INTRODUCTION
In the last few decades, archaeological excavations have revealed a great deal about the nature and organisation of Roman pottery production in the immediate post-Conquest period in eastern England.Footnote 1 In the Cambridge region, four kiln sites have been investigated, adding to our understanding of kiln distribution and kiln architecture variability, as well as shedding light on the changes which occurred to the organisation of ceramic production in the mid-first century a.d. General patterns are now emerging for the region, but evidence from new excavations continues to challenge the notion that there existed a simple or singular model of production in this period.
The results from investigations at Duxford feed directly into these ongoing debates. Excavated between July and August 2013, the site contained six early Roman pottery kilns and associated material, including one of the largest assemblages of kiln plates found in the eastern region. The kilns were all in use during the first few decades following the Roman Conquest, with hints that production was seasonal and, most significantly, focused on the specialist manufacture of flagons.
This paper examines the kilns and the pottery and discusses the organisation and output of production at Duxford. Analysis focuses on the kiln products and kiln furniture, as well as considering the likely market for vessels produced at the site. The results are set against wider emerging patterns of early Roman ceramic production in the Cambridge region and provide an opportunity to touch on broader themes of pottery manufacture and consumption during the mid- to later first century a.d.
SETTING AND SITE SUMMARY
The Moorfield Road, Duxford, excavations were undertaken by Pre-Construct Archaeology Ltd in advance of construction of a new logistics depot on behalf of Welch's Group Holdings Limited and Wrenbridge (Moorfield Road) Ltd. Measuring 0.4 ha in size and located c. 10 km south of Cambridge, the site centres on NGR TL 4810 4700 ( figs 1–2). The river Cam is c. 300 m east of the excavation area; a tributary stream flows along the site's southern boundary. The underlying geology is Nodular Chalk of the Holywell Formation, which in the south of the site was overlain by layers of alluvium from the stream, up to 0.5 m deep.
Site location showing key Roman roads and other early Roman kiln sites. (© Pre-Construct Archaeology Ltd)
Detailed plan of site showing location of kilns. (© Pre-Construct Archaeology Ltd)
The site lies within a landscape rich in archaeological evidence. During the Late Iron Age and Roman period, the surrounding area was primarily agricultural, comprising extensive field-systems, some of which are still visible as cropmarks (e.g. CHER 9741).Footnote 2 Evidence for an early Roman settlement has previously been identified very close by on the opposite (western) side of Moorfield Road, comprising a series of ditches containing concentrations of early Roman pottery (c. a.d. 50–120), in addition to surface finds of early to later Roman tile and pottery.Footnote 3 An Iron Age/early Roman cemetery and associated shrine were excavated 1 km south of the site,Footnote 4 which, although predominantly predating activity at Moorfield Road, demonstrates a significant pre-Roman and early Roman presence. The Roman town at Great Chesterford, which included a short-lived pre-Flavian Roman fort, was located 5 km to the south-east.Footnote 5 It is likely that the development of Great Chesterford had a significant impact on the surrounding communities. Likewise Cambridge Roman town had contemporary origins and also included a possible early Roman fort.Footnote 6
The site would have been located close to the route of the Icknield Way,Footnote 7 thought to have been aligned along the same route as the A505, c. 200 m north of the site ( fig. 1). Further smaller early Roman roads and route ways have been identified in Sawston,Footnote 8 2.5 km north-east of the site,Footnote 9 which are likely to have extended south towards the Icknield Way, if not beyond, possibly linking up to Great Chesterford. If this was the case, then the site at Moorfield Road would have been located very near to the crossroads where these trackways or roads intersected, which is potentially pivotal to the development of the site.
Evidence of pre-Roman activity was slight ( fig. 3). This, along with the organisation of the Moorfield Road site beyond that relating to early Roman pottery production, is not discussed in detail in this paper, with the exception of those features which had a direct relationship with the kilns. A detailed account of the remainder of the site can be found in the post-excavation assessment,Footnote 10 while a summary is provided below.
Phased plans of the site from prehistory to the late Roman period. (© Pre-Construct Archaeology Ltd)
The site was located on low-lying, seasonally waterlogged ground beside a tributary of the river Cam and was divided into a series of adjoining curvilinear enclosures defined by large ditches, the origins of which were in the immediate post-Conquest period. The intermittently wet ground conditions were indicated by both the fine silt fills in many of the features in the south of the site and the contents of bulk soil samples taken from the enclosure ditches and three of the kilns (3, 5 and 6).Footnote 11 The ditches were repeatedly recut and redefined due to the episodic flooding of the southern half of the site, creating the need to reinstate boundaries after each inundation event. The stratigraphic relationships and evidence for periodic flooding observed in several of the features provide the best evidence that the site was occupied on a seasonal basis, which in turn has important implications for understanding the nature of pottery production at the site, discussed in more detail below.
This landscape would have been ideally suited for use as pasture for livestock, an observation supported by the cattle-dominated animal bone assemblage. In addition to the animal bone, small amounts of other domestic-type waste in the ditches, including some non-kiln-related pottery, indicate that this area was agricultural ‘infield’. Significantly, finds and stratigraphic evidence show that the ditches and associated infield were already established before this area began to also be used, periodically, for pottery manufacture.
Two Roman trackways were identified: one early Roman (Trackway 1) and one later Roman (third to fourth century) (Trackway 2, fig. 3), the latter constituting the only evidence for activity on the site beyond the later first century. Trackway 1 was contemporary with at least some phases of pottery production and comprised two parallel ditches, aligned north-west–south-east, appearing to lead down towards the stream to the south of the site. It seems probable that its construction could have been connected with the need to transport fired pots off the site.
