Ancient pots, from table to grave
Archaeologists like to classify ancient pottery by type, texture and colour, but such things may offer little help with understanding how the vessels were used. A major collaborative project has considered new ways in which pots might inform our knowledge of Roman dining. Sarah Colley reports
When we imagine i ancient Romans eating and drinking, we often picture a socially and politically aware urban elite, as described in texts such as Plutarch’s Table Talk, using dining for conspicuous display. To understand everyday Roman food consumption we need to look to archaeology. Millions of Roman artefacts associated with eating and drinking have been excavated across Europe, Asia Minor and North Africa. These include ceramic finewares and tablewares such as shiny red terra sigillata, and local variations, which are so widespread and ubiquitous they are often regarded as key indicators of “Romanisation” and “being Roman”.
Archaeology generates vast tableware collections and information archives. Yet we still know relatively little about the social, cultural and symbolic significance of tablewares and their roles. Pottery research focuses on manufacture, distribution, trade and chronology; research into Roman food and drink is usually about production, supply and cooking rather than consumption. To understand social and cultural aspects of tableware use, we need to compare large samples of detailed and different kinds of information from across the Roman world. This is no easy task. Challenges of access, data quality and comparability arise from the disparate research aims and methodologies of archaeologists working in different countries.
To address such issues, Penelope Allison (professor of archaeology at the School of Archaeology & Ancient History, University of Leicester) and Martin Pitts (associate professor in Roman Archaeology at the University of Exeter) initiated the Big Data on the Roman Table research network (2015–16). This brought together 50 archaeologists from ten countries with shared interests in tablewares and consumption. The network also included informatics specialists who develop maths and computer-science
applications to analyse large and complex digital information sets (“big data”) – archaeological records of millions of pieces of Roman pottery are an excellent and interesting example.
From table to grave
Scholars have developed classification systems and typologies to describe Roman pots, including descriptors such as cups, jugs and so on. Such labels do not tell us what was served in these vessels, or about their social or cultural significance. The limits of archaeological classifications for understanding consumption are well illustrated in a cartoon by Jesús Bermejo Tirado. He has developed a new classification system for pottery use linked to eating and drinking, which he tests using archaeological data from Roman domestic contexts in Spain.
Archaeologists routinely record measurements such as vessel height and rim diameter. William Baddiley calculated “liquid capacity” from published archaeological drawings to better understand the function of typical Roman “cups”, “beakers”, “jugs” and “flagons”. He compared pottery from a legionary fortress at Usk, southeast Wales (likely used for drinking beer, though not proven) with silverwares from Pompeii associated with wine drinking. Can “liquid capacity” tell us what was being drunk? Can we identify individual and communal drinking?
Baddiley’s pilot study could not definitively answer such questions, but he discovered surprising variation in the capacity of “standard” vessel types, such as Dragendorff 27 Samian wares from Usk. That most Usk vessels held less than 300ml may relate to consumer choices as well as drinking practice. Digital innovations to automate recording and calculation of vessel shapes and volumes and big data analyses of multiple sets of information (eg find spots, vessel forms and materials) raise interesting future possibilities.
Several Big Data on the Roman Table participants studied tableware production, trade and supply on the grounds that availability can influence use. For example, black and grey Pannonian slipped wares from Croatia discussed by Tino Leleković are regarded as regional imitations of redgloss terra sigillata, blending Roman imperial taste and local Celtic tradition. The black and grey colouring is deliberate. Distinctive black/grey slipped imitations persisted when terra sigillata was later imported from Gaul and Germania. Leleković suggested reasons for this include a market preference for cheaper versions of expensive imports, and regional food and drink traditions.
Other network participants focused on social classes and other groups, for example comparing the occurrence of tablewares and associated finds on settlements and grave sites in southern England and France. Understanding differences in tableware use in different
parts of the Roman world can throw light on Romanisation and identity.
Complete pots in situ with other items are commonest in burials. Edward Biddulph compared large samples of information on locations, socio-cultural status and dates of pottery from settlements and cemeteries in southeast Britain and northern Europe. Pots in graves rarely reproduce table settings, even when laid out. At the Pepper Hill cemetery in Kent, archaeology suggests that beliefs and burial rites were more important in determining the choice of vessels for the grave than a desire to mirror their use in life.
Pottery dominates funerary offerings in Roman grave sites in northern Gaul, France. Alice Dananai and Xavier Deru analysed the function of different vessels found in burials from three tribal areas, and compared these with pots in domestic contexts on settlements. Quantitative data analysis clearly distinguishes funerary and domestic collections. The function of some vessel types used in daily life changes when they are placed in a grave; a domestic container may become a cinerary urn, for example. Archaeologists need to take into account the contexts where vessels are found to understand their varied functions.
In Roman Leicester
Three decades of major excavations of Roman Leicester have provided substantial data for broad analysis of vessel supply and use, but tightly dated groups of artefacts and food remains from specific buildings are rare. Nicholas Cooper, Elizabeth Johnson and Martin Sterry analysed three such “dining-related” archaeological collections. Two represent second century ad dining out. One is from the backfill of a cellar on Little Lane, perhaps from below a tavern, containing tablewares, drinking vessels, amphorae and flagons, alongside animal bones and oyster shells. The other is from a latrine pit on Castle Street containing amphorae, flagons, tableware bowls and dishes, but no drinking vessels, with exotic plant foods, fish and smoked shoulders of beef; this is considered to relate to a delicatessen-style take-away. The third group represents eating in, at a later third-century courtyard house on Vine Street. A wide range of animal and plant foods alongside cooking pots, bowls and dishes, and large colour-coated ware beakers, were excavated from a kitchen drain and two latrine pits.
