 9 130 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 been found in London from a variety of sites. Billingsgate Baths (early third century) produced considerable amounts of the fabric consisting mainly of brick fragments with a small number of roofing tiles (I. Betts, pers. comm.). This suggests that the fabric is likely to be a late type. Fabric 13: characterised by its sandiness. The fabric contains abundant medium sized (up to 0.5mm) white and clear quartz, but no other inclusions. Fabric 14: characterised by the sanding on the back of the tile which is very coarse (similar to fabric 2). It consists of large quartz and burnt flint nodules measuring up to 3-4mm in diameter. The fabric itself is relatively fine, and has a slightly sandy matrix with occasional white and clear coloured quartz grains. Table 2 highlights a number of interesting features. Tegulae appear abundantly in virtually all fabric types, suggesting that these tiles were produced at many, if not all, of the tile kilns represented. Fabric 1 is the most commonly found fabric present in all tile types, suggesting it was probably produced locally. However, it should be noted that the fabric is similar to one commonly found at Canterbury, attributed to the Whitehall Gardens or St. Stephen's tile kilns (Jenkins, 1956; 1960) and its presence may indicate trade between the two areas. Future work may clarify this situation. Flue tile and brick was only found in a small number of fabrics, suggesting that these tiles were specialised and were only produced at certain kilns. The fabrics which are poorly represented in the assemblage may suggest that the tile was made in a different, non-local area. Brick A small quantity of brick was retrieved from the excavation. Only one complete Bessales (Fabric 1) survived. Its dimensions are 190 x 186 x 40mm, which is relatively small according to Brodribb (1985, 34) who gives an average size of 198mm square. The other brick fragments have a thickness ranging from 45mm to 32mm, tentatively suggesting that the fragments were part of Bessales, Pedales or Lydion bricks. Only a small quantity possessed signature marks. For these see the discussion below. Imbrex Most of the imbrices were fragmentary. Only one virtually complete tile survived (Fabric 1), measuring 420 x 161 x 20mm. Tegulae Tiles have only been recorded as tegulae if the flange or other distinguishing feature, such as the cutaway is present. A total of 101 fragments was identified though only one could provide a breadth measurement (295mm tapering to 285mm). The tile also bore a signature mark (Type 2) and a lower cutaway of Type C. The flange types and their fabrics are listed in Table 3 which demonstrates 131 MARK HOULISTON TABLE 3. ROMAN BUILDING MATERIALS, TYPES, FABRICS AND QUANTIFICATION Flange types by fabric Flange Fabric 1 1,8,10,3, 2,7 2 1,8,10,13, 3,2,7 6 8,10 10 1, 8, 3, 2, 13,7 11 1,8,10 14 1 15 1 21 1,8 31 1,2,3,5,6 41 1 Quantities of cutaway types and their fabrics Cutaway type and quantity Fabric A 5 1,4,8, 10 B6 1,10,3, 13 C 7 1,8, 10, 7,2 El 8 Flue types by fabric Flue type Fabric 1 1,3 8 1 12 1,4, 3,13 13 1,3 14 1,3 Signature marks by fabric and tile type Signature mark Fabric Tile type 1 8,3, Teg, 14 brick, misc 2 1,7 Teg, brick, misc 3 1, Teg, 10 misc 5 1,4 Teg, brick, misc 20 1 misc that only flange types 1, 2, 10 and 31 are commonly represented, appearing abundantly in many fabric types. This suggests that these flanges were produced by many tile kilns, some locally and others further afield. Additionally, six tegulae displayed flanges (Type 11) which were unusually short, at approximately Hmm in height. These are not typical, considering that the average height of a flange measures up to 50mm or more. The tegulae displaying these flanges were all of the same fabric (Fabric 3), suggesting that they were produced at the same kiln and were possibly produced for a certain purpose. Lower Cutaways Four different lower cutaway types were present in the assemblage. Table 3 shows that lower cutaway types A, B and C were the most commonly used appearing in a variety of different fabrics, and that cutaway type E was rarely used. 132 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Flue Tile (Figs. 15-16, Nos. 1-10) The quantity of flue tile retrieved was small, with no complete tiles surviving. Only a small number were complete enough to provide keying patterns, and occasionally dimensions, and these are listed below. Because of the fragmentary state of the tiles it is possible that some may be allocated flue tile types when they could conceivably be voussoir tiles. A number of fragments bear the same combing as Black's Voussoir type series (Black, 1995, 1267-93) but it should be noted that it is usually only the keying, not always the dimensions, that are similar. Type 1 (Fig. 15, No. 1): a face fragment combed with a semicircle at top and base, with a diagonal cross in centre. Teeth: 5; width of stroke: 26-38mm. Type 8 (No. 2): a face fragment combed with a lattice pattern. Teeth: 5/6; width of stroke: 25-30mm, the width of the face is 107mm. f ^ ? » 1 i J i* Fig. 15 Roman flue tiles (nos. 1-6). Scale 1:4 133 MARK HOULISTON Type 12 (No. 3): a face fragment with two semicircles, one at top and one at the bottom. Teeth: 4/5; width of stroke: 25-30mm Dimensions present: height: 170mm; width of side: 123mm; width of face: 123mm (Black type M; Black, 1995: Marlowe Car Park). Type 10 (No. 4): a fragment combed with vertical wavy combing, consisting of at least two probably three bands. Teeth: 7; width of stroke: 42mm. Type 13 (Nos. 5,6): fragments with vertical wavy and straight combing. Teeth: 6/9; Width of stroke: 22-37mm. Type 14 (Fig. 16, No. 7): the fragment shows combing consisting of a diagonal cross with no border (the combing is similar to voussoir type 5 but as no side /. 8 <4f i 9 10 : Fig. 16 Roman flue tiles (nos. 7-10). Scale 1:4 134 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 survives it is impossible to identify as a voussoir). Teeth:6; width of stroke: 35mm approx; height of tile: 195mm (face). Type 20 (No. 8): (unusual combing) a fragment with a semicircle and a vertical line through it. There is also evidence of a possible cross represented by a diagonal stroke. Teeth: 6; width of stroke: 31mm. Type 21 (No. 9): (unusual combing) combed with vertical wavy bands (two or more) with a semicircle drawn through it. Teeth: 5; width of stroke: 28mm. Voussoirs There were only two tiles which could be positively identified as voussoirs and they were both corner fragments displaying the same combing. Type 5 (No. 10): a corner fragment displaying combing on both face and side. The visible keying consists of a diagonal cross on one side and a diagonal stroke probably another cross on the face. Additionally, the tile provides a height measurement which is 151mm. Tables 2 and 3 indicate that the flue tiles were generally made in the same fabrics (1 and 3). However it should be noted that these two fabrics are divided only by the amount of quartz grains in the matrix, and could therefore be variants of the same sandy fabric. Discussion The signature marks represented in this corpus are mainly incomplete and largely unclear. Only a small number of fragments bore identifiable signature (see Table 3). No signature marks have been illustrated for publication as the tiles were too fragmentary to give a good representation. Signature types 1, 2 and 5 are semicircles consisting of one, two and three bands. Signature mark 3 is a one-grooved 'loop' while signature mark type 20 is a three-grooved loop. A total of three hobnail boot impressions was present in the assemblage all on tegulae and miscellaneous fragments in Fabric 1. Additionally, a number of paw prints were present in the tile assemblage. These were all from various sizes of dog and were found on fabrics 3, 8 and 10. Most of the tile excavated from the villa site was unstratified and only a small quantity belonged to any of the site phases. Details showing the quantities and types of tile appearing in each phase are included in Table 4. It is not possible to make any meaningful interpretations of the site using the tile from any of the phases except the demolition and abandonment phase 8, which provided the largest quantity of tile. A small quantity of tile can be dated to the late Roman period (by fabric) which could tentatively suggest it was supplied for the later alterations at the villa. A good quantity of different tile types was present on the villa site and although a great deal of the material was unstratified it almost certainly derived from the villa or possibly from the timber buildings of the earlier phases. 135 MARK HOULISTON TABLE 4. QUANTITY OF EACH TILE TYPE PER PHASE Phase Tegulae Imbrex Flue Brick 1 2 3 4 5 6 7 8 9 1 1 2 4 5 - 2 22 4 1 4 1 2 1 - 3 9 5 - -- 3 - - 2 2 - - 2 - - 1 8 2 - - The material occurs in a variety of different fabrics suggesting that Maidstone was supplied brick and tile from different areas including Canterbury, Eccles and London. ENVIRONMENTAL EVIDENCE INTRODUCTION (Enid Allison) Thirty bulk samples were taken, chiefly for the recovery of charred plant remains and small bones. There was no evidence for the preservation of plant or animal remains by waterlogging in Roman layers, even in ditches and deeper features, although some moister deposits were noted in the north-east of the excavated area. Sample volume ranged from 5-53 litres with a total volume of 398 litres processed. Samples were given the number of the context from which each was taken with the addition of the suffix 7s'. Initially, bulk samples were wet-sieved to 1mm after collection of a 'washover'. The washover was obtained by swirling several litres of each sample at a time in a bucket filled with water, and then decanting off the lighter floating material onto 0.5mm mesh. The sediment remaining in the bucket was then washed onto 1mm mesh. The process was repeated for the rest of each sample. For a majority of the samples very little was being recovered in the residues, so the 1mm mesh was replaced by 2mm mesh for the remainder of the samples. Sample 4/s was treated slightly differently - of 30 litres processed, no washover was collected from 20 litres of sediment, and charred plant material was recovered only from the residue. In the majority of the samples biological material was present only 136 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 in small amounts. Charred plant remains were present in most, and were common or abundant in several, particularly in Sample 4/s. Small fragments of mussel, oyster, winkle and cockle shell were present in some samples. During processing it became apparent that land snails were present in low concentrations in many samples. Their relative paucity made collection and processing of sufficiently large samples specifically for mollusc analysis impractical, even if their presence had been recognised during excavation. Since so little other environmental data was available for the site it was decided to analyse the snail assemblages obtained from the bulk samples. The usefulness of the data obtained is limited because the sampling strategy was not specifically aimed at their recovery, however. These limitations are discussed briefly in the snail report below. Most samples contained fragmentary animal bone. The most common bones recovered were of small vertebrates, which included bones of amphibians and snakes, in addition to small mammals. These have been analysed separately from the larger bones because of their potential to produce environmental information. FAUNAL REMAINS (Robin Bendrey) The animal bone assemblage consists largely of hand-collected material, together with a much smaller sampled group. The hand-collected material consists of 409 fragments weighing a total of 5083.5g (Table 5). The assemblage is generally in good condition. Reconstruction of bones was made where possible, but fragmentation still resulted in the exaggeration of numbers. This occurred most dramatically in Phase 2. Cattle bones dominate the assemblage, followed numerically by sheep, and then by pig. Also present are bones of dog, horse and deer. Unidentified bones have been divided into cow-sized and sheep-sized where such a distinction in size has been noticeable, the rest being labelled as unidentified. Many of the unidentified fragments were too light to register on the weighing-scales, and in these cases they were given a nominal weight of 0.5g (which in turn exaggerates the weight of the unidentified fragments). Sheep and goat bones have been grouped together as ovicaprid as no diagnostic elements were found to separate them. 