A fragment of a copper-alloy military armilla converted into a finger-ring or pendant was found in one of the enclosure ditches (Ditch 4).Footnote 12 The armilla was awarded to ranks below that of centurion for force of arms in battle and those found in southern Britain form a distinctive group associated with the Conquest period; several found in rural contexts may relate to land grants to veterans.Footnote 13 Additionally, a copper-alloy coin of Nero, dated a.d. 64–8, was found in the fill of Ditch 7.Footnote 14
A possible quarry pit (Pit 3) adjacent to Kiln 1 contained a small quantity of early Roman pottery and cattle bone, as well as a semi-complete human skull and vertebrae which would have still been articulated (i.e. ‘fleshy’) when deposited. The remains are from an adult male and show evidence of sharp force trauma consistent with a double-blow decapitation, which is likely to have been the cause of death.Footnote 15 The angle of the cuts suggests an attack from above, with the change in direction suggestive of a moving target. An attack from a person on a horse, for example, could cause these sorts of wounds.
The bones were recovered from the secondary fill, suggesting that this feature had been open for some time before the remains were deposited. While it is not uncommon to find small quantities of disarticulated human bone in and around later prehistoric and Romano-British settlements, this skull would still have had soft tissue intact and been recognisably human when deposited. Although human burials are sometimes found in association with Roman kilns, apparently forming ‘foundation’ or ‘closure’ deposits,Footnote 16 the excessively violent and seemingly sudden nature of this death is more suggestive of conflict. How this related to the day-to-day activity at the site is unclear; however, that the manner of death and the deposition of these remains on the site were in some way linked to pottery production cannot be ruled out.
THE KILNS (FIGS 4–6; Table 1)
Almost all the activity summarised above took place within a short timeframe, dated on the basis of the associated pottery to the middle of the first century (c. a.d. 50–80). Within this period the development of the enclosure system and the relative positions of the different kilns in the chronological sequence are indicated by a combination of stratigraphy and spatial relationships between features. The occurrence of kiln-related material at different levels within the ditch fills is an important indicator of which kilns were contemporary with which ditches, although this does rely on the assumption that dumps of kiln material are most likely to be associated with kilns located close by. The kilns are numbered 1–6 in their perceived chronological order.Footnote 17
Plans of kilns. (© Pre-Construct Archaeology Ltd)
Sections of kilns. (© Pre-Construct Archaeology Ltd)
Photographs of kilns. (© Pre-Construct Archaeology Ltd)
SUMMARY OF DUXFORD KILNS
* Maximum internal diameter at base.
** Number of uses for which evidence was present. Other uses may have left no discernible trace due to raking-out of kilns after use etc.
*** Kiln 4 was originally the opposite way around, with the kiln chamber at the east end. This arrangement was later reversed: the former kiln chamber became the stokehole and the original stokehole was lined with clay and became the kiln chamber.
The six kilns demonstrated varying levels of preservation, determined by the depth at which they had been cut into the natural chalk and whether they had been truncated by subsequent Roman features. All six kilns date to a.d. 50–80, based on the types of pottery recovered, thus spanning the pre-Flavian to Flavian period. The pottery repertoire produced within the different kilns shows little chronological change, the exception being Kiln 6, where the vessel forms produced have a slightly later currency (up to c. a.d. 100). Further phasing of the kilns is based on their stratigraphic relationships with other features and spatial associations, particularly with dumps of kiln material (wasters, kiln furniture, fired clay lining fragments) in the adjacent ditches.
All six kilns were single-chambered and single-flued, with two basic designs identified: those with free-standing clay pedestals positioned centrally within the kiln chamber and those with projecting ‘tongue-support’ pedestals integral to the kiln lining. The kilns were all cut into the natural chalk with a thick layer of clay lining applied to the walls of the kiln chamber and flue. In the case of the tongue-support type, a projecting wall of clay was also built extending out from the midpoint of the rear wall of the chamber, across almost the full width of the firing chamber. The kilns of central pedestal type incorporated a free-standing roughly rectangular block of clay c. 250 mm high, positioned in the middle of the kiln chamber floor. Two of the free-standing pedestal kilns (4 and 5) had integral ledges moulded into the kiln lining c. 300–350 mm above the kiln chamber floor, into which the outside edges of the kiln plates would have slotted. Clay flue arches, built out from the sides of the kiln chamber where it joined the stokehole, survived in two of the kilns (1 and 4) and partially in another (5). With the exception of Kiln 5 (elongated oval), all of the kilns were ‘keyhole’-shaped in plan, a common kiln shape paralleled locally by the two ‘Type-D’ kilns at Addenbrooke's and the ‘Group-3’ kilns at Greenhouse Farm.Footnote 18 There is no indication of technological progression or development over time, with the two basic kiln designs occurring throughout the sequence.
Different firing episodes were discernible from ash deposits in the kiln chambers, in some cases separated from each other by silting or dumping layers indicating intervening periods of disuse; sometimes re-lining or repairs to elements of the kiln structure were also evident. Four kilns (2–4 and 6) demonstrate two to three separate firings, while Kiln 5 had one to two firings and Kiln 1 appears to have had a single firing event. These figures represent the minimum number of firings as evidence of other uses may have been entirely raked-out between firings. That said, it is likely that the repeated firing of the lining and pedestal and the act of breaking into or demolishing the superstructure to remove the fired pots after each firing, made the structure brittle and not suitable for continued use. After their final use, the kilns were deliberately backfilled with dumps of kiln furniture, wasters and fragments of kiln lining.
Kiln 1 [189] ( figs 4.1, 5.1 and 6.1)
Kiln 1 was an elongated ‘keyhole’ shape in plan, comprising a steep-sided linear stokehole extended north-eastwards, connecting via a surviving flue arch to a circular kiln chamber with near-vertical sides, a flat base and a broadly rectangular fired or vitrified clay central pedestal (402). The kiln chamber had a vitrified clay lining up to 110 mm thick (281), which extended out of the chamber and along part of the stokehole. The lining had a ‘ragged’ appearance in the stokehole, suggesting that it represented the truncated remnants of a lining which had been partly scoured away by raking-out of the kiln chamber. The earliest deposit in the kiln was an ash layer, present in the flue between the stokehole and kiln chamber, deriving from the kiln's last firing. This was sealed by layers representing the abandonment and initial silting-up of the disused kiln, followed by several large, deliberately dumped, deposits of wasters, kiln furniture and fired clay lining and structural fragments. Kiln 1 appears to have been one of the earliest kilns based on its stratigraphy, being cut by Ditch 2 [195].
