Abstract
The distribution of Roman pottery depended on the transportation system which moved it. Here we trace developments in these distributions during the Roman period in Britain to document how the transportation system developed and assess its impact on the island’s economy. We created a database with records from 775 excavations at 652 sites, and data on over two million pottery sherds. By analyzing the changing distributions of pottery from production centers, we are able to measure improvements in the Roman transportation system over time. These improvements seem to have been most rapid soon after conquest, with transport costs almost halving in the first century of Roman occupation. As the road network expanded and transportation technology improved and pottery gained access to wider markets, producers’ dominance over their local markets declined as rival products became more accessible, and certain industries dramatically increased their outputs. Production by small industries fell in our Middle and Later Roman periods.
A striking fact about trade and industry in Roman Britain is the ease with which the problems of transport were overcome. Wherever manufactured or produced, goods were easily transported throughout the province. (Frere, 1967).
1 Introduction
Pottery has long provided evidence for the trade of goods from the wider Roman world into Britain. Amphorae, for example, have been used to trace the movement of wine, olives, and fish sauce from the Mediterranean into Rome’s northernmost province. There was also a thriving cross-channel trade in ceramics from Gaul into Britain. However, mapping the movement of pottery and other goods within Britain is altogether more difficult and has correspondingly attracted less attention. A complication to understanding the domestic distribution of pottery is that transportation systems not only move goods but can also stimulate the manufacture and movement of goods, and so contribute to economic activity. Roads connect industries to both the raw materials they need and also the markets which consume their products. Roads also link population centers, encouraging the exchange of goods which in turn encourages the creation of industries and the development of specialization (e.g., Banister & Berechman, 2000; Docherty & MacKinnon, 2013; Helling, 1997; Rodrigue & Notteboom, 2020). The consequence is that the development of pottery industries in Roman Britain will have been entangled in the evolution of the Roman transportation system.
The Roman conquest of AD 43 transformed transport in Britain. Although there have been finds of prehistoric trackways, wheel-ruts, wooden boats, and chariot fittings in Britain, their number and extent are dwarfed by the transportation system created under the Roman empire (AD 43–410). A network of roads was quickly established across southern Britain, initially for military use (Haynes, 2002) and subsequently strengthened (Davies, 2008). Horses and cattle became widely used as pack and draught animals (Mattingly, 2006). Canals were dug, such as the Foss Dyke and Car Dyke in Lincolnshire, and the Old Tillage in Cambridgeshire (Evans, Macaulay, & Mills, 2017). Ports and wharfs were established on the coast and along rivers, many with specialized facilities for loading and storing goods. There was a thriving coastal trade, particularly along the Channel (Fulford, 2007; Jones, 2012). However, it is one thing to identify the elements of a transportation system which might have moved goods; it is something else entirely to demonstrate that certain goods were moved and to quantify the impact of transportation on their trade (c.f. Wilson, 2009, p. 214). Frere’s breezy optimism about the transportation of manufactured goods, quoted at the start of this article, stands in need of evidence.
There have been efforts to quantify the costs and impact of transport elsewhere in the Roman world, drawing on evidence surviving in Egyptian papyrus (e.g., Adams, 2007) and historical documents (e.g., Temin, 2013). Temin’s analysis of wheat prices in the early Empire period is particularly noteworthy, as he was able to use written records to reconstruct shipping costs in the Mediterranean. Drawing on surviving prices from Spain, Italy, Turkey, Palestine, and Egypt, he undertook a regression analysis which demonstrated that the price of wheat at various locations was based on the cost of grain in Rome (the location where demand was greatest), less the cost of transporting wheat to Rome. His analysis suggested the price of wheat would double for every 1,000 km it was transported by sea (Temin, 2013, Table 2.3). Thus, there was evidence of both an integrated market for grain and a relationship between distance and price.
Regrettably, it is not possible to repeat Temin’s analysis for Roman Britain, as the necessary evidence does not survive. There are currently just two documents which refer to transportation costs: one letter from Vindolanda including haulage costs to transport grain to an unknown destination (Bowman & Thomas, 1994, no. 649), and a contract drawn up in London in AD 62 giving fees for the transport of 20 loads of unstated provisions over the 30 km between London and St Albans (Tomlin, 2016, no. WT45). This is plainly far too little upon which to base any kind of analysis. However, it is possible to use the kernel of Temin’s approach, using quantities of pottery in lieu of price data. In the analysis that follows, we build upon the long tradition in British archaeology of quantifying Roman pottery by fabrics and the industries that produced them. As the locations of several dozen pottery industries have now been identified (Study Group for Roman Pottery, 2022; Swan, 1984), it is possible to estimate the distance from production to consumption locations and use these data to conduct a regression analysis of the type that Temin undertook. By looking at a large number of the industries that operated in Britain, and by examining their distributions over time, we are able to deduce changes in the overall cost of transporting goods in Roman Britain. We also observe the economic consequences of improved transport on competition amongst pottery industries, and the growth in their output.
2 Theory and Expectations
The goal of this article is to assess the impact of the evolving Roman transportation system on the British pottery industries at the level of the overall province. As there has been no previous attempt to quantify the impact of transportation on industry, our focus is on a first-order analysis of the macroscopic effects, rather than a rich account of local transportation (which doubtlessly varied somewhat from region to region across Britannia). This section outlines our expectations about how pottery distributions would have developed as the province’s transport systems were established and expanded. The following sections compare our expectations against evidence from pottery recovered from 775 excavations across Roman Britain (in addition to the patterns discussed below, Ortman et al. (n.d.) considered the effects of changing transportation costs for levels of diversity in pottery production and consumption).
As a common concern of economic analyses of the ancient world is the extent to which modern “economic rationalism” is projected onto the past, this section also sets out the key assumptions we have made about how the people of Roman Britain made economic decisions which produced the patterns we describe.
2.1 Quantifying Transport-Related Costs
In any society, transporting goods invariably involves costs in the broad sense – vehicles to transport goods; roads, canals, ports, and other infrastructure; fuel, whether for vehicles or animals; as well as time that people spend moving goods during which they cannot do other activities. Most of these costs increase with distance – more roads, more fuel, and more time on the road. The result is that goods normally become increasingly costly the further they are transported. On the assumption that the people in Roman Britain who manufactured and distributed pottery over decades did not ruin themselves in the process, it follows that they must have minimally covered the increased costs in the broad sense. Presumably they did by passing on the costs to those acquiring their pots. As a result, one would expect the number of goods transported from a given production center to have decreased with distance, as consumers became less willing to purchase more costly goods. Notice that this model presumes that there was not a standard price for pottery. It also does not presume that there was a centralized market for pottery – only that people making and moving pots tried to cover their costs, and people acquiring pots likewise did not have unlimited resources for acquiring them.
There are potentially different fall-off patterns for different kinds of goods and transport modes. In the case of Roman pottery industries in Britain, it has been nearly 50 years since Fulford and Hodder (1975) demonstrated that the typical distribution pattern belongs to one of the simplest forms: an exponential fall-off in relative frequency with distance (equation (1), c.f. Hodder & Orton, 1976). According to this model, on average, the proportion of pottery in an assemblage (P x ) is a function of the distance x from its source, a constant factor (ε) related to the level of “friction” in the transportation system, and the proportion of the pottery in assemblages at the source (P 0).