Finds from these contexts were compared to patterns already established for pottery supply and vessel use derived from 26 other rubbish deposits across the Roman town. Spatial correspondence analysis developed by Sterry revealed localised trends in pottery deposition. More vessels associated with drinking occur in the central and northern parts of the town. A greater proportion of finewares was found around the forum in the town centre than in the outer suburbs, where jars were more common. Such trends suggest we can identify zonation within Leicester that can be related back to different depositional practices and ultimately patterns of use and consumption.
The widespread adoption of massproduced terra sigillata dining vessels in first century ad Gallia Narbonensis (south-central France) represents a dramatic change in relations of production and consumption. Benjamin Luley drew on anthropology to suggest these vessels were not necessarily used in identical ways in different social contexts. He tested his ideas using quantitative analysis of large data samples from contexts where pottery had been discarded. At least three distinctive patterns of sigillata use are
apparent, rather than homogenous use throughout the province. Due to the nature of the archaeological record these patterns do not directly reflect actual dining sets or meals, but rather long-term, recurring practices among different social groups.
The Big Data on the Roman Table network encouraged archaeologists to explore new ways of visualising quantitative information using digital technology. Archaeologists frequently use multivariate statistical methods such as correspondence analysis to consider patterns in large and complex data sets. Results are usually presented as “bi-plots” that require technical understanding to read, and have limits for showing geographical and spatial information. Martin Sterry experimented with fine-grained colour representations generated by computer mapping software and interpolation to create geographic visualisations in data variability. Mapped variables can be easily changed to explore different aspects of the data, including information to help understand social aspects of tableware use and changes over time and place. Sterry also applied his method to map variations in pottery and related artefact collections from contexts across Roman Leicester.
Automating big data
Archaeologists commonly use structured tables in spreadsheets and databases to store and manipulate their data. These need to be modified to accommodate new kinds of information and to compare varied data across different projects, especially when they contain large amounts of information.
In computer science the term “ontology” is used to describe the naming and definition of categories and relationships between them, and their analysis by machine learning and related applications. Ontology components (such as “classes”, “properties” and “individuals”) can be used to consider relationships between items like pottery vessel types, and ideas and information about them, in large data sets.
As part of the Big Data on the Roman Table network, Allison reused pottery data from earlier excavations (by Werner Zanier) of a Roman fort at Ellingen in Germany, which she had collated into electronic spreadsheets for a previous project. She restructured her data so archaeologist Daniël van Helden and Yi Hong, a University of Leicester computer scientist, could analyse needs and create an ontology to support complex queries about tableware use across large amounts of linked data. Allison’s Ellingen pottery data were entered into specialist software structured to match the data model, so its value could be explored for analysing Roman tablewares in terms of consumption and between sites.
Recording and analysing archaeological information about Roman tablewares is costly and time consuming. It would help to automate some of the process. Ivan Tyukin, Konstantin Sofeikov, Jeremy Levesley and Alexander N Gorban, informatics
researchers at the University of Leicester, collaborated with Allison and Cooper to scan whole Roman pots held by Leicester’s Jewry Wall Museum, using smartphones to create a prototype system (Arch- i- Scan) for automated pottery recognition. They showed it was technically possible to classify and discriminate between objects of different types on-the-fly from a limited number of images. The technology is based on recently published research by Gorban, Tyukin and associates revealing peculiar geometric properties of finite but large samples of data in high dimension. The ambition is to create dedicated software that turns commonly available devices such as smartphones or tablets into scanners capable of classifying even small vessel sherds by form and fabric.
Archaeologists usually visually compare excavated pots with examples from physical or illustrated “type series” to divide their finds into categories of vessel “shape” and “form”. This is slow, and results can vary between projects, hindering data comparison. Martin Pitts and Jacqueline Christmas, a University of Exeter computer scientist, automated this process and used mathematical modelling to measure and analyse vessel shapes in a more standard, accurate and objective way. They first made pdf- format scans of many pages of accurate standardised drawings of vessel types, published for the well-known Camulodunum (Roman Colchester) type-series. Scanning technology located, defined the boundaries, extracted and oriented each individual pot drawing. From these processed images they extracted information about vessel width, height, interior volume, edges and the shape of the cross-section outline, and then calculated more complex measures of shape such as a pot’s “circularity”. This work offers new possibilities for classifying vessel form in a way sensitive to the physical characteristics of pots relative to other vessels in an archaeological collection, and enables archaeologists to visualise their results.
Pitts and Christmas applied their system to data about first century ad imported fineware pottery from sites previously analysed by Pitts for research into Britain's first Roman cities, published in 2014. They used pottery data that had been labelled both with forms (such as platter and bowl) and shape identifiers extracted by their digital system. Results show how pottery forms cluster for a given set of features, and how the features may be used to compare finds from different sites. This example presents a useful visual breakdown of the formal variation within the subjectively defined “bowls” category. Plotting all the forms in a notional class of vessels in this way has clear advantages in highlighting, for example, which types are not well categorised as bowls, as well as different sub-categories of bowl types that may be productively grouped in further analyses.
These are only selected examples of a much wider range of research developed by the Big Data on the Roman Table network, and published online in Internet Archaeology in 2018 with open access.
Sarah Colley co-edited and contributed a coauthored paper to the Internet Archaeology publication: see http://intarch.ac.uk/ journal/issue50/index.html. She is honorary research fellow at the School of Archaeology & Ancient History University of Leicester