Kiln 2 [277] ( figs 4.2, 5.2 and 6.2)
Kiln 2 was located in the south-west of the site and was severely truncated by Ditches 5 [278] and 14 [384]; only half of the kiln chamber and the beginning of the flue survived. It appears originally to have been an elongated ‘keyhole’ shape in plan with steep concave to near-vertical sides. The earliest kiln deposits comprised two thin silt layers, indicating a short period between the kiln's construction and first use and possibly representing minor flooding. These were sealed by a thick layer of ash which extended almost fully across the kiln chamber and the surviving part of the flue, representing at least two firing episodes. Use deposits extended beneath the pedestal, indicating that this must have been a replacement for an earlier pedestal, potentially one that had shattered. Following these firings, the kiln's flue was deliberately blocked with clay containing broken kiln plates, fired clay kiln lining fragments and pottery. This blocking is likely to have taken place during the final firing of the kiln as it showed signs of having been fired but not vitrified, which may have been a technique employed to ensure certain conditions during the final firing. Both the kiln chamber and stokehole were subsequently backfilled with dumped kiln debris, including waster sherds, broken kiln plates and fired clay lining and demolished structural fragments.
Kiln 3 [229] ( figs 4.3, 5.3 and 6.3)
Kiln 3 comprised a linear stokehole with steep concave sides and a slightly ‘bulbous’ eastern end, extending westwards into a circular kiln chamber, the base of which sloped down gently to the west, possibly as a result of repeated raking-out of the chamber floor between firings. Extending out from the rear edge of the kiln chamber and across the full width of the firing chamber was a tongue-support pedestal, constructed at the same time as, and integral to, the kiln lining. The central part of the pedestal had been destroyed by Ditch 15 [249]. Deposits of ash in the kiln chamber, interleaved with layers of washed-in silt, indicate at least two firings separated by brief periods of disuse when the kiln was open to the elements. As in Kiln 2, the flue had then been deliberately blocked with a wall made from fired clay lining fragments and some fresh clay, probably put in place during the final firing. Later deposits in the kiln reflect a combination of natural silting-up after disuse and demolition or dumping of the kiln lining, superstructure and other associated material.
Kiln 4 [153] ( figs 4.4, 5.4 and 6.4)
Kiln 4 comprised a circular stokehole feeding, via a surviving flue arch, into a circular kiln chamber with an integral ledge 20–40 mm deep, moulded into the chamber wall 310 mm up, into which the outside edge of the kiln plates forming the suspended floor would have slotted. Their inside edges would have rested on a central pedestal. Uniquely among the kilns, Kiln 4 had evidence for originally having had a different layout. The lower part of the stokehole had a surviving clay lining, which had been fired fairly hard but had not vitrified, indicating that the stokehole was originally the kiln chamber and the kiln chamber originally the stokehole. The reason for the reversal of the kiln's original layout is uncertain, although the original alignment with the stokehole facing into the prevailing wind may have resulted in too much up-draught. That the kiln was fired several times after its layout was reversed is suggested by the positions of several kiln plates, which lay in situ where they had fallen, with at least one later firing deposit built up around them. In addition to this, a repair had been made at some point to the top of the north wall of the kiln chamber utilising recycled fragments of kiln lining and some fresh clay; the hardening and discolouration of this area of repair indicates that the kiln was refired at least once after it was made. The quantity of discarded kiln-related pottery in the final fills suggests that pottery manufacture was still going on in the vicinity after the disuse of Kiln 4, presumably in nearby Kiln 5.
Kiln 5 [152] ( figs 4.5, 5.5 and 6.5)
Kiln 5 was located between the two ditches which made up Trackway 1; its positioning suggests that the trackway was no longer in use by this time. Kiln 5 was oval in plan, with an elongated stokehole with steep concave sides extending south-eastwards into a circular kiln chamber. The kiln chamber floor was largely flat but sloped down slightly towards the rear, possibly due to raking-out between firings. As in Kiln 4, the kiln chamber lining had an integral ledge 20–30 mm deep partway up, where the outside edge of the kiln plates would have rested. The ash deposits in the kiln chamber surrounded a void, indicating that a central pedestal was originally present but had been removed during the disuse and backfilling of the kiln, presumably for reuse in another kiln nearby.
Kiln 6 [258] ( figs 4.6, 5.6 and 6.6)
Kiln 6 was truncated on its north side by late Roman Ditch 18. Its stokehole was circular, joining a broadly circular kiln chamber to the west, the floor of which was stepped up a few centimetres higher than the stokehole. The kiln chamber walls and the flue where the chamber joined the stokehole were lined with clay, which formed an integral tongue-support pedestal extending from the rear wall of the kiln chamber. Kiln 6 contained evidence of at least two firing episodes, separated by an interval during which the flue (and possibly other parts of the kiln) was relined. Interestingly, this kiln also displayed evidence for the seasonal flooding described above, consisting of an alluvial layer towards the top of the kiln, once it had gone out of use. Although Kiln 6 contained relatively little pottery compared to some of the other kilns, large dumps of kiln products and waste were recovered from the adjacent slot through enclosure Ditch 5.
KILN MATERIAL By Kayt Marter-Brown
Material deriving from the six excavated kiln structures comprised a large assemblage (c. 245 kg including sampled kiln lining) of both permanent and portable kiln furnishings. Much of the kiln lining was found surviving in situ, with, in all but one instance, good preservation of the kiln pedestals. Decayed fragments of possible superstructure were retrieved and samples were taken of the extant kiln linings. Within the portable furniture recovered were a small quantity of fire bars and a single clay disc. However, it is the perforated clay floor plates which are of particular importance in providing one of the largest groups of this distinctive material type identified in the region in recent years. Despite the fragmentary condition of much of the assemblage, three complete plates were identified ( fig. 7.1), with a further 51 incomplete examples recorded ( figs 7.2–7.7).
