This article has presented the chemical analyses of over 1500 Iron Age and Roman copper alloys. These results have been considered in the light of previous analyses, metallurgical constraints and existing archaeological knowledge of the period being studied. This final section summarises and discusses some of the key issues of the research. These range from the highly specific, such as the details of the cementation process in the production of Roman brass, to the general, such as the nature of time and cultural change.
The analysis of copper alloys has included nearly 300 Iron Age, late Iron Age, and 'Celtic' samples. These results (and those of Northover and other researchers) show that only one alloy was used in the pre-Roman Iron Age: a tin bronze which occasionally contains low levels of lead, usually contains some arsenic as an impurity, and almost never contains any zinc. This alloy is uniformly found throughout those areas of Britain where archaeometallurgical research has examined Iron Age copper alloys.
The vast majority (although not all) 'Celtic' metalwork, however, is not made from tin bronze. Instead this category of metalwork is made from a range of copper alloys containing zinc, tin, and/or lead. In this respect, 'Celtic' metalwork is more similar to Roman than Iron Age copper alloys. Most 'Celtic' metalwork has previously only been assigned the vaguest of dates (on typological/stylistic grounds) and it has been conceded that some was deposited (and may have been manufactured) after the Roman Conquest (e.g. Macgregor 1976). It can now be confidently stated that the vast majority of 'Celtic' metalwork was produced after (or shortly before) the Conquest. The exact date at which 'Celtic' metalwork in Britain began to be produced in alloys other than the traditional bronze is uncertain. Brass is first used widely in Europe from the time of Augustus's coin reforms (23 BC). 'Celtic' metalwork containing significant levels of zinc (more than c.5%) was probably produced using scrap Roman metal. 'Celtic' brass is therefore unlikely to pre-date 23 BC. It is possible that the first appearance of brass in northern Britain post-dates 23 BC by some years. It may take some time for items to make an appearance far from the borders. In this respect brass is like many other items of Roman material culture which are found outside the bounds of the Roman Empire (Macready & Thompson 1984). The appearance of brass in Britain does pre-date the actual Roman Conquest (AD 43) as there is evidence for the production of brass brooches at Baldock (Stead & Rigby 1986). An exact date for the first use of brass in northern Britain is as yet unavailable.
It has long been known that Roman alloys are formed using a variety of alloying elements: zinc, tin, and lead. The definitions used for these alloys are given in Section 1. A variety of ways in which the analytical results can be represented have been explored, using graphs and tablespin. In order to deal with the large number of samples produced for this report a 3-D chart was used. The two horizontal axes represent the zinc and tin contents (as in the 2-D zinc and tin scatter charts seen in Craddock 1975, Caple 1986, and elsewhere). The third (vertical) axis of the 3-D chart used here represents the frequency of any particular combination of zinc and tin. The 3-D chart shows a number of peaks in the distribution which relate to specific alloys (probably produced to a specific recipe), a number of troughs which indicate that some compositions were never aimed for (or obtained accidentally), and a broad 'low-lying' area. The three main peaks in this distribution of alloy types represent copper (little or no zinc or tin), bronze (tin, but little or no zinc), and brass (zinc, but little or no tin). The bronze was probably produced by mixing copper and tin together whereas brass was produced by the cementation process (see Sect 10.3). The fact that these three peaks are clearly visible suggests that they are significant and were deliberate. There is, however, no absolute separation of these alloys from each other. The 3-D chart shows that the peaks are all connected to each other (through the 'low-lying' area discussed below). Any decision as to the exact limits of any peak/alloy type will be arbitrary to a certain extent. The use of a statistical method to define these alloy types, such as cluster analysis will not work as the clusters are semi-circular in their distribution.
The 'low-lying' area referred to above can be seen in extending from the brass peak to the bronze peak. This area consists of alloys containing zinc and tin, i.e. gunmetal. The fact that this area extends from the brass peak to the bronze peak (and only between these two peaks) suggests that gunmetals were formed through the mixing of brass and bronze, rather than the mixing of tin and brass. The gunmetals form a continuum with the bronze and brass peaks; there are no peaks within the gunmetal distribution. This indicates that overall the mixing of brass and bronze was not carried out to a single recipe.
