1 Introduction

This article presents the results obtained from the compositional analysis of over 1500 Iron Age and Roman copper alloy samples collected from northern Britain for a PhD thesis (Dungworth 1995). The numbers of samples and the range of information recorded for each sample (compositional data, context, date, etc.) are so large that conventional paper publication is of limited value. Publication in Internet Archaeology provides a unique opportunity for readers to interrogate the database for themselves.

1.1 Aims

The research was designed to use the analysis of copper alloys to answer wider (non-technological) questions about metal use and society in general. In particular, the selection of a limited area (northern Britain) but a long period of time (from the beginning of the Iron Age to the end of the Roman period) allowed the recognition of changes in metal use over time to be recorded. This period sees considerable social and economic changes during the Roman occupation (Millett 1990). The analysis of copper alloys from this period could act as an illustration of Romanisation. In particular, the first widespread use of brass is associated with the Roman Empire (coins and military equipment). The incidence of brass could act as an 'index of Romanisation' (cf. Millett's (1980) study of the distribution of fine wares in Sussex). The total quantity of Iron Age and Roman copper alloy artefacts from northern Britain is very high (10,000+) and so it was necessary to select a sample which was representative of that total.

1.2 Objectives

The detailed examination of Iron Age and Roman copper alloys had to be based on the systematic collection of a large number of samples from the full chronological, typological and cultural range available. This could only be achieved if a limited geographical area was selected for study. The area chosen for study was northern Britain (here defined as the region from the Trent-Mersey to the Forth-Clyde). This area was chosen as it provided a wide range of settlement types. In particular, the apparent lack of Romanisation in much of the landscape of northern Britain (in contrast to the forts and few towns) implies that there was a varying degree of social change in the Roman period. Less substantial evidence of social and economic change might be observable, however, in more portable items of material culture - such as items made from copper alloys. For these reasons the analysis concentrates on major alloying elements (zinc, tin and lead) rather than trace elements. Those alloys containing mostly zinc are known as brass, those containing mostly tin as bronze, and those containing both as gunmetal. For the purposes of this research, alloys have been described as leaded if they contain more than 1% lead. A variety of ways in which the analytical results can be represented using graphs and tables will be explored.

For the results of this archaeometallurgical research to have the widest possible relevance to archaeology as a whole, it is essential that the samples reflect the range of available material. To this end, samples were initially selected to be representative of the archaeological record rather than restrict analysis to a limited range of artefact types or a single site. The selection of sites was initially based on a survey of published excavations.

Fig.1 Map of northern Britain showing the principal sites examined

This quickly showed that some categories of sites were well represented in the archaeological record (e.g. forts) while others were less well known (e.g. farmsteads). The reasons for this are connected to the history of archaeological research and are discussed below. If the selection of sites was to be representative of the known archaeological record then I would be left with large numbers of forts and few other sites. In order to overcome this, and allow inter-site comparisons it was essential that some site categories received more attention than others.

Fig.2 Date of the samples collected

The sites have been chosen to cover all possible areas in northern Britain. The chronological range of the finds analysed is shown in Figure 2. The dating of artefacts is not without its problems. For this research project, context dating has been used as a starting point for the dating of samples. Contexts for the Roman period are usually dated by associated pottery; however, a proportion of the finds in each context are residual. In addition, the dating of pottery itself rests on associations with other artefacts. At times this method of dating runs the risk of employing circular logic. For the Iron Age, dating is less precise and is often based on broad cultural similarities. Where the commonly agreed typological dating of an artefact disagreed with the context date (owing to residuality) the earlier typological date was used.

In some cases the contextual or typological dating could be very precise (to a decade) and in others rather more vague ('Late Roman'). For the Roman period, samples were usually assigned to centuries (with the first century covering only the latter part of the century, after the Roman conquest). In many cases objects or contexts could only be assigned to less precise dates. Early Roman was used to cover the first and second centuries, Mid Roman to cover the second and third centuries, and Late Roman to cover the third and fourth centuries.