'Sheep' is used in the text for sheep/goat. Hand-collected assemblage Within the hand-collected assemblage there are ten burnt bone fragments (2.4 per cent of the total number) from a probable six bones. There are also 28 bones (39 fragments - 9.5 per cent of the total) which have been gnawed, usually by rodents, but a cattle femur caput (from G.32) was heavily chewed, possibly by a large dog. There is insufficient ageing data to produce any 137 MARK HOULISTON TABLE 5. ANIMAL BONE SPECIES DISTRIBUTION (number of bone fragments identified; and species representation by weight) Phase 1 2 3 4 5 7 8 No. Per Weight Per frag cent (grams) cent m'ts Bos Ovicaprid Sus Canis Equus Cervus cow-sized sheep-sized unidentified Total 2 . _ . - . 1 - . 3 96 32 13 15 . - 51 13 40 260 29 3 4 4 -- 21 5 4 70 3 2 . . -- 7 1 - 13 5 2 8 . 4 4 9 2 1 35 1 9 - --- 3 2 2 17 7 2 1 - -- - 1 - 11 143 50 26 19 4 4 92 24 47 409 35.0 12.2 6.4 4.6 1.0 1.0 22.5 5.9 11.5 100 3856.0 234.0 140.0 55.5 176.0 12.0 528.5 34.0 47.5 5083.5 75.9 4.6 2.8 1.1 3.5 0.2 10.4 0.7 0.9 100 meaningful conclusions. Only a small number of bones provide measurable dimensions, and these measurements are available in the archive. Amongst the cattle bones there is a high survival rate of mandibles and teeth. Although this is an expected phenomenon, it has been exaggerated by the fragmentation of the jaws and dispersal of the teeth. This occurs in G.8. where fragments constitute 64 per cent of the identified cattle fragments, and 51 per cent of the weight. The fragmentation of two cattle skulls in contexts 194 and 415 provides much of this, and has most probably produced the large number of cow-sized cranium fragments in these contexts, which makes a significant distortion of the figures. C.194 provided much of the evidence for cattle butchery. Longitudinal scratches and knife cuts along a scapula were probably produced when removing meat from the bone. A radius from the same deposit exhibits numerous butchery marks, transverse below the proximal end and descending down the exposed interior surfaces of the shaft, although the latter may possibly be modern. The initial cut, removing the proximal end, was probably executed for marrow extraction. Also of interest is a phalanx which has had its proximal and distal ends removed, though this modification also may be modern. Elsewhere, a general pattern of transverse butchery marks across a variety of long bones, often near the proximal end, is observable. Cut marks have also been observed on the rear of the ascending ramus of a mandible. Two horn cores, from the same context as the worked antler already discussed by Ian Riddler, show cut marks at their bases where they have been removed from the skull. Due to fragmentation the range of measurable bones is limited. The only complete cattle bones are two phalanges. Other measurable bones include the proximal ends of a metatarsus and a metacarpus, and the diameters of two horn core bases. 138 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Sheep bones comprise 20 per cent of the number of identified fragments, though only 5 per cent by weight. Ovicaprid measurements include the smallest depth of the olecranon of an ulna, the length of a calcaneum, and the distal end of a metatarsus. G.8 provides 64 per cent of the identified sheep fragments, by both number and weight. Within this group, sets 6 and 12 (two parallel ditches) produced all the bones. Much of this assemblage is made up of mandibles and teeth, and the calcaneum comes from here. The next most common skeletal element is provided by three fragments of metapodia (with another from G.34). From the rest of the assemblage come four fragments of pelvis, too small to be measurable. The two pieces of cranium recorded are from the same context and most probably from the same animal. No evidence for butchery has been observed. The pig bones are dominated by teeth and mandibles, as with the cattle and sheep. Two mandibular fragments, with teeth, come from the same context within G.8 and may be from the same animal. Also found were two humerii and an astragalus. One humerus from G. 11 is an unfused distal end showing clear butchery marks where it has been sawn from the shaft. The other, from G.33, has a greater length of the shaft remaining, though it is in a more friable condition and is made up of four fragments. This bone does not show any butchery marks but has undergone gnawing. The astragalus, fairly complete, exhibits an overall degradation of the surface possibly consistent with having either been corroded by soil conditions, or passed through the digestive tract of a dog. This is the only bone in the assemblage in this state, and it came from G.H. This condition annuls measurement. Bone fragments identified as dog comprise 5 per cent of the total identified number. These are distributed through four separate contexts. G.8 provides most of these bones. C.194 contains a mandible (containing P4, Ml, and M2), a loose canine, and a maxilla (with P3, P4, and Ml), all of which may come from the same animal. An intact whole metacarpus (McIII) was also excavated, from which measurements have been taken. Another dog canine and a femur come from context 415. Phase 3 provided an atlas vertebra (G.18) and another tooth (G.16). Of all the dog bones only the mandible and to a lesser extent the atlas have undergone gnawing. The two horse bones came from Phase 5 deposits. They include parts of a pelvis and a scapula (from G.38 and 40 respectively). Neither bones have observable evidence of gnawing or butchery marks. Deer is represented by a single mandible which shows neither butchery marks nor evidence of gnawing. The sampled material Twenty-two samples produced small quantities of bone. Bones of small vertebrates were quite common and these have been reported separately (see below). The majority of the remainder consist of tiny fragments of large mammal bone with smaller amounts of bird and fish bone. Identified mammal fragments include a highly fragmented cattle horn core and a dog incisor. Many of the dog remains derive from context 194. It is possible that the incisor may have come from the same animal as the mandible and maxilla (see above). A few bones and scales of fish were obtained from thirteen of the samples 139 MARK HOULISTON (896/s (G.3); 4/s and 194/s (G.8); 12/s, 452/s and 566/s (G. 14); 467/s and 889/s (G.18); 881/s (G.21); 299/s, 404/s, 466/s, 564/s). This material consists mainly of small unidentifiable fragments of fin rays. The fish bones from 466/s had been burnt. Bird bones were recovered singly from four samples, unidentifiable fragments from 13/s and 131/s, a small passerine species from 194/s, and domestic fowl Gallus gallus from 467/s. Egg shell was found in 4/s. Discussion The most notable feature which has arisen lies with the assemblage from G.8 (fills of the U-shaped boundary ditches) which came solely from the western stretches of the ditches which sloped from east to west towards the river. The vast majority of the bone came from the western part of the southern ditch and it would seem likely that it was deposited as primary waste. The number of gnawed fragments from this area suggests, however, that immediate burial of the waste did not take place (gnawing occurs at a low level in nearly all phases of the site). This is supported by the environmental sample report, which identifies the presence of carnivorous molluscs and small vertebrates, and suggests the dumping of animal remains. The bias here, towards cattle skulls, jaws and metapodia can be interpreted as waste from primary butchery of cattle carcasses. Such a pattern, but on a larger scale, has been noted in late firstcentury AD deposits from the infill of a legionary ditch at Exeter and an early second-century AD pit at Aldgate, London (Maltby, 1981, 166). The small size of the sample under study here, possibly representing a minimum of four cattle and three sheep, must be remembered. This assemblage would thus seem to reflect primary butchery discard, though there is some representation of main meat-yielding bones and secondary butchery (e.g. stripping meat from the scapula). The small size of the assemblage severely limits the conclusions that can be made. Horse and deer have a low presence, though these are typically only recovered in small numbers on Romano-British sites (King, 1978, 216-27). It has been argued that few horse bones are recovered from Romano-British sites in assemblages derived from butchery discard, probably because horse meat did not feature as an important part of the diet (King, 1978, 225; Maltby, 1981, 184). Pig is represented as a food animal by the butchered humerus. Dog is witnessed both by its own skeletal parts and by the chewing of the Bos femur caput. Cattle clearly dominate the assemblage (Table 5), followed by sheep. Both these animals (as the main meat animals) maintain a consistent presence in the various phases. The anomalous Phase 2 stands out against this background (Table 5). Consideration of the assemblage excluding G.8 (which biases any overall representation of the site fauna) shows a rather different picture. Pig attains the same quantitative importance as sheep, but most significant is the relative meat yield (Table 6), but it must be stressed that the small size of the assemblage results in large shifts in the percentages for comparatively few bones. Table 6 highlights the lower quality of meat bias of G.8, especially skull fragments of cattle and sheep, in comparison to the rest of the assemblage. 140 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 TABLE 6. RELATIVE QUALITY OF MEAT YIELD (the assemblage excluding Group 8, in comparison to Group 8 (in italics): table format from Armitage and West (1985, 112)) Cattle Sheep % weight % fragments % weight % fragments High scapula, quality humerus, 23.8 11.7 13.7 9.8 19.5 - 27.8 meat: pelvis, femur, vertebra Lesser tibia, radius, quality ulna 12.8 8.4 9.8 5.4 27.2 0.7 16.7 3.1 meat: skull, maxilla, 39.6 73.4 49.0 77.2 45.0 79.3 50.0 81.3 mandible, loose teeth Lowest metapodials, quality phalanges, etc 23.8 6.5 27.5 7.6 8.3 20.0 5.5 15.6 meat: MOLLUSCS AND SMALL VERTEBRATE REMAINS (Jessica Davies)2 Twenty-three samples containing mollusca and/or small vertebrate remains were obtained from the site, spanning much of the period of occupation from the pre-Roman phase 1 to phase 7. Phases 6, 8 and 9 are not represented in the samples. Processing methods are given in the introductory section above. Snails and small bones were recovered from both residues and 'washovers'. Samples were taken from separate, discrete contexts, not in a sequence sampled as a column at one point on the site. This limits how the data can be represented. Illustrative representation of the data, using graphical methods such as histograms, cannot be produced as the data does not in fact represent a contiguous 'time line' of events. As the samples were obtained from discrete contexts, rather than a continuous sequence, the size of the samples was not standardised to a consistent volume. Therefore, the numbers of specimens identified from different samples cannot be directly, quantitatively, compared. The molluscs recovered were in a fairly good condition, although large shelled taxa, such as Cepaea sp. were only represented in a very fragmentary form. The internal shells of slugs were present in a number of samples. The vertebrate remains were generally rather fragmentary. The majority of elements present were post-cranial, although many samples contained teeth of mouse or vole species, and one sample also contained demi-mandibles of these species. The bones themselves were generally in good condition, although some showed the characteristic etching and erosion that is expected when bone passes through the gut of an animal. The paucity of the data is consistent with the notion that deposition of the remains was the result of predator debris, 141 MARK HOULISTON rather than animals dying in situ (Evans and Rouse, 1992). This is supported by the fact that many of the deposits sampled would not have functioned as pit-fall traps for small vertebrates (for example the soil layer of phase 4). As small vertebrates, and any predators preying on them have far larger 'home