The Upper Thames valley has been a focus for extensive examination of later prehistoric and Roman settlement in advance of gravel quarrying for many years. Nevertheless, a recently completed 25-year programme of excavation in the quarry at Gill Mill is one of the largest individual projects undertaken in this region, having covered an area of c. 130 ha, of which c. 75 ha have been examined after total stripping of topsoil and subsoil. The most prominent feature of this area is a small Roman roadside settlement which, although not fully excavated, sheds light on several important aspects of the regional settlement pattern and economy. These are outlined briefly here in advance of full publication.Footnote ²⁷

Gill Mill quarry, in the parishes of Ducklington and South Leigh, surrounds the eponymous surviving house situated at SP 3795 0695, the notional central reference point for the site, probably overlying the core of the Roman settlement which is the focus of this note ( fig. 6). The quarry lies on floodplain terrace gravels of the lower Windrush valley, one of several tributaries flowing into the river Thames from the north in the area west of Oxford (some 16 km distant). At this point the north-west–south-east course of the Windrush valley is just over 1 km wide with higher ground on both sides. By contrast, a short distance to the south, the gravels are far more extensive in the parishes of Stanton Harcourt, Standlake and Northmoor, where important sites of Neolithic to Anglo-Saxon date have been examined.Footnote ²⁸

FIG. 6. Plan showing overall extent of Gill Mill excavations with location of evaluation trenches and principal archaeological features. (Drawn by Hannah Kennedy and Victoria Hosegood. © Oxford Archaeology)

Even today the valley here contains a complex system of watercourses, with two main channels of the Windrush itself and two subsidiary streams dissecting the floodplain, a situation probably not very dissimilar to that in later prehistory. There is only limited evidence for activity in the valley bottom here before the middle Iron Age, when sites of varied size and form (not all yet excavated) were strung out along its north-east side over a distance of c. 3.5 km. By the late Iron Age, however, there was only one significant settlement focus, at the south-east end of the site in the area known as Tar Farm 6. Here occupation, having evolved from an open settlement to one partly enclosed in the later middle Iron Age, continued to develop up to the early/mid-third century a.d., undergoing major reconfiguration in the later first and early second century with the development of several rectilinear enclosures and internal trackways, joined to a NNW–SSE-aligned trackway at the north-east margin of the floodplain ( fig. 7).

FIG. 7. Simplified plan of main phased features in Gill Mill Tar Farm Area 6. (Drawn by Hannah Kennedy and Victoria Hosegood. © Oxford Archaeology)

Early Roman features at the southern margin of the site (Tar Farm 7) were aligned similarly to those in Tar Farm 6, despite being on the opposite side of the Hardwick Brook, one of the subsidiary streams, but while the small enclosures in this area, probably for stock, were paralleled at the extreme north-west end of the quarry over 2 km distant, they were quite different from features seen either in adjacent Tar Farm 6 or in the main settlement area. These enclosures may have been quite short-lived and were out of use by the mid-second century at the latest. A notable feature of the extensive open space in Tar Farm 7 north of these enclosures was a small unenclosed group of cremation burials ( fig. 8) dated to the mid/late first century, an uncommon rite for the region in this period in a rural settlement context.

FIG. 8. Plan of eastern part of the nucleated Roman settlement, all Roman phases combined. (Drawn by Hannah Kennedy and Victoria Hosegood. © Oxford Archaeology)

The early second century saw the establishment of the principal Roman settlement in the valley centred almost 1 km from the Tar Farm 6 complex ( fig. 6). This eventually covered an area of perhaps as much as c. 10 ha in the floodplain. It was based around a road (Road 1) running roughly NNE–SSW across the valley. Close to its centre a large open area west of the road was surrounded by fairly irregular enclosures on its north, west and south sides, with further roadside plots on the west side of the road to the south, beyond the northern channel of the Windrush. The open area was probably entered from the east by a second road (Road 2) which ran ESE down the valley (the putative junction lies within the curtilage of Gill Mill House). What lay in the southern angle of the junction is unknown. To the north were further irregular enclosures. At the point where Road 2 is first seen it traversed an area undivided into enclosures, with to the south an important inhumation burial of an adult male in a wood-lined chamber surrounded by a ring ditch defining a mound over the burial ( fig. 8). This extremely unusual ensemble dates to about the middle of the second century. Just east of the burial mound and almost certainly respecting its location was a very different system of land allotment — an arrangement of more or less rectilinear enclosures which extended along the north side of Road 2 for more than 500 m. South of the road corresponding enclosures extended less far to the east but most seem to have been less intensively used. Another road (Road 3) ran north-eastwards towards the edge of the floodplain from the north side of Road 2 between two of the enclosures but was not traced beyond the excavated area. The metalled road surfaces are not well dated, but may have been entirely of later Roman date. They were poorly preserved in Road 2 east of its junction with Road 3 and the most easterly stretch of Road 2 was probably never surfaced, nor did it extend beyond the line of the Hardwick Brook; features beyond this point, relating to the early Roman south-eastern settlement focus in Tar Farm 6, were on a completely different alignment. It is likely that there was a connection between the two areas at a point where the stream was crossed; nevertheless the radical differences in layout and the different chronological trajectories of the two areas perhaps suggest different land holdings.

