PREVIOUS   NEXT   CONTENTS   ISSUE   HOME 

Pottery production at Heybridge by Edward Biddulph

Cite this as: Biddulph, E. 2015, Pottery production at Heybridge, in M. Atkinson and S.J. Preston Heybridge: A Late Iron Age and Roman Settlement, Excavations at Elms Farm 1993-5, Internet Archaeology 40. http://dx.doi.org/10.11141/ia.40.1.biddulph

Pottery production at Heybridge is attested by the presence of four pottery kilns. The largest volume of pottery was recovered from the paired-kiln complex in the hinterland zone of Area W. Two kilns were uncovered in the southern zone: one in Area L and the other in Area N. The latter yielded the least evidence, consisting of undiagnostic spalled sherds only. Consideration here is given to the pottery recovered from the kilns, incorporating the identification and quantification of products, dating of kiln use, and an overview of pottery production at Heybridge and its regional setting. The identified vessels have been ordered into classes following the Chelmsford typology (Going 1987, 13-54).

Descriptions of the kiln product fabrics are based on the scheme outlined by Peacock (1977c, 26-33). Munsell colour references are additionally employed. The pottery from Area W stoke-hole 1589 is fully quantified and published as a key pottery group (KPG27), including non-kiln products, and presented in the Pottery sequence section. Selected sherds from the Area W and Area L kilns were thin-sectioned at the Department of Archaeology, University of Southampton (D.F. Williams; report in archive). It was noted that the coarse ware sherds were in a reduced sandy fabric with visible quartz grains together with some small pieces of flint. A mortarium from kiln 1618 had a similar, but light-coloured, fabric, with fewer flints. Dr Williams concludes that since Heybridge is situated in an area of London Clay, covered in part by boulder clays and brickearth, the clay used in the pottery seems to reflect a certain intermingling of sources.

At Crescent Road, Heybridge, Wickenden identified an assemblage of some 4500 sherds as 'the discarded products from a nearby kiln' (1986, 46), dating their manufacture to the mid-3rd century. The material has been re-examined and the conclusion is that since the pottery was small and abraded rather than spalled or overfired, it could not be positively identified as waste products from a nearby kiln. The evidence concerning mid-3rd century production here must remain inconclusive. However, the fabric of much of this pottery was similar to that of Elms Farm kiln products, particularly from Area N, suggesting, at least, an identical clay source and, therefore, local production.

Kilns 1223 and 1618, Area W

Kiln 1223 (Group 908) yielded the least amount of pottery, 3500g, and kiln 1618 (Group 906) contained the most, totalling 27,740g. Their shared stoke-hole, 1589 (Group 910), contained just under 12,000g. Buff ware mortaria and grey ware, including many spalled and overfired sherds that have been identified as locally produced kiln products, formed the largest components of the three pottery assemblages. By weight, 69% of the total volume of grey ware was recovered from kiln 1618, as compared to 28% in the stoke-hole and a meagre 3% in kiln 1223. Buff ware mortaria were present in similar proportions; 65% was recovered from 1618, 24% from 1589, and 11% from 1223. Kiln 1618 thus produced the most extensive range of products and best dating evidence.

A somewhat limited range of forms was evidently produced, predominantly jars, including ledge-rimmed types and high-shouldered varieties with undercut rims. Folded beakers and bead-rimmed dishes were also manufactured. Mortaria sherds, comprising hammerhead-rimmed and bead-and-flanged types, were built into the structure of kiln 1618, perhaps inserted as repair or to aid firing. These, too, could originally have been fired in this kiln. In contrast to 1618, the paucity of vessels from kiln 1223 provides little clue as to what was fired in it.

The minimum vessel number was calculated for the kiln products in sandy grey ware and buff ware only, which were visually distinctive. Fine grey ware was invariably abraded and products in this fabric could not be separated easily from any non-kiln fine grey wares. An accurate count of certain fine grey ware kiln vessels, mainly B2 or B4 dishes, is likely to be higher than the six vessels actually counted.

Fabrics

Sandy grey ware (GRS)

This gritty fabric has dark grey brown surfaces (2.5Y 4/2), in which frequent small white and grey quartz grains, less than 1mm in size, and occasional flint pieces up to 5mm long, are visible. The core is darker grey (2.5Y 4/0). Frequent white, grey and black inclusions are well sorted, though larger angular clear quartz grains are also present. Totals are 488g from 1223, 10247g from 1618, and 4158g from 1589.
Forms: Dish B2/B4, jars G5.5 G25, beaker H35

Fine grey ware (GRF)

Surface and core colours are as for GRS; however, inclusions are generally finer and more frequent, and when seen microscopically, have an appearance reminiscent of Hadham grey ware (see Fabrics) No flint or quartz grains are visible on the surfaces. Examples are in poor condition and invariably abraded and powdery.
Forms: Dish B2/B4, beakers H21 H35

Buff ware mortaria (BUFM)

This fabric, which is soft and invariably powdery, varies from light green- or brown-grey (5Y 7/2, 10YR 7/2) to yellow-brown (10YR 7/6). Inclusions of fine sand, and occasional mica and iron-rich particles, pink clay pellets and flint; trituration grits of angular white, grey and black flint, 2-5mm in diameter, and sparse white quartz. Traces of a dark yellow-brown (10YR 4/6) slip are visible on the external and internal surfaces of some examples. Totals are 1611g from 1223, 10310g from 1618, and 3860g from 1589.
Forms: Mortaria D3 D11 D13

Forms

Dishes

B2/B4

Bead-rimmed dish with tapering sides and flat base. Most examples were recovered from kiln 1618, and mainly comprised rim sherds that could not be positively identified either as the shallow B2 or the deep B4 type. Principally a fine grey ware product.

Mortaria

D3

Mortarium with incurving, droopy flange and high, delineated bead. Examples vary in height of the bead and diameter, which ranges from 280 to 360mm. The form, with its rather upright flange, is akin to the D11 hammerhead-rimmed mortarium. It differs in this respect from D3 mortaria at Chelmsford, whose flanges are set more horizontally. No precise parallel, in fact, has been found at Chelmsford or Colchester. Never common, the type was nevertheless recovered in similar quantities from kiln 1618 and stoke-hole 1589. None was found in kiln 1223.

D11

Hammerhead-rimmed mortarium. Examples display considerable variation in size and shape, but are generally characterised by upright rims with flat tops. Some have curving and hooked flanges, closely resembling Chelmsford form D11.1. The smaller D11.2 is also represented. Some examples have wider 'bevels' at the top of the rim, while others are particularly large and robust, with no curve to the flange. The form was undoubtedly more common than the D3 mortarium. Kiln 1618 and stoke-hole 1589 yielded the most examples, while 1223 yielded the least.

D13

Wall-sided mortarium with small spout and grooves at the top and bottom of the flange. Just one example was recovered, from pit 1621 cut into the stoke-hole 1589. Its fabric is a little coarser than is usual for buff ware mortaria, either from these kilns or Colchester. However, it is deemed a local product on the basis of its affinities with an example recovered during the Crescent Road excavations. Both are hard and gritty, and off-white to grey in colour. The latter example was thought to be a non-Colchester product, probably dating to the early 3rd century AD (Wickenden 1986, 40, fig. 21.137).

Jars

G5

Ledge-rimmed, high-shouldered and neckless jar resembling Chelmsford forms G5.5 and G5.6. The G5.5 was made in two sizes - large (diameter of 140-180mm) and small (diameter of c. 100mm). The large G5.5 is notable for its consistent design; the ledge rim is created by a thin groove, rather than a gentle cupping, characteristic of the G5.5 at Chelmsford. No lids were found in either the kilns or stoke-hole, and it seems unlikely that ceramic lids were made to fit the form. The G5 type was made at Mucking (Rodwell 1973, 22), Grays (Rodwell 1983, 27), and Orsett (Rodwell 1974, 27), all in South Essex.

G24

Oval bodied jar with everted, slightly undercut, rim. A thin cordon appears on the short neck. Examples from these kilns are few in number, but the form was otherwise common at Heybridge.

G25

Jar with undercut rim, short neck and grooved shoulders. The type produced at Elms Farm resembles Cam 268B, a very prolific form at Colchester (Bidwell and Croom 1999, 479), and corresponds closely with Chelmsford form G25.1. Its size varies; diameters range from 140 to 180mm.

Beakers

H21

Cornice-rimmed, bag-shaped beaker, with band of roller-stamped decoration on body. Its presence is restricted to a single example. This is abraded and burnt, but traces of an external slip survive. Roller-stamped body sherds were present in numerous contexts, but could not be confidently identified as H21 beakers. The form was also produced at Ivy Chimneys, Witham (Turner-Walker and Wallace 1999, fig. 114.25)

H35

Large folded beaker with short neck, bead rim, and roller-stamped decoration on body. Examples are burnt and abraded, and only tentatively identified as local products.

H (unclassified)

A single example of a small ledge-rimmed vessel, with a short neck, was recovered from stoke-hole 1589. This was identified as a local product on the basis of fabric. The form is unparalleled, although clearly resembles ledge-rimmed jars. Diameter 92mm.

Flagons

Cam 370

Large double-handled flagon with heavily reeded rim. The single example present was set into the pedestal of kiln 1618. The rim of the flagon is warped, and its fabric identical to that of other grey ware products. The flagon may well be a waster product from a previous firing within this kiln, placed in the pedestal as repair. Alternatively, it might be from another and earlier, as yet undiscovered, kiln in the vicinity. The form is absent at Chelmsford, but present at Colchester, albeit poorly represented in a coarse oxidised ware only.

Miscellaneous

S (unclassified)

Body sherds from two examples of a curiously shaped vessel were recovered. The sherds are thick-walled, cylindrical and decorated with a band of rouletting. As no rim or base sherds were found, the form or function is impossible to discern. Its cylindrical shape and decoration suggest a beaker, although it is remarkably robust. It may have been placed in a kiln as a support or bar, but it is wheel-made and shows no signs of repeated firings. It may have functioned simply as a test-piece prior to the firing of a whole load. Its fabric, too, is typical of kiln products.

Quantification

As quantified by EVE, ledge-rimmed jars were clearly the most prolific form, and very likely candidates for firing in kiln 1618. Beakers are not well represented here and, like the Cam 370 flagon, are possibly not products of these kilns. Even assuming that all of these forms were produced here, the range is very narrow. The pottery kiln at Ivy Chimneys, Witham, yielded 9.9kg of kiln products, with seventeen separate vessel forms represented (Turner-Walker and Wallace 1999, 170). From kilns 1223 and 1618, and stoke-hole 1589, sandy grey ware alone weighed almost 15kg, yet only a maximum of seven different vessel types were produced in this fabric. The difference between volume of kiln products and range of forms is more marked at Palmer's School in Grays, where the kiln yielded over 46kg of kiln pottery and six vessel types (Rodwell 1983, 26). This situation is perhaps unsurprising. Potters responded to demand, rather than risk producing pottery that no one wanted. At Heybridge, potters produced ledge-rimmed jars simply because there was a market for them. Doubtless the shape facilitated mass production; a cut groove on top of the rim and absence of decoration meant that the form could be turned out rapidly. A high diversity of forms, then, is not necessarily indicative of a high level of output.

The amounts of grey ware recovered from both Area W kilns are perhaps lower than is typical. A kiln at Ivy Chimneys yielded 13.5 EVE of kiln products (Turner-Walker and Wallace 1999, 170). Over 40 EVE were recovered from the late Roman kiln at Inworth, and the Moulsham Street kiln, Chelmsford, produced 28.5 EVE (Going 1987, 74-84). These are similar in size to both kiln 1223 and 1618, yet the Elms Farm kilns yielded a little under 10 EVE. Again, that the size of the assemblage from the kilns is small may have less to do with the volume of actual production and more to do with volume of waste generated and the manner of disposal of such waste. The assemblage found within the kilns may represent the final dumping of pottery waste, a store of seconds or the last batch of unsold pottery, putting the kilns firmly out of use. Waste produced at each previous firing may well have been discarded away from the immediate environment of the kilns themselves, although no waster dumps were discerned in their wider vicinity.

Like grey ware, the range of buff ware mortarium forms is restricted, although there are subtle variations of shape. Hammerhead-rimmed D11 mortaria clearly predominate in this fabric and, in terms of vessel count, are more numerous than ledge-rimmed jars. Most examples of both mortarium types were built into the structure of kiln 1618. These mortaria were almost certainly locally produced (see Mortaria stamps). Like kiln 1618, pottery was inserted into the original structure of a kiln at Ellingham in Norfolk (Hartley and Gurney 1997). Both kilns show no obvious signs of repair. Hartley (1973, 143) suggests that this technique reduced the risk of kiln fracture upon initial firing. While the mortarium kiln at Ellingham, Norfolk, is 1.8m in diameter (Gurney and Rogerson 1997, 2), and that its capacity perhaps suited the wide design of its products, mortaria did not necessarily require large kilns. Mortaria were among the chief products from Colchester, yet were fired in kiln chambers sometimes only 1m across (Hull 1963, 158, 168). A high dome, rather than a wide chamber, facilitated the high temperatures required to fire the form. As both kiln 1223 and 1618 were 1.4m across, it remains feasible that mortaria could have been fired in either.

Wherever mortaria were fired, the high number of vessels in stoke-hole 1589 suggests that the kiln or kilns that produced them continued to do so after some of their products were used to construct kiln 1618. The number of vessels in 1223 is comparatively low. If mortaria were fired in this kiln, the chamber was infilled soon after it fell out of use, providing little opportunity for waste material to accumulate.

Dating

Kiln 1223 was probably built during the late 2nd century. Kiln 1618 was constructed in the late 2nd or early 3rd century AD and abandoned during the early 3rd century AD. The grey ware products in 1618 are not by themselves closely datable; ledge-rimmed jars date from the 2nd century at Chelmsford (Going 1987, 23), and production of the form up to the mid-3rd century AD is attested at Witham (Turner-Walker and Wallace 1999, 170). The forms G24 and G25 have wider date ranges, continuing into the 4th century AD at Colchester (Bidwell and Croom 1999, 479). Significantly, the G5 jar was the commonest jar form at Heybridge during Ceramic Phases 7-8 (AD 170-260).

The date of construction is best provided by the mortaria built into the structure of kiln 1618. Generally, hammerhead-rimmed D11 mortaria date to the second half of the 2nd century AD at Chelmsford and Colchester, continuing at both sites into the 3rd century AD (Going 1987, 21). K. Hartley places the production of the stamped mortaria from the kilns to the early part of the date range, AD 170-190 (see Mortaria stamps). However, construction of 1618 could be later, since unstamped mortaria were manufactured beyond this date. Early 3rd-century use is supported by the presence of folded H35 beakers and absence from the kilns of any pottery dating exclusively after this time.

The archaeomagnetic date range of AD 90-210 for the last firing of 1618 is too wide to be of any real value, but it nevertheless fits with early 3rd-century abandonment. In the absence of strong pottery dating, the archaeomagnetic date of AD 140-170 for 1223 is of greater use. Taken together, the dating evidence suggests that 1223 was abandoned before 1618, which may have been functioning into the 3rd century. Kiln 1223 may even have been built first, with 1618 being its replacement. However, 1223 remained open after disuse and continued to accumulate material until both kilns were effectively sealed. Both yielded very small amounts of non-kiln products, largely restricted to abraded body sherds, and including Nene Valley colour-coated and Hadham oxidised wares. Final closure of the entire complex is therefore likely to have occurred during the first half of the 3rd century.

Kiln 14858, Area L

Kiln 14858 (Group 714) yielded 10,654g of pottery. Eighty-six per cent of this was spalled and overfired and has been identified as waste products. This waste material was present in three fabrics; black-surfaced ware, sandy grey ware and fine grey ware, of which sandy grey ware formed the largest proportion. Large quantities of waste pottery were also recovered from three nearby pits; 14655, 14744 and 14809. Since fabric, form and date of the pottery in these features were identical to waste pottery recovered from the kiln, it has been assumed that the source of all this material is the same. By weight, pit 14655 yielded over half of the total amount of kiln waste. Pit 14809 yielded the least, just 4% of the total.

A limited range of forms was produced. Oval-bodied jars with undercut rims and plain-rimmed dishes predominated, but bead-and-flanged dishes, wide-mouthed bowl-jars, and bifid-rimmed jars were also manufactured. Notably, only bowl-jars and dishes were made in fine grey ware. The coarser black-surfaced and sandy grey ware fabrics were reserved for jars.

Fabrics

Black-surfaced ware (BSW)

This moderately hard fabric has very dark grey to black (10YR 3/1) surfaces, red-brown (5YR 4/4) margins, and light to dark grey (10YR 6/1 to 4/1) core. Moderate white and red crushed flints, less than 3mm long, protrude through the surfaces. Inclusions of frequent tiny quartz grains; larger and sparser angular white and clear grains are visible macroscopically. Totals are 3030g from kiln 14858, 7226g from pit 14655, 1060g from pit 14744, and 462g from pit 14809.
Forms: Dishes B1 B6, bowl-jar E5, jar G24

Sandy grey ware (GRS)

The fabric has dark grey (10YR 4/1) surfaces, with some lighter grey (10YR 4/2) patches. The core is usually red-brown (5YR 4/2); some examples are grey throughout. Inclusions are as for BSW, except that surfaces and core are flintier. Totals are 5720g from kiln 14858, 7005g from pit 14655, 2225g from pit 14744, and 758g from pit 14809.
Forms: Dishes B1 B6, bowl-jar E5, jars G24 G42 G (bifid rim)

Fine grey ware (GRF)

Smooth surfaces, sometimes burnished. Surface colour as GRS; some very dark grey, almost black surface patches, presumably the result of firing. Light grey (10YR 6/1 to 5/1) core. Inclusions as for GRS. Totals are 428g from kiln 14858, 905g from pit 14655, 100g from pit 14744, and 18g from pit 14809.
Forms: Dishes B1 B6, bowl-jar E5

Forms

Dishes

B1

Plain-rimmed dish with flat base. Rim is slightly bulbous and inturned. Burnishing in narrow bands on external and internal surfaces. Diameters range from 120 to 240mm. Gillam (1976, 76) suggested that this type was possibly used as a lid in conjunction with the flanged B6, forming a set akin to a casserole. The frequent appearance of both forms within kiln 14858 and associated pits, and the fact that diameters of both forms seemingly correlate lends some credence to this suggestion. Indeed, some of the more complete B1 dishes fitted the B6 dishes rather well.

B6

Bead-and-flanged dish with tapered, slightly convex sides and flat base, corresponding closely to Chelmsford form B6.2. There is variation in the shape of the flange, which is usually downturned, and can be fat and short or thin and extended. The flange is less frequently upturned. In many instances, the flange had broken off, probably during firing. Burnished bands decorate most vessels, never completely covering the external and internal surfaces. Diameter range 180 to 240mm.

Bowl-jars

E5

Round-bodied, wide-mouthed vessel with bead rim. Products correspond to Chelmsford forms E5.3 and E5.4. The former has a typical diameter of 140mm; the latter is larger with a diameter range of 200 to 240mm. Decoration is restricted to shoulder grooves and a single burnished wavy line, though this is by no means present on all vessels. The form was commonly produced elsewhere, including Moulsham Street, Chelmsford (Going 1987, fig. 35.7-9), Ivy Chimneys, Witham (Turner-Walker and Wallace 1999, fig. 113.8-9), and Mucking (Rodwell 1973, type K). This vessel type was also present in the Crescent Road kiln 'waste assemblage' (Wickenden 1986, fig. 17.39-40).

Jars

G24

Jar with oval body, undercut bead-rim, and slightly stepped shoulders. Diameter range 140-260mm. The product closely resembles Chelmsford form G24.2. There are no decorative details, but products are typically coarser and flintier than dishes or bowl-jars, and no example appears in fine grey ware. The form was produced at Ivy Chimneys, Witham (Turner-Walker and Wallace 1999, fig. 113.16), Moulsham Street, Chelmsford (Going 1987, fig. 35.10), Orsett (Rodwell 1974, fig. 8.61), and Mucking (Jones and Rodwell 1973, type J), among other sites.

G42

Short-necked 'storage jar' with wide, rounded body and thick bead rim. Examples have 'wheat-ear' stabbed shoulder decoration, and diameters of 200 and 240mm. Exact parallels are lacking, but similar vessels were produced at, for example, Mucking (Rodwell 1973, type S), Moulsham Street, Chelmsford (Going 1987, fig. 35.18) and Inworth (Going 1987, fig. 41.23).

G (bifid rim)

Jar with bifid rim and short neck. Though non-joining, the two rim sherds from separate pits are likely to form part of a single vessel. Despite this very low incidence, local production is very likely. Its fabric is identical to sandy grey ware recovered from kiln 14858, the rim sherds are spalled, and there are no close parallels.

Quantification

Measured by EVE, dishes were the most prolific class, followed by jars and then bowl-jars. The order of vessel frequency is retained when the minimum vessel number is calculated. It can be seen that, of the dishes, the B1 type is slightly more numerous than the B6, and the G24 jar far outnumbers other jar forms. Notably, G24 jars were the most frequently produced form at the Moulsham Street kilns, Chelmsford (Going 1987, 74), followed by B1 and then B6 dishes. As at Orsett (Cheer 1998, 99) and Chelmsford, B1-type dishes are better represented than B6-type dishes, although the difference at Heybridge is perhaps negligible. There is no obvious reason for this difference, except to suggest that plain-rimmed dishes were simply in greater demand than bead-and-flanged dishes. It might also be suggested that B6 dishes were always made to pair B1 dishes, but additional B1 dishes were produced for lone use. Bowl-jars never rivalled jars and dishes in terms of the quantity of production. This appears to be the case, not only here, but also at Chelmsford (Going 1987, 74) and Mucking (Rodwell 1973, 36-7). The relative popularity of all these products must derive from functional differences. Jars and dishes perhaps had multiple functions, for example, cooking, storage and tableware. Consequently, more vessels were required to fulfil these roles, and replace vessels that broke through constant use. Bowl-jars possibly served singular and less intense functions, perhaps confined to the dinner table, and therefore lasted longer.