Kiln plates: 7.1. Refitting perforated kiln plate [153] (197) Kiln 4; 7.2. Perforated kiln plate [153] (197) Kiln 4; 7.3. Perforated kiln plate [153] (197) Kiln 4; 7.4. Perforated kiln plate [153] (197) Kiln 4; 7.5. Perforated kiln plate [153] (197) Kiln 4; 7.6. Kiln plate [229] (246) Kiln 3; 7.7. Kiln plate apex [153] (188) Kiln 4. (© Pre-Construct Archaeology Ltd)
A visual examination of the kiln linings indicated that the same clay fabric was utilised both for construction of the kilns and in the manufacture of the vast majority of internal elements, such as the pedestals and floor plates. This observation is confirmed by the results of petrological analysis.Footnote 19 This main fabric comprised a basic sandy matrix with varying proportions of additional quartz, sparse angular flint and ferruginous pellets.
The clay plates were either straight or slightly tapered wedge-shaped forms; although roughly consistent in size, some variations were noted. The narrower ends, or apex ends, of the plates were rounded or squared, while the other end was bifurcated, with two sections separated by a ‘U’-shaped notch. A wide range of styles was observed in the finishing of these bifurcated ends; some examples were neatly finished with a central notch, while others were asymmetrical and show a much greater variety in size, shape and finish ( fig. 7). Given the fragmentary nature of much of this material, no attempt was made to establish a typology for these fragments.
The vast majority of the plates were perforated with one or two pre-firing holes in the centre of the plate c. 80–120 mm from the bifurcated end. These perforations were either rounded or elongated, measuring 20–30 mm in diameter or 30 by 60 mm, respectively. No correlation was observed between the overall form of the plate, bifurcate end and shape of the pre-firing holes. The plate fragments were rectangular in section and ranged in thickness from 20 to 35 mm, generally increasing in thickness and width towards the bifurcated end. The three complete examples recorded ( fig. 7.1) were in the range of 320–340 mm in length and 140 mm at their widest point and, although the complete width dimensions recorded for all fragments range between 100 and 170 mm, there was a clear concentration around 130–140 mm. In addition to the clay plates, four incomplete kiln bars with two tapering ends (Kilns 2 and 4 and Pit 5) and a single clay disc fragment (measuring 160 mm in diameter, Ditch 14) were also recorded.
Kilns 1 [189], 3 [229] and 6 [258] contained few examples and those from Kiln 2 [277] were very fragmentary and mostly central plate fragments. Although truncation may play a role, it seems probable that the fundamental differences in design between Kilns 3 and 6 (tongue-support), on the one hand, and the remaining four kilns (central pedestal), also equated to differences in the kiln furniture used. Kiln 4 [153] contained the largest assemblage of clay plates, with a minimum number of 28 plates recorded (103 fragments). There is always a possibility that the material retrieved from a structure is the result of discard from previous or subsequent firings and does not relate to the actual processes occurring in that specific kiln. The diversity of forms present, particularly in Kiln 4, could be due to reuse of clay plates from previous firings, in this or any of the other kilns. Both Kilns 4 and 5 displayed evidence of internal lips or ledges, which would have enabled the bifurcated ends of the clay plates to be held in place while the rounded ends rested on the central pedestal to create the necessary suspended floor. There is evidence in Roman Britain of kiln bars being pushed into clay linings while the linings were still wet in an attempt to fix the bars during firing, and of ledges for kiln bars,Footnote 20 in which case the internal groove would have provided a suitable support mechanism for the plates.
Within the known kiln groups in the locality (see fig. 1),Footnote 21 Duxford is unusual in this survival of a large group of kiln furniture. The best parallels for these plates are the perforated bars from Addenbrooke's,Footnote 22 although these comprise just 29 fragments compared to the 279 diagnostic pieces from Duxford. Nevertheless, the measurements given compare well with the material under discussion here. SwanFootnote 23 discusses similar perforated plates or bars, linked to ‘specialist’ production, and refers specifically to an illustration of an early Lower Nene Valley kiln produced by Artis in 1828.Footnote 24 She discusses ‘roughly-made perforated clay plates’, again from the Lower Nene Valley area, suggesting that they may be precursors to the second-century kilns of the colour-coated industry, possibly developing out of contact between continental and native potters. That said, Swan does state that, while the Lower Nene Valley kilns have strong continental links in terms of technology and typology, the perforated wedge-shaped tiles were unparalleled in Britain and on the Continent, thus questioning the validity of the interpretative illustration. The kilns at Duxford do, however, appear to closely resemble the Lower Nene Valley example, thus going some way to corroborating Artis' interpretation.
In most instances where perforated plates occur, they are recovered alongside kiln bars, as illustrated by Artis, and as found in Kiln IV at Elstow, Bedfordshire, where kiln bars and plates had been abandoned in situ.Footnote 25 By contrast, kiln bars are surprisingly scarce at Duxford, but recent experimental firing by HinesFootnote 26 has shown that perforated plates can function without the additional support of kiln bars when resting between the pedestal and the wall ledge. What is of particular significance here is the suggested continental link both inferred by Swan and suggested by the products being manufactured at Duxford.
THE KILN PRODUCTS (FIG. 8)
The complete pottery assemblage from Moorfield Road totalled 3,986 sherds, weighing 50,191 g, and representing 24.50 EVEs (estimated vessel equivalent) of which 2,216 sherds (28,051 g; 56 per cent) were identified as kiln products. Of this kiln group, 73 per cent (by count) derived from the kilns themselves, with the remaining 27 per cent originating from other features.