To estimate the decay rate, equation (1) can be simplified by taking the logarithm of both sides equation (2). The result is a simple linear function in which the logarithm of the proportion of pottery in an assemblage from a given source will be directly proportional to its distance x from that source, with the inverse of the slope of the relationship between the log-proportion and distance representing the decay rate.
This formula makes analyzing the effect of transportation costs straightforward: plotting the log-proportion of pottery from numerous assemblages against their distance from source should yield a straight line with slope −ε.
Naturally, there are many other social factors affecting pottery consumption apart from transportation-related costs. Ethnographic studies of traditional pottery industries (e.g., Kramer & Douglas, 1992; Vander Linden, 2001; Vossen, 1984) highlight factors such as consumers’ perceptions of pots’ quality or status, their need for pots, their access to money to pay or goods to barter, their personal preferences, and the prestige they attribute to the potter. Distribution also may be constrained by kinship rivalries, caste prohibitions, and administrative boundaries. All of these would have tended to push proportions in individual pottery assemblages away from a simple linear relationship between distance and consumption. However, none of these other factors is related to the distance the pottery has travelled from its source; indeed, on aggregate and across assemblages, these factors should tend to cancel one another out. As a result, so long as distribution data are drawn from over a wide area, a best fit line drawn through the assemblage data should correspond to the transport-related costs involved in delivering pottery and allow the frictional factor ε to be estimated.
In addition to putting a value on the amount of friction in the transport system, the value of ε tracks changes in transportation costs in Roman Britain over time. A high value of ε indicates pottery is not travelling far from its source, while a low value indicates a wider distribution of goods. In another study (Ortman et al., n.d.), we show that this value is interpretable as a measure of transport costs through its effect on the likelihood of movement of a good over a given distance prior to consumption. Thus, by calculating ε for different time periods, the relative change in transport costs over time can be quantified. As noted in the introduction, there are a variety of ways transportation could have improved in Roman Britain: road construction, populations moving closer to roads, replacing pack animals with wagons, barges, and ships, and building facilities to manage the loading and storage of goods, such as wharves, warehouses, and ports. Improvements in transportation would have been greatest when roads, infrastructure and new modes of transport were first introduced; later additions, such as additional roads or ports, might have further reduced the distances and handling costs somewhat, but the marginal improvements would not have been as large as the introduction of new transportation systems in the first place. These observations lead to the first two of our expected results:
Expected outcome 1: As transport networks expanded in Roman Britain, the magnitude of ε would have decreased.
Expected outcome 2: As transportation improved in Roman Britain, the effects would have been greatest as systems were first established.
2.2 The Effects of Improved Transportation
Assuming for a moment that the transport system in Roman Britain improved over time, there would have been several consequences for the distribution of pottery and other goods. For example, as transport-related costs dropped, the total cost for pottery at any given distance should also have dropped, and presumably the price at a given distance would have fallen also. As a result, the area over which it was cost-effective for ceramic producers to distribute their goods to potential buyers would have expanded. This in turn meant pottery industries could reach more consumers. However, the same would also have been true for rival producers. High transportation costs would have excluded competing products from areas around pottery production centers. But as transport improved, competitors would have been increasingly able to ship their produce more widely – even to their rivals’ production centers. This leads to our third expected outcome:
Expected outcome 3: as transportation costs fell (i.e., as ε got smaller) pottery assemblages at production sites would have received more competitors’ wares and therefore contained a smaller proportion of local products, so P 0 would have declined.
2.3 The Growth of Specialization and Decline of Diversity
As pottery producers were able to reach more consumers, they would have been able to do two things. First, the scale of production could have expanded to supply the increasing number of consumers. Second, overall levels of specialization in pottery production could have increased. This is an example of Adam Smith’s observation that “the division of labor is limited by the extent of the market” (Smith, 1776/2007, Chapter III) – a large market provides more opportunities for a division of labor, which in turn makes industries more efficient (Arrow, 1994; Kelly, 1997). Specialization by pottery industries able to access large markets would also have put pressure on less specialized producers, who would have increasingly struggled to produce pots at a competitive price. Ultimately, some of the smaller producers would have ceased production. The consequence would have been that, as pottery markets expanded in Britain, the number of industries producing pots – particularly small non-specialized industries – would have declined.
Expected outcome 4: As transport costs decreased over time, (a) the average level of non-specialist or local production would have decreased; (b) the average level of production from each specialist industry would have increased; and (c) the variation in production levels between producers would have decreased.
3 Data
To assess our expectations of the impact of transportation on pottery distribution in Roman Britain, we created a large database covering the entire province. Given the well-documented challenges in compiling and re-using Roman pottery data (e.g., Perrin, 2011; Timby, 2016), we have described the construction of this database in some detail in the Appendix. Our final database contained pottery data from 750 excavations at 663 ancient settlements. The locations of these is shown in Figure 1, left. We assigned pottery in each assemblage (excluding amphorae) to known pottery industries. Using the production dates and quantities of pottery in each assemblage, we then apportioned pottery from each excavation into three time periods – Early Roman (AD 50–150), Middle Roman (c.150–250) and Late Roman (c. AD 250–400) – using a technique known as uniform probability density analysis (see the Appendix for more details). We then calculated the proportion of each industry’s contribution to each assemblage in each of the three periods.
(left) Distribution of settlements (black) and towns (white) (N = 663) and (right) location of sourced industries (N = 49) in the dataset. Known Roman roads and potentially navigable rivers are also shown.
Using the gazetteer of Roman kilns produced by Swan (1984) and the Study Group for Roman Pottery (2022) we also identified the location of 43 pottery industries with well-documented production sites (Figure 1, right). With this information, we calculated the straight-line distance between each production site and each excavation in our database. Our reasons for using a straight-line distance rather than one based on known Roman roads is explained in the Appendix.
4 Results
For every pottery industry in our dataset, we calculated the logarithm of the proportion of each assemblage made up of its products, then plotted this against the straight-line distance between the excavation and the relevant kiln for that industry. We did this for each of our three time periods. We limited the analysis to distances of less than 200 km from production sites because beyond this distance pottery proportions become unstable (generally because the proportions are very low, even in large assemblages). Fortunately, over 95% of all the measured distances between kiln sites and excavations were less than 200 km.
Figure 2 shows the distribution of pottery from six Late Roman pottery industries, along with an interpolation of the density of each pottery type within assemblages. Figure 3 gives the corresponding scatterplots for these six industries, along with lines of best fit. Table 1 presents basic statistics for each of these six industries along with characteristics of the lines of best fit when the assemblages were plotted against distance (calculated using an ordinary least squares regression and standard errors estimated using the White correction for heteroskedasticity). As expected, the contribution of each pottery industry declined with distance, with the exception of a handful of industries for which examples were only identified in a few excavations.