ranges' than the molluscs, they are likely to represent a much larger area of the environment around the villa, not just that at the point of sampling. To identify the moliuscan and vertebrate material from the samples it was necessary to use modern field guides as well as reference collections held by the Department of Archaeological Sciences at the University of Bradford. Some problems were encountered in identifying both the moliuscan and the vertebrate material. The vertebrate remains were exceedingly fragmented, many being post-cranial elements of the skeleton. Some work has been published, particularly about the teeth of small vertebrates, but there have been few comprehensive studies of post-cranial elements of many small vertebrate species from archaeological contexts. With many elements being fragmented, and some charred, identification of species was rather difficult. Fragments of mollusc shells could not all be reliably identified to species and some specimens could only be identified to family level. Reference specimens of Ena montana and Ena obscura were borrowed from the Grensted collection, in the Department of Biology at the University of York. Results Although present in most of the samples, Cecilioides acicula is a burrowing species and thus its presence cannot be guaranteed to be contemporary with the archaeological specimens from the samples. Thus whilst specimens of this species were recorded, their presence is not considered in the interpretation of the data. The identifications made are listed in Tables 7 and 8. Phase 1. This is represented by four samples, of which only one produced mollusc and vertebrate data. Sample 296/s (G.3) contained molluscs only, 399/s and 466/s (both G.2) vertebrates only, whilst 896/s (G.3) contained both. Many of the moliuscan taxa identified from this phase, either show a tendency to be more common in damp places, or are characteristic of damp, often sheltered places. The range of the habitats of species such as Pomatias elegans and Discus rotundatus includes woods. It is unlikely that the environment on the site of a functioning villa would be such but many other situations could provide similar habitats. These include areas of disturbance, with loose rubble, as may be expected in an area where building may be taking place. Alternatively, areas that are overgrown with scrub would provide the necessary shade for these molluscs. The presence of species such as Vallonia excentrica which do not inhabit woods adds to this surmise. This species tends to prefer dry calcareous places, tolerating short grass and screes, suggesting a mosaic of sheltered situations and some more open areas. The presence of slugs of Arion and Umax species does not contradict this. It is likely that the Umax sp. is Limaxflavus which is strongly associated with human habitation, although this identification cannot be confirmed on the basis of the internal shell alone. The vertebrate data from this phase is mainly of mouse or vole species. Both of these generally prefer some amount of low growing vegetation, although specifics 142 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 TABLE 7. NUMBERS OF LAND SNAILS RECOVERED FROM THE BULK SAMPLES Phase 1 2 3 4 Species: Pomatias elegans (Mulier) Carychium tridentatum (Risso) Succinea putris (Linn6) Cochliocopa lubrica (MUller) Cochlicopidae sp. Vertigo pygmaea (Draparnaud) Pupilla muscorum (Linne) Vallonia pulchella (Mtlller) Vallonia excentrica Sterki Vallonia sp. Ena obscura (Mulier) Discus rotundatus (Mulier) Arion sp. (granules) Vitrea contracta (Westerlund) Vitrea sp. Nestrovitrea hammonis (Str5m) Aegopinella pura (Alder) Aegopinella nitidula (Draparnaud) Oxychilus draparnaudi (Beck) Oxychilus cellarius (Mulier) Oxychilus alliarius (Miller) Oxychilus ?helveticus (Blum) Oxychilus sp. Zonitidae sp. Umax sp. (internal shells) Euconulus fulvus (MUller) Cecilioides acicula (MUller) Clausilia ??bidentata (Str8m) Clausilia sp. Indeterminate fragments 1 4 2 7 1 4 7 1 3 3 14 2 1 2 23 120 1 6 1 1 1 3 1 6 6 1 9 24 1 3 1 18 90 1 1 5 14 5 1 25 19 1 11 42 7 5 16 3 261 1 1 4 1 8 1 11 1 1 1 7 3 22 10 1 3 1 1 40 I 5 111 1 3 7 9 3 7 36 7 3 10 of habitats between species vary. The presence of bank vole (Clethrionomys glareolus) suggests that there may be some thick vegetation around the site whilst short-tailed vole (Microtus agrestis) indicates that there may be rough, possibly ungrazed, grass. The identification of a snake species adds support to the idea that the site environment was somewhat roughly vegetated. It is therefore unlikely that the wider landscape around the site was wholly cleared and under cultivation. 143 MARK HOULISTON TABLE 8. NUMBERS OF FRAGMENTS OF SMALL VERTEBRATES RECOVERED FROM THE BULK SAMPLES Phase 1 2 3 4 i Small fish Rana sp, (frog) Anguis fragilis (slowworm) Snake sp. Soricidae sp(p) (shrew) ?Soricidae sp(p) Clethrionomys glareolus (bank vole) Microtus agrestris (short-tailed field vole) C. glareolus/M. agrestis Arvicola terrestris (watervole) Cricetidae sp(p.) (vole) Apodemus sylvaticus (wood mouse) Muridae sp(p) (mouse) Cricetidae/Muridae sp(p) (mouse or vole) ?Cricetidae/Muridae sp(p) Indeterminate small mammal fragments 1 1 2 10 3 1 1 2 4 ?1 2 + ?l 1 25 3 7 1 ?1+1 1 4 4 1 1 36 13 9 1 I+?2 Phase 2. Four samples represent deposits of this phase. Sample 4/s (G.8) gave only vertebrate material, 886/s (G.10) consists of moliuscan remains only, and both groups were recovered from 194/s (G.8) and 885/S (G.10). The mollusca include both generalists and more specialised taxa. Species such as Oxychilus cellarius and Discus rotundatus are present suggesting that moist sheltered habitats continued to be available on the site. Other species that are present, such as Pupilla muscorum and Vitrea contracta characteristically inhabit drier, calcareous positions. Both habitats could have been produced if there were tumbles of stony material or perhaps dry stone walls on the site. Umax and Arion slug species are also represented in this phase. Some species of Arion sp. are associated with cultivated land and gardens, suggesting that there may be some disturbance of the soil around the site. The vertebrate data indicate that the environment around the site continued to be well vegetated. Wood mouse (Apodemus sylvaticus), whose habitat is generally rather varied, tends to live in areas of more arboreal cover, suggesting that there were either wooded areas or areas of shrub-like growth around the site. Shrew species also tend to favour areas with leaf litter or thick grass cover, for the purposes of obtaining food. These conditions could have 144 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 been provided by plantings in garden areas. The presence of shrews, and the facultative moliuscan carnivores Oxychilus sp., suggest the possibility that human food refuse was being deposited on and around the site. Phase 3. Eight samples produced material from this phase. Samples 13/s (G.H), 453/s (G.H) and 460/s (G.H) gave only vertebrate material, and sample 564/s (G.H) only moliuscan data, whilst samples 404/s (G.l6), 467/s (G.18), 566/s (G.H) and 889/s (G.18) all produced quantities of both. The material from 566/s (G.H) should be considered with some caution as this deposit may have been redeposited, and the data retrieved from it may represent an earlier period of the site. Again the moliuscan data suggest that the site had damp, sheltered areas suitable for species such as Discus rotundatus, Oxychilus cellarius and Oxychilus drapernaudi. The latter two of these species are characteristic in modern gardens where there is plant derived detritus that is used as a source of food. This may suggest that there was a degree of herbaceous growth on the site, deliberately planted or otherwise, and accumulating leaf litter. The presence of Succinea putris, which prefers damp sparsely vegetated places, may indicate the presence of marshy areas on the site, although the single individual may have been inadvertently brought onto the site from the nearby river. Freshwater species could have been brought to the site if cut waterside vegetation, particularly reeds or other similar plants, were being harvested and brought to the site, or if the site itself had been subject to occasional flooding. Much of the vertebrate data accords with that from the previous two phases, suggesting that the environment around the site is likely to have remained reasonably similar. However, what is evident in the samples from this phase is that freshwater dwelling species, represented by the vertebra of a small fish and that of a frog (Rana sp.), were being deposited on the site, probably due to the exploitation of nearby water sources by birds of prey or herons, most likely the river Medway. There are several vertebrae of snake recorded from this phase. If these are from grass snakes (Natrix natrix), which dwell close to sources of water, their presence on the site may be similarly explained. This phase also boasts the only example of slow worm (Anguis fragilis) recorded from the site. This species prefers a well vegetated habitat and is characteristic of modern gardens in the south of Britain today. It is most likely that this species inhabited areas away from the villa. Again the presence of shrew bones and the snail Oxychilus sp. in the samples suggests that deposition of human refuse continued at and around the site. Phase 4. This is represented by three samples. Sample 185/s (G.l9) yielded a small quantity of moliuscan material, whilst 391/s (G.30) and 580/s (G.19) produced only vertebrate remains. The presence of one specimen of Oxychilus sp. suggests that moist situations may still be present on the site, but as so few mollusc remains were present, no substantial conclusions can be drawn from them for this phase of the site. The vertebrate assemblage contains bones of water vole (Arvicola terrestris), suggesting that species from the areas close to freshwater continued to be exploited by predators. Otherwise, the vertebrate data again indicates that the wider environment of the site had some low 145 MARK HOULISTON growing vegetation and ground cover, suggested by the various species of vole, as well as the presence of undefined mouse and shrew species. Again the evidence points towards the presence of dumped human food refuse, in the form of animal carcass remains, on and around the site. Phase 5. Sample 725/s (G.21) gave mollusca only, 767/s (G.44) only vertebrate remains, with 881/s (G.21) containing examples of both. The moliuscan data from phase 5 is no different to that seen in any of the previous phases. Therefore it is suggestive that the site continued to have both moist, sheltered places, like those which may be offered by vegetation coverage or by stone walls or piles of stony material, and drier areas, possibly on top of such piles of stones or in patches of less dense vegetation. A single individual of Succinea putris is seen in this phase. This may have been imported with material from somewhere away from the site, possibly towards the river or a marshy area. As with Phase 4, the presence of water vole (A. terrestris) suggest the continuation of predation on species living close to water. The continued detection of vole, mouse and shrew species suggests that the environment around the site continued to be one characterised by rough and probably fairly low growing vegetation. Phase 7. This phase is represented by a single sample 729/s (G.24) which contained both moliuscan and vertebrate remains. Generally the moliuscan data suggests that a moist, sheltered environment prevailed at this particular sampling spot. The sampled context is from a probable robber trench, and is described as 'a dump' of material. If the material was introduced from an area that was not subject to much disturbance prior to this deposition, the presence of Carychium tridentatum may be explained. This is a species of woodland and leaf litter, which characteristically declines sharply in abundance in reaction to human disturbance. This suggests that the dumped material may in fact have been a redeposited context, possibly an older soil from the pre-villa period prior to any disturbance taking place on the site. The vertebrate remains consisted of a single, fragmentary post-cranial bone from a mouse or vole about which little can be said. Summary and Interpretation The moliuscan remains are most likely to have been deposited within the small life-range of the species, so