Bellerophon in Boxford

12 mins read

A mythological mosaic revealed

Archaeological work near Boxford, Berkshire, has revealed a spectacular Romano-British mosaic, protected for c.1,600 years thanks to the villa it belonged to being terraced into the hillside. The land drain cutting through both the mosaic border and the villa wall to the right probably dates from around the time of the initial discovery of the villa in c.1870, but no mention was made of a tessellated pavement in the Victorian report, and to date the modern team has found no further physical evidence of these excavations. (Image: Cotswold Archaeology)
This summer, a community archaeology project in Boxford, Berkshire, unearthed a rare and beautifully preserved Roman mosaic. Experts soon hailed it as the most exciting discovery of its kind in Britain for half a century – but, as the project team reports here, the find was only the culmination of a wider five-year study of local Roman settlements.

The Lambourn valley in West Berkshire is rich farming country, and in the Roman period it lay within easy reach of Silchester, with Ermin Street providing good connections to Cirencester and the West Country. It should have been prime territory for rural settlements, but a glance at a map of known Roman sites shows much of the valley as a blank canvas. Now, an investigation in the parish of Boxford, about four miles north-west of Newbury, is beginning to fill in some of that picture.

The Roman Boxford project has its roots in the parish plan agreed by the community in 2008, which included proposals for the redevelopment of a Victorian parish room into a new heritage centre, and the production of a short book on the parish’s history. To tackle these tasks, the Boxford History Project was born, and within three years it had researched and published Boxora to Boxford: a short history of an ancient Berkshire Village. One aspect of Boxford’s past remained largely unexplored, however: three possible Roman sites, all within a mile of the village, which had been identified in the late 19th century but about which little further was known. Two, at Hoar Hill and Wyfield Manor Farm, had been simply recorded as surface find scatters. A third, evocatively known as Mud Hole, had been partially excavated in c.1870, though the resulting report was so cursory that there remained some doubt as to its true location. To investigate further, an invitation was extended to another voluntary society, the Berkshire Archaeology Research Group (BARG), to help carry out gradiometer and resistivity surveys, and five years of archaeological detective work began.

Fertile fields

It was not a simple task: all three sites lay within arable fields, limiting work to short windows of opportunity following the harvest each year. But in September 2012 we made a start on the slopes of Hoar Hill, south-east of the village. Initial geophysics were promising, hinting at the possible presence of a building c.50m long, but we would have to wait an entire year to return to probe further. During that time, though, detailed plans were developed for a small evaluation excavation to confirm the results, and to ensure the highest possible standards an arrangement was made with Cotswold Archaeology to provide a professional supervisor for the dig.

The Wyfield Manor Farm site, looking east.
mud Hole lies in the small valley to the top right. (Photo: Richard Miller)

When fieldwork resumed at Hoar Hill the following summer, resistance surveying revealed details of the buried building, showing that the structure consisted of a long central range of rooms, with wings to the north-west and south-east. Arranged around three sides of a courtyard, the whole complex measured c.50m by 20m: all the signs pointed to this being a sizeable Roman villa, probably the largest recorded so far in Berkshire. A week-long excavation quickly followed, with 45 volunteers from Boxford and BARG supervised by Matt Nichol of Cotswold Archaeology, and our three small trenches gave further credence to suggestions of a large structure, uncovering the foundations of substantial flint and lime mortar walls up to 80cm wide. In the north-west range these survived to a height of over a metre, and retained some of their dark red painted wall plaster. While we can tell that this part of the complex had been roofed with ceramic tiles, a combination of these and stone tiles were also seen in the central range. From these remains we recovered finds including pottery and coins from the 3rd and 4th centuries, as well as a fragment of window glass. So far, the project had been a stunning success. There was more to come.

At the same time as the dig, we had also been working to expand our gradiometer survey, and it was becoming increasingly clear that the Hoar Hill villa was set within a much wider landscape of features, abounding with rectilinear ditches and trackways. These findings were so detailed that we decided to devote the 2014 season solely to geophysics, covering all three sites; an undertaking that produced such encouraging results that a whole new phase in the project was proposed. The two voluntary groups and Cotswold Archaeology formed a partnership, modelled again on the principle of a small number of professionals supervising a much larger team of volunteer diggers, and this team was able to make a successful application to the Heritage Lottery Fund for a grant to support three seasons of excavation, devoting two weeks to each site. Our aim was to compare the various results in order to achieve a greater understanding of the local Roman landscape, and at the heart of this initiative there would also be a programme of training and community events, including a lecture series.

Grand designs

In 2015, we returned to Hoar Hill with an expanded programme of 13 trenches intended to more fully explore both the villa and its surrounding landscape. Careful targeting allowed us to work out something of the complex’s phasing: it appears that the two ‘side’ ranges are in fact the earliest elements, each initially forming separate structures. The north-west range began as a simple row-type house, and a portico and corner rooms were added later to produce something akin to a ‘winged-corridor’ layout. The southeast range presents more of a puzzle, though. Whilst of a similar size to its neighbour, its internal layout has a rare double-depth plan, including a suite of very small square rooms. A series of small extensions also appear to have been added to this building, but unfortunately the surviving archaeology in this area proved to be very shallow, hampering our interpretation of its likely function.