Spatial distributions of six Late Roman pottery industries: Severn Valley, Oxfordshire, South Dorset Black Burnished Ware, Mancetter–Hartshill, Lower Nene Valley, and Hadham. Pottery was transported by a mixture of road, river, and sea.
The proportion of assemblages plotted against the distance from production sites, along with lines of best fit, for the six Late Roman pottery industries illustrated in Figure 2: Severn Valley, Oxfordshire, Dorset Black Burnished Ware, Mancetter–Hartshill, Lower Nene Valley, and Hadham. Note that, on average, the proportion of assemblages made up by every industry declines with distance, reflecting mounting transport costs.
| Type | N Sites | Intercept (P 0) | SE | Slope (−ε) | SE | R 2 | F-statistic | P-value |
|---|---|---|---|---|---|---|---|---|
| Severn valley | 120 | 0.750 | 0.321 | −0.063 | 0.007 | 0.435 | 90.769 | 0.0000 |
| Oxfordshire | 341 | −1.771 | 0.200 | −0.024 | 0.002 | 0.246 | 110.446 | 0.0000 |
| Dorset black burnished ware | 164 | 0.002 | 0.299 | −0.022 | 0.003 | 0.248 | 53.355 | 0.0000 |
| Mancetter/Hartshill (midlands) | 98 | −3.811 | 0.505 | −0.015 | 0.006 | 0.104 | 11.197 | 0.0012 |
| Lower Nene valley | 288 | −1.514 | 0.158 | −0.032 | 0.002 | 0.453 | 236.572 | 0.0000 |
| Hadham | 105 | −3.331 | 0.418 | −0.026 | 0.007 | 0.153 | 18.570 | 0.0000 |
The data in Table 1 show that the values for R 2 – which measure the proportion of variance from the best fit due to independent factors – are often quite low. There are two reasons for this. First, this is because the values used are proportions rather than absolute values, and proportions are sensitive to sampling errors in the data. Second, as discussed earlier, there are many factors that affect why people choose pottery beyond transport-related costs – indeed, we should expect these many other factors to have been collectively the main source of variation in pottery acquisition, rather than just transport. However, despite all these social, economic, and geographic factors which disperse the data, there is still plainly an underlying decline in pottery consumption with distance. This indicates that transportation costs did indeed play a consistent role in pottery consumption patterns in Roman Britain.
The six industries illustrated in Figures 2 and 3 were amongst the largest which operated in Roman Britain. Most were much smaller. Figure 4 shows the total number of sites in each from which pottery from each industry was found, over our three time periods. Data from industries represented by just a handful of sites have the potential to generate significant errors in calculating the value of ε, the magnitude of the slope of the best fit line, which models the transportation costs. If the values for ε for each industry were simply averaged, then plainly all industries would make an equal contribution, but the contribution of minor industries would be disproportionate, given the small number of data points they are based on. A more robust approach is to include all of the log-proportions and distances, across all industries, in a single analysis for each time period. The results of this approach are shown in Figure 5. In this chart, we binned distances into 10 km increments, calculated the average log-proportion in each bin, and then calculated the best fit line through the average in each bin. (As before, the best fit line was calculated using a least-squares regression, and the standard error estimated using the White correction for heteroskedasticity.) Table 2 presents key figures from these charts, including the slope (−ε) of the line in each time period, and intercept where x = 0.
The number of sites in our database for each located industry and time period.
Relationship between relative consumption and distance across all industries, by period. Data are in 10 km bins and the large circles represent the mean value of each bin. Note that the slope of the relationship becomes less negative over time, by a factor of about two, which is consistent with declining transport costs. ERP = Early Roman Period (AD 50–150); MRP = Middle Roman Period (AD 150–250), and LRP = Late Roman Period (AD 250–400).
Summary of results, plotting log-proportion of pottery (P) against distance in 10 km bins
| Period | Independent | Dependent | N obs | Intercept | SE | Slopea | SE | F-statistic | P-value | R 2 |
|---|---|---|---|---|---|---|---|---|---|---|
| ERP | Distance (10 km bins) | ln[P] | 534 | −1.5704 | 0.1662 | −0.0362 | 0.0025 | 209.43 | 0.0000 | 0.2825 |
| MRP | Distance (10 km bins) | ln[P] | 1,320 | −2.5447 | 0.1139 | −0.0214 | 0.0013 | 236.22 | 0.0000 | 0.1520 |
| LRP | Distance (10 km bins) | ln[P] | 1,712 | −2.4066 | 0.1064 | −0.0205 | 0.0012 | 278.20 | 0.0000 | 0.1399 |
aDifferences in slope between ERP and LRP are statistically significant (one-sided t = −5.682; d.f. = 805.06; P = 9.301 × 10−9).
Standard errors were calculated using the White correction for heteroskedasticity.
The slope of the best fit line is steeper in our Early Roman period (ERP; ε = 0.0362) than later in our Middle Roman period (MRP; ε = 0.0214), and Late Roman period (LRP; ε = 0.0205), which confirms our first expectation. The larger value of ε in our ERP reflects a relatively high level of transport-related friction, and consequently most pottery was not being distributed far from where it was produced. There was a substantial drop in the level of friction by our MRP and a further slight improvement by our LRP. This slight later improvement is consistent with our second expectation: that the major improvements in transportation-related costs were delivered when the road transportation system was first introduced; the creation of further roads and canals later on contributed only marginal improvements at the province level.
The value of the intercepts in Figure 5 confirms our third expectation. The intercept (P 0) represents the average proportion of assemblages made up of local wares at production sites (i.e., where x = 0). In our ERP, high transportation costs meant that rival producers had difficulty shipping their products far, and consequently assemblages around production sites were dominated by local products (averaging around 22% of assemblages). However, once transportation improved in the MRP, rivals were better able to send their goods to production sites, so the proportion of local products at production sites declined to less than 10%. An example of the twin processes of territorial expansion and declining local dominance is illustrated in Figure 6a, which shows the Horningsea industry, which operated on the River Cam just north of Cambridge. Its expansion westward in our MRP was probably facilitated by the opening of the Old Tillage canal in the second century AD, linking the Cam to the River Great Ouse. This saw it enter the former territory of the Early Roman Verulamium industry (which was based in St Albans, Hertfordshire, and declined in our MRP), as well as the poorly-understood Godmanchester industry (which also operated only in our ERP). Another beneficiary of the declining Verulamium industry was the Hadham industry, whose growth in the MRP and LRP is shown in Figure 6b, taking over much of the former circulation zone of Verulamium wares in East Anglia.
Expansion of the Horningsea industry between our ERP (red) and LRP (blue), and the Hadham industry between the MRP (green) and LRP (blue). Expansion of the Horningsea industry followed the opening of the Old Tillage canal around AD 140–150 (Evans et al., 2017), allowing pottery to be transported by barge from the River Cam westward to the Great Ouse and Nene. But opening transport also brought in pottery from other industries, reducing the dominance of Horningsea wares in the Cam valley, reflected in the falling contour values in the distribution maps, and the falling proportion of Horningsea wares in assemblages within 30 km of the kiln sites in Cambridgeshire. By contrast, the Hadham industry took over parts of the territory in eastern England which had previously been supplied by the Verulamium and Colchester industries, both of which declined in our MRP. Its distribution shows a rising proportion in assemblages right across its range, with a slightly larger increase at distance from the kilns, indicating that transportation costs did improve between our MRP and LRP.