The troughs in the distribution of alloys are just as significant as the peaks. As Caple has pointed out (Caple 1986), these troughs indicate that some alloy types were never either deliberately made or arrived at accidentally. For instance, the 3-D chart shows that there are no alloys where the total zinc and tin content exceeds 30%, therefore tin (as opposed to bronze) was rarely added to brass. If tin had been mixed with brass then more alloys with, for example 20% zinc and 20% tin would be expected. A second trough occurs in the region of between 2 and 10% zinc (and less than 2% tin). There are no low zinc brasses: if alloys do contain between 2 and 10% zinc they are always accompanied by at least 2% tin. This shows that brass was never mixed with copper or recycled on its own. If brass had been commonly recycled on its own then the oxidation/volatilisation of zinc would have caused progressive loss of zinc. It is clear, then, that if brass was recycled/mixed with other alloys then it was always mixed with bronze.
Craddock (1975), Bayley (1992) and others have shown that some Roman artefacts were made to recipes that were adhered to with varying strictness. In some cases the alloy composition for a specific artefact type may have a standard deviation of 10% or less (e.g. some 1st century brooches); in other cases there is little more than a general tendency for a general alloy type to be used. In many cases there was no correlation between typology and alloy composition. This variation in the use of recipes for specific artefact types is seen in the overall alloys used. Bronze and brass were mixed to produce gunmetal, but there is no peak in the gunmetal distribution. In general, there does not seem to have been a standard gunmetal in the same way as there is a standard bronze. On the other hand, the mixing of brass with bronze was the only way in which brass was recycled - it was never recycled on its own or with copper.
The use of three alloy elements in Roman alloys produces complex quaternary alloys which make interpretations more difficult than for the Iron Age. Nevertheless, this report has continued to explore the possible 'grammar' which controlled the formation of copper alloys.
It is now widely accepted that Roman brass was produced by the cementation process (Craddock 1978; Bayley 1984). The use of this rather involved process is indicated by the composition of Roman copper alloys, experiments, ancient and medieval accounts of brass production, the copper-zinc phase diagram, and crucible remains. The maximum zinc content using the cementation process would seem to be c. 28%. (A small number of unstratified artefacts, from Roman sites, were not included in this research because their zinc contents were too high - the objects are probably post-medieval.) If c. 28% zinc were the maximum zinc content of Roman brass then it is striking that alloys with 23-28% zinc are very rare (see 3-D chart; only 1 in 1000 of Roman alloys have more than 23% zinc). This lack of brasses with the zinc percentages predicted by the cementation process can also be seen in the work of Craddock (1975) and Bayley (1992).
A number of possible solutions to this problem may be suggested. Experiments carried out to examine the oxidation/volatilisation of zinc have shown that molten brass will tend to lose a proportion of its zinc. Thus, if brass with a zinc content of (say) 28% was produced in a sealed cementation crucible, the metal would regularly be melted (possibly in an open crucible) prior to casting actual artefacts, which would then have a lower zinc content. If 1/5 of the zinc was lost during casting then the zinc content of the cast artefact would be 22.4%. Zinc contents in excess of 23% might be regularly found, however, in artefacts which have been worked from solid cementation brass (which had not been remelted).
A second explanation for the low number of brasses with zinc in excess of 23% may also be suggested. The common brasses (15-23% zinc; see 3-D chart) are often (but not always) accompanied by small amounts of tin and/or lead. The presence of either of these two elements may interfere with the cementation process. Both modern experiments and ancient and medieval accounts of the cementation process make it clear that the copper used was finely divided. This provided a large surface area for the zinc vapour to diffuse into. It has been surmised that the diffusion of zinc into the copper would continue satisfactorily only while the copper remained solid. Once the copper melted the surface area would decrease dramatically. The cementation process thus required the careful control of temperature. The temperature must be in excess of 907ºC in order to volatilise the zinc but must not exceed 1083ºC or the copper would melt. As the proportion of zinc in the copper increased, its melting point would be lowered. Eventually, the brass produced would melt and less zinc would diffuse into the molten brass. If the copper used in the cementation process was not pure but contained some tin or lead, then its melting point would already be lowered before any zinc was diffused into the copper. In this case the maximum zinc uptake into the copper would be somewhat less than 28%.
Two different explanations can, therefore, be suggested for the paucity of high zinc brasses. Both make use of the available analytical, metallurgical and experimental data, and both are plausible. This illustrates many situations (not only in archaeometallurgy) where a phenomenon may be explained by more than one cause. It may be impossible to decide that one is true and that the others are false.