Context dating is more difficult for the pre-Roman period, and for the purposes of this research project a three-part division has been used. The first period (which may overlap with the late Bronze Age) covers the earliest possible Iron Age as seen at Staple Howe and Castle Hill, Scarborough. The second period covers the Iron Age before any evidence of Roman presence or the presence of any Roman (or Gallo-Roman) imports. The largest contribution to this second group is made by 'Arras culture' burials. The late Iron Age covers the end of the Iron Age when there is increasing evidence of direct and indirect contact with the Roman Empire. This period also includes finds from indigenous sites from the latest Iron Age where the phase of occupation often continues into the period of Roman occupation (e.g. Thorpe Thewles and Dragonby). The final category of putative Iron Age material considered is 'Celtic' metalwork. Stylistically, many of these objects have their origins in the Iron Age, but most are stray finds and so cannot be closely dated. At least some are found on Roman military sites and production may have continued into the Roman period.

While every effort has been made to make the selection of samples as representative of the archaeological record as possible, this aim has had to be abandoned where the available archaeological record can be seen to be 'biased'. The Iron Age in Britain covers at least as long a period of time as the Roman but there is far less material available for study. This could reflect differences in population and the availability and use of metal, or it could reflect differences in deposition practices and modern archaeological strategies. If sampling was carried out as a strict proportion of the available material then early Roman copper alloys would dominate this research while earlier and later periods (the Iron Age and the late Empire) would be under-represented. It was therefore necessary to concentrate on some periods in order to acquire samples of sufficient size to make chronological and cultural comparisons reliable.

The sites were classified into different categories (such as military, villa, farmstead). These are categories which are widely used in accounts of Roman Britain (Frere 1987; Salway 1981). These are often formed into settlement hierarchies with forts and towns at the top and farmsteads at the bottom. These assumed settlement hierarchies are used as a framework for the interpretation of copper alloy differences. Sites were initially selected from those known from published excavation reports. Those sites which had been published in the last few decades were particularly favoured as it was often possible to determine context dating for the artefacts selected. Unpublished but recently excavated sites provided a valuable source of material where contextual information was readily available.

Fig.3 Types of sites from which samples were taken

The range of types of site from which samples were taken is shown in Figure 3. The large number of finds from military sites reflects previous research into the Roman period in northern Britain, which has traditionally concentrated on the military aspects of the occupation. This interest probably has its origins in the traditional narrative account of the Roman conquest derived from Roman texts. Excavations of forts were seen as important as they could answer questions about the date of the occupation of any particular fort. Forts were regarded as being relatively easy to excavate as they were regularly planned and excavation has continued to be popular as forts are relatively rich in small finds (including copper alloy ones). Occupation histories at different sites were then tied in to a narrative history derived from classical sources such as Tacitus (e.g. Dudley & Webster 1965; Frere 1987). Urban sites are less well represented in northern Britain than in southern Britain, and these northern sites have seen relatively little exploration. Small indigenous rural sites are fairly common in northern Britain and in upland areas they are often well preserved (Jobey 1982a); excavations, however, have rarely produced many small finds. This has not made these sites attractive propositions for further excavation and made the sites difficult to date. As this research project aims to examine Romanisation through copper alloys it was important to obtain as many samples from these rural sites as possible. It was necessary to include samples from rural sites that did not have any context dating, whereas sampling from military sites could be more rigorous.

Not all sites which were examined could be classified easily using the categories of site used in Figure 3. Settlement size and form do not always conform to a 'type', and many categories tend to merge into each other. Welton Wold, North Humberside, for example, is too large to be a farmstead but too small to be a 'larger rural settlement' (such as Shiptonthorpe). For the purposes of this analysis Welton Wold was classified as a villa (Frere 1977: 383), even though the 'Romanised', rectangular building is small, short-lived and never dominates the settlement layout (Rod Mackey personal communication). The nearby site of Old Wintringham (Stead 1976) was interpreted by the excavator as a fort, even though no defences were found, and no military equipment recovered. This site has been classified as a large indigenous rural site (cf. Dragonby, Shiptonthorpe, etc.) for this report. When dealing with forts and their vici it is never easy to be sure where the fort stops and the vicus begins (and in some ways this may be a false dichotomy - the changes in late Roman barrack blocks in forts suggests that civilians may have moved into the forts (Daniels 1980). Despite these problems of definition of individual sites, this approach does seem to offer an improvement over that used previously where context has not generally been considered.

