Evidence for Roman eating and drinking from the Mola database

I’ve blogged before on the Food for Thought blog: (https://notjustdormice.wordpress.com/2015/01/09/evidence-for-roman-eating-and-drinking-from-the-mola-database/ ) about my ongoing PhD research into “Big Data” and evidence for Roman food recovered during developer-funded archaeological work in England, as part of the EngLaId Project, and I thought I’d post a quick update now, as I’ve recently come into possession of an amazing new data-set courtesy of Museum of London Archaeology (Mola) and the excellent archaeologists who work there, including particularly Karen Stewart.

I think it’s fair to say that the database produced by developer-funded archaeological work in England is still an under-used resource, particularly in academia and the reasons for this are no doubt varied and complex. There is no space in a short blog to go into these reasons in depth; however, I think that among the main reasons is the fragmented nature of the sector, with many different organisations carrying out the work, and particularly the reliance on freelance specialists to carry out much of the finds specialist work upon which our knowledge of eating and drinking in the past relies. One consequence of this fragmentation is the great diversity of ways in which digital data is stored and managed, with specialists often tailoring their finds databases to the needs of individual projects.

At Mola though, things are a little different. Again the reasons may be complex and, I suspect, relate to the scale and complexity of the organisation, but at Mola all finds, and indeed context data is stored in a single database that includes information on all of the different finds recovered from at least 11,000 sites, some of which were excavated as long ago as 1915. The result is an incredibly flexible tool for research, which has already been used to do some fantastic work on food, including a study of the fishing industry and fish consumption in medieval London (Orton et al. 2014), which makes use of anatomical data and detailed chronological information to link changes in the nature of commercial fishing to changes in the anatomical elements present in deposits of cod remains.

The data that I’ve extracted from the Mola database relates to the period between 1500BC and AD1086 (the period covered by the EngLaId project) which covers the long period between the first wide spread permanent settlement and the origins of the modern settlement pattern. These data falls into four main categories: context data, ceramics, animal bones and charred plant remains and the idea is to plot changes in these data, both chronologically and spatially using GIS and to see (among other things) whether data analysed on this scale changes our understanding of Roman food consumption and particularly what was carried over from the Iron Age and what continued into the early medieval period. I haven’t yet crunched the numbers in a sufficiently detailed way to report on my analysis – but when I have I’ll be posting another blog to let everybody know how it went. Hopefully, the results will be interesting!

Figure 1: Distribution of Mola Sites

Dan_MOLA_dots

References

Orton, D C, Morris, J, Locker, A, and Barrett, J H, 2014 Fish for the city: meta analysis of archaeological cod remains and the growth of London’s northern trade, Antiquity, 88, 516-30.

 

Mapping pottery

Following on from suggestions (primarily by Prof. Barry Cunliffe) at our Academic Advisory Board meeting last year, we started thinking about how we might map aceramic (or minimally ceramic-using) zones through our time period. Due their general commonness and generally diagnostic nature, ceramic finds are probably the most commonly used method for dating archaeological contexts and, thus, by extension sites as a whole. As such, in areas where ceramic objects were little used, it becomes more difficult (and probably more expensive) to date sites. This, in turn, is likely to result in sites in aceramic areas being less precisely dated. This could, therefore, bias the distribution of sites of a particular period in the archaeological record, as sites in aceramic zones within a particular period are less likely to be securely dated to that period.

However, actually mapping aceramic zones is not especially easy. To do so, one must first map areas where ceramics are used, and collating data on that scale for 2,500 years of human history would almost certainly be a research project in itself on a similar scale to EngLaID as a whole. Therefore, we had to try and obtain the results of previous attempts at pottery synthesis.

We began with prehistory. The only existing national database which we could find of later prehistoric (Later Bronze Age to the Roman conquest) pottery was that created by Earl et al. (2007), archived at the ADS. The data collection for that project took place in 1995-6, so it is almost twenty years out of date, but it was the only reasonably comprehensive data source available to us. We hope that the broad brush picture will have not changed substantially in the past twenty years (albeit see below for the early medieval period), but until another such project is undertaken it is impossible to be certain.