The alloys selected for this research cover almost a whole millennium and it is possible to observe considerable chronological change. The alloys of the late Bronze Age are mostly leaded bronze, those of the Iron Age tin bronze, and those of the Roman period vary widely. The chronological changes in alloy compositions do not, however, match the conventional dates for the beginnings and ends of these periods. The alloys of the earliest Iron Age in Britain (Staple Howe and Scarborough) are little different from those of the late Bronze Age. The alloy type referred to in this report as typical of the Iron Age is largely restricted to the 'Arras' culture burials and contemporary settlements. Iron Age alloys are, however, similar to Bronze Age ones in many respects, with the only significant differences in the later period being the higher iron levels and lower lead levels. The former probably reflects a change in smelting methods (Craddock & Meeks 1987), but the latter is difficult to explain. The higher lead levels in late Bronze Age and in Roman alloys shows that such use of lead could be beneficial. If the decreased use of lead was deliberate then it appears to be a technologically retrogressive step. On the other hand the decrease in leaded alloys may simply reflect a scarcity of lead (but then the scarcity of lead itself would require explanation).
Iron Age bronze gives way to the variety of alloys used in the Roman period (brass, bronze and gunmetal) but this change does not occur with the Roman Conquest of Britain. There is abundant evidence for the use of brass artefacts in late Iron Age Britain, and for the working of brass metal, although as yet the only evidence for brass production is post-Conquest. Nevertheless the use of a range of alloy types in Roman Britain is established before the actual Conquest.
Chronological changes in copper alloys during the period of the Roman Empire might be expected on the basis of the analysis of Roman orichalcum coins by Caley (1964). The sestertii and dupondii of the Augustan reforms (and those for the following century or so) have high zinc contents (usually 20% or more). The zinc content of the later coins gradually declines until the mid 3rd century when the coins contain no zinc. Regular late Roman coins (of any kind) almost never contain any zinc. Caley (ibid.) suggested that the zinc decline resulted from a loss of the cementation process. Fresh brass was no longer produced after the late 1st century AD and later coins could, therefore, only be produced by recycling old ones. As the recycling would cause some loss of zinc, the zinc content of Roman coins would gradually decline. The overall shape of this zinc decline does not, however, match that which would be predicted by recycling losses. In addition, later coins contain progressively more tin and lead. An alternative explanation for Caley's zinc decline can be suggested which takes into account contemporary changes in the composition of the gold and silver coinages. The content of these precious coinages were carefully and deliberately manipulated. I would suggest that the composition of orichalcum was similarly manipulated by mixing fresh brass with leaded bronze to produce an alloy of desired composition. The analysis of a range of copper alloy artefacts for this research has shown that there was a chronological decline in the proportion of brass used. There was never a cessation of brass production and some typologically late Roman artefacts (e.g. late 4th century belt sets) were made of brasses which had never been recycled. The observed chronological decline in the proportion of brasses may be more apparent than real as the nature of the data set from which the samples were drawn itself undergoes chronological change. The later samples are not therefore strictly comparable with the early ones. This problem also emerges when attempts are made to compare samples from different sorts of sites.
The mis-match between the archaeometallurgical evidence and the conventional archaeological chronology casts doubt on the usefulness of the archaeological periods of the Three Age System and on the links between archaeological chronologies and historical ones. Collingwood (1993: 50) characterised such a system as viewing each era as unique and divided from each other era by a dramatic event or development. The Three Age system accentuated the role of technology and materials in defining human history. Social life may, however, have only incidental connections with contemporary technology. Earlier prehistory has already seen the weakening of the Mesolithic-Neolithic and the Neolithic-Bronze Age divisions. The late Neolithic and the early Bronze Age are now seen as having more in common with each other than the advent of metal use might imply.
The mis-match between archaeological evidence and historical evidence can also be seen in Caple's study of the production of post-medieval pins (Caple 1986). Archaeology has often been regarded as an illustration of history rather than a historical discipline in its own right. Archaeologists, however, have become increasingly confident about their evidence and no longer automatically defer to historians or classicists. Within a processualist framework such a situation is uncomfortable and archaeologists must decide whether they are right (or if it is the historians/classicists who are right). A post-processualist/post-modernist perspective, however, avoids such a crisis by admitting diversity. Truth is not absolute but is localised (Hodder 1986; Shanks & Tilley 1987), and different disciplines can examine the same problem yet develop different explanations. These differences often say a lot about each discipline's starting assumptions.
