Regional identities and the social use of ceramics Martin Pitts

Introduction In My Roman Britain, Richard Reece (1988, 32) suggested that if we excavated all the rubbish from a single farmstead, it would be theoretically possible to reconstruct the table settings of that occupation from a particular period. However, Reece described such a suggestion as being ‘light-hearted’, something to be regarded as ‘frivolous and irresponsible’ by ‘those who have to deal with pottery as a job’, as ‘we just do not do that sort of thing with pottery’ (Reece 1988, 33). Taking Reece’s idea a stage further, it is the aim of this paper to examine regional variation in cultural identities by a similar notion of interrogating the functional components of ceramic assemblages. The approach advocated combines aspects of previous research on the functional aspects of ceramic assemblages (e.g. Evans 2001), the use of correspondence analysis (e.g. Cool, Lloyd-Morgan and Hooley 1995; Lockyear 2000; Cool and Baxter 2002) and the role of pottery in social practice and the negotiation of identity (e.g. Woodward and Hill 2002; Pitts 2004). Indeed, Reece’s grail of being able to reconstruct table settings might not be so ‘frivolous and irresponsible’ as his caricatured pottery specialists would have us believe.

Why pottery? Identity manifests itself in the archaeological record in many guises. However, there are several reasons for privileging pottery in an approach to cultural identity. Firstly, certain types of Roman pottery are closely datable, providing a relatively high chronological resolution. Secondly, pottery is more or less ubiquitous on sites in the Roman period (particularly in SE England, the area chosen for the case-study), facilitating a holistic and bottom-up perspective. Thirdly, pottery is fairly resistant to taphonomic decay (unlike faunal remains) and it is less susceptible to re-use (unlike other small finds), suggesting that material found in most archaeological contexts will provide a fairly representative and reliable sample. However, an often overlooked factor I wish to dwell on in this paper is the role of pottery in social practice. I have stated elsewhere (Pitts 2004) that social practices including feasting and communal drinking, were vital to the articulation and reinforcement of power relations and identities in the late Iron Age and Roman period. Within this model, particular ceramic forms acted as a form of ‘consumption technology’, to provide a medium or vehicle for empowering forms of social practice to take place (Pitts 2004, 17). This is by no means a new realisation in archaeology, and vessel form and function are clearly of central importance in parallel research into feasting and commensality in North American archaeology (e.g. Sinopoli 1991; Dietler and Hayden 2001; Bray 2003).

Particular strategies of communal consumption are likely to require specific vessel forms to facilitate their enactment. At a basic level, drinking practices will require a vessel to store the liquid (such as jars, amphorae, barrels or skins etc), perhaps a vessel to decant the liquid (such as flagons and jugs) and finally, vessels to aid consumption of the liquid – either individually (e.g. cups and beakers) or communally (e.g. bowls, jars and larger beakers). In contrast, eating practices can be straight from the cooking pot (involving bowl and jar forms) or be served on a table (implied by shallow vessels such as platters). Added to this are vessels related to the preparation of specific dishes, such as mortaria (Reece 1988, 30-32; Cool 2004) and those associated with specific forms of cultural practice, such as the crater vessel used to mix wine and water in the Greek symposion and Roman convivium (Dunbabin 1993). Therefore, if the relative proportions of all the different vessel classes are considered for a particular phase of site activity, it should be theoretically possible to roughly characterise the functional emphasis in ceramic consumption for that period.

However, there are a number of complications with this premise. A fundamental problem is that broken archaeological assemblages cannot be directly equated with use assemblages in the past. For example, it is envisaged that some fine wares and scarce imports may be used differently and curated for longer periods than locally made coarse wares. Additionally, some vessel uses (such as storage and transport) encourage redundancy, whereas other vessel uses encourage curation and will only be thrown away through accidental breakage or intentional deposition (e.g. Hill 1995). Nevertheless, as this paper later demonstrates, once these factors have been identified they can be dealt with.

Pottery and regionality Emphasis on the role of pottery in social practice and identity formation is somewhat at odds with the current orthodoxy of Roman pottery studies. This is because traditional wisdom on pottery and regionality is typically orientated towards demonstrating economic patterning – kiln sites are located, fabrics are sourced, and site supply is considered by fabric provenance (e.g. Going 1987, Pollard 1988 and Fulford 2000). There is no doubt that such approaches are both useful and necessary, but the traditional focus on fabric and provenance (with form largely used solely as a tool for spot-dating) can ultimately only inform research on supply and technological changes in production. Locally-made pottery fabrics will always be regionally specific because of the differing geological make-up of local clays, whereas Evans (2001) has demonstrated that proportions of imported fabrics are largely determined by notions of site type and economic status rather than regionality.