The most intensive activity within the nucleated settlement was of the later Roman period, from the middle of the third century up to c. a.d. 370 (see further below). West of Road 1 the enclosures around the open area were reconfigured in this period and occupation extended further south along this road than previously, activity in an area between the Standlake Brook and the southern channel of the Windrush (Gill Mill Area 2) perhaps not commencing much before the fourth century. East of Road 1 the main enclosure complex was reworked but the overall layout was not significantly altered.

Structural evidence, quite limited in quantity overall, was mainly of this later Roman period. At least five stone-founded buildings are known. Two lay west of Road 1 in Gill Mill Area 2, but little can be said about them as this area was only subject to trenching.Footnote ²⁹ The other three were all on the south side of Road 2, two of them close together just west of the roadside enclosure complexes ( fig. 8). One was set back from the road into the edge of the second-century burial mound and probably deliberately referenced this feature, which was almost certainly still upstanding. This building and the adjacent structure were simple single-cell buildings with substantial foundations of immediately local coarse concreted gravel (‘ragrock’). A further stone-footed building (23391) lay 200 m east, only partly within the excavated area, accessed from Road 2 by a pitched stone causeway laid over the infilled roadside ditches. The question of building density is key. How many of the numerous enclosures that made up most of the site's layout contained structures of any kind, let alone ones with a clear domestic function? In the middle Roman period (c. a.d. 120–250) only three probable (timber) structures were identified within the main ‘settlement’ area, all in the area of enclosures north of Road 2. The later Roman settlement included the five stone buildings already mentioned, and a sequence of circular structures including a possible smithy north of Road 2 directly opposite building 23391. It may be no coincidence that the most clearly defined buildings were apparently grouped in pairs widely distributed across the site, but if so the significance of this is uncertain. Not all the buildings were necessarily domestic in function, though most probably were. The clearest evidence comes from building 23391, with a wide range of domestic and other debris from associated surfaces and closely adjacent pits. The nearby circular structures, however, contained a series of hearths, one of which was probably used for smithing — domestic use here is uncertain. These buildings were defined on the basis of surrounding drainage gullies; an absence of other structural features strongly suggests the use of mass wall construction, most likely in cob, as has been proposed elsewhere in this region.Footnote ³⁰ Without associated drainage features, such buildings might be impossible to identify, particularly where, as for large parts of the Gill Mill complex, post-Roman ploughing has removed all above-subsoil traces of archaeological deposits. Direct evidence for building materials is limited; tile and stone roofing material were certainly used, but quantities were modest. A small collection of tesserae from a pit west of Road 1 is anomalous but suggests the presence of one building of some architectural pretension.

The site therefore seems to lack the concentrations of buildings which would normally be considered characteristic of roadside settlements, so what did it represent and what was its role within the regional settlement pattern? Environmental evidence provides a partial answer in the form of significantly consistent characterisation of the setting of the site — all the indicators are of damp pasture land. This picture seems to have applied throughout the main periods of activity at Gill Mill, from the middle Iron Age through to the late Roman. There is no evidence whatsoever for Anglo-Saxon activity within the project area, while much of it remained damp pasture through the medieval period.

It is unsurprising therefore that there is relatively little evidence for large-scale arable production in the Iron Age and Roman settlements. Crop-processing debris, mainly of the dominant regional species, spelt wheat and barley, occurred mostly in modest quantities, while there were no ‘corn-drying’ ovens, a regular feature of the region. A total of one millstone and nine quern fragments from all the late Iron Age and Roman phases also suggests a relatively limited emphasis on cereals, in this case their conversion to flour or meal. This might indicate a small site population; equally, however, with a lack of focus on arable production, a significant proportion of the inhabitants' requirements for meal and flour might have been brought to the site already processed.