representing the local site environment. The general impression that these remains give is one of somewhat damp, shady habitats persisting throughout the occupation at the site. This does not necessarily mean that the site was covered by vegetation, although it is possible that there was some low growing herbaceous or scrub coverage. Loose rubble may produce similar conditions, and it is notable that the 'shade' species here are not necessarily those particular to leaf litter. Some species suggest the possible introduction of material to the site from waterside locations, whilst others, although probably in a re-deposited context, are consistent with preoccupation phases with little or no disturbance. The presence of small vertebrate remains on the site can be attributed largely 146 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 to predation by birds of prey and the subsequent deposition of small bones as 'pellets'. Both owls and other raptors are suggested as some of the material shows the surface 'etching' which is typical of bones regurgitated by hawks and falcons, rather than owls. The most likely scenario is that owls and other birds of prey were active around the site and may even have used suitable buildings on the site for nesting or roosting. The most likely predators are kestrels (Falco tinnunculus) and barn owls (Tyto alba), the nesting sites of which can be closely associated. Patchy woodland can also be used by both species for breeding sites, as well as providing perching sites for use whilst hunting. Voles are the best represented of the vertebrate fauna in the samples, and may have been preferentially preyed upon: they are the main prey of kestrels (Village, 1990). Species such as sparrowhawk (Accipiter nisus) tend to hunt on the wing and their pellets typically include remains of small birds, their main prey. A single bone of a small unidentified passerine species, which was not 'etched' by digestive fluids, was recovered in the samples. Short-eared owls (Asio flammeus) may have been active if areas around the villa were particularly marshy and open. The identification of aquatic species of vertebrates may suggest the activity of birds such as herons (Ardea cinerea) around the site. The presence of carnivorous molluscs and shrews in some of the assemblages suggests a plentiful supply of suitable food, possibly as a result of the dumping of human food refuse both on and around the site. The presence of gnawed large mammal bones in nearly all phases suggests that some waste remained unburied for a time, at least. Food refuse may also have been exploited by kestrels, which are known to feed on carrion. Thus the scenic environment of the site is likely to be one of fairly open countryside, probably with patches of scrub cover and long grass. There may also have been some clumps of trees, used by raptorial birds for breeding and perching during hunting. CHARRED PLANT REMAINS (Mark Robinson) The 'washovers' from twenty-eight bulk samples were scanned under a binocular microscope at the Environmental Archaeology Unit in the University Museum, Oxford, to assess the character of their charred plant assemblages. All except 781/s contained charred cereal remains and many also contained wood charcoal. It was noted that those samples from Phases 2 and 3 site with high concentrations of cereal remains were dominated by wheat grain whereas the richer assemblages from Phases 4-7 were dominated by glumes of hulled wheat. The charcoal in the flots appeared unexceptional. Funds for any more detailed work were limited. It was therefore decided to analyse one sample from each phase for charred plant remains other than charcoal and to record the charcoal observed for each phase. The identifications are listed in Tables 9 and 10. The quantity of remains in 4/s was so large that a 1/8 sub-sample of the original assemblage produced from 301 was analysed. A 1mm sieve is too coarse for the recovery of smaller weed seeds or chaff items. However, the low proportion of weed seeds shown by the 147 MARK HOULISTON 5 3 1 _ - 1 1 3.75 8 2 93 332 519 944 14 18 3 112 283 593 988 14 19 4 5 14 35 54 6 21 5 1 1 38 40 15 24 7 - 1 18 19 1 . - - 1 11 3 41 TABLE 9. NUMBERS OF CHARRED CEREAL AND WEED REMAINS RECOVERED FROM THE SAMPLES Context 296 4 467 495 881 729 Sample volume (litres) Group Phase Cereal grain: Triticum spelta L. (spelt wheat) Triticum sp. (wheat) Cereal indet. Total cereal grain Cereal chaff: Triticum spelta L. - glume (spelt - 71 47 96 84 44 wheat) T. dicoccum Shubl. or spelta L. 1 98 81 127 187 78 - glume (emmer or spelt wheat Avena sp. - awn (oat) Total chaff (excluding awns) 1 169 128 223 271 122 Cereal sprouted embryos: Weed seeds: Vicia or Lathyrus sp. (vetch or tare) Cf. Medicago lupulina L. (black medick) Rumex sp. (not acetosella) (dock) Anthemis arvensis L. (corn chamomile) Tripleurospermum inodorum (L.) Sch. (scentless mayweed) Lapsana communis L. (nipplewort) Bromus cf. secalinus L. (brome grass) Gramineae indet. (grasses) Weed indet. Total weed seeds Total no. of items per litre (excluding awns) washovers does appear to be a true reflection of this concentration in the ancient deposits because large seeds were also sparse. The only other taxa observed when the samples not analysed in full were scanned were a nut shell fragment of Corylus avellana (hazel) in 66/s (Phase 2 G.8) and a seed of _ - 0 0.4 1 1 5 298.1 H . 1 79.8 3 4 12 21.4 6 5 15 61.2 2 2 6 9.8 148 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 TABLE 10. PRESENCE OF CHARCOAL BY PHASE Phase 12 3 4 5 7 Alnus/Corylus tp. (alder/hazel) + - + + Quercus sp. (oak) - + + + + + Pomoideae tp. (hawthorn, etc.) - - - + . . Prunus sp. (sloe, etc) - . + . . . Plantago lanceolata or media (plantain) in 453/s (Phase 3 G. 14). Two samples, 4/s (Phase 2 G.8) and 404/s (Phase 3 G.16) contained large quantities of Quercus sp. (oak) charcoal. Phase 1. The few charred remains from a silty layer possibly an agricultural soil (296/s, G.3) showed no more than that there was activity on the site. Phases 2 and 3. The highest concentrations of charred plant remains other than charcoal were from C.4 (G.8) and C.467 (G.18). C.4 dated to AD 150-75 and C.467 to AD 175-225. In 4/s, there were about 300 items per litre. These assemblages were dominated by cereal grain, all of which either was or could have been Triticum spelta (spelt wheat). The better preserved grains showed evidence of having been charred in their glumes. Glumes of spelt wheat were present, although outnumbered by grain by a factor of more than 5. Weed seeds were very sparse, although they included Anthemis arvensis (corn chamomile), a typical cornfield weed of chalk soils. Allowing that chaff is more vulnerable to complete destruction by burning than grain, these assemblages probably represented the accidental burning of cleaned spikelets of spelt wheat. Phases 4 and 5. In contrast, the assemblages from C.495 (G. 19), a small pocket of material forming part of the construction backfill of one of the post-pits of the Phase 4 timber building, and C.881 (G.21), the backfill of a Phase 5 feature, possibly a post-pit, were dominated by chaff, almost entirely glumes of spelt wheat or possible spelt wheat. These deposits were dated to AD 175-225. Some grain was present, but it was mostly in very poor condition, (unlike the chaff) and only a very few grains could be identified, being again from spelt wheat. Weed seeds were sparse, the most numerous being large grass seeds which tend to stay with the grain during the early stages of seed cleaning. Unlike the other 'washovers' investigated in detail, these two also contained charred sprouted cereal embryos which in 881/s outnumbered grain. None of the grain in these samples had intact embryos but 186/s from Phase 4, another spelt chaff-dominated assemblage, contained two sprouted grains of spelt wheat. The flots from 495/s and 881/s clearly represented waste from cereal processing rather than the accidental charring of a crop. The most likely interpretation is that the assemblages were burnt debris from the rubbing of parched, malted, spelt wheat in order to remove the husks and sprouts prior to grinding then brewing. A charred deposit of spelt glumes and cereal sprouts was found in a Roman corn drier at the Bancroft Roman Villa, Milton Keynes 149 MARK HOULISTON (Pearson and Robinson, 1994, 582-3). The use of wheat to brew beer was certainly known to the Romans (Pliny, Book XVIII). Phase 7. The final flot examined in detail was from C.729 (G.24) from Phase 7 dated AD 275-325. It mostly comprised glumes of spelt-type wheat and was presumably de-husking waste. The charcoal from the site probably represented wood burnt as fuel. Quercus sp. (oak) predominated but some hedgerow/scrub taxa were also used. THE SOILS EVIDENCE (R.I. Macphail and G.M. Cruise) The foundations of the villa are cut into typical argillic brown soils formed in brickearth and fine loamy river terrace drift. Intact soil samples and adjacent bulk samples were removed from the major archaeological contexts and soil horizons found on the site (Table 11). The soil micromorphology has been studied from: (1) a 'natural' brickearth profile (samples 1,2, and 3) to examine the pre-Roman and Roman history of the soil; (2) the dark Roman soil underlying a fragmented mortar layer (sample 4); (3) a burned dump (sample 5a); and (4) the soils beneath a supposed timber-framed building in order to elucidate some use of space within the building (sample 5b, 6 and 7). In addition selected bulk sample analyses have been undertaken together with six soil pollen analyses. Thin section samples were processed at University College London and manufactured into thin sections at Stirling University using the methods outlined by Murphy (1986). They were then described and interpreted according to the criteria of Bullock et al., (1985) and Courty et al., (1989). Particle size, loss on ignition and magnetic susceptibility analyses were carried out following the methods of Avery and Bascomb (1982) and Clark (1990). Pollen samples were removed from the Kubiena tins and were prepared using the methods outlined by Moore et al., (1991) with the addition of micromesh sieving to facilitate removal of clay particles. Soil micromorphology: Brickearth Section Samples 3, 2 and 1 were taken from the side of a deeply cut pit: cut 123. Brickearth was assigned context no. 999. Sample 3: base of the brickearth section The soil is a decalcified, clay loam soil with spongy structure. The very few earthworm burrows present are infilled with burned flint and humic soil. Rare 150 TABLE 11. SOIL MICROMORPHOLOGY: BULK SAMPLE ANALYSES OO •3a . o 3 p o o •a 3 O 3o P Stratigraphy CO &. — 00 00 f> € o CO o 030 3 3 Particle size analysis o .*"<5 00 a. 3? 00 Soil micromorphology tn X o > < > 1 25 H rn o § 2 < rF > > 5 GO H § m Brickearth soil (side of Cut 123): 1 11-19 999 33-41 999 71-89 999 Junction of brown (7.5 YR4/4) topsoil and strong brown (7.5YR4/6) Bt horizon Bt horizon B(t)/C(t) layered brickearth parent material 19.30 20.93 20.74 3.1 1.9 23 34 43 (Clay loam) 20 47 33 (Clay loam) Truncated upper subsoil (Eb); highly biologically worked Clay enriched (Bt) horizon with very abundant textural features; some biological mixing Brickearth subsoil developed from fluvial deposit; occasional textural features ao •a 3- o 3 oO 3 p Stratigraphy oo s, oo C oo v 05 w Particle size <§ o 3 analysis O to oo ^ a. x« -S vP *" Soil micromorphology South face of bulk 2, by villa (Roman soil buried by fragmented mortar layer): to 0-40 38-46 9/10 10/11 &793 (18) Cut 17 (burned dump): 5a 4-12 4 (8) Reworked Roman soil and stratigraphy 40-43 cm weathered mortar surface; 43-58 cm dark brown (7.5YR3/2) Roman soil Black (2.5YR2.5/0) and red (10R4/8) burned daub and charcoal 264.17 9.8 20 20 60 Sandy clay loam Highly biological mixed soil containing upper subsoil (Eb) and topsoil; rich in charcoal and phytoliths. Burned and unburned Eb and Bt soil with much coarse and fine charcoal, burned 'daub', slag, carbonised cereal; biologically mixed. 