This functional division may also explain why fine grey ware was the least common of the three kiln fabrics (Table 26).

The amount of kiln waste recovered from kiln 14858 was very low compared to the 40 EVE, or so, collected from the kilns at Moulsham Street, Chelmsford and 20 EVE from Inworth. The overall quantity is, of course, higher, but the figure remains comparatively low. Indeed, rim circumferences were rarely more than half-complete. As in Area W, the pottery recovered here, and particularly from the kiln, seems to represent little more than the 'sweepings-up' of already broken and discarded waste material.

The similarities between the waste pottery collected from kiln 14858 and its associated pits and the pottery produced at Inworth and Chelmsford merit comment. These production sites can be linked on the basis of the crushed flint temper that the potters used, and more or less identical forms, suggesting a single potting tradition. While itinerant potters might link two manufacturing sites of comparable date, it is harder to apply to a tradition lasting almost 100 years. The mechanisms that introduced flint-tempered pottery are unknown, though strong cultural and economic links within the region doubtless had their parts to play.

Dating

These products, presumably fired in kiln 14858, were manufactured sometime during the late 3rd century or the first half of the 4th century AD. Production probably ceased before the mid-4th century AD. These assertions are made on the basis of the fabrics and range of forms produced. The sandy grey and black-surfaced fabrics represented here were tempered with crushed flint, and could be regarded as forming part of the tradition of flint-tempered fabrics in Essex. Flint-tempered pottery was recovered from the Moulsham Street kiln in Chelmsford (Going 1987, 74-8), and the much flintier Rettendon-type fabric produced at Inworth (Going 1987, 83), Sandon (Drury 1976), and Rettendon itself (Tildesley 1971). Going (1987, 89) provided a late 3rd to mid-4th century AD date for the production of these fabrics, which he viewed as forming an interconnected workshop industry.

Some of these sites also offer parallels to the range of forms represented at Elms Farm. Potters working at the Moulsham Street kiln site included all six forms, except the bifid-rimmed jar, in their repertoire. A late 3rd or 4th century kiln assemblage from Orsett (Rodwell 1974, 32) included B1 and B6 dishes, E5 bowl-jars and G24 jars. The parallels from Inworth are less exact, though most of the six vessel types are broadly represented. The B6 dish is perhaps the key to providing a date. While B1, E5 and G24 jars were produced prior to the late 3rd century, B6 dishes were only made after c. AD 260 (Going 1987, 15). Notably, the form is absent from the kiln assemblage from Ivy Chimneys, which dates no later than the mid-3rd century AD (Turner-Walker and Wallace 1999, 170).

While a late 3rd century, or later, date for pottery production is reasonably certain, the date of abandonment is less clear. Pit 14655 produced typically later 4th century pottery, including late shell-tempered ware, and a coin of Valentinian II (AD 388-392), but these could well be intrusive. The pottery, particularly, is very abraded. That production ceased before the mid-4th century is suggested more strongly by the presence in the same feature of a G24 jar in Rettendon ware and a Nene Valley colour-coated bowl with painted decoration from pit 14744. The latter is of a type that declined after the mid-4th century (Perrin 1999, 104). Both imply that waste products were deposited any time up to, but no later than, the mid-4th century AD. Notably, the kiln itself yielded nothing of exclusive mid-4th century date or later.

The archaeomagnetic dating is inconclusive, but does not necessarily contradict the pottery dating. Two dates for the last firing of the kiln are offered; AD 150-210 and AD 270-400.

Kilns 10906 and 11423, Area N

Kiln 10906 (Group 672) yielded 6346g of pottery, none of which could be positively identified as waste products. There were, however, a number of larger body sherds and diagnostic fine grey ware rim sherds in good condition among the small and abraded sherds, and are feasibly kiln products for these reasons. A total of 2113g was recovered from second kiln 11423 (Group 693). Again, no products were identified by form, although the kiln contained groups of spalled or heat-damaged sandy grey ware sherds. This small collection, amounting to 240g or 11% of the total, has been tentatively identified as kiln waste.

Fabrics

Fine grey ware (GRF)

A micaceous fabric with dark grey (5Y 5/1 to 4/1) surfaces and darker grey (2.5Y 4/0) or grey brown (2.5Y 4/2) core. Fine clay matrix with well-sorted white quartz, resulting in a speckled appearance. Occasional larger quartz grains, visible macroscopically. Surfaces have patchy burnish.
Forms: Dish B2, bowl-jar E2, jar G (pedestal base)

Sandy grey ware (GRS)

A micaceous and slightly gritty fabric. Surface colours vary in shades of grey (10YR 6/1 to 4/1). Inclusions of white and clear quartz grains, and very occasional flint. No forms identified.

Forms and dating

Three vessel types, all in fine grey ware, were identified as possible products of kiln 10906; a shallow bead-rimmed dish (B2), a ledge-rimmed neckless globular jar (E2), and a jar with a frilled pedestal base resembling Cam 207. Single examples of each were recovered.

Whether fired in the kiln or not, they suggest that the structure was abandoned probably in the first half of the 3rd century AD. Pedestalled jars were manufactured in Essex during this time, for example at Grays (Rodwell 1983, 34) and Mucking (Pollard 1983, 135), while the E2 bowl-jar dates from the late 2nd century onwards (Going 1987, 21). The dish provides a mid-3rd century ceiling. This form was produced up to or a little beyond this date at Ivy Chimneys (Turner-Walker and Wallace 1999, 170) and Orsett (Cheer 1998, 101), but was absent from late 3rd century-plus kiln assemblages at Chelmsford or Inworth (Going 1987, 78, 88). Much of the pottery recovered from the kiln deposits more strongly suggests early to mid-3rd century abandonment, and includes an H35 folded beaker and a D3 mortarium.

Kiln 11423 could well have been broadly contemporary with 10906. Pottery, including BB2, recovered from packing deposits date construction to the second half of the 2nd century AD, while bead-rimmed dishes from disuse deposits suggest that the kiln was abandoned by the mid-3rd century AD. This accords well with the archaeomagnetic date of AD 225-250 for its last firing.

Discussion

A combination of stratigraphy, artefactual evidence and archaeomagnetic dating fixes the maximum lifespans of the kilns and places them within a chronological sequence. However, establishing the actual length of time that the kilns were in use, the scale of production and the organisation of the work are altogether more challenging and are beset with problems of interpretation, but it is these answers that provide insights into the part that pottery production played in the economy of Heybridge and the daily lives of its inhabitants.

A sense of scale is achieved by estimating the kiln output. Using the assumed wastage rate of 5-10% (Cheer 1998, 99), the approximate total weights of products from the kilns in Areas W, L and N is 300 to 600kg, 290 to 580kg, and 2.5 to 3kg. respectively.

Together, these are well below the total attained at Orsett 'Cock', but still must be regarded as conservative estimates. As Cheer noted (1998, 99), the total quantity of waste pottery available from a site is an unknown factor. In any case, these figures are based on the total amount of recovered kiln products, which potentially represent any number of firings. Clearly, the output from a single firing cannot be estimated reliably, as the resultant waste material cannot be isolated. While a greater volume of kiln products were recovered from Orsett, the Heybridge kilns may have had fewer firings, but produced more pottery.

Kiln size, too, does not necessarily reflect cumulative output, though has some bearing on the maximum loads held by each kiln in a single firing. All but one of the Elms Farm kilns measured between 1.3 and 1.5m in diameter. This compares favourably to some Mucking (Jones and Rodwell 1973, fig. 2) and Orsett 'Cock' kilns (G.A. Carter 1998, 62-70), which were between 1.0 and 1.6m wide. These yielded far larger quantities and more diverse ranges of pottery than was present at Heybridge, though all these kilns may have had similar capacities. However, without the evidence of the superstructure, the exact capacities remain unknown. Measurements, listed in the archive, for twenty-four regional kilns have provided a figure of 1.3m for their average diameter.

Far more problematic is linking the pottery to the use of the kilns. While spoiled pottery can be recognised by its condition (e.g. overfired, misshapen or blown), it cannot be ascertained whether pottery recovered from the backfills of kilns were actually fired in those kilns. The presence of sherds that are clearly not local, such as Nene Valley and Hadham wares serves to remind us that the kilns, once abandoned, are, to some extent, merely ordinary receptacles for rubbish. In addition, the waste pottery recovered from a kiln is more likely to relate to its last firing, rather than reflect its entire use. While quantification of kiln products takes us towards an estimate of the scale of pottery production, it cannot so usefully indicate the life span of any kiln or the number of firings undertaken.

Instead, the life span of a kiln may be better established by identifying rebuild and repair. Taking Fulford's suggestion of one rebuild per season (1975a, 22), the absence of visible signs of repair or rebuild in all but two of the kilns suggests that they each had a life-span of no more than a single potting season. Kiln 11423 showed at least one episode of repair, suggesting continued use into a second season. The mortaria built into kiln 1618 may also be evidence of repair. The flue walls of kiln 11423 in Area N had also been repaired at least once. Identifying repair, however, is problematic, as it becomes visible only when parts of the previous structure remain extant. The original structure would leave little trace if it were replaced in one go. In addition, parts of the structure may have needed repair on a regular basis, so that, by the end of the first season, the kiln had already undergone structural changes.

A single or short period of use may help to explain the somewhat limited repertoire suggested by the Area W and L waste pottery. Potters were likely to have set up a kiln and produced pottery knowing that they would be able to sell it. What was produced during the initial period of use perhaps represented the immediate satisfaction of the market. Expansion of the repertoire came when potters became more familiar with the demands of the market or felt confident enough to introduce new shapes. Both suggest an element of longer-term production and, in the case of the latter, financial security. A narrow range of forms, then, does not necessarily suggest small-scale production, but rather an intensive and short production period.

On the current evidence, then, pottery was manufactured at Heybridge on a seemingly intermittent basis between the late 2nd and first half of the 4th century (Figure 321). However, given the view that the kilns could have been functioning for longer, the possibility that production was continuous between these dates and beyond cannot be dismissed. In addition, it is by no means unreasonable to suggest that there remain other, as yet undiscovered, kilns at Heybridge located away from the settlement and nearer to abundant fuel sources, such as woodland.

The concept of the itinerant potter is one that provides a convenient (if unconvincing) explanation for sporadic production. A potter could relocate, set up a kiln and satisfy the immediate demand before moving on. But as long as pottery was used, there was a demand for it. An itinerant potter would need to return to a location on a regular basis, perhaps every year. Competition between potters might complicate matters, with rival potters also setting up kilns. If anything, given this model, we should expect makers' marks on coarse pottery in order to differentiate products, and the presence of many more kilns. The effort required finding suitable locations with good clay and fuel sources, setting up a kiln and preparing the clay before actually making pots should not be underestimated. Once the infrastructure was in place, the potter might well be reluctant to shut-up shop, move on and start afresh. It should be noted that the kilns in Areas L and N appear to be located in domestic plots or small-holdings and were perhaps operated by the occupants/owners of these plots. An alternative model, therefore, proposes that a potter was permanently resident in the settlement, operating his kiln on a seasonal basis or during the time that was available for potting. The potter might supply other settlements by transporting the goods himself or using the services of a middleman. Opinion now tends to favour this model, based as it is on ethnographic parallels (Cheer 1998, 101). The close proximity of some kilns to domestic plots also supports the model, although such locations may be the exception rather than the rule.

What constitutes the 'potting season' also deserves some attention. Presumably, if potters were farmers engaging in agriculture as well as tending livestock, a potting season was equal to the number of months spent away from farming-related activities. Periods available for potting extended across the year, but these may have been short. Heated drying sheds would certainly be required for winter and spring production, as these months could scarcely have been advantageous to the potter, who would not be able to dry pots in wet and cold weather or easily collect dry fuel. Seasonal manufacture, predominantly during the drier and warmer months, provides the best explanation for the evidence at Elms Farm. It implies production of a surplus; potters must produce enough vessels to meet an all-year round demand. Furthermore, this suggests intensive production, which would not necessarily benefit from the uncertainties of the market, and perhaps even discourage potters from testing the market and extending their repertoires.

In summary, despite the ambiguities of the evidence, the kilns at Heybridge were probably in service for a short time, perhaps for one or two seasons only. While the impact of these kilns was restricted, pottery production is likely to have played an integral part in the local economy throughout the Late Iron Age and Roman periods. It has been argued elsewhere that most coarse wares had a probable local origin (Atkinson and Preston 2015), and the absence of kilns does not necessarily mean that no local production took place. It is evident that a single potter, even working for part of the year only, could have supplied the settlement at Heybridge, which perhaps had a population of no more than 400 people (probably nearer 200) at any one time. Greene (1993, 44) has suggested that four potters working for a period of ten years supplied all the pottery required for a legionary fortress in which the population at any given time would have certainly numbered in the thousands. Assuming a working life of 20 to 25 years, over a period of 150 years - roughly the period that the kiln evidence encompasses - we should expect just six potters to have been active (or eighteen potters to cover the entire Late Iron Age and Roman occupation). Assuming that the best place to locate a kiln is in a sheltered place with access to fuel and water, it is unlikely that excavation of a settlement itself, rather than its hinterland, would reveal many kilns. Indeed, like Area W kilns 1223 and 1618, further structures were probably located just beyond the settlement edge.

The potters themselves, perhaps settlement residents, were probably 'part-time', undertaking the work during slack periods of the farming calendar. If so, local production cannot properly be termed an industry. Unlike pottery manufacture at Heybridge, large production centres, such as those in the Nene Valley and at Colchester, were organised on more industrial lines, implying permanent production, wider distribution, and a full-time 'professional' workforce responding to and influencing the market.

Vessel function and use

Cite this as: Biddulph, E. and Compton, J. 2015 Vessel function and use, in M. Atkinson and S.J. Preston Heybridge: A Late Iron Age and Roman Settlement, Excavations at Elms Farm 1993-5, Internet Archaeology 40. http://dx.doi.org/10.11141/ia.40.1.biddulph2

Beyond issues of chronology and supply, the extensive pottery assemblage has great potential for the study of its functional aspects, providing insights into the lifestyles of the people, and their varied use of, and attitudes towards, ceramic vessels. Pottery was clearly a very significant and versatile resource, evidenced by its sheer amount and variety. The studies presented below provide an idea of the multifarious applications of pottery in daily life.

Changes in vessel shapes reflect changing uses of pottery. In some cases, changes in vessel function, and therefore the daily habits of pottery users, can be traced through time. This is exemplified here by the study of dishes and platters. We should nevertheless be wary of assigning functions to vessels on the simplistic criterion of shape alone, and acknowledge that particular types may have had varied uses. The studies of use-wear on samian vessels and of storage jar ovens clearly demonstrate this.

The functions of some vessel types tend to be fixed in the literature and the large dataset allowed many conservative notions and labels to be challenged. Amphoras are a familiar aspect of site assemblages, but rarely considered in terms of their continuing use beyond the role as transport containers. Perforated vessels are dismissed as cheese presses or colanders, or, more recently, as being 'ritually killed'. Closer analysis reveals a far wider variety of uses. Pots may have been invested with symbolism - graffiti may have been added, the form itself might have been special (e.g. face vessels), or the vessels were deliberately placed within significant deposits. Such treatment reflects attitudes towards vessels, which, in some cases, were likely to pertain to religious and sacred practices. Further discussion of these aspects is made in the sections relating to burials and structured deposits (both in and outside the temple precinct).

The studies presented focus on aspects that provided the clearest results. The very large ceramic assemblage, however, offers the potential for a wider range of studies that could not be attempted in the time available, such as investigating associations between vessel types and other artefacts. Other aspects were considered, but dismissed due to poor-quality data and lack of clear patterning; for instance, defining the relationship between lids and lid-seated jars. In contrast, some of the successful studies opened up further avenues of inquiry that could not be pursued owing to project constraints. Subsequent work using correspondence analysis on parts of the pottery data (Biddulph 2005; Pitts 2005) has highlighted the potential for further study, especially with regard to associations between vessel classes and feature types. The pottery resource provides many more possibilities for future research.

The development and function of platters and dishes

The starting point for this study was Going's statement, with regard to Chelmsford, that, 'later Roman dish forms (incipient flange-rimmed, fully flange-rimmed and deep plain-rimmed dishes) do not fulfil the same function as platters, which as a class are quite absent' (1987, 13). This idea derives from Going's observation that 'most dish forms, particularly from late Roman contexts, are comparatively small and deep', as compared to platters and earlier dishes (Going 1987, 14). The extensive pottery assemblage from Elms Farm allows us to test two basic ideas. First, that dishes and platters did not serve quite the same functions, and second, that later Roman dishes (that is, those dated mid-3rd century onwards), became smaller and deeper, probably as a result of changes in function. Samian data were excluded from the study, mainly because of its relatively short date span at Elms Farm.

Differences between platters and dishes were examined using the variables of diameter and depth. Tables 28 and 29 give the mean diameters of platters over time, based on a sample of 175 platters. Just one continental platter record belonged to Ceramic Phase 1, and so was amalgamated with Phase 2. Similarly, the single Ceramic Phase 6 local platter record was amalgamated with Phase 5. The results suggest that continental platters were consistently wider than locally produced types. The mean diameter of local platters was remarkably consistent through time, although it is significant that in Phase 5, when local platters were narrowest, the standard deviation was low. There is a narrower range of diameters at this time, suggesting that the size of platters was further standardised.

Next, the depths of sixty-eight platters were measured (using archive drawings) and the observations separated into ceramic phases. Means and standard deviations were calculated (Table 30).

It would appear, then, that platters became deeper over time. To ensure that one or two extreme values, or outliers, did not cause the difference between Phase 4 and 5 means and the relatively large Phase 5 standard deviation, two outliers (a 10mm deep platter from Phase 4 and a 49mm deep platter from Phase 5) were excluded. If anything, this confirmed the apparent difference. The Ceramic Phase 4 mean rose to 26.5mm, while the Phase 5 mean stayed firm at 34.6mm and this standard deviation was reduced to 5.8mm.

Comparing the diameters of continental platters and grog-tempered platters, it is apparent that the imported platters are wider than locally produced varieties, and it is this difference that accounts for the large means in Ceramic Phases 2 and 3. The overall mean diameter of continental platters from the original sample of 175 vessels is 289mm; that of grog-tempered varieties is 217mm. It is also clear that grog-tempered platters were not full copies of continental prototypes, certainly in terms of size (Table 31).

While local potters copied some elements of shape, the difference in diameter suggests that they may have been unable, or perhaps unwilling, to reproduce full-sized platters. The skills and equipment required to produce fine, wide, shallow platters were perhaps not sufficiently advanced during the Late Iron Age in southern Britain. The production of comparatively small vessels, while approaching the right shape, may have been the limit of the technical abilities of local potters. Rigby noted that potters local to Verulamium produced versions of imported platters, but manufacturing techniques meant that production was comparatively small scale and the forms less standardised than their continental prototypes (1989, 152). The diameters of the local platters from Heybridge suggest that production was just as standardised, at least in terms of size.

To facilitate comparison dishes were examined, again using the variables of diameter and depth. It should be noted that not all of the 825 available vessels were from closely dated contexts. In such cases, the ceramic phase nearest to the middle of the date range was selected. Thus, a dish from a context dated mid-2nd to mid-3rd century was assigned to Phase 7 (AD 270-310). The mean diameter and standard deviation for each phase was calculated (Table 32).

While both the means and diameter ranges vary to some extent, the figures do not suggest an overall and sustained change in vessel diameters. To investigate whether dishes became deeper over time, all dishes with a complete profile were measured (the internal height of the vessel from the point where the vessel wall meets the base to the rim). There were no records assigned to Ceramic Phase 4, while Phases 5 and 6 contained just three and five observations respectively and so were amalgamated. The means and standard deviations were calculated (Table 33).

It would appear that dishes were made deeper in Ceramic Phases 9 and 10 (later 3rd to mid-4th century), occurring at a time when flanged dishes had replaced bead-rimmed dishes at Heybridge (see Pottery supply). However, it should be noted that the standard deviation is highest also for these phases. In other words, the differences between the extreme values and the arithmetic mean are at their largest. As just one or two outliers may cause such differences, the extreme values in Phases 9 and 10 (one in each) were excluded. The mean and standard deviation for Ceramic Phase 9, at 42.8mm and 14.6mm respectively, thus resembled those for Phases 5-8. Phase 10, however, was largely unchanged. Such an apparent difference is likely to be statistically insignificant. Admittedly, there was a greater occurrence of deep dishes (that is, over 60mm deep) in Ceramic Phase 10 than in any other, but the majority of dishes in this phase were, in fact, under 40mm deep. There is, therefore, no conclusive evidence to suggest that dishes became smaller or deeper over time.