Perhaps the most interesting aspect of pottery production at Duxford is that it focused on the manufacture of a single vessel form — the flagon — which was produced in all six kilns. A minimum of 208 different vessels were identified, based on the number of rims, bases and handles present, although there were no complete or semi-complete vessels within the assemblage. Two basic types of flagon were being manufactured: collared flagons (also known as Hofheim-type) and ring-necked flagons. The former were seemingly the principal product, with Kilns 1–5 all appearing to produce these forms, of which two main variants were noted: those with a sharp undercut at the base of the rim ( figs 8.4, 8.5, 8.8 and 8.10), most closely paralleled with Camulodunum form (Cam.) 140BFootnote 27 (the Duxford vessels have a more developed undercut), and those without the undercut ( figs 8.2, 8.6 and 8.16), most similar to Cam. 140C.Footnote 28 At Colchester the Cam. 140s occurred most commonly during the Claudio-Neronian period,Footnote 29 though the slightly more developed nature of the Cam. 140B forms at Duxford may suggest production into the early Flavian period. However, both types were identified within Kiln 1, the earliest of the kilns, suggesting these variations did not represent chronological developments (Table 2).
QUANTIFICATION OF KILN PRODUCTS BY KILN
The ring-necked variants, comparable to the Cam. 154/155 forms,Footnote 30 appear limited to production within Kiln 6. At this point, it is worth noting that only one example of a ring-necked flagon was recovered from Kiln 6 itself, with the remaining vessels (nine MNV) coming from two associated dumps: Pit 5 (minimum of one vessel, fig. 8.15) and Ditch 5 (minimum of eight vessels, figs 8.12–8.14), both of which were interpreted as likely to have derived from Kiln 6 based on physical proximity and stratigraphy ( fig. 3). The manufacture of ring-necked flagons in Roman Britain began during the Claudio-Neronian period, thus overlapping with the production of collared varieties, before superseding them during the Flavian period. Although there are examples from other sites of both flagon forms being produced concurrently,Footnote 31 this was seemingly not the case at Duxford, with ring-necked varieties associated exclusively with the latest dating kiln, thus chronologically separate from the collared flagons and representing the final stages of pottery manufacture at the site. Further supporting evidence of this comes from the everted nature of most of the ring-necked forms at Duxford ( figs 8.12–8.14) which is more indicative of Cam. 155 vessels which have a slightly later currency than the Cam. 154 forms.Footnote 32
Kiln products: 8.1. Kiln 1 [189] (231) WW1 (see fabric descriptions below) H1 type handle; 8.2. Kiln 1 [189] (231) WW1 collared flagon rim, paralleled with Cam. 140C (rim diameter 8 cm); 8.3. Kiln 1 [189] (231) Q2 H1 type handle; 8.4. Kiln 4 [153] (193) Q5 collared flagon rim, paralleled with Cam. 140B (rim diameter 12 cm); 8.5. Kiln 4 [153] (193) Q2 collared flagon rim and H1 handle, paralleled with Cam. 140B (rim diameter 12 cm); 8.6. Kiln 4 [153] (193) Q5 collared flagon rim and H1 handle, paralleled with Cam. 140C (rim diameter 8 cm); 8.7. Kiln 4 [153] (193) Q2 H1 type handle; 8.8. Kiln 4 [153] (197) Q1 collared flagon rim and H1 handle, paralleled with Cam. 140B (rim diameter 10 cm); 8.9. Kiln 5 [152] (191) Q2 H1 type flagon handle; 8.10. Kiln 5 [152] (166) WW1 collared flagon rim with H1 handle, paralleled with Cam. 140B (rim diameter 12 cm); 8.11. Kiln 5 [152] (166) Q1 H2 type handle; 8.12. Ditch 5 [289] (284) WW1 ring-neck flagon (rim diameter 7 cm); 8.13. Ditch 5 [289] (285) Q2 ring-neck flagon (rim diameter 12 cm); 8.14. Ditch 5 [289] (285) WW1 ring-neck flagon (rim diameter 8 cm); 8.15. Pit 5 [259] (253) WW1 ring-neck flagon (rim diameter 8 cm); 8.16. Ditch 9 [399] (398) Q5 collared flagon rim, paralleled with Cam. 140C (rim diameter 10 cm); 8.17. Kiln 1 [189] (231) WW1 jar with tooled chevron decoration; probable kiln product. (© Pre-Construct Archaeology Ltd)
A large number of flagon handles were recovered, totalling 256 sherds (6,744 g), representing a minimum of 121 vessels, although it should be noted that while some of the flagons have single handles, there were examples of vessels with two handles, which would reduce the MNV count. Without any complete profiles, it is difficult to determine whether the Duxford flagons tended to have one or two handles. Within the flagon repertoire, two types of handle were identified: Handle H1 ( figs 8.1, 8.3, 8.5, 8.8 and 8.9), a rippled handle representing a minimum of 71 vessels, and H2 ( figs 8.4, 8.7 and 8.11), characterised by the thickening or beading of both edges of the handle while the interior remained flat. A minimum of 45 vessels with H2 handles were recorded ( figs 8.4 and 8.11). The H2 handles were almost exclusively produced in Kiln 5, and can therefore be seen as one of the later developments within the Duxford industry. It is of note that a single-handled pre-Flavian flagon in a coarse buff fabric was recovered from the adjacent site, which possibly reflects a kiln product.Footnote 33
In addition to the flagons, a small number of other vessel forms in kiln fabrics are suggested as probable kiln products, although these never formed a major component of the kiln repertoires. Eleven jars were identified as possible kiln products (104 sherds, 2,356 g), deriving from Kilns 1 and 5 as well as two ditches. The most commonly occurring were necked, beaded-rim variants and eight of the vessels were decorated with either rilling, tooled horizontal lines or, in one case, tooled chevrons ( fig. 8.17).
One jar had a slightly uneven rim and could be interpreted as a ‘second’. In fact, only two vessels in the entire assemblage could be considered wasters (the second sherd being a spalled sherd from Kiln 6), though of course all of the kiln sherds recovered could be considered wasters given their fragmented nature. Possible exploded vessels aside, the lack of other types of wasters may be a reflection of the skill of the potters. However, it is also possible that given the seasonal nature of pottery production, the broken products recovered from the kilns reflect deliberate discard occurring specifically at the end of a production ‘season’ — the potters not wanting or needing to take the vessels with them when production ceased.