The volume of pottery in our database – just over 32.6 tonnes and about 2.3 million sherds – was sufficiently large for us to test our fourth group of expectations. Figure 7 plots the raw weight of pottery assignable to individual industries in each of our three time periods. It shows that the median production level of specialized Roman pottery industries increased over time, as local pottery manufacturers disappeared. It also shows that that the variance in output between industries also decreased, indicating that overall, pottery industries were converging in their output. The very largest industries continued to expand, however, as producers took advantage of their expanded markets.
Changes in pottery production over time in our database. The plots illustrate the total weight of pottery assigned to each sourced industry in apportioned assemblages, and dots plot individual industries. ER = Early Roman (AD 50–150), MR = Middle Roman (AD 150–250), and LR = Late Roman (AD 250–400). Over time, the median production level increased, as did the production level of the largest industries. However, the variance in production levels across industries declined with time, indicating that production volumes of pottery industries were converging.
Finally, it is possible to estimate the change in transport costs over time in Roman Britain, by calculating the ratio of ε at each successive time periods. Using data from Table 2, the ratio of our ERP to MRP periods was −0.0362 ÷ −0.0214 = 1.691, which indicates that the costs of transport almost halved over this time. The change from our MRP to LRP was much smaller: −0.0214 ÷ −0.0205 = 1.044. This confirms our second expectation: that the bulk of improvement in transportation occurred with the initial construction of the road network, ports, traction animals, and so forth in the ERP. This result also finds some support from independent data from the Rural Settlement of Roman Britain project. Its data indicate 60% of “roadside settlements” appeared between AD 43 and 100, compared with 35% of all other settlements (derived from Allen et al., 2018), indicating Britons were quick to exploit the newly established road network. Although roads and roadside towns continued to be established after the first century AD, their effects on pottery consumption were more marginal when set alongside the wholesale transformation that followed the Roman conquest.
5 Discussion
The perishable nature of most goods transported in Roman Britain makes tracking their movement impossible. In Britain, tracing transportation networks has had to rely on proxies: chiefly pottery and stone (e.g., Fulford, 2004, p. 314). While the distribution and fall-off curves of individual types of Roman pottery have been explored several times before (e.g., Allen & Fulford, 1996; Fulford & Hodder, 1975; Hodder, 1974; Hodder & Orton, 1976, pp. 98–153), this is the first project to map all of the major pottery industries which operated in Roman Britain, calculate all their fall-off rates, and use the results to investigate the consequences for the transportation system. Our analysis found that, over time, pottery became more widely dispersed, which is apparent in the falling values of ε, reflecting improvements in the transportation system. As this happened, and competing industries were able to ship their goods more widely, pottery production centers lost their dominance over their local markets. This is reflected in the falling values of P 0 in our MRP and LRP, as assemblages at production sites included more products from competing industries. As the areas within reach of pottery industries expanded, some production centers began to specialize. In time, this allowed them to out-compete non-specialized local producers, and the latter began to disappear from the market. With the decline of non-specialized pottery producers, the proportion of pottery assemblages made up of specialized-industry products increased.
The variation in pottery consumption is not all due to transport-related costs. Leaving aside issues created by sampling bias and errors in pottery classification, other factors would have affected what pottery people choose to acquire. These might have included personal taste, prestige of particular producers, local territorial boundaries, clan or kinship rivalries, adoption or access to money, and state procurement for the military. In addition, some pottery might have been transported as containers for other higher-value goods rather than in its own right. This study in no way denies the role of these – indeed, the scale of most of the R 2 values indicates they were important, nonetheless, our figures show an underlying decline in pottery consumption with distance affecting all the large pottery industries we analyzed, and the only underlying factor they are all likely to share is the cost of transporting their wares. This study demonstrates that, despite the undoubted complexities of the Roman pottery trade, it is nonetheless possible to isolate the broad contribution of transportation.
Although this article has focused on pottery, the improvements in transportation system in Roman Britain, and the associated decline in the transport costs, would have affected many other aspects of the economy within the province. Any goods which would have been moved any distance – particularly in bulk – would have been impacted. Over time, grain, meat, timber, metal, and manufactured goods would all have been able to travel further for less total cost. And, while distances over which they moved would not have been the same as for pottery, all these commodities would have come under the same production and competitive pressures that we outlined for pottery, which would in turn have shaped their production, distribution, and price. While grains and wood have mostly decayed, and metals probably long since been recycled, there would still be traces of changes in their production, against which this prediction might be assessed. For example, data exist on the sizes and construction dates of grain stores; the rates at which animals were slaughtered; and the dates in which mines were excavated. By using pottery transportation as a framework, it may be possible to build a more integrated picture of how the Roman economy of Britain operated, and to understand how different parts interacted with one another.
That we have been able to detect even small changes in economic performance while working with noisy data attests to the potential power of “big data” approaches in archaeology. In Britain, the type of synthesis we have undertaken here is increasingly feasible – albeit with considerable amounts of data cleaning – because of the large body of gray literature now available, notably through the Archaeology Data Service. We have been fortunate to build on the work of previous synthetic studies, particularly the Rural Settlement of Roman Britain project (Allen et al., 2018), and the Defended Small Towns of Roman Britain project (Fulford, Lodwick, & Smith, 2018). We will be uploading our own data to the Archaeology Data Service in the hope that others will be able to extend it even further to yield new insights. We encourage all our colleagues working in the field to contribute to the cumulative advance of archaeological research.
Acknowledgements
Portions of this research were supported by a grant from the James S. McDonnell Foundation. We thank Luis Bettencourt, Christopher Evans, Katie Anderson, and Neil Holbrook for discussions. We also wish to dedicate this article to our co-author, Lisa Lodwick, who passed away during the final drafting of this article. It was her experience with the Roman Rural Settlement project that convinced us of the feasibility of this study. Her early encouragement was instrumental, as was her knowledge of Roman Britain, and she was a key link between the American and British collaborators in this study. The other authors are deeply saddened that an early career scholar and collaborator of such intelligence and vision is no longer with us.
-
Funding information: Portions of this research were supported by a grant from the James S. McDonnell Foundation (#220020438, to Ortman).
-
Author contributions: R.W., L.L., and S.O. designed the research; O.B. and S.O. collected the data; R.W. and S.O. performed the analyses and prepared the figures; and R.W. and S.O. prepared the manuscript with contributions from all co-authors.
-
Conflict of interest: The authors declare no conflicts of interest in the conduct of this research and preparation of this paper. The corresponding author attests that to the best of his knowledge, L.L., the deceased author, did not have conflicts of interest.
-
Data availability statement: The raw data analyzed in this article are interoperable with the Rural Settlement of Roman Britain and Defended Small Towns of Roman Britain datasets and will be deposited with the Archaeology data Service upon publication.