In summary, local platters, while becoming deeper over time, were reasonably similar to dishes in diameter. It is a commonly held assumption that size is one determinant of how a vessel might have been used; others include context, practicalities, and individualistic interpretation (Willis 1998, 113). Considering these, a model may be proposed; large vessels enjoyed communal use, while smaller vessels were more suitable for individual portions. At their widest, then, platters (mainly continental varieties) were probably placed in the centre of the dinner table, from which the participants of the meal took, or were served, their food. Over time, diners tended to eat from 'individual'-sized platters, with food perhaps being served from large bowls, such as the wide-mouthed C28-C33 types, and even samian bowl forms (see samian use-wear). The bowl was a relatively minor form during the Late Iron Age (Ceramic Phases 1-3), but became as common an open form at Heybridge as platters during the Early Roman period (Phases 4 and 5) (see Pottery supply). It is speculation as to why platters became deeper, but may perhaps indicate a reversion to native-style foods.

As a vessel class, the bowl declined in popularity from the mid-2nd century (Ceramic Phase 6) to be seemingly replaced by deep dishes, particularly the B4 bead-rimmed dish. While it is clear that dishes did not radically alter in depth over time, the large standard deviations reveal that there is a spread of values in each phase. Put simply, potters made both shallow and deep dishes throughout the Roman period. It should be noted, incidentally, that there was no 'standard' shallow or deep measurement - observations in each phase do not fall neatly into two distinct groupings - and some dishes might have been used equally for both eating and serving.

Generally, local platters had similar diameters to dishes of all periods, but were shallower, though the difference in millimetres is perhaps negligible in practical terms. Thus, the evidence suggests that platters served a similar function to dishes. The development of samian vessels appears to be analogous. Willis (1998, 113) noted that samian dishes eventually replaced samian platters, but, while shape changed, the capacities of both vessel types were similar, suggesting that there was no change to their basic functions.

The range of platters and dishes at Elms Farm perhaps allows us to trace evolving functions, beginning with the direct transference of Gallic (and ultimately Roman) uses, through to native adaptation, and concluding with dishes, and a set of functions, culturally distinct from the original uses of continental platters. Roman literature can perhaps provide an impression of these original uses of continental platters, in which it seems that social and aesthetic considerations were as important, if not more so, than practical reasons in the development of vessels. The Younger Pliny (3.1) describes a 'simple' meal of several courses. The multi-coursed dinner is alluded to in another letter (5.6). In both, the emphasis would appear to be on choice for the diner and munificence on the part of the provider. Wide, shallow, platters were ideal table vessels, in which the food was presented and admired before being served. Thus, at Heybridge, the specific and restricted social context of which continental platters were a part became devolved with the probable immediate production of platters in local wares, incorporating a more individualistic and prolific use of the form, but perpetuating continental styles of dining. These uses continued with the development of dishes.

Because dishes did not radically alter in size over time, they continued to serve the same functions until the end of the Roman period. This does not rule out the possibility that dishes acquired functions around the mid-3rd century, resulting in shape changes. The rapid typological development to which dishes were subject is critical to this argument. The dish acquired a flange by the mid-3rd century to accommodate functional changes (e.g. to support a B1 dish inverted to form a lid for cooking). By the end of the 3rd century, the incipient bead-and-flanged B5 was largely superseded by the fully bead-and-flanged B6. The mid- to late 3rd century was, then, a period in which the design of the dish was perfected to suit its new purpose. The changing assemblage composition from Ceramic Phase 8 (AD 210-260) to Ceramic Phase 11 (AD 260-410) supports this hypothesis. From the mid-3rd century, two form types, the plain-rimmed B1 and B3, became more frequent, while two new form types, the bead-and-flanged B5 and B6, were introduced. In addition, the only identifiably new jar type was a storage vessel, as opposed to a cooking vessel (see Pottery supply). Ceramic Phase 10 (AD 310-360) marks a sharp rise in the number of dishes, with the number of jars remaining steady from Phase 9. More dishes were required to meet their additional role. As if to underscore these observations, and revealing the increasing importance of dishes during this time, pottery manufacturing waste assemblages from kiln 14858, dated late 3rd to first half of the 4th century (see Pottery production above), comprise 58% dishes (both B1 and B6) by EVE, compared to 38% jars.

It is entirely possible that this new role was kitchen-based. Studying the black burnished wares from northern Britain, Gillam posited that plain-rimmed dishes were used in conjunction with flanged dishes to form 'casserole-sets' (1976, 70). Comparing the mean rim diameters of B1 and B6 dishes from Heybridge provides some statistical affirmation that these dishes, plain- and flanged-rimmed, respectively, were indeed used together. B1 dishes dated mid-3rd century onwards have an average diameter of 197mm; that of B6 dishes is 204mm. Some small difference between them is to be expected since the measurements from B6 dishes are taken from the tip of the flange on one side to that on the other. The rim diameter of the B1 dish needs only to extend beyond the bead of the B6 dish. Examples of the likely use of pairs of dishes in sets are demonstrated figuratively in Gillam (1976, fig. 6.89-91)

In this study, the development and function of platters and dishes have been examined. While there are undoubtedly size differences between the two vessel classes, it is probable that they effectively served the same functions. It is more certain that dishes did not radically alter over time, but acquired, rather than changed, functions. Future studies could examine whether flanged dishes were consistently deeper than plain-rimmed dishes during the later 3rd and 4th centuries. If so, this would link shape and size and strengthen the belief that flanged dishes were made for a specific purpose. Examination of signs of cooking, for example burning and residues, might also prove instructive.

Samian use-wear

As well as the examination of form itself, the study of use-wear upon ceramic vessels is informative as to vessel function. Colour-coated vessels provide the best evidence of use-wear. Breaks in the slip caused by abrasive processes accompanying cooking, consumption and other activities can be clearly seen. Intensive use of a vessel eventually removes the slip. Where the same regular process is applied to a vessel, a wear pattern will emerge. Identification of wear patterns is therefore crucial for assigning functions to specific vessel types. For the purpose of this study, samian vessels were examined. The range of samian vessel classes is sufficiently wide to reveal information about a variety of household settings. In contrast, other colour-coated vessels found at Elms Farm, for example those manufactured at Colchester and in the Nene Valley, mainly comprise beakers, which tend to show use-wear, if any, less clearly. However, analysis of this sort remains possible by using a wider range of products from these industries not available at Heybridge. Wear patterns were noted in too few vessels to be able to gain spatial or chronological insights. This is not to say that processes that caused the patterns were very infrequently practised at Heybridge, but rather that few vessels could be sufficiently reconstructed to identify patterns sufficiently. Practices were likely to have been more widespread than the small number of vessels studied here suggests. Perhaps the most common sort of wear to affect samian vessels was that on the underside of foot-rings and the top of rims. However, this is deemed to be incidental, as it reveals little about vessel function per se.

Cups

Two cup forms, f27 and f33, were particularly strong in wear-pattern evidence, with many examples displaying very consistent patterns. That noted on f27 cups is similar to the wear seen on bowls, see below. In Figure 340, the wear pattern can be seen to be circular, and the stamp is almost entirely worn away. A more developed pattern, probably achieved through prolonged or more intensive use, is seen in Figure 341. Here, the wear extends to the lower wall, forming an even boundary around it. Again, the base interior retains very little slip.

Form 33 cups tend to have a ring of wear at the junction of the base and wall. In Figure 342, the centre of the base can be seen to be unworn, but the base is very worn at the edge. A thin ring of slip is retained in the basal angle, but a second wear-ring is present at the bottom of the wall. Just as with f27 cups, a progression can be observed (Figure 343). Prolonged or intensive use generally results in a thicker wear-ring radiating towards the centre of the base. The corner between the base and wall, too, has become worn.

The remarkable aspect about both forms is their exclusive wear patterns. Very few f27 cups that displayed use-wear had patterns other than those described above. Similarly, f33 cups either showed no wear at all or a wear-ring at the junction of the base and wall. This suggests f33 cups were used differently from f27s. The pattern on the f27 bears some similarity to that on certain bowls, such as f38, and thus some functions might also have been shared. Products placed into f27 cups seem likely to have required further processing. In the kitchen, spices or herbs might have been placed inside and ground. In the dining room, the cup may have contained foodstuffs, such as curds, yoghurt or a sauce, requiring the use of a spoon. This does not preclude f27 cups from being used as drinking vessels, so long as the liquids that the cups contained required no processing from which wear would result. Stirring would not concentrate the wear in the centre of the vessel, nor result in a pattern extending across the entire base.

There are two possible causes of the very distinctive ring-effect wear pattern observed in f33 cups. Perhaps the least likely cause is residual traces of wine in unwashed cups. The acidic properties of the liquid might, over time, have eaten away at the surfaces. This explanation is not wholly convincing, since the damage consistently occurs in the wrong places - just above the junction of the base and the wall, but not within it. However, experimental work might be usefully undertaken to assess the effects of residual wine, say, on pottery surfaces. More likely is the suggestion that the f33 was repeatedly subjected to a single process, such as stirring. The unworn ring in the corner of the base and wall suggests that the stirring implements could not meet the corner, perhaps because they tended to have slightly bulbous use-ends. A spoon, naturally, comes to mind, but stirring rods, which have narrower rounded ends not unlike spoons, were probably more usual. The need to stir indicates that the liquid required processing within the cup. Beverages may have been sweetened with honey (perhaps implying a hot drink), or ingredients added to the cup to create something akin to a modern cocktail. The regular occurrence of the wear-ring in the f33 cup suggests that the form was indelibly associated with a specific function at Heybridge, just as a modern wine glass, or whisky tumbler, is defined by shape and tends to contain only certain selected beverages.

Bowls

Wear patterns were noted in four forms, namely f37, f38, f44 and Curle 11, and affected mainly Central Gaulish pieces. Three distinct patterns were observed. Figure 344 shows the base of an f38 bowl with multiple linear scratch marks. These extend internally across the base and part-way up the vessel wall, stopping approximately at a point level with the flange. The scratches are short and have no set orientation. The stamp is slightly worn, but is otherwise not scratched. It is likely that a sharp pointed implement caused these scratches, probably a knife. The size and shape of the bowl would not allow extended cuts to be made, so items placed within the vessel were probably small. Assuming that the vessel enjoyed domestic use, the scratches may well have been made in the kitchen, with the vessel being used during the preparation of ingredients and when cutting small items, such as fruit. Alternatively, the vessel might have been placed on the dinner table and used by an individual diner in the same manner as a plate.

A second type of wear again affects an f38 bowl (not shown). Internal wear is restricted to the centre of the base. A circular pattern has formed, with the stamp almost completely worn away. Figure 346 shows an example of a developed form of this pattern. The wear extends to the lower part of the wall, with the base almost entirely devoid of slip. This sort of pattern was also noted on f37, f44 (Figure 347) and Curle 11 bowl types.

These use patterns are generally heavy, but the condition of the vessels where the slip remains is good. This suggests an intense, but directed usage of such vessels. In addition, a very even boundary between slip and wear can often be seen on the lower part of vessel walls, suggesting repeated processes and exclusivity of function. Again, use in the kitchen may be suggested, with vessels perhaps serving as mixing bowls or mortars. The vessel in Figure 346, with its more developed pattern, may have been used for longer, or more intensively.

Mortaria

Preservation of samian mortaria at Elms Farm was poor. Although many mortaria sherds were recovered, no vessel was complete enough for wear patterns to be fully identified. Sherds are worn to various degrees, but nearly always on the interior surface, and range from barely used to heavily worn. In well-used examples, wear generally extends towards the top of the interior surface level to the base of the collar, forming an even boundary around the vessel (Figure 348). The interior surface is often devoid of slip, and the trituration grits, when present, are often smooth with use. Figure 349 shows a Nene Valley colour-coated ware copy of samian mortarium f45. As with examples of its prototype, the wear extends to a level just below the collar. The interior surface has not been entirely utilised, however; the centre of the vessel is less worn than the side, and usage seems to have been concentrated in an area of the vessel wall towards the spout. This is consistent with its possible use as a mixing bowl. When mixing ingredients, such as flour and water, modern experience reveals that there is a tendency to tilt the bowl. In any case, the restricted pattern suggests repeated action of a single process.

Dishes

Although no particular patterns were observed in this class, the wear on one vessel was unusual, and warranted further attention. This incomplete vessel comprised a Central Gaulish f31 base. Apart from the wear patterns, the base is in good condition with reasonably fresh surfaces. A ring of 'kiln grit' is present on the interior or top surface, indicating that some of this surface was virtually unused. There are three areas of wear evident. First, the underside of the foot-ring is worn, with very little slip remaining. The slip on the sides of the foot-ring is still present and in good condition. Secondly, a roughly circular pattern of wear is present on the underside of the base, within the foot-ring (Figure 350). The zone of wear almost meets the junction of the inner side of the foot-ring, while a ring of slip is present centrally within the prominent 'kick'. A third wear pattern can be seen on the interior surface of the base (Figure 351). Two opposing triangular patches of wear are present, maximum height 35mm, base lengths 25mm and 15mm. The apices of both patches almost meet in the centre of the base, extending out towards the outer edge. In the centre, at the peak of the 'kick' and covering the middle of a stamp of Priscinus (see Samian section) is a very small wear circle, 7mm in diameter.

The underside of the base has been used intensively. The good condition of the inner surface and the abraded foot-ring suggest that the vessel was used conventionally for only a short time prior to its possible deliberate breakage and rough trimming to remove the vessel sides. That the modified vessel rested on the peak of its internal kick is clearly shown by the resulting wear. The cause of the wear on the underside remains unknown, but its sub-circular pattern suggests circular movement on the part of the user. The triangular patterns on the inner surface support this explanation. These very specific patches suggest a rocking or swivel action, as if the piece had been fixed, restricting movement to a short turn clockwise and anti-clockwise. This is a very unusual pattern, for which no explanation is immediately apparent.

Discussion

These wear patterns inform us that samian vessels were functional at Heybridge. Against the conventional view that samian was valued and looked after (cf. Willis 1998, 86), the wear-marks reveal that some vessels were not rarely seen heirlooms - delicately handled or brought out on special occasions - but served robust and mundane functions. Bowls and cups were not the exclusive preserve of the dinner table, but were used also for preparation and could become just as worn as mortaria. Members of ostensibly the same vessel class did not necessarily serve the same functions and were undoubtedly employed in a variety of household contexts. However, the wear patterns on all of the vessels in this study resulted from repeated processes, suggesting that these vessels served singular purposes. A bowl used for mixing was, by and large, always used for mixing, its function fixed by custom and practice.

There are a number of implications to be drawn from these observations. Some fine wares, and perhaps all, are likely to have been treated no more carefully than coarse wares. Distinctions between fabrics made through a microscope or hand lens do not necessarily reflect the functional or practical distinctions of a past society. Modern experiences, too, result in differing perceptions of ancient use. The f27 cup, for example, is a drinking vessel according to British archaeologists, but a 'sauce-dish' on the continental mainland (e.g. Schucany 2000, 123). Such labels are modern constructs, and without the evidence of wear-patterns or, better still, residues, assigning actual, as opposed to perceived, functions to vessel types is problematic.

It should not be taken for granted that all cups were used for drinking and bowls for eating. Potentially, the range of functions that vessels might have served and the social contexts in which the vessels could be employed can be increased. This can affect the interpretation of, say, burial assemblages. The function of a suite of accessory vessels recovered from a cremation burial cannot be pre-assigned. A suite comprising the seemingly standard flagon, cup and bowl should not be automatically given simplistic eating/drinking interpretations, but instead a range of activities may be represented.

With one exception, samian has formed the basis of this study. However, all colour-coated fabrics can be easily studied for use-wear. Clearly, the more complete the vessel, the more identifiable a pattern becomes. Wear on fabrics that were not originally slipped cannot be observed so clearly, but inferences drawn from the samian might also apply to copies of samian prototypes manufactured in non-slipped fabrics. As a matter of course, any signs of use on these copies should be noted to ascertain whether function, as well as form, was being copied. Experiments involving a range of processes and conditions could be undertaken to replicate wear patterns. Ultimately, a study incorporating a much wider dataset than that present at Heybridge is required in order to establish whether the trends established here are mirrored elsewhere.

Graffiti

Relatively large numbers of graffiti were noted during pottery recording. All fabric and vessel types are represented, although the preferred vessel classes seem to be dishes, beakers and jars. Graffiti can be loosely divided into two types, literate and non-literate, and both can be applied either before or after firing. Graffiti incised after firing, however, appear to be the most numerous. Archive drawings were produced, as a matter of course, for all graffiti and other marks noted on the coarse pottery. The samian was isolated from the pottery assemblage before recording work commenced on the coarse pottery, and the identification of graffiti on samian sherds was carried out by Brenda Dickinson. Archive drawings, therefore, were not produced for these.

At least 134 examples of graffiti were recorded, and further possible examples, recognised as faint marks or part-letters, were also noted. Full details can be found in the archive. Most of the graffiti (91 examples) occurred on coarse wares and regional traded wares of both Late Iron Age and Roman date; thirty were recorded on samian, three on amphora, two on micaceous terra nigra and one on Pompeian-red ware (Fabric 5). There is no evidence that graffiti appear more often on a particular type of pottery, for instance on samian, which has long been thought to be the case. At Heybridge, as many graffiti occur on black-surfaced ware as on samian. It may be notable, though, that graffiti on amphora sherds were restricted to three examples. Comparative data can be found in the archive. Differences in the frequency of certain graffiti on samian and other pottery types do not necessarily indicate illiteracy, since post-firing graffiti on all pottery types are more likely to have been made by the users of the pottery, rather than at source.

For the purposes of this study, non-literate graffiti, in the form of crosses and notches, were selected for detailed consideration. Literate graffiti, numerals and other symbols, though briefly summarised here, are noted in the archive and form a resource for further study. Those on coarse pottery are illustrated (Figures 322, 323 and 324, nos 1-16).

Literate graffiti

At least twenty-three examples of literate graffiti were recorded, comprising letters, numerals and groups of letters, some of which are part-examples. Eleven of these graffiti were recorded on samian and twelve on other pottery types. Single letters are difficult to interpret, especially if they are at the edge of sherd breaks, and may represent numerals or the remains of names. Ten examples of groups of letters were recorded, with the remainder comprising probable numerals, such as V, VI and VII. Literate graffiti comprising three or more letters are normally published in Britannia. A single samian graffito from Heybridge was published in Britannia 27 (Hassall and Tomlin 1996, 442, no. 5), but, unfortunately, a further three were omitted. These read MELVA[ on a dish base, FIR[ and FII[ both on f31 dish bases. A single amphora graffito, on a Dressel 20 body sherd, reads VAΛ. The graffiti on coarse pottery were published in Britannia 32 (Tomlin and Hassall 2001, 394, nos 23-33). The 'three-letter' rule was waived, since several graffiti are inscribed on pottery of apparent Late Iron Age date.

Non-literate graffiti

Simple 'X' graffiti

Forty-three simple 'X' graffiti were selected for study. This assemblage comprises only crosses that have been cut into the vessel. Burnished crosses, typically placed on the internal surfaces of open forms, were regarded as decoration and thus excluded from this study. The size of this graffito assemblage appears to be large enough to show trends, though data are required from other sites to verify the significance of those from Elms Farm.

All forty-three 'X' graffiti provided general sherd information, such as fabric, broad vessel class, and ceramic phase. Twenty-eight examples, principally those extracted for illustration, were examined in closer detail to ascertain on which part of the vessel they were cut, whether they were cut before or after firing, their positioning, and whether they were large or small. Based on size, it was possible to identify two types. Type 1 comprises graffiti with relatively long strokes cut into the underside of the base. These strokes, which can each measure more than 40mm, usually extend from one edge of the base through the centre to the opposite edge. The strokes are typically cut before firing, though post-firing examples are also known. Ten of the twenty-eight examples are Type 1 graffiti. Type 2 is more numerous, consisting of seventeen examples. These 'X' graffiti are relatively small, comprising scratches of less than 20mm in length. They are usually cut after firing and located either on the body of the vessel or on the underside of the vessel base. The sole remaining graffito from the twenty-eight examples was made pre-firing and comprises a long, slightly curving stroke and a second stroke that ends just after it intersects the first stroke (KPG24, Figure 256, no. 16). Fitting neither type closely, it is perhaps more likely to be an accidental manufacturing mark.

A wide range of vessel classes had 'X' graffiti, with dishes, jars and beakers predominating. In the sample of twenty-eight graffiti, Type 2 graffiti are found to a greater extent on dishes, and a lesser on beakers and jars. Type 1 graffiti are found on beakers and jars, but are absent from dishes, and therefore exclusively associated with narrow-based vessels. These bring to mind an encircled cross graffito cut into the side of a jar from Chelmsford, which has been interpreted as a wheel symbol with apotropaic properties (Going 1992a, 108, fig. 58.7). Type 1 graffiti are in some ways similar: the intersecting strokes meet and are encircled by the edge of the base. In plan, the strokes, too, resemble the spokes of a wheel (cf. Black 1986, 224). Unlike the Chelmsford 'wheel', Type 1 graffiti were incised mainly before firing. If they did serve superstitious functions, then this role was ascribed at the point of manufacture, a factor of which the buyer must have been aware. Interestingly, two of the four Type 1 graffiti that were incised after firing were regional wares, made in Oxfordshire and Colchester respectively. This hints at a restricted regional practice. The function to which the vessels were put was perhaps significant enough at Heybridge for local potters to include 'special pots' in their repertoires.

Simple 'X' graffiti were inscribed throughout the Roman period. A single example was dated, by the presence of grog-tempered pottery only, to the Late Iron Age. The lack of Late Iron Age examples cannot be for want of pottery, and the practice of cutting crosses into pottery must be a predominantly, if not exclusively, Roman phenomenon at Heybridge. Table 35 suggests intermittent X-cutting activity, with peak periods of deposition occurring during the mid-1st century AD, the late 2nd and first half of the 3rd century, and second half of the 4th century. So, rather than the practice continuing throughout the Roman period, it is one that gained most significance during the mid- and late Roman periods. On this evidence, it is difficult to place the practice (involving both types of graffiti) at Heybridge within a 'deeply rooted pre-Roman Celtic milieu' (Going 1992a, 108).