Macroscopically, a total of six fabric groups were recognisable as kiln products (Q1–Q5 and WW1). The groups encompass fabric types with the same composition and range of inclusions (for fabric descriptions see Appendix 1), although there was slight variation between groups in terms of the sorting and occasionally coarseness and frequency of inclusions. Microscopically, the six fabrics could be further reduced to just two groups, with petrological analysis revealing that five of the fabrics (Q1–Q4 and WW1) comprised broadly the same clay matrix,Footnote 34 thus supporting the view that the same natural resources were being exploited over the period of pottery production at the site. The major differences within the fabrics are the frequency and coarseness of the sand tempers, although this is unlikely to result from conscious actions by the potters. Only Fabric Q5 stood out as being distinctively different in terms of composition, containing very fine, highly calcareous inclusions, thus indicating a different clay source.Footnote 35
There does not appear to be any chronological patterning to the fabrics, with all six fabrics present in at least two kilns (Table 3). Petrological analysis indicated that the clay was likely to have been procured in the local area, with the calcareous clays likely to have come from a primary marine sedimentary deposit, available locally within the Holywell Nodular Chalk Formation and the New Pit Chalk Formation found around Duxford. Several of the fabric samples also had added sand tempers, which in some cases contained glauconite.Footnote 36 Although originally deriving from Lower Greensand,Footnote 37 the rounded nature of the clasts suggests that many of the sand tempers may have originated from alluvial deposits rather than till, the nearby river Cam being a likely source.Footnote 38
QUANTIFICATION OF KILN PRODUCTS BY FABRIC GROUP
All of the kiln fabrics were oxidised and, with the exception of WW1, the fabrics occurred in a range of colours including buff and whitewares. There is no correlation between fabric group and vessel form. Even the ring-necked flagons associated exclusively with Kiln 6 were produced in Fabrics Q1, Q2, Q5 and WW1.
The remaining 44 per cent of the complete pottery assemblage derived from ditches and pits, totalling 1,949 sherds weighing 21,272 g, of which 30 per cent comprised probable kiln products. The largest single assemblage of material was from Ditch 5, totalling 409 sherds weighing 3,949 g, 85 per cent of which was identified as probable kiln products (349 sherds, 3,237 g), thought to be associated with Kiln 6, based on its physical proximity. This included 28 flagons, including handles and ring-necked variants, the latter occurring almost exclusively with this feature. Nearby Pit 5 contained 164 sherds weighing 1,799 g and representing a minimum of 23 vessels. Of these, 28 sherds (355g) were identified as kiln products, and included the only examples of ring-necked flagons outside of Ditch 5, which may further reinforce an association with Kiln 6.
Kiln products aside, the assemblage was dominated by Romano-British coarsewares, which represented 99 per cent of the non-kiln vessels. Unsourced coarsewares were the most commonly occurring fabrics, with finewares totalling just 23 sherds. The only evidence of activity after the late first century was from the late Roman trackway, which contained pottery of third- to fourth-century date, including two Nene Valley colour-coated beakers, one Hadham black burnished straight-sided dish, one Nene Valley whiteware mortarium sherd and one sherd from an Oxfordshire red-slipped vessel. The only other sourced wares identified comprised a single Hadham black burnished ware sherd and a Verulamium whiteware sherd. Six imported wares were identified (85 g): one Baetican amphora sherd, one terra rubra sherd and four South Gaulish samian sherds. The composition of fabrics suggests that occupants of the site had limited access to and/or limited need for vessels from outside the local area in the early Roman period.
Unsurprisingly, flagons dominated the assemblage, representing 51 per cent by MNV, although by EVEs count this figure was reduced to 38 per cent. This discrepancy can be explained by the presence of a large number of flagon handles, which are easily identifiable, yet not included in the EVEs count. Jars were moderately represented, totalling 34 per cent (MNV 119). The remaining vessel forms were poorly represented, with five or fewer examples of beakers, bowls, cups, dishes, platters and lids. The lack of a diverse range of vessel forms is not surprising given that the site's primary function was pottery production. However, the composition of the remainder of the assemblage is indicative of some domestic activity taking place in the vicinity of the site.
DISCUSSION
The discovery of six early Roman kilns at Duxford adds to a growing corpus of pre-Flavian kilns in the Cambridge area, at Cherry Hinton, Greenhouse Farm, Addenbrooke's and SwaveseyFootnote 39 ( fig. 1), which have been excavated and published over the last 25 years. In terms of kiln morphology, the Duxford examples can be regarded as fairly typical of the period and region, four of the kilns having central pedestals while the remaining two contained tongue-supports.
Although as a group they display less variability in overall architecture than those on other contemporary sites in Cambridgeshire,Footnote 40 there were contrasts in the details of each individual kiln, be it in the design of the pedestal or the evidence of reuse, namely re-linings and repairs. Such idiosyncrasies, however, are proving to be a feature of early kilns. These possibly reflect the choices and preferences of individual potters, technological developments, or perhaps the need to adapt kiln design in relation to the specific products, and perhaps even the individual load, being fired. That said, the perforated kiln plates are a more unusual aspect of kiln technology for this region, seemingly reflecting continental influence, with the only published parallels being that excavated by Artis in the Lower Nene Valley in the early nineteenth century and Kiln IV from Elstow, Bedfordshire.Footnote 41
Also typical is the wider setting of the kilns at Duxford, in terms of their peripheral position with regards to settlement and the haphazard arrangement of the kilns on the site. At a very basic level, the availability and access to the raw materials required to produce and fire pots was a prerequisite for all kiln sites and the immediate landscape around the site clearly provided these resources, including ready access to water via the river Cam and its tributary stream. In addition to natural resources, the location of the site is likely to have been influenced by the proximity of the Icknield Way, thought to have been aligned approximately along the route of the A505, c. 200 m north of the site.Footnote 42 Additionally, evidence for two minor Roman roads has been identified in Sawston, 2.5 km north-east of the site.Footnote 43 These were aligned broadly north to south and thus are likely to have extended south to the Icknield Way, if not beyond, possibly linking with Great Chesterford. The kilns would therefore have had a roadside position next to a crossroads and along a major route way, which would have been vital for transporting the vessels to consumers.