Appendix
A Materials and Methods
The analysis presented in this article is based on a large database of Roman pottery, compiled specifically for this project. The final database is available from Archaeology Data Service. This Appendix describes the construction of the database, and particularly the choices we made in selecting and categorizing material.
A.1 Scope
Unlike many existing syntheses of Romano–British pottery, which are regional in scope, this study deliberately spanned the entire Roman province. There were three reasons for this. First, the inclusion of a large number of pottery industries clearly made isolating the contribution of transport easier to identify than a smaller sample. Second, as pottery specialists themselves acknowledge, it is possible to misidentify the source of individual pottery sherds. However, over a large dataset, these random misidentifications will tend to cancel out. Third, as Figure 2 illustrates, the very largest pottery industries – and those with the most data to work with – also spanned large parts of Britannia; a regional study would have made understanding their distribution impossible. Fourth, a larger dataset clearly has greater statistical power than a small one, even after we simplified some of the data. And finally, the distribution and arrangement of Roman roads across Britain clearly indicates that the Romans planned for long-distance movement within Britain – a point underscored by references in letters from Vindolanda to trade with locations as distant as London.
We focused primarily on “civilian” settlements and did not include military sites (unless they developed into towns). This was partly because such sites were not included in the original Rural Roman Settlement database, and because procurement of pottery for the military appears to have been centrally organized and distinct from the self-organized systems that would have supplied farms, villas, and towns (Allen & Fulford, 1996).
We excluded several pottery types from our analysis (although data for them were recorded in our database). Amphorae were invariably transported with goods in them rather than empty, so their distribution would have largely reflected demand for their contents. We also did not assess transport costs of continental wares, as sea portage would have featured prominently in their overall transport costs, and these are estimated at just a few percent of overland costs during the Roman period (e.g., Duncan-Jones, 1974, pp. 366–369), and so would have significantly biased our results.
A.2 Assembling the Database
Our starting point was the pottery dataset created for the Rural Settlement of Roman Britain project (RSRB, Allen et al., 2018), which had 3,652 records. We merged this with the database for the Defended Small Towns of Roman Britain project (STRB, Fulford et al., 2018), which had data from 209 selected sites. Both datasets are available from the Archaeology Data Service. The data in these sources are generally limited, consisting of a brief summary of the pottery assemblage, the total number and weight of sherds, and separate counts of the amphora, mortaria, and samian sherds. There is no breakdown of pottery fabrics in the RSRB database, and only limited fabric information in the STRB database. The pottery data and excavation reports from which these summaries derive also vary considerably in their quality, coverage, and chronological resolution.
To enrich the data from these sources, we added the individual pottery fabric codes and quantifications used in the primary excavation reports cited in the RSRB and STRB databases. These were a mix of grey literature, published articles, and monographs. Fortunately, many of the reports used by the RSRB had been scanned and uploaded to the Archaeology Data Service. These were supplemented by manual searches of UK county-level journals and the libraries at the Universities of Cambridge and Colorado. We also incorporated data from previous studies of South Dorset Black Burnished Ware (Allen & Fulford, 1996) and Horningsea Ware (Evans et al., 2017), and we added pottery data from excavations in seven primary towns (Caerwent, Cirencester, Colchester, Exeter, Leicester, Silchester, and Wroxeter). The list of primary sources we consulted for these towns is presented in Table A1. The primary sources for the other sites are all available with the source data.
Primary data sources
| Rural settlements | Roman settlement of Roman Britain | Allen et al. (2018) |
| Small towns | Defended small towns of Roman Britain | Fulford et al. (2018) |
| Primary towns | Caerwent | Wessex Archaeology (2009) |
| Cirencester | Holbrook (1998), McWhirr (1986) | |
| Colchester | Crummy (1984, 1992), Gascoyne and Radford (2013), Symonds and Wade (1999) | |
| Exeter | Bidwell (1979), Rippon and Holbrook (2021a,b) | |
| Leicester | Connor and Buckley (1999) | |
| Silchester | Creighton and Fry (2016), Fulford (1984), Fulford, Clarke, and Eckardt (2006) | |
| Wroxeter | Barker, White, Pretty, Burd, and Corbishley (1997), Ellis (2000), Webster (2002), White et al. (2013), | |
| South Dorset black burnished ware | Allen and Fulford (1996) | |
| Horningsea ware | Evans et al. (2017) |
Excavation reports summarize the pottery finds in a variety of ways: some by different areas of the site, some by groups of features, and some by chronological periods. We entered the data in whatever divisions were used in the reports, and subsequently aggregated the figures to create a total for the site. Pottery quantities were presented in a variety of ways. Since our focus was transportation, the attribute of greatest relevance was weight. Where excavation reports gave only sherd counts, we multiplied these counts by 15 g (the average sherd weight in our database) to create an estimated weight. If only estimated vessel equivalents were reported, we treated these as counts and converted them to estimated weights. As some of the pottery assemblages in these lists are small, and could have unduly biased the analysis, we chose to exclude all sites with fewer than 500 sherds, or 4,000 g of pottery when only the weight was reported.
A.3 Rationalizing the Fabric Codes
At the start of the project, we had envisioned using our database to track the distribution of specific wares from individual industries. As the database took shape, however, it became apparent this was unrealistic. Our final database contained over 6,000 separate pottery fabric codes – an absurd and unworkable number given that fewer than 100 pottery production sites are known in Britain, and the productions sites of many industries remain unknown. It would only have been possible to track individual wares by using a very much smaller – and consequently less reliable – dataset.
This chaos of fabric codes is a byproduct of how Roman pottery analysis has developed in Britain, and the resulting diversity of pottery “standards.” The latter include:
National Roman Fabric Reference Collection (NRFRC, Tomber & Dore, 1998) developed in the 1990s – but by no means used nationally and now increasingly out of date as many more pottery industries have been identified
codes developed by individual contracting units, based in cities and counties around the UK (e.g., the Gloucester City Fabric Type Series, the Bedfordshire Fabric Series, and Museum of London Archaeology Fabric Codes)
codes created for major excavations (e.g., the Colchester Fabric Type Series), subsequently re-used for other projects.
specialist- or site-specific codes.
As tracking individual wares was not feasible on any scale, we instead set about identifying all products from pottery industries. So, for example, all wares produced at Verulamium (course white slipped, fine white slipped, mica dusted, mortaria, and marbled ware) were assigned a single industry code, based on the NRFRC. Fortunately, many excavation reports did include details of the codes they used – but a great many did not and lacked any mention of the fabric series employed or the publication in which they were defined. Our growing database also contained many identical codes which plainly referred to different fabrics. Assigning fabric codes to known industries involved many iterations between our database and excavation reports over several months, using known codes from particular counties or used by individual contractors to identify unknown codes. In our final database, just under 40% of pottery by weight could be assigned to identifiable British and Continental fabrics: 90 fabrics to a specific British industry and 29 imported fabrics. The remaining 60% were assigned to 43 generic fabrics (e.g., “oxidized,” “reduced,” “coarse grey ware,” etc.). Most of the fabrics in this last group were probably the output of small-scale local production, although it doubtlessly also includes some pottery codes we were unable to translate to a known industry.