Simple 'X' graffiti were evenly distributed across the settlement. Twelve graffiti were recovered from the northern zone, which converts to a ratio of 126kg of pottery per graffito. The central zone yielded fourteen examples, or 152kg per graffito, while seventeen examples were recovered from the southern zone (133kg per graffito).

The practice of inscribing crosses into pottery was widespread, but infrequent, within the eastern region. This widespread occurrence appears to be genuine and not just the result of a regional reporting bias, although under-reporting elsewhere, where Xs tend to be dismissed as denoting ownership, may lead to a perceived lack of these marks. Around fifteen 'X' graffiti from Chelmsford have been published (Going 1987, 102; 1992a, 108), mainly mid- and Late Roman. Just one example, incised before firing on the exterior of the base of a later 4th-century beaker (Going 1992a, fig. 49.12), is most likely to be a Type 1 graffito. A more certain example, also pre-fired, was found at Woodham Walter (Rodwell 1987a, fig. 24.162). In all of the above, the incised lines extend to the edge of the base. Another probable Type 1 graffito was recovered from Folly Lane, Verulamium (Lyne 1999, fig. 78.106); in this case, a vestigial foot-ring encircles the pre-fired cross. A samian f33 cup from Brightlingsea, Essex, was found to have an 'X' graffito (Martin 1996b, fig. 8.1). Here, the cross is encircled by the junction where the top of the foot-ring meets the base. The assemblages at both Heybridge and Chelmsford far outnumber that at Colchester, from which, given the very large amount of pottery in total recovered, just four examples with graffito are published in the Roman Colchester pottery volume (Symonds and Wade 1999).

Despite these occurrences, the motivations behind the practice cannot be established with any certainty, and, given the apparent differences in types of crosses, no single explanation is likely to fit all the evidence. Traditional explanations, such as batch marks, owners' marks, or symbols denoting capacity (cf. Going 1992a, 108), have, perhaps rightly, fallen out of favour. Inscribed during manufacture, pre-fired crosses could never be owner's marks, since the vessels had yet to be owned. More generally, it is not the ubiquity of this type of graffito which renders these explanations unlikely (pace Going 1992a, 108), but their intermittent chronology. We might otherwise expect more occurrences in all periods of occupation.

That these graffiti at Heybridge appear mainly on dishes or beakers may suggest deliberate selection. Perhaps crosses marked out vessels that were used for specific purposes. A spindlewhorl from Chelmsford (Going 1992a, fig. 46.2) made from a reused samian base provides an intriguing possibility. The base has a post-fired 'X' graffito that meets the edges of the base in a similar fashion to our Type 1 graffiti. Where the lines intersect, a hole has been drilled, perhaps to utilise the sherd as a spindlewhorl. It is possible, then, that a pre-marked cross enabled the hole to be centred more accurately, ensuring correct balance during spinning. Obviously, Type 1 graffiti cut before firing were not done so on the off-chance that, once broken, the vessel could be reused in this manner. In any case, all but one of the numerous spindlewhorls from Heybridge created from pottery sherds have no such crosses. Simple Xs on sherds were thus unlikely to be aids for the production of spindlewhorls.

Allusion has already been made to the possibility of the graffito resembling a wheel that had both religious and magical connotations. These took several forms. The wheel is strongly linked to Fortuna and appears in a number of stone reliefs depicting the deity, for example from Netherby (Webster 1986, 171). The wheel thus has connotations of luck and fate - the 'wheel of fortune'. Wheels depicted on reliefs tend to have six spokes, but four-spoke wheels are also known, for example as depicted on a relief from Wiesbaden, Germany (Webster 1986, 131). Chariot wheels depicted on Colchester colour-coated beakers also tend to have four spokes (e.g. Figure 298, no. 81). If Type 1 graffiti did represent wheels symbolic of Fortuna, then it remains unclear whether an 'X' graffito gave the vessel and its contents luck, or imbued the user with good fortune. Whole vessels, though portable, would be somewhat impractical as amulets. Instead, it could be imagined that these vessels were used as part of a regular, perhaps daily, rite within the household. Eating or drinking from Fortuna's vessels ensured good fortune. If the graffiti made vessels lucky, then a more prosaic use might have involved pastimes for which good luck was required. Beakers, for example, may have been used as dice-shakers in gambling, the graffito ensured good luck on the throw. However, if acquiring good luck was only a matter of scoring a cross on a pot, then many more crosses, especially post-fired examples, should perhaps be expected.

The wheel was not just an attribute of Fortuna, but was central to celestial iconography of pre-Roman cultures of northern and eastern Europe, the solar overtones perpetuated into the Roman period. The wheel was also an attribute of the sky-god Jupiter, who was identified with the Celtic deity, Taranis 'the Thunderer' (Green 1995b, 155-6). It has been suggested that wheels were offered as votives (Webster 1986, 130). Exceptional evidence of this nature, which included four copper-alloy wheel models, was found at Wavendon Gate in Buckinghamshire, identified as a possible solar-cult centre (Green 1995a, 116). By scoring vessels after firing with 'X' graffiti, or wheels, the vessels, or parts of vessels, might become makeshift votives. None of the vessels with post-fired graffiti was placed in obviously ritualised contexts, although the vessels could have been set above ground.

The purpose of Type 2 graffiti at Heybridge is just as open to speculation. Since most of the vessels with this type are dishes, the purpose of the graffiti and the function of the form might be closely linked; the graffito might denote the special role of the vessel during dining or cooking. Vessels with similar crosses have appeared in funerary contexts, for example at Kelvedon (Rodwell 1988, figs 88.G25, 89.G101, 89.97c) and Ospringe, Kent (Whiting et al 1931). The occurrence of vessels at Heybridge outside funerary contexts, however, suggests that the graffiti did not have an exclusively funerary purpose if, indeed, at all.

Complex 'X' graffiti

Complex 'X' graffiti were identified on twelve vessels. This type of graffito superficially resembles the simple 'X' graffito, but incorporates further scratches and marks. All graffiti were inscribed into the bases of locally produced platters, dishes, jars and beakers. Although the dataset is small, the indications are that complex 'X' graffiti share trends with so-called simple 'X' graffiti. The dates of both have a mid- and Late Roman emphasis. No two graffiti are exactly alike, but some groupings based on style may be tentatively suggested.

Some graffiti are formed by a series of intersecting strokes (e.g. Figure 326, nos 25-27). No. 25 might represent a poor attempt at producing a cross or star similar to No. 27. Alternatively, the former could represent a somewhat crude chi-rho symbol (M. Hassall, pers. comm.), and, conceivably, the latter represents a neater version. While the date of No. 25 lies within the second half of the 4th century, No. 27 dates to the second half of the 2nd century, which would seem to preclude its interpretation as a Christian monogram. Exceptionally among these complex graffiti, No. 27 is a pre-fired graffito, and the strokes seem to extend to the edge of the base. It is possible that this example is a complex version of a Type 1 'X' graffito, and its function may well be identical. If these graffiti represent wheels, then No. 27 is an example of a six-spoke wheel. No. 27 may be another poorly executed cross, or perhaps may even represent letters, now indecipherable.

Other graffiti that comprise a single line extending in one direction, intersected at right angles by at least two other strokes (e.g. Figure 326, nos 28-31) form a second grouping. Unfortunately, none of the bases is complete, so it is possible that the motifs extended to show more detail. These graffiti have all been incised after firing. The purpose is just as unclear, but similar interpretations to those suggested by the simple and other complex 'X' graffiti may be suggested.

Notched graffiti

Twenty-eight sherds marked with incised or notched graffiti were recovered from Elms Farm. These graffiti typically take the form of a series of short incisions made after firing in the rim or base of the vessel, though cuts made both in the body and before firing are also represented. Each example was examined to ascertain the maximum number of notches present. Since multiple notches tended to be equidistant, a group of notches was deemed to be complete if there were no further notches visible at the expected distance beyond the last extant notch. If a sherd had broken before a position where a notch might be expected, the notches visible were deemed to comprise a minimum number. Based on these criteria, a summary is presented in Table 36.

In complete examples, groups of three notches are commonest. Even with incomplete sherds, groups of at least four or more appear infrequently. The single example of a group of five is clearly exceptional. So, mainly groups of up to three notches were cut into vessels, with three appearing to be the optimum number. None of the graffiti was found on complete parts of vessels, and it remains a possibility that some formed part of repeating patterns extending around the edges of rims or bases (cf. Symonds and Wade 1999, fig. 6.92.57), though not necessarily comprising equidistant notches.

The collection of twenty-eight notched graffiti proved large enough to extract clear trends. Notches were cut into the angular parts of vessels, twenty-three of which were rims. Just four examples were cut into the base - usually externally at the junction of the base and wall. A single example, a graffito on a samian f33 cup, was cut on the body, though even this was at the sharp angle of the carination. All but five graffiti were cut into jars or dishes, which share roughly equal numbers of examples. Two graffiti were cut into platters, although given that dishes and platters could well have served similar functions (see above) the motivation governing the choice of platters was probably the same as that governing the choice of dishes. Two graffiti were cut into cups (both samian f33), and one into a beaker. The majority of graffiti (sixteen) were cut into locally produced coarse wares (fabric codes BSW, GRS, GRF and GROG). Eight were cut into regional coarse wares (fabric codes BB, BB1, BB2, LSH, PORD, RET and HAB), while four were found on samian ware. This proportionality probably reflects ceramic supply patterns, and there appears to be no selection made on the basis of fabric. Determining whether the graffiti were made before or after firing was particularly difficult, since they amount to little more than shallow nicks in most cases. Sixteen were cut after firing, one before firing, and the remainder are undetermined.

Dating is potentially problematic. Most sherds are small and undiagnostic, and the assumption is that they are contemporary with the dated pottery in their contexts. This assumption is crucial to dating the practice; the single grog-tempered example was recovered from an Early Roman context and so is dated to that period, but it may well be residual. Taking the dating at face value, the practice of cutting notches into pottery was undertaken throughout the Roman period at Heybridge. It does not appear to have taken place during the Late Iron Age, despite the earliest evidence for the practice providing a date shortly after the conquest. On the contrary, with the majority of examples recovered from Ceramic Phase 11 (AD 370-410+) contexts, the practice appears to have been predominantly Late Roman.

As with 'X' graffiti, distribution of notched graffiti seems to have been reasonably evenly spread across the settlement. Seven examples were recovered from the northern settlement zone, ten from the central zone and eleven from the southern zone. Expressed as proportions of the total amount of pottery recovered from each zone, this equality remains. For every 217kg of pottery in the northern zone, one notched graffito was recovered. In the central zone, the ratio is 212kg per graffito, while the southern zone has the lowest incidence at 227kg per single graffito.

Notched graffiti are not restricted to Heybridge. At least six notched graffiti have been recovered from Chelmsford (Going 1987, 102; 1992a, 108). Five of these (four dishes and one jar) are either 2nd century or 4th century in date. The remaining example, a platter, is dated mid-1st century AD. The groups of two or three notches were scored mainly on rims. Four notches were cut into a samian dish foot-ring. Examples were also recovered from Colchester. One of these has already been alluded to above. Another is a dish whose rim is scored with four notches (Symonds and Wade 1999, fig. 6.43.83). A third is a cup with at least five notches, also on the rim (Symonds and Wade 1999, fig. 6.94.60). At least five vessels with notches were recovered from Brightlingsea, Essex (Martin 1996b). Just one example shows a complete set of notches: a group of three cut into the carination of a jar (Martin 1996b, fig. 8.18). The wider distribution of notched graffiti is a harder to gauge. While these graffiti appear only occasionally outside the eastern region, it is impossible to establish from published reports alone whether this absence reflects a genuine regional phenomenon.

Explanations for this practice range from the mundane to the ritual. Batch, capacity, or owner's marks remain remote possibilities. It has also been suggested that the notches represent Roman numerals (e.g. Martin 1996b, 315). Though a variety of notch combinations exist - groups of two to five-plus and double groups of two and three notches - this idea is not entirely convincing. Considering the apparent Late Roman emphasis of notched graffiti at Heybridge, these explanations are among the least likely. For the same reasons, other explanations can also dismissed, though perhaps not rejected altogether. These include the possibility that notched vessels were associated with specific contents. Notches perhaps acted as a visual label or enabled identification of the vessel by touch if the user had impaired vision through poor light or blindness. If notched vessels contained harmful substances or were associated with unpleasant activities, then positive identification of these vessels was absolutely necessary. That notches were cut mainly into rims, the most prominent part of the vessel, supports this idea. It is worth noting that dishes and jars were usually the ones chosen, and it may be no coincidence that this practice was principally a Late Roman phenomenon. It has already been suggested that, during this period, dishes and jars shared functions (see above), and the evidence of notched vessels by no means contradicts this view. Whatever functional significance notched vessels had, the function could be served equally well by both dishes and jars.

We cannot fail to consider the possible religious or superstitious connotations that notched vessels may have had. The vessels themselves are mainly ordinary. They were locally produced and common pots. The appearance of the vessel, rather than its functional qualities, does not seem to have played a significant role. These vessels are part of the everyday, and, likewise, any ritual use they may have had was also regular and part of everyday life. Rather than the notches acting as a tactile or visual indicator of vessel contents, they may have reminded the user to give thanks for the items that the vessels held or for the service that the vessels provided. The act of notching might itself have formed part of the ritual, and linked to prayers or sequences of incantations. That the number of notches is commonly three is likely to be significant. Many religious practices involve the number three. In pagan Roman religion, this is perhaps most notable in terms of the Capitoline Triad, comprising Jupiter, Juno and Minerva. However, there is certainly no direct link between the use of notched pottery and formal religious practice. Notably, none was found within the temple precinct. Just one notched vessel was found at the religious complex at Ivy Chimneys, Witham (Turner-Walker and Wallace 1999, fig. 114.29). This was not recovered from the temple complex itself, but instead found within a deposit of pottery kiln waste.

Returning to Heybridge, two examples were found in structured deposits. Boundary ditch 25027 yielded a jar rim with a minimum of three notches, along with a complete flagon, a face-mask flagon and personal items. The second example came from pit 20008. This yielded a collection of eleven complete vessels and substantial portions of at least three others. A jar rim with at least two notches was also recovered (Figure 322, no. 5). The ritual significance of both deposits is discussed elsewhere (see Structured Deposition). In both contexts, seemingly ordinary and mundane broken pottery was included, presumably discarded as rubbish. The notched vessels were incomplete also, and so should probably be regarded in the same light. Considering, too, that no other notched graffito was found in an obviously ritual context, there is no strong reason to assume that the graffiti from ditch 25027 and pit 20008 were deposited in a structured manner. The possibility cannot be ruled out that notched pottery enjoyed special significance during the vessel's life, but lost this connotation when it became discarded and/or broken.

The use and reuse of amphoras

Of all the vessel classes, there is one for which the function can be stated with some confidence, the amphora, whose primary function was as a container for the transport of bulk commodities, such as wine, olive oil, fish products and fruit syrup. This function is borne out by epigraphic evidence in the form of painted inscriptions on many of these vessels, which provide details of their contents and/or destinations. Amphoras, and their contents, were imported into Britain from early in the 1st century BC until late into the Roman period, and also into post-Roman Britain. At Heybridge, small numbers of Dressel 1 wine amphoras first appeared in contexts dated c. 50-30 BC and increased greatly in quantity from c. 25 BC. It is thought that these vessels may have had a particular status or significance in the Late Iron Age, beyond their function as containers for wine, or perhaps, as an extension of that function. This status may have been conferred upon the vessel because of its origins and content, with its symbolic significance as a commodity central to social and political aspirations.

In order to establish whether there is such a status attached to Dressel 1 amphoras, their deposition was examined mainly in relation to that of Dressel 20 amphoras. These were globular carriers for olive oil, and were generally later in date, found in larger numbers from the late 1st century AD. Deposition of the intermediate form Dressel 2-4, mostly used for the transport of wine, was also examined. These vessels superseded Dressel 1, although their numbers at Heybridge are low compared to either Dressel 1 or Dressel 20, as is the case elsewhere. Data are drawn from the stratified assemblage, which includes residual material, but excludes that from cleaning and machining layers. For all pottery types deposition throughout occurred principally in pits and, as the southern settlement zone is characterised by pitting, this is where the highest percentage of pottery was deposited. Very few amphoras were found in positions suggesting continued or secondary vessel use following their primary function as transport containers, which may be surprising given the number of storage jar ovens recorded, see below.

Dressel 1 and Dressel 2-4 amphoras

As can be seen from Tables 38-9 below, the deposition pattern for both Dressel 1 and Dressel 2-4 is similar. Both appear in the same number of feature types (twelve) and a similar percentage of amphoras occur in each feature. A difference can perhaps be seen in the percentages deposited across the settlement. A slightly higher percentage of Dressel 2-4 occurred in the hinterland, due in part to the presence of three of these vessels in pyre-related features. Deposition in all types of funerary feature for both vessel forms is almost identical at 6% for Dressel 1 and 5% for Dressel 2-4.

Conversely, the graph showing deposition by ceramic phase (Figure 328) indicates that Dressel 1 was entering the archaeological record much later than its presumed period of importation. These vessels were imported into Britain during the 1st century BC with a terminal date for their import of c. 10 BC. As can be seen, many of the vessels were deposited in the late 1st century BC (CP1) but deposition in equally large numbers occurred at least until the mid-1st century AD (CP3). This is unlikely to be accounted for fully by the factors governing deposition and residuality. It is noteworthy that a similar situation occurred at Camulodunum, where a large number of Dressel 1 amphoras were found in contexts dated to as late as c. AD 61 (Hawkes and Hull 1947, tables pp. 277-81). The graph for Dressel 2-4 (Figure 329) shows the expected pattern, with maximum deposition occurring during the currency of the form. It is possible that their contents were not always immediately consumed, but it also seems plausible to suggest that once the vessels had been emptied, Dressel 1 amphoras, and to some extent Dressel 2-4, still held a special significance resulting in continued or secondary use as containers.

Direct evidence for Dressel 1 reuse, though, is sparse, with just two potential examples. Firstly, there is an abraded handle stub from pit 15049, although this is residual, perhaps accounting for the abrasion. The second is an amphora spike in pyre-debris pit 15417, the broken tip of which has clearly been smoothed in antiquity (Figure 235, 27). This may indicate continued use as a container, following the original loss, or deliberate removal, of the spike tip. The presence of this vessel, along with two others, in a pyre-related deposit (see Religious and ritual activity) might also indicate a degree of reverence, whereby a damaged amphora was made to look new again before incorporation into the funerary rite. Only the pyre sites and their related pits appear to attest directly to the importance of Dressel 1 amphoras and their contents. Reuse need not have been restricted to funerary or similarly solemn occasions. Continued secondary use simply as a liquid container is feasible, and not necessarily for wine alone, although wine could also have been decanted into the vessel for the purposes of display/status. Import of wine in barrels along with amphoras is attested, but perhaps wine proffered from amphoras was more highly regarded.

Dressel 20 amphoras

As can be seen from Table 40, the deposition pattern for Dressel 20 is unlike that for either of the above types (Tables 38-9). Dressel 20, at twenty-five examples, appeared in twice as many feature types as Dressel 1 (twelve), although this is probably a reflection of the nature of later occupation at Heybridge. Ground disturbance increased through the Roman period, thereby increasing the number of places in which pottery could be deposited and redeposited. This has had the effect of masking the trends of pottery deposition, so that there is difficulty in establishing the length of time vessels remained in circulation. Fewer Dressel 20 amphoras were recovered from pits than either Dressel 1 or 2-4, and consequently more appeared in other features including hearths and foundation deposits. In addition, a higher percentage of Dressel 20 was recovered from the central settlement zone. The reasons for the differences in deposition are not clear.

It is possible to infer that reuse of Dressel 20 might include more mundane activities than for early wine amphoras, besides storage of either liquids or dry goods, including use as hardcore. The fact that Dressel 20 appeared in many more feature types, coupled with an atypical chronological deposition pattern, lends weight to this suggestion. Figure 328 shows that minimal amounts of Dressel 1 amphoras were being deposited after CP3, whereas Figure 330 shows a steady higher level of Dressel 20 deposition, with peaks occurring in CP4, CP8 and CP11. The graph for Dressel 2-4 (Figure 329) accords with that for Dressel 1. It could also be inferred that both Dressel 1 and Dressel 2-4 amphoras were deposited in features that subsequently remained deliberately undisturbed. The location of pyre-debris pit 15417, in particular, seems to have been respected throughout the Roman period.

Amphora body sherds, mainly Dressel 20 types, with large drilled holes, are frequently found on Roman sites, although only a single example, a spindlewhorl, was identified at Elms Farm. These drilled holes are generally considered to have been tapping holes whose purpose was to make the extraction of the contents easier. While many holes may have been made for just this purpose, it is beginning to be recognised that repair of amphoras was also taking place. Sherds with in situ lead plugs were discovered in the civil settlement at Caerleon (Evans 2000a, 296, nos 179, 181). In his discussion, Evans cites an example from London (Marsh 1981, 227), where a vessel comprising 289 sherds held together with 613 rivets was still found to hold water. Amphoras were perhaps designed for storage of liquids, wide-mouthed storage jars being less useful and therefore not selected if amphoras were available. Conversely, amphoras could presumably also have been used to store dry goods.