2 > jo co r GO zo 00 £? •a a p O CTl •a 3" O 3 n o3 ST X P Stratigraphy 'oo 2 oo c 00 V \% 00 t - o 00 3 Particle size analysis Q g» 2? *? Soil micromorphology rn Xo <> > H zo co > H H raca 2 o H JO O 2 z> < r>- 2 > 5 GO H O z m Deposit below possible timber-framed building: 5b 237 Beneath dark stained (3) centre of floor 237 Soil beneath possible (3) beam slot 237 Marginal area adjacent to (3) stained floor deposit and possible beam slot Compacted Eb also containing dusty clay coatings, charcoal and rare possible phosphatic void infills Brickearth soil with major iron and clay staining; iron phosphatic infillings; biologically mixed. Brickearth soil with iron staining and possible iron phosphate void infills; some humic soil with very dusty dark brown coatings; biologically mixed MARK HOULISTON calcium carbonate has been redeposited in void spaces. Occasional textural features of various types are present here; (1) limpid clay coatings, (2) impure clay and silt coatings, (3) limpid clay and dark humic coatings, (4) microlaminated coatings. Sample 2: argillic (Bt) horizon of brickearth soil This soil has a more open structure with few earthworm burrows, vughs, channels and chambers. Iron stained clay is ubiquitous in the matrix and textural features are very abundant. Several types of textural features occur here: (1) limpid clay coatings, (2) microlaminated coatings comprising clay and dark, humic fine soil, (3) dark, dusty microlaminated coatings, (4) dark, dusty coatings occurring within the fine fabric. Sample 1: top of brickearth section An homogenous, decalcified clay loam soil which has undergone much biological mixing. It is poor in organic matter and has very few dusty clay coatings. Probably a truncated upper subsoil (Eb) horizon. Two major points are of interest here, the depth of the argillic (Bt, clay enriched) horizon and the juxtaposition of varied textural features within the Bt horizon. The relative shallowness of the clay enriched Bt horizon was noted during field examination of the soils and it was suggested that these were truncated having lost a significant part of the upper sub-soil and topsoil. The depth of the Bt horizon at around 10 cm can be compared with depths of around 40cm more normally encountered in argillic brown earths (Avery, 1990). In thin section the uppermost part of the soil profile (sample 1) appears to be a highly biologically mixed upper subsoil horizon (Eb) which contains little clay and organic matter. Further work will be needed to determine if biological mixing of this subsoil was a response to truncation and loss of the topsoil, and if so whether it corresponds to the Phase 1 truncation suggested above. Although some earthworm burrows have penetrated into the Bt horizon (sample 2), textural features are generally well preserved and it is these features which provide a record of the soil history at the site. They can be compared to those reported in the published literature (Avery, 1990; Murphy et al., 1985) and to reference thin sections including at least three full Roman brickearth soil profiles from London. The textural features found in undisturbed clay loams can be expected to be both more uniform and extensive than those found at the Mount (Macphail, 1980; Macphail and Cruise, unpublished). It is well established that the juxtaposition or superimposition of varied textural features can be related to a number of separate pedological events (Fedoroff et al., 1990). The features associated with cultivation for example(Jongerius, 1970; 1983; Macphail etal, 1990; Gebhardt, 1992), can become superimposed upon the micromorphological features of undisturbed argillic horizons. A micromorphological study of similar soils at Selmeston, East Sussex, enabled the interpretation of a series of coatings as follows; limpid coatings (undisturbed woodland?), dusty coatings (forest clearances from the Neolithic onwards?), and coarse-grained coatings (Saxon and medieval agriculture?) (Scaife and Macphail, 1983). 154 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Roman soil under fragmented mortar layer (C. 793; G. 18 clayey loam layer west of aisled building) Sample 4 Both decalcified upper subsoil (Eb) and humic topsoil are present here within a highly biologically mixed soil. It is very rich in charcoal and phytoliths, and also contains some bone and shell. Inclusions of mortar, brickearth and hammerscale also occur. The mixture here of both humic topsoil and Eb soil material together with well mixed phytoliths and charcoal all point to biological mixing of a number of soil types and materials. This mix is unexpected in an in situ undisturbed soil. It is probable that this soil has many characteristics in common with the dark earth at a number of sites which have been studied using soil micromorphology. For example, the dark earth at Winchester Palace was shown to have developed from the dumping of waste prior to building construction (Macphail, 1994). Burned dump (C.4; G.8 upper backfill northernmost of two U-shaped boundary ditches) Sample 5a This is a heterogenous soil containing a mixture of both upper and lower parts of the brickearth subsoil (Eb and Bt horizons). It appears that both burned and unburned soil fragments are present here and that these have been thoroughly mixed by subsequent earthworm activity. Coarse charcoal is ubiquitous, sometimes with burned topsoil attached, and much appears to be breaking up in situ. Many inclusions such as slag, daub-like material, organic-rich dung/ daub and a charred cereal grain occur here. Here the wide range of inclusions suggests that soil materials were being used in a variety of processes on the site. In particular the inclusions of possible slag suggests that materials were being used for industrial processing. Other materials such as the mix of burned and unburned Eb and Bt could derived from hearths and floors. A possible burned cereal grain fragment would also suggest the inclusion of domestic waste. Further work will have to be undertaken before of the elements of this deposit can be identified, along with any indications of cultivation it may have undergone. In previous studies of the Roman sites of Culver Street, Colchester and Wittington Avenue, London, it was found that the microfabrics contained features of cultivation which had not been apparent in the field (Macphail, 1994). Soils beneath the possible timber aisled building (C.237; G.3 silty clay layer) Sample 5b This is a compacted brickearth soil (Eb) containing well mixed charcoal, a little bone and shell. Of note are the few features of iron staining and very few possible phosphatic void infills. Dusty clay coatings are occasionally present. Sample 6 This soil differs from sample 5b by having a more open and spongy structure indicative of major biological mixing. It is also noticeably different by containing common iron-staining of void infills and some of these could be iron phosphatic residues also containing dusty clay coatings. 155 MARK HOULISTON Sample 7 The soil appears to have intermediate characteristics between samples 5a and 6, by containing few possible iron phosphatic infills. It is primarily a biologically mixed soil containing both humic soil with frequent dark brown dusty coatings, and more iron rich (and possibly containing more phosphate) soil. The field appraisal suggested that the dark colour of context 237 could have been caused by staining from surfaces associated with the timber building located above, even though these surfaces had not survived. This present study found no obvious soil micromorphological or palynological indications of stabling, for example, such as those known from Butser Ancient Farm (Allen et al., Macphail and Goldberg). This relative paucity of evidence, however, could relate to the sampling as the samples were removed from the subsoil immediately below the postulated occupation floor. It is to be expected then, that the evidence for floor activities would be diluted at depth in the subsoil. In spite of these problems, there are important variations between the three samples (5a, 6 and 7). In particular, iron staining, possible iron phosphate void infills and textural features are variable with maximum amounts occurring in samples 6 and 7. Further examination of these variations will be needed to distinguish possible variations in space within the aisled building. An absence of slag in the thin sections suggests an absence of industrial activity. On the other hand, the micromorphology and chemistry (J. Heathcote, pers. comm.) both indicate the presence of phosphate here. The results of the phosphate analysis are included in the Phase 3 discussion above. Data from the bulk sample analyses are presented in Table 11. The burned dump soil is distinguished from the natural brickearth soil by containing more sand, and by having higher magnetic susceptibility and greater organic content (loss on ignition). In contrast the brickearth soil is a clay loam but is noticeably lacking in organic matter which would be expected if a humic topsoil were present here. Soil pollen was found beneath possible aisled building (C.237; G.3 silty clay layer). The pollen and spores observed in samples 5b, 6 and 7 are listed in Table 12. The pollen spectra are dominated by Lactuceae (e.g. Taraxacum type) and Pteridium (bracken). These are resistant taxa which reach especially high frequencies in archaeological soils where pollen preservation is poor. At the Mount the pollen concentrations are extremely low, and it is almost certain that these taxa are over-represented as a result of differential preservation. It is surprising that the relatively thick-walled pollen grains of Lactuceae, Sinapis type and Anthemis type are very badly preserved while the remaining taxa including the relatively thin-walled Poaceae (grasses) appear to be well preserved. These differences in pollen preservation might suggest at least two different phases of pollen deposition. Unfortunately the soil micromorphological evidence for earthworm burrowing and biological mixing of the soil, strongly suggests that the soil pollen could be composed of a mixed assemblage^) of which part could derive from post-depositional mixing processes. 156 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 TABLE 12. SOIL MICROMORPHOLOGY: POLLEN AND SPORES Soil Sample Depth (cm) in Kubiena tin Trees and tall shrubs: Crataegus type Herbs and dwarf shrubs: Lactuceae Trifolium type Centaurea scabiosa Plantago undiff. Sinapis type Anthemis type Poaceae Umbelliferae Calluna Ferns: Pteridium Polypodiaceae 5b 1 * * * 5b 4 ** * * * ** * 6 1.5 * ** * # * ** 6 5 * ** * * * 7 1 ** * * ** 7 5 ** * * * *** GENERAL DISCUSSION One problem which bedevils the study of Romano-British villas is that the quantity of the excavated evidence far outweighs its quality. In this regard the villas of Kent are no exception. Although the number of known sites has increased since Sir Mortimer Wheeler undertook his comprehensive study for the Victoria County History (Wheeler, 1932), very few have been investigated using modern archaeological methods; of these even less have been totally excavated. Furthermore, no excavations have extended much beyond the immediate hinterland of the villa, and no thorough investigations, such as those at Gorhambury (Neal et al., 1990) and Barton Court Farm (Miles, 1986), have been made of land use in the surrounding estate ox fundus. In the Medway valley only the large villas at Eccles (Detsicas, 1983)3 and Snodland (Ocock and Syddell, 1967; Birbeck 1995) have been the extensively excavated. Despite these deficiencies the villas of Kent have formed part of a number of interesting studies,4 though further work - for example local studies of areas such as the Medway valley - is required before a more detailed understanding of their role within the local settlement hierarchy can emerge. For the purpose of this report the results of the 157 MARK HOULISTON recent excavation at the Mount will be examined against the work that has already taken place. The Medway valley occupies an important geographical position within the landscape of south-east England. It links the wooded heartland of the Weald, dominated by its heavy clay soils, to the Thames estuary and the North Sea. In doing so it passes through two important geological formations, the Hythe Beds of the Maidstone area, and the North Downs south of Rochester. The valley offers many advantages favourable to settlement: a rich variety of welldrained soils, a constant supply of water, good transportation links and plentiful supplies of building stone, namely flint and ragstone. Settlement in Roman times appears to have been relatively intense, for despite the problems with the quality of the excavations and the relative absence of low status sites from the archaeological record, a regular distribution of sites (mainly villas) can