Access to a market would also have been critical and the likely location and nature of the market for the Duxford products is discussed below. However, the economics of these determining factors had to be balanced against the concerns of the wider community and it appears that the favoured location for kiln sites in Cambridgeshire was on the periphery of settlements and, notably, some distance away from the emerging urban centres at Great Chesterford, to the south, and Castle Hill, Cambridge, to the north. In fact, all five of the identified kiln sites in the Cambridge environs were located on what appears to be the edge of settlements, utilising existing enclosures and boundary systems. Of course, there are likely to be pragmatic reasons for this, such as avoiding the potential risk of fire breaking out within settlements, or the need for space in which to work. Although open fires would have been the norm across all settlements, those associated with industrial processes might perhaps have been perceived as being more dangerous and their associated emissions and residues more polluting.
These points aside, beyond a general sense of the site providing a focus for kilns and pottery production, it is difficult to identify any spatial pattern in the arrangement of the structures, which were scattered and shared no consistent orientation. Nor is there any obvious sense of a chronological sequence to their construction and use. All that can be deduced on this issue is that the kilns were broadly contemporary, with Kiln 6 possibly reflecting the latest phase of pottery manufacture based on the flagon form being produced. That being said, the lack of coherency in their arrangement makes it unlikely that they were a planned group used simultaneously, as a greater degree of order and regularity would be expected. It is, though, possible that a number were in use at the same time, which of course begs the question of what the temporal relationship was between the use and abandonment of different kilns. Admittedly, the evidence to hand affords only a partial understanding of this topic, but it does have clear implications for how we think about the organisation and duration of production at Duxford.
Crucial in this respect is the identification of dumps of pottery and kiln waste in different fills within the ditches, indicating punctuated episodes of firing and clearing-out over the period in which these boundaries were infilling. This, coupled with the small number of extant stratigraphic relationships between kilns and ditches, points towards the site being used for production on an episodic basis during the mid- to later first century. Yet, the period of production need not have been more than a decade or so, as the area was susceptible to flooding and boundary infilling appears to have been rapid, requiring regular renewal and reordering of ditches as ground conditions changed. Indeed, the unstable and shifting ground conditions in this location may account for the haphazard arrangement of the kilns — the area suitable for construction is likely to have changed year on year. There is certainly evidence that once abandoned, some of the kilns were inundated. It is also probable that further kilns were located outside of the excavation area, an observation particularly suggested by a dump of kiln-type pottery in Ditch 6, some distance away from any of the identified kilns.
While it remains difficult to be exact about the timeframes involved, it seems likely that production was seasonal, of limited scale and duration, and probably saw the use of only one or two kilns at any one time. This interpretation is consistent with that reached for other major pre-Flavian kiln sites in Cambridgeshire,Footnote 44 and given the strong similarities in the evidence, invites the parallel conclusion that production was possibly in the hands of specialist itinerant potters serving a local market.
Unique to the Duxford kilns, however, is the fact that output was focused almost exclusively on collared (Cam. 140 derivative) flagons, with limited production of ring-necked types, whereas all the other regional production sites investigated to date produced a variety of vessel forms. In short, in Cambridgeshire, it is currently the only pre-Flavian multi-kiln site known to have made a single product. This was undoubtedly specialist production, and given the suggestion of continental links in the character of the kiln furniture and products and reinforced by the early post-Conquest date of production, it seems highly likely that the potters themselves had a continental background. Likewise, the market being served from Duxford was almost certainly different to those being provided for by production at the other published sites in the area. The latter principally catered for a local market, with a focus on domestic consumption; the range of vessel types being made occurs in most contemporary domestic assemblages from excavated rural settlements.Footnote 45
In stark contrast, flagons are scarcely represented in most contemporary domestic assemblages from the area. Typically, percentage representation falls below 1 per cent, and in many cases the forms are entirely absent. A number of large-scale, early Roman rural settlement sites have been investigated around Cambridge in recent years (e.g. Addenbrooke's, Clay Farm, North-West Cambridge, BabrahamFootnote 46 ) which have produced large contemporary pottery assemblages demonstrating that the low instance of flagons seems to constitute a genuine trend on ‘typical’ rural sites in the region. The question is, then, where was the market for these vessels?
To answer this we must consider the nature of the products themselves. Flagons were, first and foremost, specialised utensils for carrying and serving beverages, principally wine. Like native pedestal jars and tazzas, or imported forms from the Roman world such as amphorae and fineware cups, flagons were not regular components of the domestic repertoire of early post-Conquest communities in Cambridgeshire. Rather, their use appears to have been limited to particular social arenas or social groups (elites?) and rarely filtered into the rural domestic sphere. Around Cambridge, flagons are perhaps best represented in funerary contexts, be it isolated gravesFootnote 47 or formal cemeteries, from small groups, such as North-West Cambridge, where 13 of the 64 grave goods comprised flagons,Footnote 48 to larger burial grounds such as Guilden Morden.Footnote 49 The shrine on Castle Hill, Cambridge, provides a further example, which although slightly later in date (second to third century), was noted as having a significantly high proportion of flagons, with a minimum of 256 vessels identified.Footnote 50 However, while some vessels may have made it into local cemetery sites,Footnote 51 it seems unlikely that the products manufactured at Duxford were aimed specifically at the ‘funerary market’. With this in mind, production at Duxford is significant and suggests that these products were destined for a very specific, specialist market, unlikely to have been open to all members of the community.