The approach advocated here is simple. Returning to the conception of pottery as a vehicle for social practice, it is the form of the vessel, rather than what it is made of, that becomes important. Applying this idea to the problem of regionality, it should be possible to gain some idea of the relative emphasis being placed on different aspects of eating and drinking habits in a region by mapping the proportions of different form types from a range of sites in the said region. Thus, it should be theoretically possible to extrapolate intra-regional variations in social practice, and in particular the degree of heterogeneity or homogeneity across a wide geographical area. The next section will consider the methodological implications of this approach, with particular reference to a specific study area and period – south-eastern England, c. 50 B.C. – A.D. 70. The study area within south-east England was selected because it represents one of the first in Britain to undergo both prolonged and intense continental contact. Furthermore, the chronological parameters were chosen to examine the differential use of pottery in the generations before the cultural and social repositioning that attended annexation from the Flavian period onward.

The Trinovantes and Catuvellauni a regional case-study The main problem to be addressed in this case-study is whether or not it is possible to use evidence for ceramic consumption and eating and drinking practice to distinguish between two historically attested regions or tribes – the Catuvellauni (roughly equating to modern day Hertfordshire) and the Trinovantes (roughly equating to modern day Essex). This example was specifically chosen to use archaeological data to ‘test’ the largely accepted regional constructs of tribal territories and civitates.

The evidence for these two late Iron Age tribes is fundamentally historically determined. Caesar (The conquest of Gaul, V. 20–22) mentions the Trinovantes being at war with the assumed leader of the Catuvellauni, Cassivellaunus. In addition, the extent of their respective territories has largely been derived from the inscribed coinage of historically attested figures. For example, De Jersey’s recent work (2001) on Cunobelin’s silver seemed to suggest distinctions between the two tribal territories in terms of the distributions of different contemporary coin issues. However, using Iron Age coins to establish the basis of regions is somewhat problematic. Firstly, late Iron Age coinage was scarce and probably only used by the highest echelons of society for articulating relations of clientage (Creighton 2000). Secondly, the intrinsic value of coins (especially the gold and silver issues) means that a significant proportion would have been melted down for re-use in newer issues (Whyman and Perring 2002, table 6).

Haselgrove’s (1982) core-periphery model for LPRIA south-east England provides a more archaeologically driven approach to regionality, the core zone being defined on the basis of the distribution of distinctive burials and coins of Cunobelin. This model characterises large elements of south-east England as the main component of a single polity (incorporating the Trinovantes and Catuvellauni) based on Camulodunum which exploited its immediate periphery for slaves and raw materials to fuel the continued influx of prestige goods from the continent. Although the general distributions of burials and coins allowed Haselgrove to establish the boundaries of what appears to be a meaningful region (in archaeological terms), it remains to be seen to what extent this trend is upheld on a more site-specific resolution. Before examining the ceramic data for the region I was expecting one of two broadly different outcomes. The first and most unlikely scenario would envisage the two tribes as separate social entities, being differentiated through emphasis on their respective styles of consumption, reinforcing the scant historical evidence and the coin distributions. Conversely (and more likely), if the coin distributions merely indicated political divisions rather than separate cultural-regional identities, one would expect more similarities in consumption habits between the two neighbouring areas, in line with Haselgrove (1982).

Selection and analysis of ceramic assemblages, c. 50 B.C. – A.D. 70 The analysis discussed below is based upon a selection of published and unpublished sites with quantified ceramic assemblages from both modern counties, with about half from Hertfordshire and half from Essex (see figure 1). The sites, all with pottery groups broadly dating to the LPRIA and earliest Roman period, comprised Baldock (Stead and Rigby 1986), Ermine Street, Braughing (Potter and Trow 1988), Wickham Kennels, Braughing (Partridge 1982), Skeleton Green (Partridge 1981), Lockleys, Welwyn (Ward-Perkins 1938), Folly Lane, Verulamium (Niblett 1999), Park Street (O’Neill 1945), Nazeingbury (Huggins 1978), Great Dunmow (Lavender 1997), Braintree (Garwood and Lavender 2000), Elms Farm, Heybridge (Atkinson and Preston, forthcoming; Atkinson and Compton pers. comm.), Ardleigh (Brown 1999) and Sheepen, Camulodunum (Niblett 1985). The data used from this list of sites ranged from a small number of large, well-stratified pottery groups (e.g. Ermine Street and Braintree) to larger numbers of stratified groups which could be split into sub-phases (e.g. Elms Farm and Baldock). Since both modern counties had employed quite different systems of recording pottery types over the years, a simplified and overarching form series was applied to the data to eliminate any possible bias. This form-series was a hybrid of several type-series (Hawkes and Hull 1947, Hull 1958, Bidwell and Croom 1999, Thompson 1982 and Going 1987) currently in use by field archaeology units in the region (details available in Pitts (in prep.)). All the assemblages were originally quantified by [minimum] number of vessels represented (except Elms Farm, which was quantified by estimated vessel equivalent), and these values were subsequently converted to percentages for analysis.