A.4 Production Dates and Apportioning Assemblages into Chronological Periods
An important issue for our project was assigning quantities of pottery to specific time periods. Unfortunately, there is no consensus document on the production dates for pottery industries and individual wares. The dates we used were a compromise between four sources: Tyers’ Roman Pottery in Britain (Tyers, 1996/2003) and his Potsherd website (Tyers, 2022); a list of fabrics produced by the Museum of London Archaeology (MOLA, 2019); and the gazetteer of Roman kilns produced by the Study Group for Roman Pottery (2022). These collectively agreed on broad dates of operation for the industries we analyzed. For more details on the production spans we used for each industry, see Ortman et al. (n.d.).
In our analysis, we assigned pottery to one of these broad date ranges: Early Roman (AD 50–150), Middle Roman (AD 150–250), and Late Roman (AD 250–400). While these dates do not correspond exactly to the normal chronological division of Roman archaeology in Britain, these periods correspond to the start and end dates of key industries, including imported Samian ware, Lower Nene Valley ware, Oxfordshire ware, New Forest ware, and SE Dorset Black Burnished ware. Naturally, pottery production in some industries started before or after these cut-off dates and continued across the period boundaries. In this case, we apportioned the assemblage using a technique known as “uniform probability density analysis” (Ortman, 2016; Roberts et al., 2012), using a script made available by Matthew Peeples (https://github.com/mpeeples2008/UniformProbabilityDensityAnalysis). This assumes that pottery was produced at a uniform rate for entire period the industry was active, giving a uniform distribution across the production span. This is obviously not likely to be true in detail, but in the absence of more robust information on the output histories of individual kilns, this approach introduces fewer biases into the analysis than other theoretical distributions, such as the normal distribution, would. The quantity of pottery from a given industry in an assemblage is multiplied by the corresponding uniform distribution, and the results summed and then divided by the total assemblage size to create a summed probability density distribution. In this distribution, the height for a given year is the sum of the estimated amount of pottery of each industry deposited at the site during that year. These annual distributions were then summed across groups of years to create sub-assemblages for our three chronological periods: Early, Middle, and Late. (This method for apportioning pottery to time periods is related to the similar summed probability density approach which has become widespread in the demographic interpretation of radiocarbon dates, e.g., Bird et al., 2022; Downey, Bocaege, Kerig, Edinborough, & Shennan, 2014; Rick, 1987; Robinson, Bocinsky, Bird, Freeman, & Kelly, 2021; Shennan et al., 2013.)
A.5 Locating Production Sites
Our analysis depended on us being able to measure the distance between sites where pottery was produced and sites where it was eventually deposited. However, UK archaeologists have not yet identified the kiln sites for all known Roman pottery industries. Of the 90 Romano–British pottery industries we identified, we located production sites for 49, drawing primarily from the gazetteer of Roman kilns produced by Study Group for Roman Pottery (2022) based on Swan (1984), and supplemented with information in Tyers (1996/2003) and the NRFRC (Tomber & Dore, 1998).
Using these data, we calculated the straight-line distances between production sites and excavations. Where pottery was produced at several dispersed sites, we used the distance to the nearest kiln. We used a straight-line distance rather than distance via road, canal, or sea-lanes, as our knowledge of the latter is very far from complete. While the major public roads (viae publicae) in Britain have been largely traced (e.g., Davies, 2008; Margary, 1955), smaller roads created by local town councils (viae vicinales) and private landowners (viae privatae) – routes which would have shortened some travel distances considerably – have mostly been lost. Also, the network of recognized routes does have gaps in it, so algorithms in geographic information systems used to calculate distances via roads sometimes create artificially long routeways to circumvent these gaps – something no Roman wagon-driver would have done in practice. We did compare the difference between straight-line distances and distance via the incomplete road network as calculated by GIS. The average distance by road was around 50 km longer than a straight-line distance. We judged this was an unacceptable difference for our analysis, as we had calculated that 95% of pottery travelled less than 200 km from its source. Using road-based distances also produced a much greater dispersal of our data, suggesting that current knowledge of the road system is too incomplete for our purposes. Our final assessment was that straight-line distances were probably a closer approximation to actual distances travelled than one based on the surviving fragments of the Roman road network.
A.6 The Completed Database
Our final database held pottery data from 775 excavations on 652 distinct settlements. It contained 1,408 pottery groups organized into our three chronological periods. They contained a total of 21,691 quantified entries organized by industries (or generic fabrics where we were unable to identify an industry). Table A2 summarizes the number of settlements in our dataset, organized by size.
Number of settlements in the dataset, organized by size
| Settlement size | Total |
|---|---|
| Small rural (<1–3 ha) | 308 |
| Medium rural (4–8 ha) | 82 |
| Large rural (9+ ha) | 44 |
| Fortress1 | 3 |
| Small town | 14 |
| Primary town2 | 9 |
| Uncertain | 192 |
| Total | 652 |
Notes: 1) The database includes assemblages from the fortresses at Wroxeter and Exeter. 2) The assemblage from Exeter has also been divided into early and late phases.
A.7 Final Comment
Pottery is a fundamental material to archaeologists investigating Roman Britain. But the lack of agreed fabric codes, production dates, and even methods for quantifying pottery gravely reduces its usefulness beyond the reporting of individual sites. It is no surprise to us that the Roman Rural Settlement project included no pottery analysis in its three seminal monographs (Allen, Lodwick, Brindle, Fulford, & Smith, 2017; Smith, Allen, Brindle, & Fulford, 2016; Smith et al., 2018). The current state of pottery reporting is largely a result of the “atomization” of commercial archaeological practice by development-funded excavations and the planning system they operate in. These place primacy on individual sites, with little consideration for the long-term re-use of data. While there is no real prospect of retrospectively amending the existing corpus of pottery reports, we would urge all those responsible for excavation, analysis, and curation to develop and adhere to a national system of pottery recording.