Expert analysis of the Mud Hole mosaic’s iconography suggests that it depicts episodes from the story of the heroic monster-slayer Bellerophon, as well as other mythological figures such as Hercules. (Images: Richard Miller / Lindsey Bedford)








The site saw further transformation around the middle of the 3rd century, when the two separate buildings were linked together by a narrow central range, c.32m long. Later still, this appears to have been subdivided into separate rooms and extended along half its length. The resulting courtyard villa had its peculiarities: in its final form there was no obvious ambulatory passage to connect the two side ranges, while the entry points are also uncertain, though a small square room connected to the courtyard wall may have played a role. And for all the effort that must have gone into creating it, this most grandiose phase of the villa was relatively brief. Pottery and coins from collapse layers in the central range suggest this section had been abandoned by the mid-4th century, and yet amongst the remains of its destruction we found clues to how impressive it must have been at its peak. Amongst the debris in this area there was more painted wall plaster, including some fragments large enough to suggest a possible design scheme, with octagonal and circular frames in a combination of red, white, and yellow that might seem somewhat garish to modern eyes, but was the height of Roman fashion.

Trenches in the wider landscape also revealed new information. About 30m downslope from the north-west range, a strong magnetic anomaly proved to be a detached bath house, measuring c.15m by 5m. Around two-thirds of the building was exposed by our excavations, including, at one corner, a projecting cold plunge pool. The hypocaust system under the warm room also seemed intact, though the uppermost floor layer has perhaps been lost to the plough – a few loose tesserae found in the ploughsoil here prompted some diggers to lament that it was the closest they would get on the project to a mosaic floor (O ye of little faith – little did they know what was to come!). Although the bath house was only revealed in plan, a few sherds of 1st- and 2nd-century pottery found in this area might point to an early Roman origin for the building, and perhaps the wider site. The ploughsoil here also gave up two arrowheads, both potentially Roman in date. These, together with a ballista bolt-head and a possible military fitting from the main villa, form an unusual (but not unheard of) assemblage for a domestic villa.

In the land around the villa, the system of ditches and enclosures covers an area of 5ha or more. Limited excavation of the ditch intersections suggests at least two phases in the Roman period, with an earlier field system replaced in the 3rd or 4th century by a highly designed landscape. Two ditches c.44m apart seem to have defined a formal approach route or ornamental area in front of the villa, while two side trackways appear to have provided more workaday access, including possible funnelling arrangements for managing livestock. Each track might also have served the opposing side ranges of the building. Could this be a hint of multiple occupation? Intriguingly, we have also seen a magnetic anomaly in the geophysics that looks tantalisingly similar to the one that marked the location of the bath house, in a comparable position downslope from the south-east range – might we have two bath houses? The site is so vast that many questions inevitably remain unexplored for the time being.

Crop of finds

Our second site, excavated in 2016 at Wyfield Manor Farm, provided a real contrast. It sits on a small plateau above a dry valley and has suffered from both the plough and quarrying activity over recent centuries, so choosing where to dig was a challenge. Mapping surface finds revealed a widespread but low-density scatter of Roman brick and tile across the site, while geophysics proved of limited help – it showed a series of mainly rectangular ditched enclosures, the largest measuring c.100m by 75m, but no obvious domestic focus.

This aerial photo shows the large building, probably a barn, that accompanied the Mud Hole villa. The partly exposed wall towards the south-east end (right, in this image) marks the original end wall before the building was extended. (Photo: Richard Miller)

To try to unpick these ambiguous clues, seven small trenches were placed over various ditches, revealing that the large enclosure ditch had been re-cut at least twice in the Roman period. One of those episodes also saw a remodelling of the site, with the addition of a smaller sub-rectangular enclosure. The most promising targets seemed to be two separate clusters of ‘noisy’ magnetic anomalies inside the main enclosure, and a larger trench was opened over each. The first revealed a series of irregular shallow pits containing middle/late Iron Age and 1st-century Roman pottery, which seem to have been dug to quarry chalk for marling, presaging the much later use of this area. One circular pit of uncertain date was possibly a storage pit or well, and near the trench edge we found a row of post holes and an associated gulley that were likely to be from a Roman timber building.

The second section of magnetic anomalies proved to be an area devoted to crop processing, where a group of seven small ovens had later been superseded by a row of post settings for a Roman building. The probable gable end of this structure had flint foundations surviving only to a single course, but more robust remains lay inside this building, where a sturdily constructed corn drier had been inserted, retaining evidence of its surrounding superstructure and parts of the drying floor. Environmental sampling from its stokehole produced evidence for grain germination, suggesting that at least in its final stage the drier had been used for malting/ brewing. It was possibly in use into the mid/late 4th century, but most of the pottery from Wyfield Manor Farm points towards a 2nd- and 3rd-century peak of activity. Taken together, the settlement appears to have been a substantial Roman farmstead, albeit one where the main domestic focus was probably just outside our trenches.