Figure 1. Selected sites in Hertfordshire and Essex, c. 50 B.C. – A.D. 70 (after Jones 1997, 37). The first stage of analysis was to examine the basic functional characteristics of the ceramic assemblages using Evans’ method (2001) of plotting sites by the relative proportions of different functional categories of pottery, such as cooking vessels (i.e. jars and storage jars), drinking vessels (i.e. cups and beakers) and tablewares (i.e. bowls, platters and dishes).

Figure 2. Selected sites by percentages of jars and drinking vessels, c. 50 B.C. – A.D. 70 (based on [minimum] no. of vessels represented).

Figure 2 is a plot of jars against drinking vessels the sites marked by open squares are in Essex, and the sites marked by black squares are in Hertfordshire. An interpretation of this graph was that it did appear to show differentiation between the two modern counties. Barring a couple of outliers that did not fit the pattern, most notably late first century B.C. Baldock at the top-left extreme and Camulodunum at the bottom-right extreme, it appeared that the Hertfordshire sites had higher proportions of drinking vessels and lower proportions of jars than the Essex sites. This pattern is almost identical when considering jars against proportions of tablewares (see figure 3). First century B.C. Baldock and Camulodunum again stand out, but the rest of the sites are grouped in a similar way to the drinking vessels – with the Hertfordshire sites having higher proportions of table wares and lower proportions of jars than the Essex sites.

However, if these patterns are examined in more detail, it is apparent that the situation is more complex than immediately suggested. Considering the sites with high proportions of tablewares and drinking vessels, they appear to be either sites which are chronologically later (such as Camulodunum, Wickham Kennels and post-conquest Baldock) or earlier sites known to be of high-status (such as Folly Lane and the pre-conquest occupation levels from two first century A.D. villas – Lockleys and Park Street). In contrast, the sites with lower proportions of drinking vessels and tablewares appear to be either chronologically earlier (such as first century B.C. Baldock, Elms Farm and Braintree) or sites not known to be of high status – (such as Ardleigh and Nazeingbury). Going back to the raw data, it is also apparent that sites towards the middle of the spectrum (such as pre-conquest Baldock and early first century A.D. Elms Farm) clearly have elements of high status consumption, but these were set against a background of more mundane activity. Thus, patterning that initially appeared to be reflecting two regions differentiated by consumption practice seems to be more reasonably explained by factors such as site status and chronology.

Figure 3. Selected sites by percentages of jars and table wares, c. 50 B.C. – A.D. 70 (based on [minimum] no. of vessels represented).

To examine this patterning further it was necessary to undertake a complete functional analysis of the data used above, using correspondence analysis (hereafter referred to as CA). To avoid a repetition of the results obtained above, the background noise of the main jar and storage jar forms was removed. Jars occurred on all sites in the late Iron Age and Roman period in south-east England in vast quantities, so much so that their numbers would inevitably reduce the significance of patterning in other less frequently occurring functional categories. It is likely that jars were produced in significant quantities for relatively short life-spans of use in everyday tasks such as cooking and food storage, encouraging redundancy rather than careful curation. Amphorae were also not included as their small rims in relation to the overall size of the vessel does not facilitate adequate counting or quantification (Peacock and Williams 1986, Sealey forthcoming). This was not a significant problem as amphorae were relatively rare in this period, which meant that their presence in assemblages used for analysis could be easily noted.

CA was chosen as it provides a quantitative means of summarising the associational traits between find types and contexts – in this case pottery forms and different sites and sub-phases. This method has already had a number of useful applications to finds data (e.g. Barclay et al. 1990, Cool et al. 1995, Lockyear 2000, Cool and Baxter 2002) in addition to pottery data (Pitts 2004). This paper will concentrate on the potential archaeological meaning of results generated for this specific case-study, although full mathematical explanations of the technique are available elsewhere (e.g. Greenacre 1993 and Shennan 1997).