In contrast to the slender evidence for arable-related activity, the site produced a substantial assemblage of animal bone (c. 33,600 refitted fragments were analysed), mostly of middle Roman and (particularly) late Roman date. The Roman phase groups were dominated by bones of cattle, which formed an increasingly large proportion over time. The late Roman phase group has a higher proportion of cattle than any comparative site in the vicinity, except for specialised butchery deposits such as that at Chester Street in Cirencester.Footnote ³¹ Sheep were the second most common taxon, but horses were also important (and in some phase groups were more numerous than pigs). Overall the animal bones indicate a heavy emphasis on cattle raising, but this activity presumably went beyond simply producing animals for local consumption, and a more extensive pattern of exploitation of these animals may be suspected (see below).

The artefact record provides limited evidence of other activities within the site which might have complemented stock raising. Transport-related objects are well represented and include ten linchpins, other vehicle fittings and also part of an oak cart wheel from a late Roman pit. There is relatively little indication of craft activities, whether in the form of tools, debris or finished objects. Except in the vicinity of building 23391, the richer and more varied finds assemblages tended to be recovered from the numerous pits and over 20 small wells that dotted the site. Some of these had partly waterlogged fills containing organic material, though the latter was not generally well preserved. The importance of pits at this site is emphasised by the fact that just over 40 per cent of all the pottery (by sherd count) came from pit fills — an unusually high proportion by comparison with most sites in the region, where assemblages tend to be dominated by material from ditches. The pottery provides both general and some very specific indicators about aspects of site character. The early Roman assemblage from Tar Farm 6 includes some distinctive material quite anomalous in a rural settlement milieu but readily paralleled in military contexts in the region. These exotica apart, the pottery assemblage (c. 60,000 sherds weighing just over one tonne) contains little that is intrinsically remarkable. In the second and third centuries it is dominated by products of a local ‘west Oxfordshire’ industry, while the Oxford industry became a more important source in the late Roman period. Further aspects of note are a relatively high representation of black burnished ware, the latter amounting to 15.5 per cent of total REs, reflecting an increasingly observed pattern of high levels of this ware at certain sites in the Upper Thames valley, contrasting with the rapid fall off in these levels east of Cirencester mapped by Allen and Fulford.Footnote ³² It is notable that among a large number of ‘marked’ vessels (57 in total) from the site, 40, including the only two literate graffiti, were on black burnished ware, suggesting a perceived need that these vessels, in particular, should be individually identifiable.

The pottery evidence suggests that occupation in the main Gill Mill settlement effectively ceased before the end of the fourth century. This is borne out by the substantial list of 1,047 coins, of which 764 could be assigned to issue periods.Footnote ³³ A scarcity of coins of the House of Valentinian and an absence of later material support the suggestion that activity after c. a.d. 370 was at a very low level. There was a notable peak of later third-century coin loss, unusual in a rural context. Overall, the assemblage profile is difficult to parallel closely and it may incorporate some votive material. Religious activity is directly represented by four pieces of worked stone: a plain small altar, a female deity, a Genius and a weathered horse-and-rider relief incorporated in a nineteenth-century building; but whether these originally derived from a single shrine or temple structure or from more diverse contexts is unknown. Other small objects of religious significance include fragments of three Venus pipeclay figurines and a fine jet crescent pendant which may have been an item of priestly regalia.

In sum, the main site area has several characteristics indicative of a minor nucleated settlement, of which the most obvious are the simple network of paved roads with associated enclosures, although nothing is known of the probable focal area. However, several aspects, already outlined, suggest that this was not a typical roadside settlement. First is the topographical setting, in a location that must have been vulnerable to periodic flooding. Second is the scarcity of buildings — arguments relating to the archaeological invisibility of cob-built and comparable structures can only be pushed so far — and the spatial arrangement of those that are known in the later Roman period. The possibility remains that there were further structures in the unexamined central area and it is a legitimate conjecture that a shrine or temple, a consistent feature of minor roadside settlements and ‘small towns’ in this region, as elsewhere, was located here. Perhaps more remarkable is the number of enclosures and the systematic nature of the layout of at least some of them. Many would have been entirely suitable for stock holding and it is clear that cattle loomed large in the activities in and around the site while arable agriculture, for which the floodplain was not suitable, was not an important aspect of the site's economy. Although the animal bone assemblage is dominated by cattle, it is probable that these only represent a fairly small proportion, effectively the remnant, of a much larger population; it is suggested that the principal concern of the site's occupants was with cattle herding and particularly with accumulation of stock to be driven to market, arguably at major regional centres such as Cirencester, if not beyond. Whether a more far-reaching connection to official/military supply networks was involved is perhaps possible, but unknown. Either way, it seems likely that a proportion of the site's inhabitants, drovers, could have been absent for significant periods. Is it possible that, as has been suggested for some Iron Age sites in exactly comparable ecological niches, occupation was only seasonal?Footnote ³⁴ In the present case this is only suggested for part of the population — the infrastructure of the settlement surely indicates some permanent occupation — but a partly mobile population would at least help to explain the scarcity of structures. Perhaps the ‘settlement’ should be seen as the location of seasonal cattle round-ups, with associated fairs and religious festivals, at which times incomers, both local and perhaps also from further afield, occupied temporary accommodation in the form of tents and shelters erected in some of the enclosures, while others held large numbers of animals awaiting the drive to market. The associated activities generated quantities of rubbish which were then buried in the numerous pits that dotted parts of the site.