References
Adams, C. (2007). Land transport in Roman Egypt: A study of economics aand administration in a Roman Province. Oxford: Oxford University Press.10.1093/acprof:oso/9780199203970.001.0001Search in Google Scholar
Allen, J. R. L., & Fulford, M. G. (1996). The distribution of South-East Dorset Black Burnished Category I Pottery in South-West Britain. Britannia, 27(1996), 223–281.10.2307/527045Search in Google Scholar
Allen, M., Blick, N., Brindle, T., Evans, T., Fulford, M. G., Holbrook, N., … Smith, A. (2018). The rural settlement of Roman Britain: An online resource [data-set]. York: Archaeology Data Service [distributor]. doi: 10.5284/1030449.Search in Google Scholar
Allen, M., Lodwick, L., Brindle, T., Fulford, M. G., & Smith, A. (2017). The rural economy of Roman Britain (New Visions of the Countryside of Roman Britain, Vol. 2; Britannia Monograph Series No. 30). London: Society for the Promotion of Roman Studies.Search in Google Scholar
Arrow, K. J. (1994). The division of labor in the economy, the polity, and society. In J. M. Buchanan & Y. J. Yoon (Eds.), The return to increasing returns (pp. 69–84). Ann Arbor: University of Michigan Press.Search in Google Scholar
Banister, D., & Berechman, J. (2000). Transport investment and economic development. Abingdon: Routledge.Search in Google Scholar
Barker, P., White, R., Pretty, K., Burd, H., & Corbishley, M. (1997). The Baths Basilica Wroxeter: Excavations 1966–90 (Archaeological Reports 8). London: English Heritage.Search in Google Scholar
Bidwell, P. T. (1979). The Legionary Bath-House and Basilica and Forum at Exeter (Exeter Archaeological Reports 1). Exeter: Exeter City Council.Search in Google Scholar
Bird, D., Miranda, L., Vander Linden, M., Robinson, E., Bocinsky, R. K., Nicholson, C., … Freeman, F. (2022). p3k14c, a synthetic global database of archaeological radiocarbon dates. Scientific Data, 9(1), 27.10.1038/s41597-022-01118-7Search in Google Scholar
Bowman, A. K., & Thomas, J. D. (1994). The Vindolanda writing-tablets (Tabulae Vindolandenses II). London: British Museum Press.Search in Google Scholar
Connor, A., & Buckley, R. (1999). Roman and Medieval occupation in Causeway Lane, Leicester. Leicester: University of Leicester Archaeological Services.Search in Google Scholar
Creighton, J., & Fry, R. (2016). Silchester: Changing visions of a Roman Town (Britannia Monograph Series No. 28). London: Society for the Promotion of Roman Studies.Search in Google Scholar
Crummy, P. (1984). Excavations at Lion Walk, Balkerne Lane, and Middleborough, Colchester, Essex (Colchester Archaeological Reports 3). Colchester: Colchester Archaeological Trust.Search in Google Scholar
Crummy, P. (1992). Excavations at Culver Street, the Gilberd School, and other sites in Colchester 1971–1985 (Colchester Archaeological Reports 6). Colchester: Colchester Archaeological Trust.Search in Google Scholar
Davies, H. (2008). Roads in Roman Britain. Oxford: Shire Books.Search in Google Scholar
Docherty, I., & MacKinnon, D. (2013). Transport and economic development. In J. P. Rodrigue, T. Notteboom, & J. Shaw (Eds.), The sage handbook of transport studies. London: Sage.10.4135/9781446247655.n13Search in Google Scholar
Downey, S. S., Bocaege, E., Kerig, T., Edinborough K., & Shennan, S. (2014). The Neolithic demographic transition in Europe: Correlation with Juvenility index supports interpretation of the Summed Calibrated Radiocarbon Date Probability Distribution (SCDPD) as a valid demographic proxy. PLoS ONE 9(8), e105730.10.1371/journal.pone.0105730Search in Google Scholar
Duncan-Jones, R. (1974). The economy of the roman empire: Quantitative studies. Cambridge: Cambridge University Press.Search in Google Scholar
Ellis, P. (2000). The Roman Baths and Macellum at Wroxeter: Excavations by Graham Webster, 1955–85 (Archaeological Reports 9). London: English Heritage.Search in Google Scholar
Evans, J., Macaulay, S., & Mills, P. (2017). The Horningsea Roman pottery industry in context (East Anglian Archaeology 162). Bar Hill: Oxford Archaeology East.Search in Google Scholar
Frere, S. (1967). Britannia: A history of Roman Britain. London: Routledge & K. Paul.Search in Google Scholar
Fulford, M. (2004). Economic structures. In M. Todd (Ed.), Companion to Roman Britain (pp. 309–326). Oxford: Oxford University Press.10.1002/9780470998861.ch17Search in Google Scholar
Fulford, M. G. (1984). Silchester: Excavations on the defences 1974–1980 (Britannia Monograph Series). London: Society for the Promotion of Roman Studies.Search in Google Scholar
Fulford, M. G. (2007). Coasting Britannia: Roman trade and traffic around the shores of Britain. In C. Gosden, H. Hamerow, P. de Jersey, & G. Lock (Eds.), Communities and connections: Essays in honour of Barry Cunliffe (pp. 54–74). Oxford: Oxford University Press.10.1093/oso/9780199230341.003.0012Search in Google Scholar
Fulford, M. G., Clarke, A., & Eckardt, H. (2006). Life and labour in Late Roman Silchester: Excavations in Insula IX since 1997 (Britannia Monograph Series 22). London: Society for the Promotion of Roman Studies.Search in Google Scholar
Fulford, M. G., & Hodder, I. R. (1975). A regression analysis of some late Romano-British pottery: A case study. Oxoniensia, 39, 28–35.Search in Google Scholar
Fulford, M. G., Lodwick, L., & Smith, A. (2018). Defended small towns of Roman Britain [data-set]. Archaeology Data Service [distributor]. York. doi: 10.5284/1046751.Search in Google Scholar
Gascoyne, A., & Radford, D. (2013). Colchester, fortress of the war god: An archaeological assessment. Oxford: Oxbow Books.10.2307/j.ctvh1dvq3Search in Google Scholar
Haynes, I. (2002). Britain’s first information revolution. The Roman army and the transformation of economic life. In P. Erdkamp (Ed.), The Roman army and the economy (pp. 111–126). Amsterdam: J. E. Gieben.10.1163/9789004494374_007Search in Google Scholar
Helling, A (1997). Transportation and economic development: A review. Public Works Management and Policy, 2(1), 79–93.10.1177/1087724X9700200108Search in Google Scholar
Hodder, I. (1974). Regression analysis of some trade and marketing patterns. World Archaeology, 6(2), 172–189.10.1080/00438243.1974.9979600Search in Google Scholar
Hodder, I., & Orton, C. (1976). Spatial analysis in archaeology. Cambridge: Cambridge University Press.Search in Google Scholar
Holbrook, N. (1998). Cirencester: The Roman town defences, public buildings and shops. Cirencester excavations V. Cirencester: Cotswold Archaeological Trust.Search in Google Scholar
Jones, J. E. (2012). The maritime landscape of Roman Britain: Water transport on the coasts and rivers of Britannia (BAR British Series 556). Oxford: BAR Publishing.Search in Google Scholar
Kelly, M. (1997). The dynamics of smithian growth. The Quarterly Journal of Economics, 112(3), 939–964.10.1162/003355397555398Search in Google Scholar