Evidence for Dressel 20 vessel reuse at Heybridge is also sparse, other than the occurrence of the top half of an amphora buried upside down in pit 4582. The rim had been carefully trimmed off before the top half of the vessel had been placed inverted in the pit. It is not clear exactly what purpose this served, but further examples have been recorded elsewhere (Crummy 1984, 135). As many of those vessels had probably been inverted in the ground, it has been suggested that they may have been utilised as soakaways. Another vessel found at Colchester had been placed upright, minus the neck, in a pit excavated in the passageway between two houses (P. Crummy 1992, 105, fig. 3.54). A further example from Colchester was discovered in 1959 during excavation of the kilns (Hull 1963, 38, figs 10, 15; 134, fig. 74.2). The top half of a Gaulish amphora (Cam 188) was found buried upside down in the west bank of the stoke-yard wall. Its neck had been plugged with lead, and Hull surmised that 'it had no doubt been placed there as some convenience to the potters in their work', although nothing in the contents gave any clue as to the purpose. However, the insertion of a lead plug in the neck might preclude its use as a soakaway. A further published example occurred at Causeway Lane, Leicester (Clark 1999, 122, fig. 66.118); although this vessel seems to have been finally discarded in the back-filled remains of a ditch after reuse (Clark 1994, 10). That it had once been inverted was inferred from the internal staining (Clark 1994, 11).

Discussion

Consideration of the deposition of amphoras in funerary features (Tables 38-40) highlights some differences. There are five occurrences of Dressel 1 in pyre-related features, although three of these appeared in the fills of the pyre-debris pit 15417. In contrast, the appearance of Dressel 20 in funerary features is confined to extraneous sherds, one with inhumation 10776, and another in Roman cremation pit 9665. In all of these occurrences the amphoras are represented in the main by burnt sherds. Complete and unburnt Dressel 1 amphoras were interred with Welwyn-type burials at, for instance, Welwyn Garden City (Stead 1967) and Hertford Heath (Hüssen 1983), both in Hertfordshire. Stead sees these Welwyn-type burials as the graves of Celts impressed and enriched by their contact with Roman (more likely Gallic) merchants (1967, 49). Three amphoras included with the Dorton Mirror burial, Buckinghamshire (Farley 1983), were incomplete; all had had their necks removed in antiquity and the edges had been smoothed. In addition, the broken spike of one vessel had been worn smooth. This treatment is echoed on the spike of the vessel in pyre-debris pit 15417, although this deposit is not an example of a conventionally defined Welwyn-type burial (Stead 1967, 44).

The presence of burnt amphora sherds in funerary features is very nearly unrecorded in Late Iron Age Britain (P. Sealey, pers. comm.), although the practice is attested on the continent, for instance, at Clemency (Metzler et al. 1991). Numerous amphora sherds were recovered from the Lexden tumulus (Foster 1986a, 124), although there is no indication in the published report that these had been burnt. The other grave goods were fragmentary, with the appearance of having been through the pyre; further detailed examination of the amphoras might reveal more clues. The single definite British example of burnt amphoras in a funerary context occurs at Folly Lane, Verulamium (Niblett 1999, 51), although these are from a later burial (c. AD 55) and are Dressel 2-4 (Williams 1999, 193).

If, as seems to be the case, Dressel 1 was often associated with funerary activities, then this practice continued with Dressel 2-4. As the latter was current at the peak of funerary activity at Heybridge, this is unsurprising, but it is all the more notable that Dressel 1 was present in such features. There are two definite occurrences of Dressel 2-4 in pyre-related features (pyre sites 2490, 2908) at Elms Farm. In both cases only well-burnt fragments were recovered, and each pyre site produced a different vessel. Dressel 2-4 occurs elsewhere in funerary contexts, besides the ceremonial site at Folly Lane, Verulamium. At least eleven vessels of this form were recovered from the Lexden tumulus (Williams 1986, 131) and two from the Dorton Mirror burial (Farley 1983, 291), both alongside Dressel 1 types. Further discussion of the presence of amphoras in funerary features at Elms Farm can be found in the amphora report.

Deposition of Dressel 1 in contexts much later than the currency of the vessel has been discussed in detail by Sealey (1985, 101-8) to explain the occurrence of numbers of these vessels at Sheepen, a site founded more than a decade after the demise of the form (Sealey 1985, 26). Several reasons have been proposed, including use as a prestige item for gift-exchange and the ageing of the contents to produce a vintage. All the reasons given are plausible, since occupation in the late 1st century BC at Sheepen can be discounted (Sealey 1985, 105). The problem was also considered by Hawkes, who proposed secondary storage of liquids in order to explain the later deposition of Dressel 1 at Camulodunum (Hawkes and Crummy 1995, 73).

The number of vessels found in funerary contexts is too few to fully account for the phenomenon; just seven, representing approximately 10% of the total Dressel 1 and Dressel 2-4 amphoras found at Elms Farm. At least twelve Dressel 1 amphoras occurred in Welwyn-type burials (Stead 1967, table 11; Rodwell 1976, appendix IIA), with a further nine occurring in less well-defined burials in Essex (Rodwell 1976, 318-20). Examples of both Dressel 2-4 and south Spanish amphoras from similar burials are also listed. Inclusion of amphoras in burials is one manifestation of the high regard these vessels (or their contents) seem to have commanded. This manifestation is highly visible, but does not tell the whole story. Dressel 1 amphoras, in particular, must have had many highly regarded roles for so many to be found in contexts as much as half a century beyond their currency. These roles remain archaeologically undetectable. The evidence at Heybridge, with a start date of c. 25 years before the foundation of Camulodunum, parallels the situation there. The reasons applied at Camulodunum (Sheepen) to explain the phenomenon are equally valid at Heybridge.

Similarly, reuse of Dressel 20 remains archaeologically undetectable, although examples, when found, perhaps demonstrate the more mundane nature of such reuse. Few instances were recovered at Elms Farm, but the pattern of deposition (Figure 330) indicates that Dressel 20 amphoras were not being deposited during the currency of the vessel. A large Dressel 20 body sherd was recovered from shallow pit 13845, unfortunately poorly recorded, but apparently partly lining the cut. The location of the pit (in the central zone) and the absence of other finds, apart from a few iron objects, give no clues as to the purpose. Dressel 20 sherds were utilised as a hearth at Camulodunum (Hawkes and Hull 1947, 107), and the bottom of a globular amphora had been reused as a hearth or furnace base at Causeway Lane, Leicester (Clark 1994, 11). The sherd from Heybridge does not appear to be burnt, however, but is badly fragmented and has shattered and laminated. This could more easily be seen as the result of frost damage.

In conclusion, the evidence suggests that amphoras were likely to have been put to secondary uses once the original contents had been consumed. In the absence of definite examples, it may be inferred that Dressel 1 and Dressel 2-4 had more significance as containers for wine, and their deposition in funerary features reflects this status. Whereas all amphoras might have been reused as containers, either for liquids or for dry goods, Dressel 20 seems to have been less important than Dressel 1 and Dressel 2-4. None was found in funerary features, nor in contexts associated with religious activity - in the temple area for instance. On the contrary, the highest incidence of Dressel 20 amphoras was in the area with the largest concentration of storage jar ovens (see below).

Storage jar ovens

Storage jars are ubiquitous at Heybridge, as elsewhere. These large, heavy vessels were manufactured locally in both grog- and sand-tempered fabrics, although examples made in north Kent and Surrey were occasionally recovered. Not easily moved even when empty, storage jars must have remained immobile when full. Within domestic settings the vessels served as containers, storing dry goods and liquids. The jars might also have been found within industrial environments, storing raw materials or items of trade and manufacture. No storage jar was found in situ within these settings, although a grey ware jar may have been set upright within pit 16108. Storage jars were additionally used as seats of fire for food preparation. Over thirty so-called 'storage jar ovens' have been excavated, the majority of which were located in the central settlement zone. A few were found in the northern zone and on the northerly edge of the southern zone.

The dating of storage jar ovens is problematic. Storage jar fabric - the ware in which these vessels were usually produced - is not intrinsically datable, and, by itself, a body sherd in this fabric could conceivably be placed anywhere within the Roman period. Dating very largely depends on stratigraphy and presence of associated artefacts. The latter alone cannot indicate the period of use of storage jar ovens. In most cases, material found within and perhaps immediately around the oven must have been deposited when the oven fell out of use. However, the earliest dated material can still provide useful chronological pointers. The majority of ovens lie within a late 1st to early 2nd century AD date range, though all were not necessarily in use at the same time. There is none for which a mid-1st century AD date is certain, although oven 6939 utilised a grog-tempered jar (Cam 271) and would appear to be among the earliest of these features, with the fabric being manufactured no later than c. AD 70 (see Pottery supply). It is conceivable, however, that the use of the jar as an oven may have been significantly later than its date of manufacture. Two ovens, 6958 and 5948, contained bead-rimmed dishes dated no earlier than c. AD 125. Oven 6958 also yielded Colchester colour-coated ware, a fabric that did not reach Heybridge in any great quantity until the mid-2nd century AD. The actual use of these ovens need not be dated so late and could still be contemporary with the others. However, the predominance of Phase 6 pottery gives them a distinctly later slant.

Most, and potentially all, ovens fall within Ceramic Phase 5 (AD 80-125), and so use of the feature type is restricted chronologically. Storage jar ovens have been found elsewhere, but these tend to show a limited distribution and short-lived practice. One example, dating to the 2nd century, has been recently found at Witham (A. Robertson, pers. comm.). Three such features were excavated at Chelmsford (Wickenden 1992, 32), and dated there to the mid-2nd century. However, the associated pottery recovered from Hearth 733 at Chelmsford is undiagnostic grey ware, which could in fact date to before this time. The storage jar itself is a Going G44 type - the most commonly represented form of the Heybridge storage jar ovens. While there may be stratigraphical grounds for assigning to it a mid-2nd century date, Hearth 733 reflects the earlier-dated practice at Heybridge.

All but two ovens incorporated sand-tempered storage jars. Apart from the grog-tempered jar noted above, the exception was a shell-tempered jar (Cam 258, Monaghan class 3D4) manufactured in north Kent. That grog-tempered vessels are virtually absent is because production of the fabric terminates before the main period of oven use. The lack of shell-tempered jars is less easily explained, but a reason may lie in the peculiarities of supply from the north Kent industry. Jars of this type appeared relatively infrequently at Elms Farm. That early shell-tempered ware is utilised even once must in itself be unusual, and may reflect the convenient presence of the shell-tempered vessel and a temporary unavailability of otherwise ubiquitous sand-tempered vessels. The commonest jar form utilised is the thick bead-rimmed G44 type - the earliest typologically of the range of storage jars made in the sand-tempered fabric. The hook-rimmed G45 type was also used, though less frequently. This form did not become current until the 2nd century AD (Going 1987, 27).

In order to be utilised as ovens, storage jars were laid on their sides within shallow scoops that ensured stability. Although sections from ceramic vessels are known to have formed hearth or furnace bases, e.g. an amphora at Causeway Lane, Leicester (Clark 1994, 11), most jars at Heybridge were likely to have been placed whole in the ground. This can be attested in some examples. More than half of the early shell-tempered jar in oven 10501 was present and both the rim and base of storage jar 6398 in oven 6399 were complete. Once the jar was in place, it was probably fixed in situ by soil, which might have also aided insulation. Storage jars varied in height, ranging from 0.38 to 0.97m, and averaging 0.62m. Since all vessels were incomplete, volume cannot be accurately measured. However, where the lengths and breadths of jar sections were recorded (measurements being the widest distances) some idea of surface area can be gained by simply multiplying these measurements (Figure 331).

Figure 332 shows that the jars tend to fall into two size ranges clustering on 0.3sq.m and 0.5sq.m. The relatively few occurrences at 0.4sq.m may well suggest deliberate selection of specifically sized jars. It seems unlikely that relatively few jars were made that would fall into the 0.4sq.m bracket while jars immediately smaller or larger were common, and normally jars of this size should be expected. A more even spread of values can reasonably be expected if the distribution reflected usual sizes of manufacture. But what purpose this size differentiation may have had remains unknown. An obvious suggestion is that different sizes suited certain foodstuffs. This cannot be substantiated, of course, without the evidence of residues within the hearths.

With around thirty ovens altogether, the dataset divided into respective areas is perhaps too small to be meaningful. Even amalgamated, Areas D and F in the northern settlement zone comprise just four entries. Area K stands alone in the southern zone and contains three ovens, although it should be noted that these ovens were positioned alongside the road adjacent to the central zone. Areas H and I, in the central zone, contain fourteen and seven respectively.

Rim diameters of storage jars chosen for utilisation as ovens appear to have been reasonably standardised (Figure 333). The diameter of over half the number of measured storage jar rims lies between 350 and 450mm. There was otherwise a wide spread of values, with a difference of 260mm between the narrowest and widest jars. When we compare the mean diameter of storage jars used as ovens to the overall mean diameter of all storage jars in the assemblage, it would seem that larger than average storage jars were deliberately chosen for oven use. The mean for storage jar ovens is 416mm and the overall site average is 323mm. The reason for this is likely to be entirely practical; larger jar mouths were preferable, to facilitate ease of access when cooking. However, it should be noted that the storage jar mean is based on just twenty-three examples, while the site average is calculated from 155 examples.

Storage jar ovens are presumed to have been used to heat food. The inner surfaces of a number of vessels were burnt black, indicating the use of fires, or at least, hot embers. Pebbles were certainly used as pot-boilers (naturally occurring stones used as a secondary heat-source), but were perhaps less likely to burn the sides of the vessel. The floors of ovens 4534 and 6441 were burnt in the centre from rim to base, showing that the heat source was placed along the entire length of the vessels. More unusual is oven 6308, whose sides are burnt internally. A storage jar in oven 6958 is burnt across the width of the top half, with a smaller patch in the bottom corner. Direct heat, as supplied by charcoal or fire, can be assumed. A griddle may also have been used. It has been suggested that bread dough could have been stuck onto the underside of the roof of the storage jar. Once baked, the bread could have been gently prised from the roof, perhaps requiring the use of a paddle-like implement or peel. Oven 16345 comprised a jar that was burnt in the centre. It also yielded pot-boilers, as did two other ovens, 9506 and 13521. Notably, 9506 showed no signs of burning.

Internal discolouration of the storage jar is an immediate effect of the heating process. Over a longer period, or perhaps even after a single firing, the fabric of the storage jar might begin to laminate as a result of the heating/cooling process. Storage jar remains 6192 and 6310, from ovens 6189 and 6323 respectively, are fissured. Heat-induced shattering is less apparent in other jars, but fragmented pottery sherds are perhaps ample evidence of heat damage. Heat-shattering might well have been the root cause of most jars breaking up into sherds. Once use of the jar was no longer viable, the weakened roof of the oven was probably broken up and removed. In some cases, the floor of the jar appears to have remained in situ while the roof and other debris were cleared away. Oven 6462 provides a good example. It contained three individual vessels representing three successive phases of use, damage and replacement.

With little trace of soil and debris between the successive jars of oven 6462, care seems to have been taken upon destruction of one vessel to create a 'clean' surface on which to place a successive jar. It seems somewhat anomalous, then, that lower halves of previous vessels stayed in the ground. These seemingly redundant halves may have acted as ceramic 'plates' to aid insulation and ward off the effects of heating and cooling. But no such 'plates' were found underneath single-phase ovens, such as 5846, 6399 and 16345. If this were indeed their function, 'plates' were required, or opportunistically employed, only to insulate successive phases. Replacement jars would have raised the height of the oven, requiring further soil for packing. It is still possible that, in other instances, damaged vessels were removed in their entirety. Ovens with one apparent phase of use, evidenced by the remains of a single storage jar, may have had, in fact, multiple phases - the jar present representing the final phase of use. Thus, storage jar ovens may have been more numerous. Large amounts of storage jar fabric present in some rubbish deposits might be the cleared remnants from ovens. It is impossible to tell whether their use as ovens represented their primary or secondary function, though, given the damage that the jars would have sustained from repeated heating; it is unlikely that former ovens were put to any other use subsequently. Lastly, the concentration of these ovens in the central settlement zone perhaps indicates that they were not fulfilling a normal domestic function. Proximity to the temple might suggest a religious connotation or, at least, association. Ovens with halved storage jar bases were recorded in the vicinity of the temple at Springhead, Kent (Penn 1964, 173-5). Bread-making in association with the temple is surmised (see Atkinson and Preston 2015).

Face vessels

Ceramic vessels with applied or moulded faces from Elms Farm form a small but distinctive assemblage. Face vessels have been the subject of extensive research (Braithwaite 1984, 99-131; Dovener 2000) and it is not intended to recount their development and distribution here. Instead, a catalogue is presented listing the vessels, followed by a brief discussion of their possible significance and function. The terminology used for vessels with faces is as varied as the faces themselves. Some of the more common labels applied to flagons include moulded-face flagons (e.g. Munby 1975), face-neck flagons (e.g. Braithwaite 1984, 99) and face-mask flagons (e.g. Young 1977, 150). This last seems to have most common currency and is used here. Other vessel types with faces are known collectively as face-pots (e.g. Bidwell and Croom 1999, Cam 288; Braithwaite 1984, 99-131). The label is retained here, but a distinction is made between jars and beakers, separated by shape and fabric. Thirteen vessels were recovered, the majority of which are flagons (seven). Of the remainder, four are jars and two are beakers. Most of the vessels are in fine fabrics, either red or buff. The bias towards a later Roman date for nearly all of the pieces is notable.

Jars

Fill 8737, Pit 8736, Group 902, late 4th century (CP11), fabric GRS (not illustrated)

Just the nose survives from this vessel. A reconstruction of the whole face is impossible, but, at 30mm long from bridge to tip, the nose is of similar size to noses from complete jars from London and Colchester. This face-pot was probably produced locally. The nose is well shaped and naturalistic, with a decidedly 'Roman' profile.

Fill 9180, Pit 9165, Group 783, mid-2nd/early 3rd century (CP7), fabric HAWO (Figure 334, no. 1)

This jar was manufactured at Hadham in a white-slipped oxidised ware. Two body sherds survive; fortunately the face is almost intact, but the shape of the vessel itself is otherwise unknown. The facial features most closely resemble those on examples representing Braithwaite's south-eastern regional groups (1984, 105-10). The eyebrows are heavy and meet in the centre of the face at the top of the nose. The somewhat sleepy eyes are set close to the nose. There is no mouth, but the incised chevron and punched square decoration below the nose might represent a beard.

Fill 16073, Pit 16088, Group 559, early/mid-3rd century (CP8; KPG29), fabric BUF (Figure 334, no. 2)

The face pot fragment has an applied ear and, trailing from the ear lobe, stabbing which represents a beard. The ear is rather more naturalistic than the ears depicted on face-pots of Braithwaite's south-eastern regional groups (1984, 105-10), but resembles that on a buff ware vessel from Colchester (Symonds and Wade 1999, fig. 6.23.658).

Layer 16236, Group 573, late 4th century (CP11), fabric GRS (Figure 334, no. 3)

This sherd is in sandy grey ware, although there are traces of white slip along the rim frill. No face is extant, but there is a V-shaped piece of clay applied just below the rim. The applied clay is broken at the base of the 'V' and must have extended onto the shoulder. The sherd may have derived from a so-called 'smith-pot', with the V-shaped clay representing part of a tool, probably tongs (G. Braithwaite, pers. comm.). Even so, the tool is unusually placed underneath the rim; an arrangement around the vessel girth is more typical (cf. Braithwaite 1984, fig. 10.5; May 1930, fig. 3).

Beakers

Layer 9425, Group 1301, early/mid-3rd century (CP8), fabric HAX (Figure 334, no. 4)

With its narrow neck and frilled rim, this form corresponds to the H17.1 beaker (Going 1987; Bidwell and Croom 1999, Cam 290). In overall size, the vessel may have matched fig. 8.5 in Braithwaite 1984, which is 220mm in length. The face varies slightly from typical examples (e.g. Going 1999a, fig. 5.56.152). The frill does not appear to continue around the entire rim circumference, but instead trails a short way from the bridge of the nose to the eye, and probably represents an eyebrow. The ear-like lug is closed, and cannot seemingly have been used effectively for handling or suspension.

Fill 6314, Ditch 6313, Group 4001, early/mid-4th century (CP10), fabric COLC (Figure 334, no. 5)

This vessel, of which body and base sherds remain, is in Colchester colour-coated ware and thus residual in its context. A bust, incorporating the head and shoulders, is applied above the girth of the vessel. The face itself is plain and nondescript and the shoulders merge seamlessly into the vessel wall. The figure appears to be wearing a cap. The remains of barbotine decoration can be seen to one side of the bust, indicating that the vessel was more fully decorated. This additional decoration and the curve of the vessel profile are suggestive of the H23/H24 range of beakers (Going 1987; Bidwell and Croom 1999, Cam 392). Alternatively, it may be a copy of a cantharus or beaker produced at Trier. A series of faces representing the seven planetary gods are applied around the girths of such vessels (cf. Symonds 1992, pl. 49 and fig. 38.702). Given the relationship between Rhenish wares and the development of Colchester colour-coated ware, this second possibility may well be the correct one. Indeed, the cap itself immediately brings to mind a deity. Mercury is one possibility (C. Wallace, pers. comm.), depending on whether an undetermined detail on the peak of the cap is a wing. Other identifications are equally likely, perhaps more so if the 'wing' is disregarded. Attis, Jupiter Dolichenus and Mithras are all depicted wearing peaked caps (G. Braithwaite, in litt.). Of interest in this respect, is a vessel from Vignes, France. A cap- or bonnet-wearing bust is applied to the girth of the vessel, along with a medallion, a dog, and a snake; all are said to evoke the Mithraic cult (Tuffreau-Libre 1992, 128). Generally, small moulded faces are not unknown, but these tend to belong to jars and decorate the shoulder, neck or rim. An Oxfordshire colour-coated folded beaker found at Catsgore, Somerset, is a rare exception. The moulded face is said to depict Sol Invictus (Young 1984, 26).