1 earl_pot_density_ALL
Density of later prehistoric pottery records

Simply plotting the density of records in this database shows a distinct bias in the distribution of later prehistoric pottery towards the southern and eastern half of England (essentially, Cyril Fox’s “lowland” zone of Britain), with the exception of a notable lack of pottery in the Weald and on the South Downs, and small peaks of pottery in western Cornwall, East Yorkshire, and County Durham. North Devon and large swathes of the West Midlands and the north west show a distinct lack of ceramic usage (or at least recovery by archaeologists).

2 earl_pot_density_phased
Density of LBA to EIA vs MIA to Conquest period pottery records

We can nuance this picture slightly by looking at change over time.  Following discussion with the prehistoric experts on the team, I split the data temporally into two broad time periods: Late Bronze Age to Early Iron Age, and Middle Iron Age to the conquest. The pattern that results seems to show a movement (of the peak in density) away from Wessex and northwards into the East Midlands, which could be the result of any number of factors (population growth, environmental change, etc.).

3 earl_pot_density_UnspecPrehist
Density of unspecified prehistoric pottery records

However, there are also large numbers of unspecified later prehistoric records in the database (especially in East Anglia), so temporal patterns should not be too heavily emphasised.

4 earl_pot_sherdcount_ALL_overEI
Sherdcounts of prehistoric pottery over density of prehistoric records in EH’s Excavation Index

Many of the records in the database also record sherd counts of the assemblages recorded, which helps to nuance the picture further. In an attempt to see if the patterns produced when mapping the database records simply stemmed from where archaeological work takes place (which inevitably they must to some extent), I mapped the records against the density of later prehistoric events recorded in English Heritage’s Excavation Index. As the map above shows, there does appear to be a fairly strong correlation. However, there are low peaks in the density of events in the north west which are not represented in the pottery database, so the pattern is not entirely determined by modern archaeological practice.

5 earl_pot_sherdcount_ALL_over14Cprob
Sherdcounts of prehistoric pottery over modeled radiocarbon probability for same period

To take this further, I also mapped the sherd counts against a modeled surface of radiocarbon probabilities for the same period (see previous post). This seems to show that there are areas of relatively high radiocarbon probability in apparently aceramic zones, suggesting that activity was taking place in those areas at that time. This helps to suggest that our aceramic zones, although partially biased by patterns of modern archaeological practice, are reasonably likely to be real. For later prehistory, then, it does appear that there was less use of pottery in the north west, the West Midlands, and in north Devon.

Moving on to the Roman period, the best source of national level data which we could find is Paul Tyers’ excellent Potsherd website. Naturally, collating sherd count level data for the Roman period would be an immense task (due to the incredible amount of ceramics deposited on Roman sites): as such, Tyers maps pottery by ware type on a presence / absence basis (by 10x10km square). His maps are all dated 2004, so we assume that the data mapped is around ten years out of date. Again, it is assumed that broad brush patterns will not have changed immensely, although proving that would be difficult.

Tyers provides encyclopaedic detail on his website, but does not offer direct downloads of his data. Fortunately, his maps are all relatively high resolution and all constructed in the same way, so it is possible to perform various trickery on them in order to study them further in GIS. It then becomes feasible to sum Tyers’ maps together and produce a map of variability in pottery wares across Romano-British England. As such, this is not directly comparable with the later prehistoric maps discussed above, as we are mapping the number of different ceramic wares deposited across England for the Roman period, rather than the density of records (i.e. site assemblages) for prehistory.

6 tyers_ware_density
Roman pottery variability

The map above shows the overall variability in Roman pottery across England, based on Tyers’ data. Dark blue areas have no pottery (the aceramic zones we sought) and red areas have many different types of pottery. The results are quite interesting: the greatest variability in pottery wares is in a similar region to the greatest density of later prehistoric pottery records, i.e. in the south and east of England. However, the zone covered is significantly larger and there are also further significant peaks in otherwise “quiet” areas, particularly around the Roman cities and military sites.

7 tyers_ware_density_domestic_imported
Roman pottery variability: domestic vs imported

We can, however, take this further. Comparing variability in domestic and imported wares, we can see that the areas with greatest variety in imports were around the major settlements and, in particular, around the Thames estuary. By contrast, the greatest variability in domestic wares was more widespread.