This market was clearly stable in the region. Had the evidence implied a single season of production, then this would present a picture of a ‘trial’ run of flagon manufacture that proved unsuccessful. That production at the site appears to have been repeated over several years, albeit seasonally, suggests that the market for the flagons was persistent enough for repeated production. Ultimately we can only speculate about the nature of this market, but given the scarcity of these wares on rural sites, it is perhaps more likely that the appetite for the consumption of some of these products was focused elsewhere and, to a lesser extent, in the funerary market and/or the emerging Romanised centres of the region. However, while these are likely destinations for some of the ring-necked variants, collared flagons, seemingly the primary product at Duxford, have a much more specific distribution, which to date has been almost exclusively linked to the Roman army and military sites, with very few examples from either domestic or funerary contexts.Footnote 52
A ready market for the material culture of the Roman world was undoubtedly created by the presence and stationing of the Roman army and the possibility that the Duxford kilns served the needs of local garrisons cannot be ruled out, particularly since the emerging local centres at Great ChesterfordFootnote 53 and Castle Hill, Cambridge,Footnote 54 had a military basis.Footnote 55 It may well be that the potters themselves travelled with the merchants and traders that followed in the wake of the Roman invasion. To date, Longthorpe and Morley St Peter are the only kiln sites in the region with definite military connections.Footnote 56 While there are parallels with Duxford in that both of these sites produced collared and ring-necked flagons, these were just two components of a varied repertoire of forms, rather than manufacture being limited to a single vessel type.
What is evident is that ultimately pottery production came to a halt in the later first century and, with the exception of a late Roman trackway, there was no evidence for further activity at Moorfield Road. The reasons for the decline may be complex. However, it is tempting to suggest that, with the military identified as the main consumer of the Duxford wares, the end of production was connected with the movements of the army.Footnote 57 While this would be a plausible explanation, it is relevant to note that the abrupt end to pottery production is mirrored at Green House Farm, Cherry Hinton, Addenbrooke's and Swavesey,Footnote 58 suggesting that wider changes were afoot. The exact reasons are, as yet, unclear. However, the decline of these five small-scale production sites in the Cambridge environs may well be linked to the growth of larger regional pottery industries, the closest and most relevant of which is the Horningsea pottery industry.Footnote 59
APPENDIX 1: FABRIC DESCRIPTIONS (MICROSCOPIC AND MACROSCOPIC) with Dr Patrick QuinnFootnote 60
(Q1) Macroscopic: moderately fine sand with common to frequent very small sub-rounded quartz, occasional to moderate poorly sorted rounded limestone (up to 0.5 mm) and occasional rounded quartz up to 0.2 mm.
Microscopic: fine to medium sand-sized quartz and limestone inclusions in a fine clay matrix. The sand-sized inclusions appear to have been added as temper given their fairly well-sorted grain size and the relative absence of silt-sized inclusions. The sand material is dominated by sub-angular to well-rounded, equant and elongate monocrystalline quartz and micritic limestone inclusions in roughly equal proportions. Other less common inclusions include polycrystalline quartz, chert, siltstone, probable oxidised glauconite, shell and opaques. Sparse fine angular silt-sized quartz and mica are present in the base clay, as well as rounded opaques.
(Q2) Macroscopic: as Q1 but with sparse to occasional silver mica and rare black iron ore(?) up to 2 mm.
Microscopic: petrographically closely related to those in microscopic Fabric Q1, in that they contain sand-sized quartz and limestone temper in a fine clay matrix.
(Q3) Macroscopic: as Q1 but with degraded or leached-out limestone and significant secondary calcite.
Microscopic: the same general petrographic composition as the microscopic Fabrics Q1 and Q2.
(Q4) Macroscopic: as Q1 but with occasional calcite voids and rare to occasional rounded black iron ore up to 0.5 mm.
Microscopic: this does not significantly deviate from Q1. It has partially degraded limestone inclusions, sparse shell, a vitrified clay matrix and significant secondary calcite in voids and patches.
(Q5) Macroscopic: fine to medium quartz sand, with rare limestone in a highly calcareous clay matrix.
Microscopic: different from all other samples. Characterised in thin section by the presence of sparse sub-rounded to well-rounded, fine-medium sand-sized inclusions of quartz, polycrystalline quartz, chert, weathered feldspar, rare limestone and glauconite in a fine, highly calcareous clay matrix with foraminifera microfossils. The large rounded inclusions are highly likely to have been added as temper in the form of loose, polymict sand.
(WW1) Macroscopic: fine sandy whiteware with well-sorted very small rounded quartz and common small sub-rounded red inclusions (<0.1 mm). Occasional small rounded black (iron ore?) and rare chalk.
Microscopic: petrographically similar to the samples classified within microscopic Fabrics Q1, Q2, Q3 and Q4. It contains quartz and limestone sand inclusions within a well-fired fine clay matrix with secondary calcite. The sand temper inclusions are less abundant and finer than in some of the above samples.
ACKNOWLEDGEMENTS
The fieldwork and post-excavation analysis were generously funded by Welch's Group Holdings Limited and Wrenbridge (Moorfield Road) Ltd, and PCA would like to extend particular thanks to John McHale (Wrenbridge) and Jim Welch for their support. The project was managed for PCA by Mark Hinman and the fieldwork directed by Tom Woolhouse and Dr Daryl Stump. Andy Thomas of Cambridgeshire County Council Historic Environment Team monitored the work on behalf of the local planning authority. The authors would like to thank the site team: Dave Curry, Karl Hanson, Jan Janulewicz, Tom Learmonth, Matt Lees, Lawrence Morgan-Shelbourne, Sîan O'Neill, Mary-Anne Slater and Aileen Tierney, for their hard work in an exceptionally hot summer. Figures accompanying this article were prepared by Mark Roughley and Hayley Baxter of PCA's CAD Department; illustrations were drawn by Cate Davies. In addition to those specialists who have contributed directly to this article, the following are thanked for their analyses of the finds and environmental remains from the site: Dr Barry Bishop (lithics), Dr Matthew Brudenell (prehistoric pottery), Nina Crummy (small finds and metalwork), Kevin Rielly (animal bone), Val Fryer (plant macrofossils) and Dr Patrick Quinn (petrological analysis). Finally thanks to Victoria Ridgeway, Mark Hinman and Matt Brudenell for comments on the paper and with assistance in editing.