be observed all the way from Teston in the south to Rochester in the north. Such sites also appear along the River Loose, a tributary which joins the Medway at Maidstone. Four villas located in the Medway valley were significantly larger than the rest. Two of these, Barton Road (Roach Smith, 1876) and the Mount, were located on the great semicircular arc formed by the Medway in the area of modern Maidstone, whilst Eccles and Snodland were located in the central Medway area, roughly half way between Maidstone and Rochester (Durobrivae). The Mount, Barton Road, and Eccles (which was the largest) were situated to the east of the Medway, while Snodland was to the west. Apart from these at least fourteen other smaller villas, all with masonry foundations, have been identified. Most of them have been classified as farms (Detsicas, 1983,92-6), though future excavation may place some in the same category as the four large villas mentioned above. The existence of other villas is suggested by the presence of isolated temples, and bath-houses such as that at The Slade, Boughton Monchelsea (Wheeler, 1932, 105). The greatest concentration of villas appears to have been in the Maidstone area; there are at least eight within a 5 km. radius of the Mount (Fig. 17). Closer to the site itself numerous burials, coins and other artefacts have been discovered. This has led to the suggestion that a 'small town' existed at Maidstone (Webster, G., 1975), a view reinforced by the discovery of a possible Roman building at the corner of Week Street and High Street, Maidstone. However, in the absence of well-excavated sites all that can be said is that there was an intensification of activity along the routes of the Medway and the main Rochester road in the Maidstone area. Of the numerous burials found in the area, three groups are partic- 158 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 r • o A (jT\ ^ c • ^ * \ _ ^ G Cobham /nBw/ O Hf ^ ^ Durobrivae (Rochester) /P Snodland (& • >200ft 1 • Eccles 1 > 400^ A \ \ /\ V-\^ \ o East Malling /I ^^^fcH W e s t b o r o u g h ^ ^ 1 East Banning J^-^ W Barton Road ^ A i ^ A Church y \ \ ^v • / \ > > Jll \» Park Wood ) Boughton Monchelsea ft*-*».^J ^j&i^^ «* ( quarrtes -.._ „. _. r^^ y The Slade | Villa or other masonry building S "V__ 1 Walled cemetery ) o 5km Zf /i ' ' LI 1 Fig. 17 The Geography of the Lower Medway Valley in Roman Times 159 MARK HOULISTON ularly interesting since they were enclosed by walled cemeteries. These cemeteries, at East Barming church, Boughton Monchelsea and Westborough, are a relatively localised phenomenon though isolated examples have been found elsewhere in Kent (Detsicas, 1983, 151), and in larger groupings in north-west Europe (Luik, 1994). Although romanized the cemeteries may have had their origins in the burial practices of the late pre-Roman Iron Age. As each of them was associated with a specific villa site it is possible that the villa at the Mount also had its own walled burial ground. Many villa studies have stressed the importance of good connections to the regional road and river networks, and in particular the importance of riverine links for the cheap transport of bulk goods (Percival, 1976, 158-9). Specific attention has been drawn to the close correlation between such links and the villa sites of North Kent, including those of the Medway valley (Sheldon et al., 1993). The villa at the Mount was especially well positioned, since not only was it adjacent to the Medway but it was positioned at the point where the Medway came closest to the main road from Rochester to the Hastings area. Occupation at the Mount Roman villa began around AD 130-175. Prior to this human activity in the area was probably either low level or located at some distance from the site. The absence of evidence for later pre-Roman Iron Age settlement appears to conform to a pattern seen elsewhere in West Kent, for despite the identification of a relatively large number of Roman villa and farmstead sites5 few appear to have had their origins in the pre-Roman period. However, the problem may be caused by the lack of excavation; the early levels of more sites need to be examined before a clearer picture can emerge. In the Medway valley the earliest known Roman site is the palatial residence at Eccles, which dates from AD 65. This building was special in a number of ways, for not only was it built soon after the conquest but there is also evidence for military influence in its design. Although this does not necessarily suggest that the state had a role in its construction, or indeed that it had an imperial function such as the official residence of a procurator (Marsden, 1994, 83), the occupants of the building undoubtedly belonged to a family of considerable local importance. The villas at Snodland (Ocock and Syddell, 1967; Birbeck, 1995) and East Mailing (Detsicas, 1983, 94) appear to date from the end of the first century, as does the lower status site at Cobham (Tester, 1961). Nevertheless, where dating exists, all the remaining villas of the area appear to have been constructed during the second century AD. Animal husbandry and spelt wheat production were probably taking 160 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 place at the Mount by AD 150-175, though at this time there is no evidence of associated habitation. An aisled building was constructed during the period AD 175-225. Enough evidence exists to suggest this building conformed to the model proposed by J.T. Smith (1963), with a lower 'open hall' end, and an upper 'domestic' end. The excavated 'open hall' appears to have been used for livestock housing, and possibly for the storage of wheat. The upper end may represent a masonry rebuild of an original timber structure. Aisled buildings are often seen as reflecting native methods of social organisation, their origins being sought both in the long halls of the Low Countries and as a Romanized expression of the traditional Iron Age round house (Millet, 1990, 201). Local examples have been found in the west of Kent, at Keston (Philp, 1991) and Darenth (Philp, 1973), and in the east at Wingham (Dowker, 1882; 1883). The southern end of the aisled building was replaced by a similarly constructed timber building around AD 175-225, yet whether it, too, had aisles is not clear. Occupation in the northern part of the building may have continued into this period. Subsequently the remains of the aisled building were replaced with a large villa constructed with substantial masonry footings. This change appears to represent an alteration in the function of the structure, for while the earlier building can be assigned to the lowest 'farmhouse' class according to the classification proposed by Rivet (1969, 209-214) the masonry building clearly belongs to Class A. The suddenness of this change goes against the normal pattern, which is for early villas to be enlarged and embellished but not totally rebuilt. Such a change does not necessarily imply an equally sudden upturn in the fortunes of the owners, or indeed, a change of ownership, since the necessary wealth could have been accumulated over many years. Furthermore, although villa construction has been seen as an affirmation of the standing of the local elite, rather than as a direct expression of their wealth, the construction of the masonry villa at the Mount does not necessarily represent a change in the status of the owners. This is because the late second century has been seen as a time when the ruling elites of Britain were starting to display their social positions in the countryside, through the construction and adornment of villas, rather than by public munificence within the towns. Therefore the construction of the masonry villa, although abrupt, may be no more than an early reflection of this trend. As a corollary to the above discussion it should be noted that although the Mount was one of a number of villas which surround the town of Durobrivae (Rochester), very few have been found near the civitas capital at Canterbury or the provincial capital at London. If the 161 MARK HOULISTON growth of villas such as that at the Mount can be seen as reflecting the changing aspirations of the social elite at Rochester similar patterns have yet to be observed around the larger centres. Alterations were made to the main villa building during its lifetime, though these do not amount to major rebuilds or significant additions. Changes in the northern courtyard were significant since they seem to indicate that the villa underwent a contraction during the second quarter of the third century (Kelly Phase 2: c. AD 225-250) when the original bath suite was allowed to fall into disrepair. This was followed by expansion in the late third or early fourth (c. AD 275-325) when the bath suite was rebuilt (Phase 6, Kelly Phase 3). New rooms were also added to the southern and northern sides of the main villa building at about this time, and the eastern portico was rebuilt. Later in the same time period substantial buttresses were added as supports to the corners of the southern and northern wings (c. AD 275-325). It seems most likely that they were made necessary by ground movement caused by springs in the area, though alternatively they may have been added to support an upper floor and if so this would conform with a wider trend, since the creation of very large residencies is one of the features of fourth-century villa construction in Britain. The sudden demise of the building, apparently at the very end of the third century or at the beginning of the fourth (c. AD 275-325), is slightly at odds with the wider trend, since the general decline in villa numbers does not begin until the middle of the fourth century. Locally, where dating is available, most of the Medway villas appear to have survived well into the century. In particular the larger villas at Eccles (Detsicas, 1983, 182) and Snodland (Ocock and Syddell, 1967, 193) were probably still in use after AD 367. Special circumstances might have prevailed to bring an early end of occupation at the Mount. The positions of the Medway villas in the trade networks of the region have been largely unexplored, though this situation is likely to be remedied by the analysis of the large material assemblages retrieved during the excavations at Eccles. Some relationships have already been noted. Inferential evidence for the transportation of iron ore and smelted iron from the Weald comes from a number of sites including the Mount. Although the origins of the industry apparently lie in the late pre- Roman Iron Age it grew dramatically during the Roman period and production levels were possibly influenced by the activities of the Classis Britannica (Cleere, 1974). The main road between Rochester and the Hastings area would have been an important route for the industry. 