The total number of sites examined for this report form a small proportion (approximately 10%) of those excavated and an even smaller proportion of those known (from fieldwork, etc.). There are, however, considerable differences in the 'coverage' of different sorts of sites. The number of forts examined is relatively small, and the selection of samples from those forts selected is correspondingly small (20% or less). The number of vici included in this study is small compared to the total number known, as relatively few have been subjected to extensive excavations. There are only four possible towns known in northern Roman Britain and two of these (York and Carlisle) are included. Aldborough and Corbridge were not examined because there are no recent excavations and the status of the latter site is uncertain. The category Large Nucleated Rural Sites covers all sites between towns and vici on one hand, and villas and farmsteads on the other. Nucleated Rural sites and villas are relatively well-known in southern Britain but are rare in the north. The 'coverage' of such sites for this report is therefore very high (probably in excess of 50%). Small rural sites (farmsteads) which display little sign of Romanisation and hillforts are very common in northern Britain. The farmsteads and hillforts examined represent a very small proportion of those known (probably less than 1%). Such sites have not received as much attention as forts and towns, and so almost all excavated sites have been included. There are relatively few caves from northern Britain known to have been occupied in the Roman period. The examination of the Settle caves represents nearly all of the available evidence from caves. Cemeteries should be associated with every major Roman settlement but only two are known from recent excavation - Brougham and Trentholme Drive, York. Trentholme Drive was not included as the number of copper alloy artefacts recovered was very small. The hoards examined for this report represent a large proportion (approx 50%) of those known from northern Britain.

In order to ensure that a representative range of artefact types from each site was analysed, the typology of objects was taken into consideration. Sampling could not take place on the basis of each typologically unique artefact category as the number of different types (where typologies have been constructed) was too great, and in some cases there exists no established typology for classifying some Roman material. The construction of typologies has not been seen as a priority in Roman studies as they are often used by prehistorians for 'cultural' analysis, and the Roman period in Britain is frequently regarded as being a historical period. Those objects which have been the subject of typological analysis are usually those which it is hoped may serve as dating evidence (e.g. brooches). For the collection of samples, an interim classification of artefacts was used which assigned objects to broad categories on the basis of their perceived function (see Figure 4). This approach has seen increasing use in the publication of small finds reports (e.g. Crummy 1983; Mould 1991). Thus, artefacts made from a variety of different materials (copper alloy, bone, jet, etc.) are grouped together, e.g. all pins are dealt with in the same section by the same specialist. This approach provided a framework for the selection of samples from individual sites (especially as the number of recorded copper alloy artefacts from some sites exceeded 1000). All of the finds from a site to be examined were divided into the functional categories and (initially) a 20% sample selected from each category. This 20% sample was not used where the number of finds from a site were very large (e.g. the extensive excavation of a Roman fort, town, etc., where a smaller sample was selected), or very small (e.g. farmsteads which often yielded a single artefact). The functional categories were not intended to be used rigidly (i.e. all samples evenly distributed across the six categories) but their use did ensure that certain artefact types and categories did not dominate the analysed data base.

Fig.4 Categories of objects sampled

There are problems, however, in assigning objects to functional categories when the exact function of many items is unknown. There are few contemporary written accounts which describe the use of the sorts of artefacts that are dealt with here. Function is usually deduced by comparison with modern practice. In some cases, however, there are no modern parallels and function can only be guessed at. The function of some artefacts is often ambiguous, e.g. button-and-loop fasteners have been interpreted as items of personal dress and as horse harness fittings. A large proportion of finds (such as fragments of sheet and wire) could not be assigned to any functional category. The above classification was used in the early stages of the collection of samples; however, it is a rather crude and arbitrary method of classification and proved of little further use because of the ambiguities mentioned.

While every effort was made to collect samples which illustrated all aspects of the archaeological record there remain some biases. These biases are caused by a number of different factors (such as excavation strategies, deposition environment, ancient deposition practices, ancient metal use practices). There are concentrations of sites around Dunbar, the Tees valley, and in East Yorkshire, while other areas have fewer known sites (see Figure 1). Some artefact types are common while others are rare. It is not easy to determine which of these biases are representative of original metal use and which have arisen through depositional and post-depositional factors. Some of these biases are discussed in later chapters.

© Internet Archaeology
Last updated: Thu Apr 3 1997