Flagons

Fill 10361, Segment 10657, Ditch 25027, Group 838, mid-/late 4th century, fabric OXRC (Figure 334, no. 6)

This is the most complete example of the two Oxfordshire face-mask flagons. The female face has been luted onto the front of the rim. The marks from tools used to smooth the clay are clearly visible. The face closely resembles Young 1977, fig. 53, type C11.4, for which an AD 350-400+ date range has been provided.

Fill 5160, Pit 5179, Group 442, late 4th century, fabric OXRC (Figure 334, no. 7)

The second Oxfordshire face-mask flagon comprises the face only. Its facial features match those of Figure 334, no. 6, above.

Fill 20469, Post-hole 20468, Group 4020, late 4th century, fabric HAX (Figure 335, no. 8)

Of the four oxidised Hadham ware face-mask flagons, this is the most complete. The face is typical of the Hadham form, and is paralleled at, for example, Burgh Castle in Norfolk (Johnson 1983, fig. 39.46) and Colchester (Symonds and Wade 1999, fig. 5.56.155-61).

Fill 10296, Segment 10406, Ditch 25027, Group 838, late 4th century, fabric HAX (Figure 335, no. 9)

A sherd showing face only was the second face-mask flagon to be recovered from Ditch 25027. The face is similar to Figure 335, no. 8, above, if a little narrower. A piece of shell is lodged in the clay above the hair. Shell is not a natural inclusion of the clay from the Hadham area, and so this piece may have been placed deliberately, along with others, for decorative effect.

Fill 14204, Pit 14203, Group 720, late 4th century+, fabric HAX (Figure 335, no. 10)

This fragment represents one of the two strips of 'hair' placed at either side of the head. The piece is clearly residual, coming from Building 65, a Saxon feature.

Layer 21619, Group 446, late 4th century, fabric HAX (not illustrated)

The fourth piece, which might conceivably belong to any of the above Hadham flagons, is a fragment of the 'false handle' normally luted onto the back of the rim (cf. Johnson 1983, fig. 39.46b). The top of this decorative feature rises above the rim slightly, perhaps mirroring handles from metal flagons (cf. Wickenden 1986, fig. 28.23) and may be intended to represent hair at the back of the head.

Layer 10800, Group 8007 unstratified, fabric ?HAX (Figure 335, no. 11)

Most of the face is present, but there is no trace of the hairstyle. The face incorporates both Hadham characteristics (the eyes) and Oxfordshire characteristics (the chin). The relatively coarse fabric is oxidised, but there are no traces of the slip. On balance, the form and fabric is closest to a Hadham product, but it may represent a local version of the face-mask flagon form.

Discussion

This collection of face vessels forms a significant assemblage in terms of variety, and perhaps even in terms of quantity, as is suggested by the comparative data in Figure 336.

With one face vessel for every 260kg of Roman pottery, Heybridge does seem to have a higher than usual number, except for Colchester. It should be noted, however, that these ratios are a crude indicator only, and based on rounded figures, and so the difference between Heybridge, Great Dunmow and Ivy Chimneys is perhaps not particularly significant. More significant, though, are the differences between Heybridge and Colchester at one end of the scale, and Heybridge and Chelmsford at the other. The reasons cannot be determined from these figures alone, but the role of Colchester as a colonia is likely to be crucial. An urban character alone, such as that displayed at Chelmsford, would otherwise seem to have played little part.

Face pots

Face-pots in Roman Britain are considered to have religious significance (Figure 337). Vessels tend to be found in ritual deposits, although they are largely absent from burials (Braithwaite 1984, 124). At Folly Lane, Verulamium, for example, the majority of face-pots, some complete, were recovered from ritual shafts (Lyne 1999, 300-1). None of the examples at Heybridge was found in overtly ritual contexts. Indeed, superficially, the pottery found along with the face-pots remains steadfastly domestic. This much is clear when assemblage composition is compared.

Assemblages recovered from the Folly Lane shafts tended to incorporate more open forms (dishes, bowls, etc.) than jars. Beakers, too, form a large component. This trend is also displayed for Baldock with the assemblage from that ritual shaft (Lyne 1999, fig. 87). In contrast, jars predominate in domestic and industrial contexts at Folly Lane, while beakers take a noticeably reduced share (Lyne 1999, fig. 87).

A similar exercise can be conducted using Elms Farm data. The principal difference is that flagons, cups and other forms (e.g. mortaria and bowl-jars) take a smaller share of the assemblage in 'non-ritual' groups, increasing the overall shares of the remaining forms. Otherwise, the ratios between the major classes are reasonably identical. That flagons form a large proportion of the face-pot assemblages is explained by the presence of a complete flagon rim in the top fill of ditch segment 6313. A ritual element to at least one deposit cannot be entirely disregarded. No. 5 was recovered from ditch segment 6313 along with a near-complete beaker and two vessels whose profiles were complete. But even bearing these in mind, there must be some doubts as to the contemporaneity of the assemblage. The complete beaker, an H41 type in fine grey ware (GRF), is part of a small element of 4th-century pottery, along with an E1 bowl-jar and a G42 storage jar. The remaining pottery is stubbornly late 2nd/early 3rd century (CP7/8) in date, including the Colchester colour-coated ware face-pot. The pottery from this deposit appears to derive from at least two sources, and the deposition of the complete beaker and face-pot cannot be linked with certainty.

The lack of obvious ritual elements in the manner of their disposal does not preclude the ritual use of face-pots prior to disposal. Face-pots that have ritual usage may lose ritual connotations once broken, at which point the vessels are discarded along with mundane rubbish. However, much uncertainty remains about their use, and ritual functions cannot be clearly established. Indeed, Braithwaite (1984, 125) regards their functions as being mutable, changing through time and space. By and large, face-pots are seen as having apotropaic properties (Lyne 1999, 301); they protect the space around them and the contents inside them. The face, too, might be retained after the actual pot has broken in order to be carried around as a keepsake or talisman. That face-pots at Heybridge were recovered from mundane deposits suggests 'everyday' above-ground use. Propitiatory or funerary functions, and other ritual acts requiring burial, can be reasonably ruled out.

Two face-pots at Heybridge may be directly associated with specific deities; No. 5, possibly Mercury or another cap-wearing deity, and No. 3, tentatively attributed to a smith-deity. Manufacturers of the remaining face-pots may have had no particular god in mind, and that, on production, the pot had no fixed superstitious connotations. The pot was, then, available for people to acquire and give their own meanings, perhaps attributing it to a local deity, or a spirit connected to the home. The H17 type beaker made at Hadham (e.g. No. 4), which has widespread distribution across the East Anglian region, may well suggest this. If the face applied to this type represents a deity, then it is one with an equally widespread recognition. Alternatively, the face is a 'blank', ready for whoever acquires the vessel to personalise with the meaning/deity of their choice.

Face-mask flagons

If the evidence for the ritual use of face-pots is unconvincing, then that for face-mask flagons is clearer. That surrounding four of the six flagon examples is particularly strong. Numbers 6 and 9 were recovered towards the eastern terminal of major boundary ditch 25027. Material evidence from this feature is prolific. Excavated segment 10296, which contained No. 9, yielded a collection of twenty-five coins, mainly 4th century. A diverse array of metal, glass and stone items was also recovered. Segment 10361 contained a smaller, but nevertheless similarly diverse, material assemblage. Accompanying No. 6 was a complete Nene Valley colour-coated ware flagon (Figure 299, no. 99) as well as further glass, stone, and metal objects.

Their placement close to the ditch terminal flanking an approach road to the centre of the settlement, and to the temple complex, is perhaps significant. The plethora of small portable and personal objects within the ditch suggests concerted and defined activity, and the deposition of personal effects on approaching (or leaving) the religious focus may be seen as a propitiatory act. The complete Nene Valley ware flagon can be included with this rite, but the deposition of the face-mask flagons is a subtle variation, because these vessels were not whole. Number 9 comprises the face only, while 6 has the complete flagon rim and upper part of the handle only. Careful examination of the other pottery within their respective contexts produced no sherds that could conceivably belong to these vessels. It is a strong possibility, then, that the faces were deliberately removed from the main body and deposited. That the depiction of the face was significant in the past has long been discussed. Braithwaite (1984, 126) emphasises the apotropaic properties of faces, some of which continue to resonate in modern times. S. Esmonde Cleary (pers. comm.) has suggested that face vessels deposited into deep, wet, ground-penetrating features (wells and shafts) stood proxy for human skulls, and were thus linked to funerary rituals. In any case, the ceramic face either takes on naturalistic abilities (the eyes 'see' and keep watch), or it is imbued with the spirit or character of a particular individual. The face-masks deposited in the boundary ditch which, at the time, may have demarcated land over which the temple had jurisdiction, perhaps guarded against bad fortune or evil spirits transgressing the boundary. This function may be seen as being different from that governing the deposition of coins, bracelets and the like. These are overtly personal objects, and represent personal investment and the sacrificing of material wealth, perhaps for the return of good fortune.

Unlike face-mask flagons, face-pots at Heybridge have few associations with personal items, such as coins and jewellery. Face-pot assemblages are weighted more towards consumable items, dominated by food and structural waste. This seemingly underscores the differences between the deposition of face-pots and face-mask flagons at Heybridge, and emphasises differences in use: face-pots having above ground, whole-vessel significance, and face-mask flagons additionally having a ritualised manner of disposal.

Two face-mask flagons are associated with structural remains. Number 8, of which most of the rim and neck is present, was recovered from post-hole 20468. No other material was found in association. Its relative completeness suggests that the flagon was carefully and deliberately placed into the constricted aperture of the post-hole. It takes a considerable stretch of imagination to regard its deposition as being accidental, and the lack of other finds suggests that the flagon did not derive from a rubbish-rich backfill deposit. Face vessels as foundation deposits are well attested, and tend to be associated with religious buildings (Braithwaite 1984, 123). No clue about the use of the structure that post-hole 20468 formed part of can be provided, but there is sufficient reason to suggest that this flagon, too, represents a foundation deposit. Like the faces that watched over the settlement centre, the face here watched over and protected the building and its occupants.

The ritual significance of No. 10 is less convincing. It was found within the backfill of Building 65. It is not immediately apparent that the face-mask fragment was deliberately selected for deposition, not least because the feature was an Early Saxon sunken-featured building and the pottery found within it would have been deposited during the post-Roman period. If deliberate selection had occurred, then this demonstrates either continuity of a Late Roman ritual practice, or selection using other, purely Saxon, criteria. The latter is more likely. No part of the face actually remains and the piece could not have functioned spiritually in the way that others with faces might (e.g in ditch 25027). But, associated with the flagon fragment was a near-complete stamped bowl (Figure 296, no. 35). Together, the bowl and flagon may have been regarded by Saxon incomers as odd-looking pottery and curated for the connotations that such pottery may have brought. That the vessel was once a face-mask flagon was irrelevant and perhaps even unknown to the Saxons. Of course, its presence may well be accidental. The fragment is a piece that was applied to the side of the face. There is an inherent weakness in the method of its fixing, and thus might be one of the first pieces to break off the vessel (and be thrown away) during use, and its presence in the pit could simply be explained through the usual taphonomic processes of rubbish disposal and relocation.

The two remaining face-mask flagons were both recovered from the temple area; this fact alone might give them particular significance. One of the flagon fragments was found in a layer, and it is not possible to tell whether it derived from an underlying feature or was simply deposited in the layer itself. Number 7 was recovered from pit 5179, located on the northern precinct boundary, which contained a range of other 'votive' material, again bringing to mind the role of spiritual protection.

Face-mask flagons were not eminently practical, and users of face-mask flagons placed in domestic settings might have experienced difficulties. There is no neat spout over the coiffure of no. 6, and the application of the mask has resulted in an internal bead that would have displaced the flow of liquid. Liquid poured from this flagon would be as much over the table as in the cup. The scalloped head-dress or diadem rising above the rim of no. 8 cannot have enhanced pouring. Hadham face-mask flagons like this one have a handle on each side of the vessel. Conceivably, liquid could have been poured from the back of the flagon, with the applied 'false-handle' or hairpiece functioning as the spout.

In conclusion, there appears to be a major difference in the way that face-pots and face-mask flagons were deposited. If both vessel types had similar sorts of significance when complete and above ground, then for face-pots these ended when the vessels were broken. The sherds were discarded along with the rest of the household waste. Face-mask flagons, in contrast, retained their significance, or perhaps acquired new meanings, and were deposited carefully and deliberately. For most vessels, however, this significance is more likely to be superstitious than religious, and there is no clear evidence to suggest that the vessels were used in formal religious activities at Heybridge.

Pottery with perforations

Vessels that have been pierced either before or after firing (and before or after breakage) have long been recovered from excavations. Perforations can occur as either single or multiple instances; the latter can be grouped or spread over the surface of the vessel. Very often these perforations have been dismissed either as repair holes or the like, or vessels have been given catch-all labels such as sieve/strainer, colander or cheese-press. With little further thought, they are occasionally named 'incense burners' or even 'wine coolers'. Analysis, particularly of vessels with post-firing holes, is informative and a variety of alternative interpretations is proffered below, with reference to the Heybridge examples. A large number of vessels with either pre- or post-firing holes were recovered, enabling the gathering of sufficient data for a wide-ranging study. In particular, the number of vessels with post-firing holes, in excess of 150, presented the statistical data for a comprehensive investigation of the phenomenon. The following study is based on the data for vessels in coarse fabrics and selected fine wares; data for samian vessels were not used.

Pottery with one or more piercings can be divided into several groupings, the principal division being vessels with holes made before firing and those made post-firing. Perforations made while the clay was malleable probably indicate that the vessel was manufactured to perform a specific function. Those made after the pot had been fired might have been for a variety of reasons and examples of these can be found below. Many works were consulted during this analysis and opinions as to the functions for pierced vessels are diverse. A list of comparanda appears in the research archive, along with full details and measurements for all of the vessels found at Elms Farm. A report by Dr Paul Sealey on the spouted strainer-bowls, and a set of general notes on cheese-presses and funnels by Colin Wallace, can also be found in the research archive. This study has benefited from their contributions.

Frequency and zonation

The number of vessels with post-firing perforations far exceeds those that were purpose-made. This might indicate difficulty in obtaining function-specific vessels, but more likely demonstrates the ease with which locally made pottery could be adapted to suit the purpose required. The distribution of both purpose-made and utilised vessels follows a similar pattern (Table 41). Three-quarters of the total were found in the southern and northern settlement zones. Few were recovered from either the central zone or the hinterland, probably indicating that the activity represented by the use of pierced pottery was largely domestic in character.

Pre-firing perforations

Strainers and strainer-bowls

Vessels with multiple holes pierced before firing are perhaps easiest to assign to a function. In general, such holes are small in diameter and closely spaced, perhaps indicating the ease with which soft, or even leather-hard, clay can be pierced with a pin/needle, sharpened twig or nail. The majority of these vessels may be considered to be purpose-made strainers, comprising a group of mainly bowl-shaped vessels with a panel of holes that is usually confined to the base (Gillam 1970, Type 348; Cam 298; Going 1987, Type M2). A second group, allied to these simple strainers, comprises spouted strainer-bowls (Cam 322 and 323; Going 1987, Type M1). These vessels incorporate a pierced strainer-plate, which is applied to the vessel wall behind the spout. Once separated from the parent vessel by breakage, these strainer-plates can cause difficulties in identification owing to the similarity with the panels from M2 strainers. Generally, M1 spouted strainer-plates are roughly finished on the side covered by the spout, as, during manufacture, the panel appears to be pierced after its attachment to the vessel, and is an internal surface not meant to be visible. In contrast, the panels from M2 strainers are usually well finished, internally as well as externally, as both sides of the strainer-panel are on view. Confusion with funnels (see below) can also arise when spouts become detached. This problem is harder to resolve, unless identifying details survive.

In her survey of Late Iron Age pottery, Thompson (1982) describes two forms of strainer, both in grog-tempered fabric; S1 Strainers (1982, 559) which are equivalent to Going's M2, and S4 Strainer-spouted bowls (1982, 567) equivalent to Going's M1. Fragments of both types of strainer were found at Elms Farm. As a class, M1 strainer-bowls comprise two vessel styles represented by Cam 322 and Cam 323 (Hull 1958, fig. 121; 1963, fig. 105). Although there is substantial variation in detail within both styles, the broad categories are useful because they derive from two quite different metal prototypes, a rounded and a carinated form respectively (Sealey 1999, 119, 121). Carinated spouted strainer-bowls are current mainly during the Late Iron Age but the devolved, rounded form continued into the 2nd century. Sherds of both forms were identified at Elms Farm, plus fragments of strainer-plates that could have come from either vessel type, amounting to a minimum of eleven vessels, all of which are grog-tempered. Spouted strainers are not common, although examples have been found elsewhere in Essex, notably at Sheepen (Niblett 1985, fig. 33) and Ardleigh (Sealey 1999, fig. 82). A rounded strainer-bowl was found at Langford Junction, Heybridge, in the 19th century, and incorporated into the Fitch Collection (Wickenden 1986, fig. 26.31). The vessels from Elms Farm are all from 1st century AD contexts, and the two rounded strainer-bowls (Figure 338, nos 1, 2) are the only examples of the Cam 322 family to come from Late Iron Age contexts in Britain (P. Sealey, pers. comm.). An example of the rounded type was found at Chelmsford in a context dated AD 120-40 (Going 1987, 35, fig. 26.189).

Spouted strainer-bowls have long been regarded as British accessories at wine-drinking ceremonies, but recent evidence has demonstrated, rather, that a form of so-called Celtic beer was being prepared (Sealey 1999, 122-3). This drink is attested in texts such as the Vindolanda Tablets (Bowman and Thomas 1983, 91, no. 12), and known in antiquity as cervesia. The lack of associations between strainers and wine amphoras has been documented (Sealey 1999, 123), lending weight to the view that the drink served was a native brew. Only three of the ten Elms Farm contexts with spouted strainer-bowls also had wine amphoras; indeed most were accompanied by other grog-tempered vessels. It is also significant that there were no spouted strainers among the impressive suite of imported fine wares, Dressel 1 wine amphoras and grog-tempered vessels from pyre-debris pit 15417. Recent work using correspondence analysis has additionally demonstrated an interesting association between spouted strainer-bowls and butt beakers (Pitts 2005, 155; fig. 4b). Butt beakers would have been ideal for drinking ale, and were a very common 1st-century AD vessel class at Heybridge, as elsewhere.

Class M2 strainers are found more frequently, although there were no intact or even substantially complete examples at Elms Farm. These vessels usually take the form of shallow wide-mouthed bowls, sometimes with round or sagging bases, but more often flat-based. The type-vessel for Cam 298 (Hull 1958, fig. 67.89) has a flat base, an out-turned rim and grooves on the shoulder. A similar vessel, with a sagging base, appears in May's catalogue of Roman pottery from Silchester (1916, 119, Type 70), and the vessel illustrated by Gillam (1970, Type 348) has a hemispherical profile with a thickened, slightly out-turned rim. The piercings of the latter extend beyond the base to cover the bottom half of the vessel. The recovery of intact vessels is relatively rare and, unless fragments of the strainer are found, full identification is difficult. Isolated rim and upper wall sherds would probably be classed as bowls. Unfortunately, fragments of strainer-panel do not in themselves indicate the full form of the vessel. This is the case at Elms Farm, where eleven fragments of strainer-panel were found. Ten of these are flat, some with the lower wall of the vessel still extant, and the eleventh is curved. Enough survives of the walls of six examples to indicate that they came from bowl-shaped vessels. The curved sherd could have come from either a sagging base, such as the Silchester Type 70, or from a hemispherical vessel like Gillam's Type 348. Although this type of strainer is relatively common, there are few published examples, perhaps demonstrating the probable fragmentary nature of such vessel finds. There is an incomplete example in the Fitch Collection from Heybridge (Wickenden 1986, fig. 26.38) and a grog-tempered bowl, half-complete, came from Ditch CF101 at Woodham Walter (Rodwell 1987a, fig. 22.146). M2 strainers are a long-lived vessel class, occurring throughout the Late Iron Age and Roman periods.

Funnels

A vessel class that may be allied to strainer-bowls, and which can also be misidentified if fragmentary, is the funnel (Going 1987, 35, Class N). A detached funnel spout could be mistaken for the spout from a strainer-bowl, rimless spouts probably mistaken for flask fragments or flagon necks, and rim and upper wall sherds could be classed as bowls. Their function seems to be straightforward, but funnels could have been used in conjunction with cloth, or other organic material, to act as strainers. Going divides the vessel class into four types, although the N1 funnel depicted could be a handled bowl (1987, fig.18) and there is very little typological difference between the remainder (N2-N4). Other published examples are rare; a near-complete example, in Verulamium Region parchment ware, was found at King Harry Lane (Rigby 1989, fig. 36.12). A second near-complete funnel was found at Baldock (Rigby 1986b, fig. 134.386). Interestingly, this vessel is in mica-dusted ware, normally taken as imitation of a metal prototype. A variation of the form, in hard buff ware, was found at Richborough (Cunliffe 1968, pl. lxxvi, no. 594). This vessel has an internal perforated strainer-plate attached at the junction between the body and the spout. One of the funnels from Elms Farm (Figure 338, no. 5) has an attached strainer-plate, but this example is in grog-tempered fabric. No complete funnels were found, but fragments from thirteen different vessels were identified, one in Colchester buff ware, six in grog-tempered fabric and the remainder in black-surfaced ware. It seems that funnels had their origins in the Late Iron Age, although Thompson (1982) does not include funnels in her corpus. The form seemingly continued late into the Roman period.