162 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Analysis of pottery assemblages from the Mount and other Medway villas such as Snodland (Seager Smith, 1995) has emphasised the predominance of sandy grey wares, probably products of the industries of north-west Kent and south Essex. The Thames estuary and River Medway would have provided an easy, and relatively direct, link for the transportation of such wares. Pottery also appears to have arrived from up-stream locations, East Sussex Wealden ware, in particular, having been identified at the Mount as well as the villas at Thurnham and East Mailing. The trade in this ware is particularly interesting given its postulated links with the iron working industry (Lyne, 1994,313-4). Ceramic bricks and tile may also have been traded over long distances before arriving at the Mount, though no evidence of the stamped tiles often associated with iron production in the Weald were noted. Large quantities of a tile type often found in London, as well as similar amounts from the villa at Eccles were retrieved during the excavation. The most common fabric was similar to those manufactured at Canterbury, though the source of these may have been located elsewhere. The most important resource in the Maidstone area was probably the ragstone of the Hythe Beds; as well as being used in the construction of the Mount Roman villa and a significant number of other buildings in the Medway valley the stone was used extensively on sites throughout south-east Britain. A cargo of ragstone was found in the wreck of a Roman ship found at Blackfriars in London (Marsden, 1994). Quarrying seems to have been taking place soon after the conquest, for ragstone is used in the pre-Boudican temple of Claudius at Colchester and similarly dated levels in London, and in the early building at Eccles (c. AD 65). It was used extensively in the late first and early second centuries for the construction of public buildings in London and Canterbury, and on an even larger scale during the late second and third centuries when major fortifications were built at London, Canterbury, Reculver, Richborough, and Lympne (Worssam and Tatton-Brown, 1993). Marsden has estimated that 1,750 shipments of a similar size to that found at Blackfriars would have been required to complete just one of these fortifications, the defensive wall at London; because of this he suggests that the trade may have been imperially controlled. Whatever the relationship between the villas and quarries of the Maidstone area, there is no doubt that large-scale quarrying would have had a significant effect on the local economy. The imaginative reconstruction shows ragstone, perhaps from quarries in the Boughton Monchelsea area, being loaded onto barges in the low lying 163 MARK HOULISTON area south of the Mount. A wharf existed at that point during the last century, though there is no evidence for the presence of a Roman predecessor. There appears to have been a decline in the use of ragstone after the construction of the Saxon Shore defences at the end of the third century. If the Mount had a close connection with the trade this may explain its apparently early demise compared with the other villas of the area. We should, however, not lose sight of the fact that expenditure on the maintenance of a villa is evidence of consumption and not of income. The recent suggestion (Milne, 1996) that the Blackfriars boat may have been built by the Classis Britannica in the timber and iron rich areas of the Weald suggests a possible connection between the ragstone, iron, tile and perhaps even pottery industries of the area. Possibly the small quantities of iron and pottery found at the Mount may have travelled around the coast of Kent courtesy of the Roman navy rather than along the more direct land route. As well as evidence for regional and local industries material was recovered during the recent excavation indicating that a range of domestic handicrafts took place. The production of textiles is attested by the presence of spindle-whorls, and domestic items such as furniture mounts by lathe-turned cylinders made from antler bone. Evidence for antler working was also recovered from Kelly's excavations. In common with other sites in the Medway valley soils in the area of the Mount were light and well drained. Unfortunately direct evidence for the agricultural exploitation of the area has not been found. It is not yet possible to view the villa as part of an agricultural system. Indirect evidence comes from the charred grains of spelt wheat and the butchered animal bones recovered from the dump deposits and ditch backfills of the early phases of villa development; the wheat appears to have been dried in the area prior to its storage. Cereal drying ovens are well known features on rural sites in lowland Britain. Good local examples having been found at the villa at Keston (Philp, 1991, 87-90). Wheat also appears to have been used in the production of beer. The local environment of the villa seems to have been somewhat damp and shady throughout its occupation, and was possibly characterised by rough and fairly low growing vegetation. Human food refuse may have been dumped in the area as well as waste from the animal butchery and wheat drying that was taking place. The presence of owls and other birds of prey is attested by the recovery of the 'etched' bones of voles and other small mammals. 164 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 It can be seen from the above discussion that despite the restrictions placed upon the recent excavation much of importance was discovered. Modern archaeological techniques were employed, and as a consequence results of'quality' were produced to add to the great quantity of material already available from earlier excavations. It is hoped that the deposits which remain unexcavated (in the area located immediately east of the villa and the area of the surrounding fundus) will survive for the attention of future archaeologists. They may then fill some of the gaps in our knowledge of the history of this important building. ACKNOWLEDGEMENTS Thanks are extended to all those whose efforts contributed to the success of the excavation. In particular to Dr John Williams and Lis Dyson of Heritage Conservation, Kent County Council, Planning, to Peter Brightman the resident Project Engineer and his colleagues, to the staff of Miller's and Gallagher's, contractor and sub-contractors on the project respectively, and to all those who worked on the site, both professional and volunteer and principally the members of the Maidstone Area Archaeological Group and the Dover Archaeological Group. The author would like to express his particular gratitude to David Kelly, formerly of Maidstone Museum, to the late Dr Alec Detsicas and to Dr Tom Blagg who all made invaluable comments on earlier drafts of this report. Their input into the project has been greatly appreciated. Thanks are extended to Ivan Lapper, who made such a splendid job of the reconstruction painting and to Maidstone Museum for allowing us to photograph and reproduce the Thomas Charles paintings of the 1843 excavation, and for allowing access to pottery held in the museum. Dr Richard Pollard is thanked for providing a copy of his archive notes on Kelly's earlier excavation and Maggy Taylor for her assistance with the identification of samian wares. The pottery, tile, and small finds were drawn and mounted by Cathy Tutton and Mark Duncan. The photographs were taken and prepared by Andrew Savage. NOTES 1 Those with P added have petalled knops of some form and demonstrate that other types were generally favoured by makers using this spring system. The date range is very much in line with the main run, perhaps not quite so emphatically early, and certainly with a fairly clear end by 175: the last example may well have been residual. All items dating definitely to the later third century and later were excluded. 165 MARK HOULISTON 2 Identification and analysis of the material was undertaken at the University of Bradford in part fulfillment of requirements for the degree of B.Sc. in Archaeology, and was supervised by T. P. O'Connor. 3 Interim reports on Eccles appear in Archaeologia Cantiana, lxxviii (1963) - xciii (1977). 4 For example, Detsicas, 1983; Percival, 1976,91-4. 5 The farmstead sites of West Kent have been described by Philp, 1963. BIBLIOGRAPHY Addyman, P. V. and Priestley, J., 1977, 'Baile Hill, York: a Report on the Institute's Excavations', Archaeological Journal, 134, 115-56. Allason-Jones, L., 1991, 'Objects of amber, jet and shale' in Holbrook and Bidwell, 271-4. Allen, M. J., Bloomfield, C, Macphail, R. I., and Wyles, S. F., 1990, 'A record of soil and moliuscan changes resulting from landuse management at Butser Archaeological Farm: some preliminary results' in P. J. Reynolds (Ed.), Butser Ancient Farm Project 1990, Butser Ancient Farm (in press). Anderson, A. C, 1980, A Guide to Roman Finewares, VORDA Research Series no. 1, Highworth. Armitage, P.L. And West, B., 1985, 'Faunal evidence from a late medieval garden well of the Greyfriars, London', Transactions of the London and Middlesex Archaeological Society, 107-36. Avery, B. W., 1990, Soils of the British Isles, Oxford. Avery, B. W. and Bascomb, C. L. (Eds.), 1982, Soil Survey Laboratory Methods. Soil Survey Technical Monograph No. 6, Harpenden: Soil Survey of England and Wales. Bedoyere, G. de la, 1991, The Buildings of Roman Britain, London. Betts, I. M., 1992, 'Roman tile from Eccles, Kent, found at Colchester' in P. Crummy, Excavations at Culver Street, the Gilberd School, and Other Sites in Colchester 1971-85, Colchester. Birbeck, V., 1995, 'Excavations on a Romano-British Villa at Churchfields, Snodland, 1992-4', Archaeologia Cantiana, cxv, 71-120. Biro, M. T., 1994, The Bone Objects of the Roman Collection, Catalogi Musei Nationalis Hungarici, Series Archaeologica II, Budapest. Bishop, M., 1995, Finds from Roman Aldborough, Oxbow Monograph 65, Oxford. Black, E., 1995, 'Voussoirs and Wall-tiles' in Blockley et al, 1267-93. Blockley, K., Blockley, M., Blockley, P., Frere, S. S. and Stow, S., 1995, Excavations in the Marlowe Car Park and Surrounding Areas, The Archaeology of Canterbury V, Canterbury. Brailsford, J. W., 1962, Hod Hill, I, Antiquities from Hod Hill in the Durden Collection, London. Branigan, K., 1989, 'Specialisation in villa economies' in K. Branigan and D. Miles (Eds), The Economies of Romano-British Villas, Sheffield, 42-50. Brodribb, A. C. C, Hands, A. R. and Walker, D. R., 1971, Excavations at Shakenoak Farm, near Wilcote, Oxfordshire, Part II: Sites B&H, privately printed. 166 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Brodribb, G., 1985, Roman brick and tile, London. Buckley, D. G. and Major, H., 1983, 'Quernstones' in N. Crummy, 73-6. Bullock, P., Fedoroff, N., Jongerius, A., Stoops, G. J. and Tursina, T., 1985, Handbook for Soil Thin Section Description, Wolverhampton. Bushe-Fox, J. P., 1913, Excavations on the Site of the Roman Town at Wroxeter, Shropshire in 1912, Reports of the Research Committee of the Society of Antiquaries of London, No. 1, Oxford. Bushe-Fox, J. P., 1949, Fourth Report on the Excavations of the Roman Fort at Richborough, Kent. Reports of the research Committee of the Society of Antiquaries of London, No. 16, Oxford. Calkin, J. B., 1972, 'Kimmeridge Shale objects from Colliton Park, Dorchester', Proceedings of the Dorset Natural History and Archaeological Society, 94, 44-8. Charles, T., 1847, 'Excavation of the Roman Villa at the Mount near Maidstone', Journal of the British Archaeological Association, ii, 86-8. Clark, A. J., 1990, Seeing beneath the soil, London. Clarke, G., 1979, The Roman Cemetery at Lankhills, Pre-Roman and Roman Winchester II Winchester Studies 3, Oxford. Cleere, H., 1974, 'The Roman iron industry of the Weald and its connexions with the Classis Britannica', Archaeological Journal, 131, 171-99. Cool, H. E. M., 1990, 'Roman metal hair pins from southern Britain', Archaeological Journal, 147, 148-82. Cotton, M. Aylwin, 1947, 'Excavations at Silchester', Archaeologia, 92,121-67. Courty, M. A., Goldberg, P. and Macphail, R. I., 1989, Soils and Micromorphology in Archaeology, Cambridge. Crummy, N., 1983, The Roman Small Finds from Excavations in Colchester 1971-9, Colchester Archaeological Report 2, Colchester. Crummy, N., Crummy, P. and Crossan, C, 1993, Excavations of Roman and later Cemeteries, Churches and Monastic Sites in Colchester, 1971-88, Colchester Archaeological Report 9, Colchester. Cunliffe, B. W., 1968, Fifth Report on the Excavations of the Roman Fort at Richborough, Kent, London. Cunliffe, B. W., 1971, Excvations at Fishbourne, Vols I and II (London). Cunliffe, B. W., 1975, Excavations at Portchester Castle. Volume 1: Roman, London. Detsicas, A. P., 1983, The Cantiaci, Gloucester. Dool, J., Wheeler H. et al, 1985, Roman Derby: Excavations 1968-1983, The Journal of the Derbyshire Archaeological and Natural History Society, 105. Dowker, G., 1882, 'The Roman Villa at Wingham Part I', Archaeologia Cantiana, xiv, 134-139. Dowker, G., 1883, 'The Roman Villa at Wingham Part II', Archaeologia Cantiana, xv, 351-357. Down, A., 1974, Chichester Excavations II, Chichester. Down, A., 1978, Chichester Excavations III, Chichester. Down, A., 1989, Chichester Excavations VI, Chichester. Drury, P. J., 1988, The mansio and other sites in the south- eastern sector of Caesaromagus, Chelmsford Archaeological Trust Report 3.1, C.B.A. Research Report No. 66, London. 