8 tyers_ware_density_coarseware
Roman pottery variability: coarsewares
9 tyers_ware_density_fineware
Roman pottery variability: finewares
10 tyers_ware_density_terrasigilata
Roman pottery variability: terra sigilata / Samian ware
11 tyers_ware_density_mortaria
Roman pottery variability: mortaria
12 tyers_ware_density_amphorae
Roman pottery variability: amphorae

Further patterns emerge when looking at more specific groups of wares. Coarsewares are quite well spread; finewares largely restricted to the south; terra sigilata is very clustered; mortaria are well spread and possibly rural in character; amphorae are very tightly clustered into small areas.

13 tyers_ware_density_overtime
Roman pottery variability: over time (smaller version here for tablets etc.)

We can also look at change over time, which also shows some interesting patterns, with the peak of variability being most widespread (albeit largely southern) in the 3rd and 4th centuries. The strong 5th century peak in Cornwall is caused by imported wares from the eastern Mediterranean.

Overall, the patterns produced by mapping Tyers’ data in this way can potentially tell us interesting things about pottery supply in the Romano-British period, in particular in regard to economic factors (as availability of different ceramic wares must be linked to economic conditions / opportunity to some extent). Also, although we have mapped somewhat different things, certain comparisons can be made with the later prehistoric data: areas with less ceramics in the Romano-British period were less widespread than in later prehistory, but in generally the same places, especially if you mentally factor out the influence of military garrisons.

14 Vince_Blinkhorn_C9_pot
9th century AD pottery industries (after Vince 1993; Blinkhorn and Dudd 2012)

Moving finally to the early medieval period, we struggled to find any datasets of anything like the degree of comprehensiveness of either the Earl et al. or Tyers data. The best source discovered was a fairly old article by Alan Vince (1993), which mapped the major pottery industries of the 9th century AD. However, it does appear that this map was now quite out of date, as his zone of Ipswich Ware (highlighted in red above) was much more restricted than the areas recorded recently by Blinkhorn (2012) (black dots and shaded in grey: it is assumed that the grey shading is record density by modern administrative region, but the map had no legend). This also only really covers the very end of our period, when wheel thrown pottery came back into production in England: we have no data for the mid-5th to 8th centuries. As such, it is hard to draw any conclusions at all about the early medieval picture.

In conclusion, largely aceramic zones probably existed in later prehistory in the north west, the West Midlands and the south west. These largely persisted into the Roman period, albeit with ceramic using areas around the military installations and larger settlements. In the early medieval period, we do not have enough data to reach even tentative conclusions, but we might assume that the same areas continued to use less pottery than in the south and east? Or that might be plain conjecture.

Chris Green

References:

Blinkhorn, P. 2012. The Ipswich Ware Project: Ceramics, Trade and Society in Middle Saxon England. Medieval Pottery Research Group Occasional papers.

Earl, G., E. Morris, S. Poppy, K. Westcott, T.C. Champion. 2007. Later Prehistoric Pottery Gazetteer. http://dx.doi.org/10.5284/1000013

Tyers, P.A. 2014. Potsherd. http://potsherd.net/atlas/potsherd

Vince, A. 1993. “Forms, Functions and Manufacturing Techniques of Late Ninth- and Tenth- Century Wheelthrown Pottery in England and their Origins.” In D. Piton (ed.), Travaux du Groupe de Recherches et D’Etudes sur la Céramique dans le Nord – Pas-de-Calais; Actes du Collque D’Outreau (10 -12 Avril 1992). Numéro hors-série de Nord-Ouest Archéologie, pp.151-64.

Addendum – 12/01/2015:

In an attempt to see if the aceramic zone in the north of England in later prehistory was genuine or an artefact of modern archaeological practice, we mapped hillfort excavations prior to 1997 recorded in English Heritage’s Excavation Index (mapped as green diamonds) against hillfort ceramic assemblages recorded by Earl et al. (up to 1996). The results do appear to show that, on the whole, hillfort excavations do produce pottery in the southern half of England, but largely do not in the northern half, with the notable exception of northern Northumberland. This suggests that this is likely to be a genuine aceramic zone:

15 earl_pot_sherdcount_Hillforts_overEI
Hillfort excavations recorded in EH’s Excavation Index (pre-1997) against hillfort ceramic assemblages