Considering the lack of published examples, and thus implied rarity as a vessel class, it is noteworthy that so many examples were found at Heybridge. The number of funnels, in conjunction with the relatively large number of strainer-bowls, suggests perhaps that ale-making was a common activity. There is a fragment from a curious cone-shaped vessel (Figure 338, no. 10), which may have been crudely made to perform the function of a funnel. Presumably, fragmentary vessels, such as flagon necks, could have been used as makeshift funnels, too. A coarse grog-tempered jar found at Kelvedon (Rodwell 1988, fig. 84.139) had the rim and base removed and was then carefully filed down perhaps to make a funnel. An enigmatic sherd from Colchester in coarse oxidised ware, illustrated by Symonds and Wade (1999, fig. 6.28, no. 822), may belong to this vessel class. There is no description, but the illustration implies use as a funnel at least after breakage, if not before. Funnels, and makeshift funnels, may have been used to transfer liquid from barrels into narrow-necked vessels, or for decanting wine from amphoras into flagons. Jars with holes made in their bases after firing (see below) could also have been utilised in this way, either as strainers or funnels, or as a combination of both.

Cheese presses

A further vessel type that has pre-firing holes is the cheese-press or cheese wring (Gillam 1970, Type 350; Cam 199), six examples of which occurred at Elms Farm. These are usually thick-walled and flat-bottomed, with internal concentric ridges and channels in the base, and a high central point. They are pierced at intervals along the bases of the channels, and also in a row along the mid-point of the vessel wall. Flat, correspondingly pierced and ridged discs, known as cheese-press lids, were also made. Neither type of cheese-press is common in Britain, so the six identified examples represent a relatively large collection, in Essex second only to the number found at Colchester. There are four instances of the vessel itself, two in black-surfaced ware and two in Colchester buff ware, the latter pierced with square holes. There are two examples of the cheese-press lid, also in black-surfaced ware.

Seven cheese-presses, all in buff ware, were found at Camulodunum (Hawkes and Hull 1947, tables pp. 277-81). None was retrieved from pre-Neronian contexts (Hawkes and Hull 1947, 256). Part of a typical base in red-buff ware, with concentric ridges and perforations, was found in the kiln excavations of 1959 (Hull 1963, 134). Further examples have been found in the colonia, mainly in buff ware and most commonly in 2nd century levels (Symonds and Wade 1999, 476). Most cheese-presses made in the Colchester kilns are pierced with circular holes (Symonds and Wade 1999, fig. 6.27, nos 797-802), although some are pierced with square holes, probably by using a nail (fig. 6.27, nos 803-8). It would appear that these vessels were manufactured specifically for the inhabitants of the colonia, probably indicating a military association. The main period of use was between the mid-1st and early 3rd centuries, perhaps demonstrating that the cheese-press was a fully Romanised vessel class. Cheese-presses were also made at Ardleigh, near Colchester, mainly in Roman grey ware (Going 1999b, fig. 103, nos 137-40).

Examples of cheese-presses from elsewhere in Britain mainly follow the mid-1st to early 3rd century dating of the Colchester vessels. At Dragonby, Lincolnshire, Swan noted that, in the latter part of the 1st century AD, the use of ceramic cheese presses was not intrinsic to the local native traditions of preparing and serving food (May et al. 1996, 576). Pottery cheese presses seem to have been an early Roman, probably military, introduction, and the evidence at Elms Farm does not conflict, since the stratified examples come from contexts of mid-1st to mid-3rd century date. Later production of the form is attested at the Oxfordshire kilns (Young 1977), but their distribution does not seem to have been widespread. There seems little reason to doubt their function as cheese-presses. The diameter and hole-spacing of Roman pottery vessels accords with those of the post-medieval chessil or chesset (Vince 1982, 46-7), although post-medieval vessels are wooden. The probability is that the majority of Roman cheese-presses were also wooden, perhaps thus explaining the apparent rarity of pottery examples.

The purpose-made cheese-press fragments found at Heybridge demonstrate the diversity of form and detail that can occur with these vessels. The vessels may have been used to produce small 'family-sized' cheeses, perhaps using milk from goats or sheep. It seems unlikely that the flat, pierced discs commonly known as lids would have performed such a function, since the vessel holes allow excess liquid to drain off while a weight is applied to the cheese from above. It is possible that the discs were used to press cheese in conjunction with other vessels, jars or wooden bowls perhaps, or used for stacking small cheeses during pressing. The necessity for pottery cheese-presses to have internal ridges cannot be satisfactorily explained. The comparable wooden, post-medieval chessets appear to have plain, smooth, internal surfaces. Frequent cheese-making at Heybridge was highly likely, given the obvious presence of sheep/goats (see Animal bone report), and the range of domestic, manufacturing and preparation activities attested in this essentially agricultural settlement.

Other vessels

Other vessel types with pre-firing perforations are uncommon site finds, and none was definitely identified. These vessel types include triple vase cups some of which have a pierced plate between the cup and the base-ring (Gillam 1970, Type 344; Cam 494), and bowl-shaped vessels with an internal pierced flange, made in fine fabrics, variously described as colanders, 'wine-coolers' or 'incense-burners' (Gillam 1970, Type 349; Cam 387). The Cam 387, in particular, is usually found on sites with Roman military associations and, coupled with the rarity of this vessel type, it is unsurprising that none was found at Elms Farm. The choice of fabric, coupled with its rarity, might indicate specialised tableware not in common usage.

Post-firing perforations

In comparison with vessel types with pre-firing holes, those that have been pierced after firing are perhaps more difficult to allocate a definite function. They are also more frequent. The excavations produced more than 150 examples of vessels pierced in this way, for which detailed data are held in the archive. Conclusions cannot be drawn from the many pierced sherds that are now isolated from the parent vessel, giving little clue as to the reasons for perforating them. However, the assemblage contains sufficient substantial fragments and reasonably complete vessels to allow several observations to be made. These observations go some way towards an understanding, or at least an appreciation, of this phenomenon. Most vessels are pierced through the base (65%), the fewest through the neck, below the rim (8%). Of the holes that appear in body sherds, very few could be assigned to the shoulder of the vessel (seven out of forty-two examples), although the highest individual hole-counts recorded were in the shoulders of jars (twelve and thirteen holes respectively). The probable reason for this high figure in vessel shoulders is discussed below.

Single holes were the most frequent at 72% of the total pierced vessel assemblage, with single holes in bases being the most common of these. The fragmentary nature of the assemblage is demonstrated in Table 42 below, where it can be seen that most body and rim sherds recovered have single holes only, giving no indication of the number of holes once present in the complete vessel. Entire bases with multiple holes are relatively rare, with only twelve definitely having three or more holes, although the large number of fragmentary bases makes the calculation of numbers difficult. At least forty of the part-bases examined could have had more holes than were extant, and a significant proportion, almost half of the total, obviously did. Bases with multiple holes are thus certainly under-represented in the sample. The number of holes per vessel, as found, and their position on the vessel, is shown in Table 42. The figures represent the holes in complete vessels or whole bases; for these the exact number can be given. Brackets around a figure indicate the number of holes extant in an incomplete base or in a sherd, i.e. the minimum likely to have been present when the vessel was whole.

Where measurable, hole diameters seem to show some patterning (Table 43). Holes in bases tend to range between 2mm and 10mm, with occasional examples in the range 11-20mm. Two clusters form around 5mm and 10mm. Holes in body sherds are heavily biased towards the lower end of the range, with 26% measuring 3mm, and an additional even spread between 5mm and 19mm. Few holes made in the neck of a vessel exceed 3mm, although only twelve examples of pierced necks were recorded. Very nearly all perforations (c. 90%) seem to have been drilled from outside the vessel, with very few instances from the inside. For a complete vessel, drilling from the inside must have proved difficult, unless the vessel had a wide mouth. Just ten of the vessels in the study had holes made from the inside, eight of which are single and central in bases. Holes made from the inside are perhaps more likely to indicate utilisation after vessel breakage.

Besides circular, or near-circular, holes there are twelve examples of jars with a single, large, sub-square perforation made in the base. These perforations vary in dimensions from as little as 16 x 17mm to as much as 40 x 48mm, with an average of 22 x 27mm. Quite how these holes were made is unclear, and it is not known how easily a hole can be knocked through a jar base without breaking the vessel. The base may have been drilled first, in which case some of these sub-square holes could have been formed through breakage and the vessel may have had several holes originally. However, this possibility appears not to be the case for every example; there is evidence that the majority of these holes have been made carefully with a pointed instrument. Even within the single category of holed bases, there is likely to have been a diversity of intended use. This is demonstrated in the wide range found in count, size, shape and configuration of the perforations.

The distribution of pierced vessels over the settlement is uniform, but with slightly more appearing in the easterly part of the southern settlement zone than elsewhere. This is accounted for, in part, by the number of pierced jars (at least twelve, discussed below) that were found in the fills of a single pit, 17177. Table 44 shows the distribution of post-firing perforated vessels by settlement zone. It can be seen that the highest proportion by far occurred in the southern settlement zone, which is characterised by large numbers of pits and thought to be essentially domestic in nature.

Pierced sherds occur in a number of different feature types, with more than half coming from pit fills and 15% coming from ditches and gullies. Within these two classes of feature, more than half occurred in top or single fills, with the highest proportion within intermediate fills of pits. Layers and spreads accounted for a further 15% of pierced sherds and another 13% were found among the unstratified material. The variety of feature types coupled with the concentration of pierced sherds in pits of the southern zone suggests that the deposition of most pierced vessels is likely to be associated with domestic rubbish. The fragmentary nature of the pottery, the number of grog-tempered pierced vessels found in contexts of later date than the currency of the pottery, plus the high percentage recovered from top and single fills, lends weight to this suggestion. The distribution through time reveals a higher percentage of pierced vessels in the Late Iron Age and Early Roman period, with 77% either occurring in contexts dating up to the late 1st century AD or comprising grog-tempered vessels of intrinsic Late Iron Age date.

Vessels with post-firing holes are common site finds, with a widespread distribution in Britain, but very many are dismissed with little further comment (e.g. at King Harry Lane; Rigby 1989, 203), particularly if not interpreted as repairs. A large number of vessels were recorded in Roman Colchester (Symonds and Wade 1999), where they are described as sieves or strainers (e.g. 1999, 416; fig. 6.82). At Kelvedon several pierced vessels are illustrated, including three recovered from 1st and 2nd-century graves. All three contained the cremated bone, one jar had four holes drilled in the base and the other two jar bases each had a single central hole made after firing (Rodwell 1988, figs 87 and 88). Further afield, vessels with post-firing holes were found in the Late Iron Age cemetery on the route of the A27 bypass at Westhampnett, West Sussex (Mepham 1997, 130). The jar in Burial 20384 had two holes just under the rim, one either side of a dunting crack. Another, from Burial 20029, had a row of repair holes on either side of an ancient break. Finds of pierced vessels found in graves are commonly assigned a ritual function in themselves (e.g. Great Dunmow; Going 1988, 23), although closer study might reveal a more mundane reason for the piercings. This is evidently the case with the pots from Westhampnett, at least, where repair is indicated.

It can perhaps be seen that pierced pots in burials do not necessarily hold any particular significance. The selection and deposition of functional vessels in graves is common practice. At King Harry Lane, the obviously repaired vessels outnumber those pierced with just one or two holes each. Even these holes could have been functional; those near the rim used for suspension, for instance. This is true of the bowl from Burial 20451 at Westhampnett (Fitzpatrick 1997a, fig. 91); the vessel could quite easily have functioned as a cheese-press before deposition in a grave. The perforated cinerary containers at Kelvedon (Rodwell 1988, fig. 87, G5, G14; fig. 88, G74a) may also have held no additional ritual significance as a result of being pierced. There is evidence that both cinerary containers and accessory vessels were not necessarily new, unused vessels. Many have external sooting, or interior lime-scale, and the fact that repaired vessels are routinely found in graves may demonstrate a prior role. That a pierced vessel may have had a mundane function did not preclude its use in a funerary context, although pre-existing holes may have influenced its selection. By analogy, it can be suggested that the piercing of any vessel had little religious or superstitious significance, although ritual aspects cannot be entirely ruled out.

Although large numbers of pierced vessels are found on sites throughout Britain, detailed studies have not normally been carried out. A variety of functions has been postulated in several reports, ranging from the mundane - strainers, sieves, funnels - to the more imaginative - chafing dishes, braziers, flower pots and beehives. Pierced jars found in an abandoned fish-pond at Shakenoak, Oxfordshire, were thought to have been associated with fish-farming (Brodribb et al. 1978, 18; Hands 1993, 154). More recently, at Brightlingsea, Essex, the lower half of a flagon with a pierced base and regularly spaced holes along the girth was thought to have been 'ritually killed' (Martin 1996b, 313). However, this modified vessel might just as easily have functioned as a cheese-press or strainer. Establishing ritual activity is very difficult unless it can be demonstrated that holes were made at the time of deposition, or other factors are present that would indicate a ritual deposit. That pottery played a part in ritual deposition is not in dispute, although the frequency of this activity is not likely to be high. Of the 150 or so pierced vessels identified at Heybridge, only six vessels (less than 4% of the total) could be reasonably interpreted as having been holed ritually.

A deposit with undoubted ritual connotations has been found at Dovehouse Field, Cressing Temple, Essex (Bennett 1999, 218). A number of vessels, along with loomweight fragments and apparently selected animal bones, were found in the terminal of a Late Iron Age ditch. A large, near-complete tazza-bowl (Cam 210) from the group has a square hole cut in the centre of the base. The vessel had evidently been well used; the exterior is sooted and the interior is coated with lime-scale. The hole had been cut through the lime-scale, but the edges of the hole remain scale-free. The pot had clearly been deposited not long after the hole had been cut through the base. The composition of the deposit, coupled with its location in the terminal of the ditch, indicates a likely ritual function both for the deposit and for the pierced pot. A tentative case could be made for suggesting that single, large, sub-square holes cut into the centre of vessel bases at Heybridge are deserving of a similar explanation. Three of the jars found in ditch 25274, see below, each had a single, large, central hole in the base (Figure 354, nos 1-3), and each was likely to have been intact when buried. The single central perforation in the base of the cinerary container in Grave 74 at Kelvedon (Rodwell 1988, fig. 88) appears to be large, and a section of the rim is also absent (Rodwell 1988, 119). An accessory vessel in the same grave, flagon 74b, has a large circular hole in the side (Rodwell 1988, fig. 88). There is difficulty in assigning a mundane function to these perforations and ritual defacement could be considered for the holes in both the flagon and the urn.

The consensus from a number of reports seems to be, however, that the majority of vessels with post-firing perforations in the lower body were used as strainers or in cheese-making. Jars with bases pierced with a single central hole could perhaps have been used as funnels. The contextual evidence at Elms Farm certainly supports a view that the vast majority were use-vessels; there are few examples where a ritual function could be substantiated. Exceptions that might have ritual significance are the two pairs of jars from ditch 25274, a jar from pit 20008 (KPG17), and a near-complete Cam 204 jar recovered from pit 14579 (see Catalogue below). The latter has a large irregular hole at mid-girth, which appears to have been made deliberately rather than accidentally (Figure 339, No.26). The pedestal may also have been deliberately trimmed away, although pedestals were luted onto the body and this may simply have become detached.

The deposit of two pairs of jars in ditch 25274 is noteworthy, however. Each pair was buried in an upright position, approximately 2m apart, apparently intact except for the large central piercing in the bases of three of them, and the small hole to one side of the base of the fourth. It is not certain that the perforations were part of the procedure that led to the jars being placed in the ditch. Three sets of paired ceramic vessels, also buried at 2m intervals, in a ditch at Orpington, Kent (Merrifield 1987, 38, pl. 13), indicate that this is not an isolated phenomenon. Most of the pierced vessels from Woodham Walter, Essex, had single central holes in the base, including two in the large ceramic assemblage recovered from Ditch CF101 (Rodwell 1987a, fig. 22, nos 133, 141). This assemblage is now thought to represent a closure deposit or rite of termination (Wallace 1989, 172), although Rodwell considered that the deposit may have been the result of a 'domestic calamity' (1987a, 39). The two pierced vessels appear to have lost separate ritual significance against the mass of other pottery buried with them.

Re-cataloguing of the Silchester Collection in Reading Museum (Fulford and Timby 2001) has added to the debate, since most of the vessels in this collection were complete or semi-complete. Most of the perforated vessels were jars, flagons or beakers, and a large number were pierced through their bases. However, because of the completeness of the vessels, it could be seen that many had a hole or holes at the mid-girth. A variety of uses for vessels pierced in this way are suggested, specifically noting that vessels with holes were not rendered useless. The authors argue against piercings having ritual connotations, preferring mundane uses for most instances (Fulford and Timby 2001, 297).

Although many of the previously postulated functions for holed vessels are unsatisfactory, specific functions for the pierced vessels found at Heybridge are hard to define, mainly because of the lack of conclusive evidence once pots are discarded. Repair holes are perhaps the easiest to categorise, although the smaller the vessel fragment, the more difficult this becomes. When a large part of the vessel is present (for instance Figure 339, nos 24, 25), pairs of holes on either side of a crack are easily identifiable as repairs. The shoulder hole in the jar from pit fill 4140 (Figure 338, no. 23) is probably from a repaired vessel, although only a small part of the jar remains. A hole pierced in the wall of a terra nigra platter (cleaning layer 8500, not illus.) is also likely to have been for a repair. These holes are close to a break, which may be along an ancient fracture, the sherd with the opposing repair hole was not recovered. Holes in rim sherds could also have served to repair dunting cracks in whole pots (cf. the repaired vessel from Westhampnett), although suspension is another possibility. It may not be a coincidence that most holes found in the body or neck of a vessel measure 3-4mm. This hole diameter could well be the optimum size for the use of organic ties in repairs. As many vessels are functional, their repair might be seen as allowing their function to continue and that it might be less trouble to repair a pot than invest in a new one. This is likely to be an explanation for the more commonly recorded repair of samian vessels. The supply of samian into Britain was not constant and periodic difficulty in obtaining replacement vessels is likely to have been encountered.

Larger holes in the body, some measuring as much as 12mm, are just as difficult to interpret. Similar holes in the bodies of amphoras have been described as tapping holes for ease of extraction of the contexts (Evans 2000a, 297). Single holes in small sherds provide very little information, unless the sherd has also been shaped (for instance spindlewhorls, see below). A flagon body sherd (pit fill 7128, not illus.) has had several small holes drilled from the inside. As this can only have been done with extreme difficulty to an intact flagon, the sherd must have been pierced after breakage, perhaps to serve as a strainer or as a weaving-tablet. Another example (not illus.) from pit fill 9569 has had a single hole drilled from inside the base. The interior surface of this vessel is very worn, possibly as a result of stirring. Most holes, as noted above, are pierced from outside the vessel.

The difficulty in allocating functions is highlighted by the recovery of at least twelve jars with post-firing holes from the fills of pit 17177, part of a complex of intercutting pits in the easterly part of the southern zone. No pierced vessels were found in any of the other pits in the complex, but two further examples came from contemporary ditch 25179. A ritual function seems unlikely as the vessels are fragmentary and, although large parts of some pots are present, most seem to have been deposited when already broken. Also, the other contents of the intercutting pits are typical of a normal domestic assemblage. As far as can be ascertained, the pierced vessels all have holes cut through the base only, three have central sub-square perforations, two have single, off-centre round holes and five have multiple holes. Three bases are sufficiently complete for illustration (Figure 338, nos 18-20), but not enough survives of the remainder. The number of pierced vessels found together perhaps deserves mention but, unfortunately, nothing in the deposit provides a clue as to their function.

Spindlewhorls

Most pottery spindlewhorls appear to be examples of selected sherds pierced after firing in order to perform a specific function, far divorced from the function of the original vessel. Eighty-two suitably modified sherds were recorded, more than 80% of which are in grog-tempered fabric of Late Iron Age date. Their distribution follows that for pierced vessels, i.e. the highest proportion occurs in the southern settlement zone (Table 45). The sherds identified as whorls, two-thirds of which are complete, vary greatly in diameter (29-69mm) and weight (8-50g) ranges.

Spindlewhorls should fall within certain weight and diameter ranges in order to function properly, and it follows that a broken vessel might provide only a few sherds that would fulfil the conditions for such reuse. In her study of the spindlewhorls from Colchester, Crummy (1983, 67 and 94; following Wild 1970a) suggests that the overall diameter should be no greater than 50mm, although the weight of the whorl is not stipulated. Weight is also necessary in order to determine the size of the thread that would have been produced (Barber 1991, 52). A too-heavy spindlewhorl would break the thread it was meant to spin. The complete whorls from Elms Farm have a mean weight of 26g and this may be the optimum for production of a substantial woollen thread. Pierced sherds that are much heavier than the required weight range for producing thread must have served different functions, perhaps as weights. Larger pierced sherds could also have been used as lids during cooking, for instance. Indeed, a number of the purpose-made lids recovered at Elms Farm were provided with a vent-hole made before firing.