167 MARK HOULISTON Dunning, G., 1968, 'The Stone Mortars' in Cunliffe, 1968, 110-14. Evans, J., 1996, 'The Roman Pottery' in D. S. Neal, Excavations on the Roman Villa at Beadlam, Yorkshire, Yorkshire Archaeological Report No. 2, Leeds, 69-93. Evans, D. R. and Metcalf, V. M., 1992, Roman Gates, Caerleon, Oxford. Evans, J. G. and Rouse, A. J., 1992, 'Small-vertebrate and moliuscan analysis from the same site.' Circaea, 8 (2), 75- 84. Farrar, R. A. H., 1973, 'The techniques and sources of Romano-British black-burnished Ware' in A. P. Detsicas (Ed.), Current Research in Romano-British Coarse Pottery, C.B.A. Research Report No. 10, London, 67-103. Fedoroff, N., Courty, M. A. and Thompson, M. L., 1990, 'Micromorphological evidence of palaeoenvironmental change in Pleistocene and Holocene paleosols' in I. A. Douglas (Ed.), Soil Micromorphology: a Basic and Applied Science, Elsevier, Amsterdam, 653-66. France, N. E. and Gobel, B. M., 1985, The Romano-British Temple at Harlow, Gloucester. Frere, S. S., 1962, 'Excavations at Dorchester on Thames', Archaeological Journal, cxix, 114-49. Frere, S. S., 1972, Verulamium Excavations Volume I, Oxford. Fulford, M., 1977, 'Pottery and Britain's foreign trade in the later Roman Period' in Peacock, 1977, 37-84. Fulford, M. and Allen, J. R. L., 1992, 'Iron-Making at the Chesters villa, Woolaston, Gloucestershire: Survey and Excavation 1987-91', Britannia, xxiii, 159-216. Gebhardt, A., 1992, 'Micromorphological analysis of soil structural modification caused by different cultivation implements' in P. Anderson (Ed.), Actes de la Table Ronde CNRS: Prehistoire de I'agriculture: nouvelles approches experimentales et ethnographique, Valbonne: Monographie du CRA 6, CNRS, 373-383. Gillam, J. P., 1970, Types of Roman Coarse Pottery Vessels in northern Britain, 3rd edn., Newcastle-upon-Tyne. Goodburn, R. and Grew, F., 1984, 'Objects of Stone' in S. S. Frere, Verulamium Excavations Volume III, Oxford. Greep, S. 1995, 'The worked bone, antler and ivory' in Blockley et al, 1112-52. Harrison, L. 1998, 'The Roman Brick and Tile' in A.J. Hicks, 'Excavations at Each End, Ash, 1992', Archaeologia Cantiana, cxviii, 91-172. Hawkes, C. F. C. and Hull, M. R., 1947, Camulodunum, First Report on the excavations at Colchester, 1930-1939, Reports of the Research Committee of the Society of Antiquaries of London No. 14, Oxford. Holbrook, N. and Bidwell, P. T., 1991, Roman Finds from Exeter, Exeter Archaeological Reports 4, Exeter. Houliston, M., 1993, 'Maidstone Baracks', in Canterbury's Archaeology, 1992-3. Houliston, M., 1995, 'Maidstone Roman Villa', in Canterbury's Archaeology, 1993-4. Jenkins, F., 1956, 'A Roman tilery and two pottery kilns at Durovernum (Canterbury)', The Antiquaries Journal, xxxvi, 40-56. 168 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Jenkins, F., 1960, 'Two pottery kilns and a tilery of the Roman period at Canterbury', Archaeologia Cantiana, lxxiv, 151-61. Jessup, R. F., 1939, 'Further excavations at Juliberrie's Grave, Chilham', Antiquaries Journal, xix, 260-81. Johnston, D. E., 1972, 'A Roman building at Chalk, near Gravesend', Britannia, iii, 112-48. Jones, C. E. E., 1987, 'A note on Roman bone hinges from the City of London', Transactions of the London and Middlesex Archaeological Society, 38, 19-21. Jongerius, A., 1970, 'Some micromorphological aspects of regrouping phenomena in Dutch soils', Geoderma, 4, 311-31. Jongerius, A., 1983, 'Micromorphology in agriculture' in P. Bullock and C. P. Murphy (Eds.), Soil Micromorphology, Berkhamsted, 111-38. Kelly, D. B., 1992, 'The Mount Roman villa, Maidstone', Archaeologia Cantiana, ex, 177-235. King A., 1978, 'A comparative survey of bone assemblages from Roman sites in Britain', Institute of Archaeology Bulletin, 15, 207-32. Lawson, A. J., 1976, 'Shale and jet objects from Silchester', Archaeologia, cv, 241-75. Leech, R., 1986, 'The excavation of a Romano-Celtic temple and a later cemetery on Lamyatt Beacon, Somerset', Britannia, xvii, 259-328. Leveson-Gower, J., 1995, 'Objects of shale and jet' in Blockley et al, 1184-93. Luik, M., 1994, 'Ein kOrpergrab mit spielsteinbeigabe aus kOngen, kreis Esslingen' Fundberichte aus Baden-Wurttemberg, 19, 357-81. Lyne, M., 1994, Late Roman Hand-made wares in South East Britain, (unpublished thesis of The University of Reading). MacGregor, A., 1976, 'Finds from a Roman sewer system and an adjacent building in Church Street', The Archaeology of York, The Small Finds 17/1, London. MacGregor, A., 1978, 'Roman Finds from Skeldergate and Bishophill', The Archaeology of York, The Small Finds 17/2, London. MacGregor, A., 1985, Bone, Antler, Ivory and Horn. The technology of skeletal materials since the Roman Period, London. Mackreth, D. F. 1995, 'Pre-Roman and Roman Brooches' in Blockley et al, 955-82. Mackreth, D. F., 1996, 'Brooches' in R. P. J. Jackson and T. W. Potter, Excavations at Stonea, Cambridgeshire, 1980-85, London. Macphail, R. I., 1980, 'Report on a soil in a Romano-British context at Lloyds Merchant Bank, London (LL078)', Ancient Monument Lab. Report 3045, London, English Heritage. Macphail, R. I., 1994, 'The reworking of urban stratigraphy by human and natural processes' in A. R. Hall and H. K. Kenward (Eds.), Urban-Rural Connections: Perspectives from Environmental Archaeology, Symposia of the Association for Environmental Archaeology, Oxford, Oxbow Monograph, 13-43. Macphail, R. I. et al, 1990, 'Soil micromorphological evidence of early agriculture in N.W. Europe.', World Archaeology, 22(1) , 53-69. 169 MARK HOULISTON Macphail, R. I. and Goldberg, P., 1999, 'Recent advances in micromorphological interpretations of soils and sediments from archaeological sites' in A. J. Barham and R. I. Macphail (Eds.), Archaeological Sediments and Soils: their Study, Interpretation and Management, London (in press). Macphail, R. I. and Cruise, G. M., 1993, 'Assessment of three thin sections of brickearth (coord. 95/45) from the arena area, Guildhall, London (GYE92)', Report to MoLAS. Macpherson-Grant, N., 1996, 'Flint-tempered ware from Maidstone Roman Villa', unpublished CAT archive report. Maltby J. M, 1981, 'Iron Age, Romano-British and Anglo-Saxon husbandry - a review of the faunal evidence' in M. Jones and G. W. Dimbelby (Eds.), The environment of man: the Iron Age to the Anglo-Saxon Period, B.A.R. British Series, 87, 155-203. Manning, W. H., 1985, Catalogue of the Romano-British Iron Tools, Fittings and Weapons in the British Museum, London. Marsden, P., 1994, Ships of the Port of London, first to eleventh centuries A.D., Archaeological Report 3, English Heritage. Mcllwain, A., 1980, 'Quernstones' in D. M. Jones (Ed.), Excavations at Billingsgate Buildings, Lower Thames Street, London, 1974, London and Middlesex Archaeological Society Special Paper No. 4, London, 132-4. Meates, G. W., 1987, The Roman Villa at LuUingstone, Kent, vol. IL The Wall Paintings and Finds, Monograph Series of the Kent Archaeological Society, vol. Ill, Maidstone. Miles, D., 1986, Archaeology at Barton Court Farm, Abingdon, Oxon., C.B.A. Research Report No. 50, London. Millet, M., 1990, The Romanization of Britain, Cambridge. Milne, G., 1996, 'Blackfriars ship 1; Romano-Celtic, Gallo-Roman or Classis BritannicaeV, The International Journal of Nautical Archaeology, Vol. 25, Numbers 3 & 4, August & November 1996. Monaghan, J., 1987, Upchurch and Thameside Roman Pottery, Oxford. Moore, P. D., Webb, J. A. and Collinson, M. E., 1991, Pollen Analysis, Oxford: Blackwell Scientific Publications. Murphy, C. P., 1986, Thin Section Preparation of Soils and Sediments, Berkhamsted. Murphy, C. P., McKeague, J. A., Bresson, L. M., Bullock, P., Kooistra, M. J., Miedema, R. and Stoops, G., 1985, 'Description of thin sections: an international comparison' Geoderma, 35, 15-37. Neal, D. S., Wardle, A. and Hunn, J., 1990, Excavations of the Iron Age, Roman and Medieval Settlement at Gorhambury, St Albans, English Heritage Report no 14, HBMC. Norris, N. E. S„ and Burstow, G. P., 1950, 'A Prehistoric and Romano- British site at West Blatchington, Hove', Sussex Archaeological Collections, 89, 1-56. Ocock, M. A. and Syddell, M. J. E., 1967, 'The Romano-British buildings in Church Field, Snodland', Archaeologia Cantiana, lxxxii, 192-217. Parkhouse, J. and Evans, D. R, 1992, 'Objects of stone' in D. R. Evans and V. M. Metcalf, Roman Gates, Caerleon, Oxbow Monograph 15, Oxford, 191-4. & 170 EXCAVATIONS AT THE MOUNT ROMAN VILLA, MAIDSTONE, 1994 Peacock, D. P. S., 1977, Pottery and Early Commerce. Characterization and Trade in Roman and Later Ceramics, London. Peacock, D. P. S., 1980, 'The Roman millstone trade: a petrological sketch', World Archaeology, 12.1, 43-53. Pearson, E. and Robinson, M. A., 1994, 'Environmental evidence from the villa' in R. J. Williams and R. J. Zeepvat, Bancroft, a late Bronze Age/ Iron Age settlement, Roman villa and temple/mausoleum, 2, 565-84, Aylesbury. Penn, W. S., 1957, 'The Romano-British settlement at Springhead; excavation of the bakery, site A', Archaeologia Cantiana, lxxi, 53-102. Percival, J., 1976, The Roman Villa, London. Philp, B., 1963, 'Romano-British West Kent A.D. 43-100', Archaeologia Cantiana, lxxviii, 74-82. Philp, B., 1973, Excavations in West Kent, 1960-1970, Dover. Philp, B., 1981, The Excavation of the Roman Forts of the Classis Britannica at Dover, 1970-77, Dover. Philp, B., 1991, The Roman Villa Site at Keston, Kent, Dover. Philp, B., 1996, The Roman Villa Site at Orpington, Kent, Dover. Pliny, Natural History, book XVIII (trans. H. Rackham, 1968, London). Pollard, R. J., 1987, 'The other pottery' in Meates, 1987, 164- 305. Pollard, R. J., 1988, The Roman Pottery of Kent, Monograph Series of the Kent Archaeological Society, vol. V, Maidstone. Pollard, R. J., 1992, 'The Mount Villa, Maidstone: the evidence of the Roman pottery for trade and the economy' in Kelly, 214-26. Pollard, R. J., 1995, 'The mid and late Roman pottery' in Blockley et al, 690-717. Richardson, K. M., 1960, 'A Roman brooch from the Outer Hebrides, with notes on others of its type', The Antiquaries Journal, 40, 200-213. Richardson, B., 1986, 'Pottery' in Miller et al, 96-138. Richmond, I., 1968, Hod Hill, volume two: Excavations carried out between 1951 and 1958 for the Trustees of the British Museum, London. Rivet, A. L. F„ 1969, The Roman Villa in Britain, London Roach Smith, C, 1876, 'On a Roman villa near Maidstone', Archaeologia Cantiana, x, 163-72. Scaife, R. G. and Macphail, R. I., 1983, 'The post-Devensian development of heathland soils and vegetation' in C. P. Burnham (Ed.), Soils of Heathland and Chalk Grassland, Ashford: Wye College on behalf of South East England Soils Discussion Group, 70-99. Schadler, U., 1995, 'XII Scripta, Alea, Tabula - new evidence for the Roman history of "Backgammon"' in A. J. de Voogt (Ed.), New Approaches to Board Games Research: Asian Origins and Future Perspectives, International Institute for Asian Studies, Working Papers Series 3, Leiden, 73-98. Seager Smith, R., 1995, 'Pottery' in Birbeck, 97-109. Sheldon, H., Corti, G., Green, D., and Tyers, P., 1993, 'The distribution of villas in Kent, Surrey and Sussex: some preliminary findings from a survey', London Archaeologist, Vol. 7, no. 2, spring 1993. Smith, J. T., 1963, 'Romano-British Aisled Houses', Archaeological Journal, cxx. 171 MARK HOULISTON Stead, I. M. and Rigby, V., 1986, Baldock, The Excavation of a Roman and Pre-Roman Settlement, 1968-72, Britannia Monograph Series No. 7, London. Stead, I. M. and Rigby, V., 1989, Verulamium: the King Harry Lane Site, English Heritage Archaeological Report No. 12, London. Tester, P. J., 1961, 'The Roman villa in Cobham Park, near Rochester', Archaeologia Cantiana, Ixxvi, 88-109. Tester, P. J. and Caiger, J. E. L., 1954, 'Excavations on the site of a Romano-British settlement in Joyden's Wood, near Bexley', Archaeologia Cantiana, lxviii, 167-83. Thompson, I., 1982, Grog-tempered 'Belgic' Pottery of south- eastern England, BAR(B), no. 108, Oxford. Ulbricht, I., 1978, Die Geweihverarbeitung in Haithabu, Die Ausgrabungen in Haithabu Band 7, Neumunster. Village, A., 1990, The Kestrel, London. Webster, G., 1975, 'Small Towns without Defences' in RodweU, W. and Rowley, T., Small Towns of Roman Britain, BAR (B), no. 15, Oxford. Webster, J., 1975, 'Objects of shale' in B. W. Cunliffe, 226-8. Wedlake, W. J., 1982, The Excavation of the Shrine of Apollo at Nettleton, Wiltshire, 1956-1971, Reports of the Research Committee of the Society of Antiquaries of London "No. 40, Dorking. Wheeler, M. A., 1932, 'Romano-British Remains' in Victoria County History, Kent, iii, London. Wilson, T., 'The lithic assemblage from Maidstone Roman villa', forthcoming. Worssam, B. C, 1963, Geology of the country around Maidstone (Memoir for 1 -.50,000 Geological Sheet 288), London. Worssam, B. C. and Tatton-Brown, T., 1993, 'Kentish Rag and other Kent building stones', Archaeologia Cantiana, cxii, 93-125. Young, C. J., 1977, Oxfordshire Roman Pottery, BAR(B), no. 43, Oxford. Zenkiewicz, J. D., 1986, The Legionary Fortress Baths at Caerleon. Volume II: The Finds, London. 172