Kramer, C., & Douglas, J. E. (1992). Ceramics, caste and kin: Spatial relations in Rajasthan, India. Journal of Anthropological Archaeology, 11, 187–201.10.1016/0278-4165(92)90021-3Search in Google Scholar
Margary, I. D. (1955). Roman roads in Britain. London Phoenix House.Search in Google Scholar
Mattingly, D. (2006). An imperial possession: Britain in the Roman Empire, 54 BC–AD 409. London: Penguin Books.Search in Google Scholar
McWhirr, A. (1986). Houses in Roman Cirencester. Cirencester Excavations 3. Cirencester: Cirencester Excavation Committee.Search in Google Scholar
MOLA. (2019). Roman Pottery Codes (online resource). https://www.mola.org.uk/roman-pottery-codes.Search in Google Scholar
Ortman, S. G. (2016). Uniform probability density analysis and population history in the northern Rio Grande. Journal of Archaeological Method and Theory, 23(1), 95–126.10.1007/s10816-014-9227-6Search in Google Scholar
Ortman, S. G., Bulik, O., Wiseman, R., Lobo, J., Bettencourt, L. M. A., & Lodwick, L. (n.d.). Transport costs and economic change in Roman Britain. Under Review. European Journal of Archaeology.Search in Google Scholar
Perrin, R. (2011). Research strategy and updated agenda for the study of Roman Pottery in Britain. Study Group for Roman Pottery Occasional Paper No.1.Search in Google Scholar
Rick, J. W. (1987). Dates ad Data: An examination of the Peruvian Preceramic Radiocarbon record. American Antiquity, 52(1), 55–73.10.2307/281060Search in Google Scholar
Rippon, S., & Holbrook, N. (2021a). Roman and Medieval Exeter and Their Hinterlands: From Isca to Excester (Exeter: A Place in Time 1). Oxford: Oxbow Books.10.2307/j.ctv2npq99t.13Search in Google Scholar
Rippon, S., & Holbrook, N. (2021b). Studies in the Roman and Medieval archaeology of Exeter (Exeter: A Place In Time 2). Oxford: Oxbow Books.10.2307/j.ctv2npq9kzSearch in Google Scholar
Roberts, J. M. Jr., Mills, B. J., Clark, J. J., Hass, W. R. Jr., Huntley, D. L., & Trowbridge, M. A. (2012). A method for chronological apportioning of ceramic assemblages. Journal of Archaeological Science, 39(2012), 1513–1520.10.1016/j.jas.2011.12.022Search in Google Scholar
Robinson, E. R., Bocinsky, K., Bird, D., Freeman, J., & Kelly, R. L. (2021). Dendrochronological dates confirm a Late Prehistoric population decline in the American Southwest derived from radiocarbon dates. Philosophical Transactions of the Royal Society B: Biological Sciences, 376(1816), 20190718.10.1098/rstb.2019.0718Search in Google Scholar
Rodrigue, J-.P., & Notteboom, T. (2020). Transportation and economic development. In J-.P. Rodrique (Ed.), The geography of transport systems (5th ed.) (pp. 90–101). Abingdon: Routledge.10.4324/9780429346323-3Search in Google Scholar
Shennan, S., Downey, S. S., Timpson, A., Edinborough, K. Colledge, S., Kerig, T., … Thomas, M. G. (2013). Regional population collapse followed initial agriculture booms in mid-Holocene Europe. Nature Communications, 4, 2486. doi: 10.1038/ncomms3486.Search in Google Scholar
Smith, A. (1776/2007). An inquiry into the nature and causes of the wealth of nations. Amsterdam: Metalibri Digital Library.10.1093/oseo/instance.00043218Search in Google Scholar
Smith, A., Allen, M., Brindle, T., & Fulford, M. G. (2016). The rural settlement of Roman Britain (New Visions of the Countryside of Roman Britain, Vol. 1; Britannia Monographs 29). London: Society for the Promotion of Roman Studies.Search in Google Scholar
Smith, A., Allen, M., Brindle, T, Fulford, M. G., Lodwick, L., & Rohnbogner, A. (2018). Life and death in the countryside of Roman Britain (New Visions of the Countryside of Roman Britain, Vol. 3; Britannia Monographs 31). London: Society for the Promotion of Roman Studies.Search in Google Scholar
Study Group for Roman Pottery. (2022). The Pottery Kilns of Roman Britain by Vivien Swan [website]. https://romankilns.net/about/.Search in Google Scholar
Swan, V. (1984). The pottery kilns of Roman Britain (RCHME Supplementary Series 5). London: Her Majesty’s Stationery Office.Search in Google Scholar
Symonds, R. P., & Wade, S. (1999). Roman Pottery from excavations in Colchester, 1971–86 (Colchester Archaeological Reports 10). Colchester: Colchester Archaeological Trust.Search in Google Scholar
Temin, P. (2013). The Roman market economy. Princeton: Princeton University Press.10.23943/princeton/9780691147680.001.0001Search in Google Scholar
Timby, J. (2016). Review of current practices in roman pottery analysis. Approaches to the investigation, analysis and dissemination of work on Romano-British rural settlements and landscapes: A review. https://cotswoldarchaeology.co.uk/wp-content/uploads/2017/08/RRS-Methodology-Paper-4-pottery.pdf (Accessed April 2022).Search in Google Scholar
Tomber, R., & Dore, J. (1998). The National Roman Fabric Reference Collection: A handbook. London: Museum of London Archaeology Service.Search in Google Scholar
Tomlin, R. S. O. (2016). London’s first voices: Roman London’s first voices: Writing tablets from the Bloomberg excavations, 2010–14 (MOLA Monograph 72). London: MOLA.Search in Google Scholar
Tyers, P. (1996/2003). Roman Pottery in Britain. London: Routledge.10.11141/ia.1.6Search in Google Scholar
Tyers, P. (2022). Potsherd: Atlas of Roman Pottery [website]. https://potsherd.net/home/.Search in Google Scholar
Vander Linden, M. (2001). Social dynamics and pottery distribution in the Faro department, northern Cameroon. African Archaeological Review, 18, 135–151.10.1023/A:1010976304972Search in Google Scholar
Vossen, R. (1984). Towards building models of traditional trade in ceramics: Case studies from Spain and Morocco. In S. E. Van Der Leew & A. C. Pritchard (Eds.), The Many dimensions of Pottery. Ceramic Archaeology and Anthropology (pp. 339–397). Amsterdam: Institute for Pre- and Proto-history.Search in Google Scholar
Webster, G. (2002). The Legionary Fortress at Wroxeter: Excavations by Graham Webster, 1955–85 (Archaeological Report 19). London: English Heritage.Search in Google Scholar
Wessex Archaeology. (2009). Caerwent Roman Town, Monmouthshire, South Wales. Archaeological evaluation and assessment of results. OASIS ID 68736.01. York: Archaeology Data Service.Search in Google Scholar
White, R. H., Gaffney, C., & Gaffney, V. L. (2013). Wroxeter, The Cornovii and the urban process: Final report on the Wroxeter Hinterland project 1994-1997 (Vol. 2: Characterizing the City). Oxford: Archaeopress.10.2307/j.ctv17db2sjSearch in Google Scholar
Wilson, A. (2009). Approaches to quantifying Roman trade. In A. Bowman & A. Wilson (Eds.), Quantifying the Roman Economy (pp. 213–249). Oxford: Oxford University Press.10.1093/acprof:oso/9780199562596.003.0009Search in Google Scholar
© 2023 the author(s), published by De Gruyter
This work is licensed under the Creative Commons Attribution 4.0 International License.