CA was thus conducted on selected assemblages from the sites outlined above, the results being displayed in figures 4 and 5. Figure 4 shows the various pottery forms by their relative occurrence at the different sites, and figure 5 shows the sites by their proportions of different form types. The points on one plot directly correlate to the other, hence the term correspondence analysis’. The following discussion of the results is based on interrogating each numbered quadrant of the CA plots in turn (see figures 4 and 5). Figure 4. Ceramic form types plotted by their proportional deposition at selected sites and sub-phases. Numbers denote quadrants referred to in the text (corresponding with Figure 5).

Figure 5. Sites and individual phases plotted by their proportions of different ceramic form type. Numbers denote quadrants referred to in the text (corresponding with Figure 4).

Looking at the first quadrant in each plot, five form-types (pedestal beakers, flagons, mortaria, cups and platters) correspond with six sites (both phases of Camulodunum as well as Skeleton Green, Braughing Ermine Street, Braughing Wickham Kennels, and Great Dunmow). Concerning the sites and phases, with the exception of Great Dunmow, they all appear to have had some kind of direct Roman influence. The material used from Camulodunum is mostly post-conquest, with the Sheepen material coming from the supposed industrial area serving the legionary fortress (Niblett 1985). The Braughing complex (including Skeleton Green) was a settlement with much evidence for direct continental links, including evidence for pre-conquest literacy shown by the amount of graffiti on pottery sherds (Partridge 1981, 351). Conversely, examination of the corresponding group of associated pottery forms reveals a group of vessels of the sort classically associated with Romanisation. Mortaria, flagons, cups and platters are all obvious Roman imports to Britain, but there are also pedestal beakers which are clearly derived from the crater (or krater) (see Hawkes and Hull 1947, plate 54). Re-examination of the raw data for Great Dunmow revealed that this site was plotted in the first quadrant because of a high proportion of Gallo-Roman cups and platters rather than the other more ‘Roman’ diagnostic forms.

Figure 6. Cultural assemblage suite 1: eating and drinking vessels with direct Roman ancestry (after Hawkes and Hull 1947, cups after Webster 1996). Not to scale.

Thus, the most distinctive elements of the pottery assemblages from the Camulodunum and Braughing complexes highlighted by CA are overt markers of what are typically regarded to be Roman consumption practices (e.g. Woolf 1998, 186) (see figure 6). Mortaria and crater derivatives are clear markers of a Mediterranean habitus, both being involved in quite specific forms of cultural practice on the continent – mortaria in the preparation of food, and craters in a specific drinking ritual. Furthermore, the emphasis on cups and flagons rather than larger beakers and jars (as will be highlighted below) seems more orientated to the consumption of wine as opposed to the larger quantities of potentially less potent local beer suggested at other sites in this period (e.g. Sealey 1999).

Turning our attention to the second quadrant in each plot, four form-types (platters, girth beakers, butt-beakers and dishes) correspond with four sites (the early first century A.D. phases at Elms Farm, LPRIA Folly Lane, and the pre-Roman occupations from Lockleys villa and Park Street villa. Again, all these sites seemed to exhibit common traits. They all dated broadly to the first half of the first century AD, and were also probable locations for the indigenous elite. Folly Lane is well known for its conquest period conspicuous burial, Lockleys and Park Street both went on to become 1st century villas, whereas the pre-conquest phase at Elms Farm (EF2) is the same phase in which the so-called event-pit was dated, containing at least three wine amphoras and a significant proportion of other imported vessels. Although the functional emphasis of the corresponding pottery was again on eating (platters and dishes) and drinking (butt and girth beakers), the cultural emphasis of these vessels seemed be along second-hand Gallo-Roman lines, rather than the more direct and contemporary Roman links implied by the previous correspondence.

Figure 7. Cultural assemblage suite 2: eating and drinking vessels with Gallo-Roman ancestry (after Hawkes and Hull 1947). Not to scale.

Accordingly, the core of the second strategy of cultural consumption comprises eating and drinking vessels of Roman character that had reinterpreted in the genesis of Gallo-Roman society (see figure 7). The emphasis on so-called tankard forms (butt and girth beakers) rather than smaller cups possibly suggests the use of locally produced beer as a substitute for imported wine (Sealey 1999). Wine was clearly a rare and prized commodity in northern Gaul and south-east England at this time, evidenced by both classical writers and archaeological evidence. Diodorus Siculus famously wrote about the Gauls desiring wine to the extent that they would exchange a slave for a jar of wine, ‘getting a servant in return for a drink’ (Diodorus Siculus V.26), whereas the presence of wine amphorae is well-attested in the highstatus burials of this period (e.g. Stead 1967, Foster 1986 and Niblett 1999). However, the sheer quantity and distribution of these tankard forms (as opposed to cups) in Essex and Hertfordshire (especially butt-beakers) suggests that wine consumption represented only the highest status end of a wider and more socially inclusive practice of alcohol consumption across the whole region in the LPRIA.