Mud Hole’s mosaic

This summer’s final excavation took place at Mud Hole. As its name suggests, parts of the field are susceptible to waterlogging and it was during drainage works that the site was first identified as being of archaeological interest, in c.1870. Under our project, a combination of geophysics and mapping surface finds confirmed that this was probably home to another villa with at least two substantial buildings.

Unlike the Hoar Hill complex, however, here the main villa building was relatively modest, measuring c.26m by 13m at its zenith, and the layout appeared to be fairly simple with a row of rooms bordered by front and back corridors. There were no projecting wings, although a small extension of c.3m at the northwest end does seem to have allowed the insertion of a bath suite. Initial impressions of the site might therefore be unprepossessing – but we soon uncovered signs that this was no ordinary villa. Clues emerged almost immediately in a trench designed to define the south-eastern end of the building, when a plain red tessellated pavement was noticed just below the ploughsoil during initial machining. We promptly switched to excavating carefully by hand, and as this painstaking work progressed it became clear that this was only the border of a much more highly decorated mosaic (see CA 332). The reality quickly dawned that the floor had been well protected by its terraced position (something that had also preserved sections of its walls up to c.60cm in height) and the covering of collapsed building material, and that much of its design had survived intact.

Volunteer numbers grew over
the course of the project, with more than
70 taking part this year – some of the 2017
fieldwork team are pictured here. (Photo: Joy Appleton)

Some of the country’s leading mosaic experts were contacted at this stage, but an immense challenge still remained: to expose and clean the full 6m by 2m strip of mosaic within the trench in the time available. Many volunteers worked long hours, reluctantly going home only when darkness fell – it turns out uncovering a mosaic can be addictive. As the mosaic specialist Anthony Beeson describes in more detail (see box below), the design uncovered contains elements thought to be unique within Britain. Whoever commissioned it seems keen to have displayed a knowledge of Greek mythology, and it has been suggested the floor might even have replicated a page from an illustrated manuscript. More than half the mosaic remained tantalisingly beyond the trench edge, but from the section that we did manage to reveal, we can tell that it depicts scenes from the story of Bellerophon, a heroic monsterslayer (and son of either the sea deity Poseidon or Glaucus, king of Corinth) who is best known for taming and riding the winged horse Pegasus, and killing the fearsome Chimaera.

The discovery of the mosaic was undoubtedly a highlight of the excavation, but the site contains much else of interest too. The second building, also boasting substantial masonry foundations and measuring c.18m by 6.5m, had been extended at least once and seems most likely to have been a barn. It was connected to a perimeter wall that includes a pair of sizeable gateposts, and the entranceway is aligned with the villa building, albeit off-centre. About 30m outside the perimeter wall, moreover, we found a rectangular cut feature measuring c.13m by 6m that might be a rare example of a Roman sunken featured building – something more commonly associated with the Anglo- Saxon period. In the wider landscape, two parallel ditches c.110m apart define the limits of the site, while two anvils previously found by detectorists came from the area of another possible building, perhaps a workshop. Also found nearby was a small copper alloy handle of a spatula, probably used for smoothing wax writing tablets. The handle depicts Minerva, goddess of learning, holding just such a tablet. Might it have once belonged to our literary-minded villa owner?

Community roots

Work on dating the Mud Hole villa is ongoing but there are stylistic clues that the mosaic could belong to the mid/late 4th century, and the extension with the bath house may have been added at this time too. If so, it shows that a significant investment was being made here at a time when the other two Boxford settlements were already in decline. That said, all three sites seem to have been contemporaneously occupied for a time and, despite their varied topographical situations, all were laid out on a similar south-west-facing alignment – we hope that future research will shed more light on how they related to each other.

Above all, the project has shown the benefits that can flow from professionals and ‘amateurs’ working together. Not only has new archaeological information been revealed, but new skills and capabilities have been developed. Professional supervision allowed the community groups to tackle sites of a size and complexity that would otherwise have been somewhat daunting. With this support much of the work has been volunteer-led, including the geophysical surveys, finds processing, logistical planning, aerial photography, photogrammetry, and report-writing.

Volunteer numbers also grew over the course of the project, with more than 70 taking part this year. They spanned all ages, with many getting their first taste of archaeology on one of our sites. Several Sixth-Formers who joined the early excavations were inspired to do archaeology or a related subject at university, coming back to help us during their holidays, and two of our volunteers even went on to work for Cotswold Archaeology, returning as professional supervisors in the project’s later years. Community archaeology has been a positive experience for many of the professionals, too, providing a welcome change from the routine of commercial work. Although future funding has yet to be secured, we all hope the project may yet have another chapter. After all, we do have the rest of that wonderful mosaic to uncover…


Joy Appleton and Janet Fuller (Boxford History Project); Lindsey Bedford and Steve Clark (BARG); Duncan Coe and Matt Nichol (Cotswold Archaeology).

This article appeared in CA 333.

1 Comment

Leave a Reply

Your email address will not be published.