If this scenario is plausible, what is the settlement context in which it might have occurred? A potentially long-lived tradition of seasonal activity within the Thames floodplain might provide a background. LambrickFootnote ³⁵ has certainly argued for a carefully integrated use of the closely adjacent territory around Stanton Harcourt combining agricultural practice and an older ceremonial landscape within a communal organisational framework. Survival of this structure into the Roman period is possible, but the scale of activity at Gill Mill is greatly larger than that postulated for the earlier period, perhaps suggesting other possibilities.

A feature of much of the settlement pattern of the Upper Thames valley is a significant reorganisation of the landscape in the early second century,Footnote ³⁶ reflected in transformation of settlement plans, abandonment of some sites and a marked increase in the visibility of systems of trackways.Footnote ³⁷ It is arguable that these developments in turn reflected radical changes in patterns of land tenure, perhaps most clearly seen at sites like Claydon Pike,Footnote ³⁸ probably established as an estate compound, even if the centre of that estate (possibly, though not certainly, villa based) was located elsewhere. Agricultural specialisation, with an emphasis on hay production and herding, is suggested for this site.Footnote ³⁹ The inception of the main Gill Mill settlement was approximately contemporary with these changes and is therefore quite likely to have been a consequence of the same factors which drove the other developments in the early second century, of which changes in patterns of land ownership and tenure may have been the most important, albeit the most difficult to grasp from archaeological evidence. Placing Gill Mill within the framework of a large integrated estate might be the best (though not necessarily the only) way to account for the provision of infrastructure for the specialised economic focus on cattle raising, in a site which did not recommend itself as a centre for a large population. The location of such an estate centre is unknown. It may have been at some distance; the nearest known villa site, leaving aside a very uncertain example at Ducklington itself,Footnote ⁴⁰ is at Shakenoak, 7 km north of Gill Mill.Footnote ⁴¹ However, the occurrence of the first-century military-style pottery in Tar Farm 6 might also be relevant. If this indicates, for example, the presence of a former auxiliary soldier, perhaps of local origin, this man, now a Roman citizen, would have been an individual of consequence in the community and potentially in a position to enhance his family's standing in the local settlement hierarchy,Footnote ⁴² the family consequently enjoying the sort of status that enabled them to take advantage of the changed circumstances of the early second-century reallocation of landholding. Furthermore, nineteenth-century finds of lead-lined coffins on the rising ground barely 150 m north of Tar Farm 6 indicate a relatively high-status cemetery and therefore hint at the presence of an equivalent associated settlement in the near vicinity. There is no certainty that such a settlement was formally linked to the Gill Mill complexes, and the scenario involving these possible associations is highly speculative, though plausible in general terms, even if not in detail.

The intensive late Roman activity on the main Gill Mill site suggests that its economic focus was maintained through most of the fourth century. However, as indicated above, there is very little evidence for activity anywhere within the site in the last quarter of the fourth century. This is unusual in the region, where rural sites occupied in the early/mid-fourth century generally remained in occupation up to the end of the period, as can be seen very clearly at places like Shakenoak. Perhaps the very particular character of the Gill Mill complex suggested here made it vulnerable to some specific vicissitude, possibly environmental, but it may be as likely to do with disruption of the organisational structure which had successfully maintained it for some 250 years. This could have involved changes to the principal markets served from the site, but might perhaps have been caused by changes in personal or family fortunes. It is unclear from the archaeological evidence whether abandonment of the site was a gradual or rapid process; once out of use, however, its only legacy in the local landscape was as the location of the later mill site.