A search through the literature for comparanda shows that pottery spindlewhorls are not particularly common. Twenty pierced roundels were recorded at Colchester (Crummy 1983, 94) and only eight of these were considered to be true spindlewhorls. At the Airport Catering site, Stansted (Major 2004b , 169), twenty-six ceramic whorls were recorded, only four of which are worked from pot sherds. Forty-four spindlewhorls were noted at Dragonby (Barford 1996, 333), of which only six are utilised pot sherds. The variation in dimensions and sheer number of the spindlewhorls recorded at Elms Farm may throw doubt on the accuracy of the identification in many cases. It should be noted, however, that few spindlewhorls in other materials were found and it may be that those in pottery were the preferred choice locally.

Normally, holes in sherds that have been reworked for use as spindlewhorls appear to have been drilled from both sides of the sherd, perhaps implying that sherds without a hole were chosen for modifying. Presumably, the hole would be drilled centrally first and then the edges trimmed into the circular shape required for ease of spinning. Once a shaped and utilised sherd has subsequently broken, however, its function can be obscured. As diameter and weight measurements demonstrate, not every pierced sherd can be a spindlewhorl and sherds with holes that probably performed different, but equally important, day-to-day functions are thus always under-represented. Full measurements, including the weight, might provide a method of segregating pierced vessel sherds from genuine spindlewhorls.

Conclusion

The majority of pots with both pre- and post-firing holes were likely to be functional vessels performing everyday domestic tasks. This is borne out by the evidence of their deposition in rubbish pits, although disposal as rubbish obscures specific function. Complete, or near-complete, vessels might yield more information. Even so, identification of the function of such vessels is fraught with difficulty. More in-depth analysis of pierced sherds, such as examination of residues or even just the recording the presence of lime-scale, would help in understanding the function of the vessels from which they derive. Lime-scale deposits that cover the edges of a hole might indicate a different function for a vessel than one where a hole is cut through the deposit. The determination of when in the life of the vessel it might have been pierced may be a small detail, but it is potentially important.

The designated functions performed by purpose-made vessels, while seemingly straightforward, can nevertheless be open to question, as can be shown for spouted strainer bowls. So-called cheese-presses are just such a vessel, although ideas for alternative uses for this enigmatic vessel class are not forthcoming. Full recording of the vessels themselves, the vessel associations and the deposits in which the vessels are found, may prompt further understanding both of the vessels and the processes that encouraged their production.

Even so, the uses to which pierced vessels might have been put seem multifarious. The range of functions posited is large and all suggestions are valid. Use as strainers and cheese-presses is particularly viable, especially in the light of the apparent scarcity of the purpose-made vessels. Mundane uses for complete or near-complete vessels, such as braziers or funnels, carry equal weight. Holes made under the rim could be used to secure lids, as well as for suspension of the pot. Even vessels with single, large, central holes, postulated as having more ritualistic potential, could have an equally down-to-earth use. The basal hole is reminiscent of modern-day flower-pots, and the lime-scale deposit in some vessels might indicate prolonged use as something as basic as that. Even protracted use as a cooking pot would not preclude a further function as a flower-pot, once a hole was cut through the base. Vessels used for cooking, and discarded because of rancid internal remains, would be eminently suitable for such reuse.

Ad hoc utilisation of the sherds from broken vessels is also likely. One flagon sherd, pierced from the inside after breakage, has already been mentioned. Sherds, whether pierced or not, lend themselves to further uses, separate from their original function as part of a vessel. The most obvious of these uses is as spindle whorls, but sherds might be pierced for use as plumb-bobs and in braid-making. Jar bases might be used to fit inside the necks of large vessels and then pierced to facilitate pouring, for example. These need not necessarily have been re-shaped, especially if opportunistic use was made of such a base.

Finally, use in a ritual context cannot be discounted, although the evidence at Elms Farm indicates that the likelihood of this, in the vast majority of cases, is remote. Mundane, domestic functions are the more realistic option. The number and range of vessels that catered for those functions gives an insight into the activities and industries undertaken by the inhabitants of Heybridge. It perhaps also gives an insight into their inventiveness and willingness to reuse and recycle vessels, and parts of vessels, on a daily basis.

The structured deposition of pottery

Determining whether ceramic vessels had certain ritual significance - that is, whether a vessel had been accorded a meaning beyond its mundane usage and then deposited with due care and ceremony, distinct from the usual means of rubbish disposal - is by no means an easy task. In many cases, the components of a ritual are severely 'truncated' by poor preservation (e.g. of organic items). In addition, the conceptual processes (motivations, symbolism etc.) are not recoverable archaeologically. Some forty-odd contexts can be selected as potentially having significance as structured deposits. This selection includes deposits with complete or near-complete, intrinsically interesting vessels found in association with other unusual objects, and carefully deposited material (the complete list is contained in the archive).

Most vessels are ascribed a ritual significance on the basis of their being complete or near-complete. Consideration of formation processes at Elms Farm suggests that the average pottery deposit comprises material that has been middened prior to burial. A midden may include material collected from multiple sources, accumulating and abrading over time. With its relocation, the already broken pottery may be spread between a number of open sub-surface features. A complete vessel, then, should be regarded as having escaped the normal processes of deposition. If broken, the vessel was probably deposited directly after breakage. If complete, it was deposited directly from the point of presumed premature disuse, or entered the feature as a special deposit.

The relationship between structured deposits and domesticity is a recurrent theme of the discussion below. The informal and domestic nature of structured deposits is emphasised by the paucity of structured ceramic deposits that lie within and around the temple area. Just three deposits from the temple area contain pottery that was probably deposited as part of a ritualised act. A face-mask flagon fragment (Figure 334, 7) was recovered from pit 5179 and was probably deposited for its intrinsic, 'magical', properties. A second vessel (Figure 355, no. 1) was buried upright in a small cut, 18578, in, or under, the floor of Late Iron Age shrine, Building 8, within the temple complex. A small tazza (Figure 355, no. 2), placed in post-hole 21745 at the eastern end of the temple, forms a third structured pottery deposit, and a likely foundation deposit.

Some of the listed pottery may itself have held special significance because of its intrinsic qualities. The possible religio-magical properties of face-mask flagons, such as those in ditch 25027 (Figure 334, 6; Figure 335, 9) and post-hole 20468 (Figure 335, 8), may have determined their use as amulets or votives.

A number of trends can be drawn from the dataset of possible structured deposits.

Immediately evident is a distinct later Roman emphasis, with almost 80% of ritual deposits (in which pottery forms a principal component) dating to the mid-2nd century or later. The single highest proportion falls within a Late Roman date range. Expressed as a ratio of the total number of pottery-yielding contexts per total number of ritual deposits incorporating pottery, the later Roman bias is retained. This follows the general pattern established elsewhere. Marking pottery with both notch and 'X' graffiti, for example, is predominantly a Late Roman phenomenon.

Table 47 appears to show that pits were favoured for structured pottery deposition. However, this distribution more-or-less reflects the proportions of features present at Elms Farm, with pits predominating. It is worth noting that structured deposits are likely to have been placed above, as well as below, ground. Unfortunately, above-ground deposits are archaeologically undetectable; the evidence rarely survives in situ (and none was identified at Heybridge). It is feasible that objects placed above ground, perhaps as votives within the temple, were then deliberately cleared and buried during a second ritual once fulfilling their tasks, but above-ground structured deposits subsequently found below ground are also difficult to detect. Again, none was identified at Heybridge.

What is notable is the relatively small number of structured pottery deposits recovered from building-related features. The predominance of pits and ditches for structured deposits suggests that pottery did not play a large role in foundation deposits. Just four deposits are directly associated with buildings. The face-mask flagon (Figure 335, no. 8) recovered from post-hole 20468 was clearly deliberately placed, and it is difficult to escape the conclusion that the vessel was a foundation deposit for a poorly defined structure. Further discussion of this deposit is presented above. Two vessels were placed within construction slots for the walls of buildings (Building 54, Figure 355, no. 3; Saxon Structure 56, Figure 355, no. 4). The vessels in this case are complete and, like the face-mask flagon recovered from post-hole 20468, are likely foundation deposits.

The distribution of structured pottery deposits reveals a northern zone bias, which is emphasised when expressed as a ratio of contexts within each zone per structured deposit.

Despite the presence of the temple complex, the central zone did not become a principal focus for structured pottery deposition. Instead, the distribution perhaps highlights the less formalised practice of structured deposition and its integration within the domestic environment through superstitious/magic practices.

As to choice of vessels, jars were preferred, accounting for 40% of ceramic vessels in recognised structured deposits. Beakers and dishes/platters are also well represented. Vessel precedence here largely reflects that of vessels recovered from Elms Farm generally. The pattern of vessel use obtained from the securely dated pottery groups shows that jars were the commonest vessel class in Roman Heybridge, followed by dishes and beakers (see Pottery supply). The selection of vessels suggests that mundane aspects of everyday life were reflected in ritual practices. Interpretation of vessel selection can be approached in two ways. Favourite vessels, or vessels in common use within the household, perhaps represented the influence of the individual taking part in the ritual. The vessels were valuable for practical or sentimental reasons, and their ritualistic use meant tangible material loss, but, conversely, investment and anticipated gain. With their use in religio-magical acts, a connection between the ritual and the individual was enacted for spiritual and, it was hoped, material benefit. Alternatively, the choice of vessels reflects what pottery was available, and perhaps what was dispensable. One or two, out of half a dozen or so, non-essential vessels could be removed from the household but could also be replaced easily if required. It may therefore be that the commoner, more easily obtainable, and less valuable vessels on a personal level were favoured for structured deposition.

There are some obvious exceptions to this practical/pragmatist view of vessel selection. Deposit 4148 comprised two samian f30 bowls, along with a grog-tempered jar and lid (Figure 353, nos 1-4). The jar has pre-firing notches on the shoulder and a large hole punched through the base after firing. The lid is unusually large. Clearly it did not just cover the jar, but was perhaps intended to 'seal' the collective deposit. Unfortunately, the feature is poorly understood, and so this aspect remains speculative. Parallels for the structured deposition of decorated samian are almost unknown or, perhaps, unrecognised, but two near-complete vessels were found in Ditch 4 at 201-11 Borough High St, Southwark (Bird et al. 1978, 64-5). Both vessels are Neronian; one is a decorated f29 bowl, stamped by Primus, and the second a f27g cup, stamped by Licinus (Bird et al. 1978, 95-6, nos 7, 15). A cranium from an adult female lay just to the north of the vessels, but other finds in the ditch fills were few. Samian is present in just one other Elms Farm structured deposit - well 6280, context 16083 (KPG26), so merely the presence of samian is sufficient to suggest differences in practice and motivation between these and the remaining deposits. Among structured deposits, it is only in graves that samian forms a significant contribution. It is possible that these two structured deposits are in some way connected with funerary practices. Brenda Dickinson (pers. comm.) has tentatively suggested that the samian bowls in deposit 4148 formed part of a cenotaph. But, assuming a funerary connection, these samian bowls remain unusual simply because they are decorated; such bowls are hardly ever found in funerary contexts. This risks a circular argument, so it is perhaps safer to suggest that deposit 4148 is unlikely to represent a cremation burial, minus the cremated bone. Possibilities involving other funerary-related (or funerary-like) practices remain open, however.

A further pattern concerns the number of vessels deposited within each feature. Table 50 shows that deposits involving single vessels are by far the most common, and that the frequency of multiple-vessel deposits decreases with the increase in the numbers of vessels deposited.

The deposition of single vessels was the predominant practice and, again, would seem to represent something of an individual act, carried out by or on behalf of the individual or perhaps for a single household. Even so, these vessels would not appear to be treasured keepsakes, but simply anonymous pots in normal everyday use. The vessels have not been personalised in a way that has survived, and so they exude little sense of the individual. In addition, we cannot be sure that the vessels were deposited in isolation.

Compared to the mass of single or double-vessel deposits, the collection of at least eleven intact or semi-intact vessels from purpose-made pit 20008 is clearly unusual (KPG17, Figure 247). This deposit comprised six jars, two beakers, a platter, a lid (all probably manufactured locally), and a Colchester buff ware flagon. Resembling the composition of non-ritual deposits of the same ceramic phase (Ceramic Phase 4, see Pottery Supply), this assemblage is functionally orientated towards storage/preparation; although one of the jars is narrow-mouthed and probably functioned as a liquid-containing flask. In addition, substantial proportions of three other vessels (two bowls and a jar) were present. These vessels were buried with miscellaneous sherds, some of which joined sherds recovered from neighbouring pit 20010. The profiles of a number of vessels could be reconstructed as a result, and a beaker, most of which was in pit 20010, was thus near-complete.

Apart from the intact vessels, there are other aspects that seemingly confirm the fill of pit 20008 as a structured deposit. Five of the vessels appear to have been 'killed' by having chips removed from along the rim, or by deliberate perforation of the vessel walls (Figure 247, nos 1, 7, 12, 14). While this sort of evidence is largely recovered from cremation burials, affirming a ritual motivation behind the practice, some deliberate damage need not have been undertaken for religious or superstitious reasons. While the two holes punched through the walls of one vessel (Figure 247, no. 7) would appear to have served no practical function, that in the base might well have once served as something more mundane. The small, semi-circular indentations around the edge of the large hole suggest that it was once a collection of smaller, strainer-like holes. Of course, these smaller holes may have facilitated the making of the large perforation. Nor is there any indication that all three large holes were contemporaneous. One jar (Figure 247, no. 5), albeit incomplete, has two notches cut into its rim. On the evidence of other notched pottery, this act was also unlikely to have been a precondition for its burial. Finally, the internal surface of the platter (Figure 247, no. 1) is scorched, suggesting that direct heat had been applied. While it is tempting to imagine the platter playing a ritualistic role immediately prior to burial - offerings burnt to deities or substances heated to alter ambience or perception - the possibility that the vessel was used simply for domestic cooking remains equally valid, albeit unusual. It should not, therefore, be presumed that all such characteristics of the pottery were connected with the ritual practices leading to and carried out during burial.

The placement of the vessels, and their sequence of deposition, provides much firmer evidence for this assemblage being a structured deposit. Stratigraphically, pit 20008 was filled during a single event. The pottery, however, was deposited during the course of at least five separate actions. First, the perforated jar was inverted and placed at the bottom of the pit. Six or so vessels, not all complete, were placed around the jar. The pit was perhaps partially backfilled, with the soil covering the pottery, although not completely - the perforated jar base poked through the soil and was visible during the subsequent deposition. The remaining pottery was placed within the pit on top of the deposit of soil and around the visible base of the perforated vessel, which remained central. The pit was subsequently fully backfilled. There were many conjoining sherds from the neighbouring pit 20010, suggesting that the pottery intended for special treatment, incorporating both complete vessels and miscellaneous sherds, was spread over the two features, with most of the complete pots going into pit 20008.

In terms of vessel treatment, pit 20008 resembles cremation burials. Like pottery from the funerary features, some of the vessels from pit 20008 were deliberately mutilated and one vessel was inverted. Similar evidence was recovered from the Elms Farm burials. Presumably, the motivation behind such treatment, discussed with the burial evidence, is similar, too. It is perhaps here that the similarities end. Notwithstanding the absence of human bone, which does not necessarily preclude the deposit from being funerary-related, the number of vessels involved is atypical of funerary features of this period. None of the earliest burials at Heybridge, dating to the second half of the 1st century AD, incorporated more than three vessels, including the cinerary vessel. Nor is pit 20008 located near to these contemporaneous burials. In functional terms, the assemblage from pit 20008 is weighted towards preparation and storage. However, eating and drinking functions are still better represented in the combined assemblage of pit 20008 and pit 20010 than in the combined assemblage of the remaining well-dated Ceramic Phase 4 (non-ritual) pottery groups (see Pottery supply).

The differences between the assemblage composition of pit 20008 and that of the Ceramic Phase 4 average, in conjunction with the condition and treatment of the pottery, are sufficient to suggest differing processes of selection and deposition. However, the similarities between pit 20008 and typical burial assemblages are not strong enough to identify 20008 as a burial. Other funerary interpretations, for example a cenotaph, remain possible, but comparison between assemblages from other funerary contexts and pit 20008 is necessary in order to substantiate such interpretations.

Aside from pit 20008, which is clearly exceptional, a number of more typical structured deposits warrant close consideration. Motivations determining their form and function lie well within the realms of speculation, but some clues rest with the vertical and horizontal location of the deposits within their respective features. No strong patterns emerge from the dataset, although tentative biases towards top and bottom fills in pits, intermediate fills in ditches and lower/bottom fills in wells are apparent.

Excavated segments 401 and 568 of ditch 25274, each contained a pair of coarse ware jars, complete save for perforated bases and dating to the second half of the 1st century AD (Figure 354, nos 1-4). The jars were recovered from the intermediate fills. Some care seems to have been taken when depositing the jars by placing them in pairs in an upright position, although excavation details are sketchy. The vessels were apparently selected for deposition, comprising typologically similar G23 necked jars, each with a single large perforation through the base. Even if these perforations were not made for the purpose of deposition, perforated jars had been selected so that form and treatment were virtually identical. A similar occurrence of three sets of paired ceramic vessels in a ditch at Orpington, Kent (Merrifield 1987, 38, pl. 13), indicates that this may have been a recurring, though perhaps seldom recognised, phenomenon.

Further potential ritual deposits were recorded at Elms Farm. Two complete beakers were placed at the end of probable boundary ditch 25270. Bottom fill 9895 of well 9421 contained a complete jar, as well as significant animal remains. Complete/near-complete vessels were recovered from the first disuse fill of well 6280. In all of these instances, the ceramic vessels deemed to have ritual significance were placed at the bottom or ends of features originally dug for purposes other than rubbish disposal. Considering vessel location and feature type, the rite of closure would seem to be an apposite interpretation. The four jars in ditch 25274 and the two beakers at the terminal of ditch 25270 perhaps commemorated these boundaries passing out of use. The material placed in the bottom of the wells reinforces the ending of an essential, life-giving, function - the drawing of water. This interpretation works equally well with pits, assuming that the pit in question served a purpose, such as storage, prior to its being filled with rubbish. The depositing of multiple complete vessels, as well as mundane rubbish, perhaps articulated the gratitude that the inhabitants may have felt for the pit fulfilling its duty, or more simply, the deposit formally ended its original role.

A rite of termination is perhaps an inappropriate interpretation for some intermediate fill deposits. Examples include the complete jar from well 17155 and the two complete vessels (a bowl-jar and a beaker) recovered from the fourth fill of pit 23012. Since the lower fills also contained pottery, the pit was already functioning as a rubbish pit prior to the deposition of the complete vessels. It is possible, then, that the feature itself was, in some cases, less important than the act of deposition and the vessels themselves, or more correctly, the contents of those vessels. If the complete vessels contained food or drink and were to be offered to a deity, perhaps the feature into which the vessels were placed was largely irrelevant. All cut features are ground-penetrating and potentially make some connection with the spiritual world. Nor should it matter whether the feature was already partially filled, even with rubbish. So, the complete vessels, particularly the beaker, in pit 23012 may have held libations to be poured into the pit, followed by the deposition of the vessels themselves. Indeed, the presence within the same fill, or feature, of 'mundane' rubbish may have added significance to the deposition of complete vessels. Deriving from middens by way of 'life-enhancing' compost, refuse material may have carried its own 'liminal qualities' (Fitzpatrick 1997, 79), although pottery sherds were unlikely to be, by themselves, significant.

Assuming that jars were used essentially for cooking foodstuffs or for storage, the relative paucity of beakers and flagons, and complete absence of cups, suggest that libations were infrequently offered. However, as discussed above, the choice of vessels was likely to have been determined by what was available, rather than what was necessarily most appropriate for the task. So, jars may well have held liquids, because no beakers were available for disposal. In a similar vein, samian vessels were not commonly deposited, because they were not as readily disposable as coarse ware jars. While samian was fitting and appropriate for human burials, it was deemed less necessary for structured deposition of a religio-magical nature.

Finally, a more mundane interpretation might be that complete or near-complete pottery represents domestic clearance - the throwing-out of older, disused household material, and its replacement with new items, or, simply, the discarding of damaged or broken pots. Vessels may well have become rancid through use and thus undesirable, to be thrown away despite their being whole (Rigby 1986b, 259). Doubtless, too, the inhabitants of Heybridge discarded material simply because they tired of its presence, or changes in personal circumstances led them to regard the material differently. Interpretations such as these are rightfully considered for complete vessels recovered from above-ground layers or dumps, and intermediate, single and top fills of seemingly ordinary rubbish pits, including pit 20008. Even pottery given 'special' treatment and deposited with deliberation need not have 'ritual' connotations. A case in point is a Dressel 20 amphora recovered from pit 4582. The top half of the vessel was extant; its rim had been neatly removed before being inverted into the pit. Similar deposits are known elsewhere (e.g. P. Crummy 1984, 135; 1992, 105; Clark 1999, 122) and have been regarded in purely functional terms, for instance as soakaways. A similar interpretation here is just as likely.

A number of trends, then, emerge from this brief study. Structured pottery deposition was predominantly an informal, almost domestic, concern and unconnected with formal religious practice focused upon the temple complex. The pottery was selected on the basis of availability and, perhaps, sentimental value. The deposits were small-scale, intermingling with 'mundane' rubbish, and almost disappearing into the features in which they were placed. While the deposits mainly represented individual interventions into the spirit world, the uniformity of the practice, as suggested by the predominance of the single utilitarian vessel deposit, suggest common motivations. For some deposits, the reasons behind them seem clear. But for the majority, in the absence of their contents and 'conceptual' data, little can be said of the intentions and beliefs that underpinned their creation.

 PREVIOUS   NEXT   CONTENTS   ISSUE   HOME