Examining the final quadrants (three and four) in each plot, six form types (lids, bowls, beakers, tazze, pedestal jars and flask-jars) correspond to seven sites (Ardleigh, Braintree, three phases at Baldock, and late 1st century B.C. Elms Farm). Although the correspondence is slightly less clear-cut, it is still possible to highlight some coherent patterning. The site closest to the vertical, pre-Flavian Baldock, is understandably verging towards the Romanised end of the spectrum, so can be effectively excluded. The remaining sites are essentially dated to the first century B.C. (i.e. Braintree, Elms Farm 1 and Baldock 1), or lower status sites (i.e. Ardleigh).

The functional emphasis of the corresponding group of pottery seemed to be on early forms like pedestal jars and tazze, as well as forms perhaps more associated with native consumption practices (e.g. narrow-necked-jars and other forms of bowls and beakers).

Figure 8. Cultural assemblage suite 3: 1st century B.C. vessels with continental ancestry (tazza and pedestal jars after Stead 1967 (figs. 7 and 8), La Tène A vessels after Hawkes and Dunning 1930 (figs. 3 and 4). Not to scale.

Therefore, the most distinct vessels in this final cultural assemblage are pedestal jars and rarer tazze of the first century B.C. (see figure 8). These vessels were linked typologically to much older continental vessels, which were in turn derived from contemporary Mediterranean bronze flagon forms (Hawkes and Dunning 1930), and were probably associated with drinking practice (Hill 2002, 147). It is debatable whether this typological link implies a parallel transmission of Mediterranean social practice, but the association of such vessels with Italian wine amphorae in the Welwyn graves of this period (e.g. Stead 1967) does suggest a similar use in drinking ritual. Indeed, this interpretation is reinforced by Okun (1989, 47–9), who demonstrated a similar phenomenon of pre-Roman assemblages being characterised by high proportions of tall drinking forms in the upper Rhineland. Therefore, given the likelihood of alcohol consumption as a deeply embedded practice in LPRIA society, it seems likely that the high-status accredited to imported vessels derived not from their inherent value as prestige objects, but rather through their ability to create hierarchy in social practice (i.e. feasting, communal drinking and other ritual). As Murray (1990, 5) stated, ‘pleasures exist within a social context, and will be subject to manipulation and development for social ends.

Conclusions To conclude, the study did not pick out any meaningful regional distinctions in consumption between Essex and Hertfordshire, and by inference the tribes of the Trinovantes and Catuvellauni. This outcome seems to resonate with the possibility that the Trinovantes and Catuvellauni were in fact not two tribes but one big dynastic family with individuals occasionally bickering with each other and named successors having their own sub-territories (Creighton 2000, 78–9; also pers. comm., February 2004). However, instead of isolating regional distinctions in consumption practice, three inter-regional and chronologically overlapping ‘suites’ of consumption were apparent across both supposedly different tribal territories and modern counties. This includes the ‘directly’ Roman influenced assemblage components from Camulodunum and the Skeleton Green-Braughing complex (see figure 6), the Gallo-Roman assemblages from high-status Hertfordshire sites and Elms Farm in Essex (see figure 7), and finally, the earlier assemblage components deriving from longer-term diffusion of continental ideas towards the end of the first millennium B.C. (see figure 8). Moreover, it is clear that the core components of all of these strategies are ultimately derived from elements of continental practice. However, instead of characterising these trends as Romanisation before conquest, I would prefer to see this patterning as the culmination of a broader, longer term process by which south-east England became increasingly ‘continentalised’ up to the first half of the first century AD.

On a final note, it may never be possible to grasp the Reeceian grail of reconstructing the table-settings from a site in a particular period with complete confidence. However, this paper has demonstrated that methods do exist to facilitate this kind of approach, perhaps not to reconstruct table-settings, but to characterise consumption patterns at a high resolution. I believe that this approach is not ‘frivolous and irresponsible’ (Reece 1988, 33), because we can do that sort of thing with Roman pottery.

Department of Archaeology University of York

Acknowledgements I thank Dominic Perring for supervising my doctoral research and for his constructive commentary on the various drafts of this paper, and Steve Roskams and Tania Dickinson at the University of York for providing additional guidance during my research. I am very grateful to English Heritage, Mark Atkinson, Joyce Compton at Essex County Council FAU and Paul Sealey at Colchester Museum for giving up their time and providing the necessary data and information prior to publication. Finally, I thank the AHRB for funding my doctoral research programme.