During the period studied for this research northern Britain underwent considerable cultural change. The most dramatic changes occur with the expansion of the Roman Empire and the Conquest of Britain. The Roman Empire was an alien social and economic system which had the capacity to transform both the societies it conquered and those which remained outside its borders. In some cases such transformation may have been deliberate; Rome seems to have been keen to devolve the day-to-day running of conquered societies into the hands of trusted locals (often members of the elite groups who ruled the societies prior to conquest). Local elites will have been encouraged to conform to Roman behaviour norms (at least while dealing with Romans). Rome was also capable of carrying out physical and economic warfare against neighbours who were regarded as potentially threatening. In other cases transformations occurred which were not deliberate. Comparisons with modern colonial and imperial situations suggest that the presence of the Roman army and the Mediterranean economy may have effected cultural changes which were not necessarily intended. The nature and causes of all of these changes have been of fundamental importance to Romano-British archaeology since Haverfield's (1912) Romanization of Britain.
The use of archaeometallurgy as a means of examining Romanisation is attractive as copper alloy artefacts are usually non-utilitarian and so may reflect ideology or personal choice, and they are small and so can be widely exchanged. This is in contrast to the majority of the evidence for the indigenous inhabitants of northern Britain during the Roman period. Most evidence consists of cropmark or earthwork sites (settlements and field systems). The landscape may suffer from 'cultural inertia' and so be relatively unchanged by the Roman Conquest (or the Anglo-Saxons or the Vikings). Copper alloy artefacts may be a better reflection of such changes. The potential usefulness of copper metallurgy in this way is strengthened by the appearance of brass in Europe in the Roman period. Brass was produced in considerable quantities for Roman coins and military equipment and it has often been assumed (e.g. Grant 1946) that Rome maintained a monopoly over the production of brass. Samples were therefore collected from a range of different sites to examine the exchange and diffusion of brass from the core of the state to its margins (particular effort was expended to ensure that small un-Romanised rural settlements (farmsteads) were included in the data set even though such sites are relatively poorly known in northern Britain).
The examination of the incidence of brass on a range of Roman sites has, however, produced unexpected results. While a range of Roman sites have similar proportions of brass, farmsteads emerge as the one class of settlement to have high proportions of brass. These are the sites which, according to conventional explanations of Roman Britain, occupy positions at the base of the socio-economic system. The high incidence of brass on these sites challenges the assumption that the inhabitants of such sites were poor, powerless, and had little control over their lives (let alone the province as a whole). The high incidence of brasses in the archaeological record of farmsteads may not necessarily be a direct reflection of the use of brasses on such sites, however. The work of Hill (1994) and others shows that the archaeological record is not necessarily a passive reflection of past activities. As such, brasses may have been preferentially deposited rather than used on farmsteads.
Further work can be suggested in two major directions: an expansion of the geographical horizons of study, and the detailed examination of single sites. A number of highly specific research projects are clearly needed, e.g. the analysis of more examples of early (i.e. pre-Neronian) and late (Septimius Severus and later) orichalcum coins, and cementation experiments using impure copper to examine the maximum zinc absorbed by such metal.
On a broader scale, there is still very little known about the Iron Age copper alloys of Scotland, Ireland or the continent. Similarly, relatively little is known about Roman copper alloys of other provinces of the Empire. Such data bases will help to clarify to what extent the alloys of northern Britain are peculiar to northern Britain.
While research can be recommended on a broad front, there is also a need for small-scale examinations of single sites in order to examine the formation process which produce the archaeological record. This report has identified a number of difficulties in interpreting the range of alloys found on different archaeological sites. Social rules may have restricted the deposition of certain alloys in some contexts and encouraged it in others. The possible role of taphonomy in determining which copper alloys are recovered from particular archaeological sites and contexts should be investigated through detailed examination of individual sites. Suitable sites (i.e. those with large area excavations) should provide large quantities of copper alloys from a range of different contexts. The interpretation of copper alloys should also consider the formation processes behind the contexts from which they originate, and the relationships between different contexts.
©
Internet Archaeology
Last updated: Thu Apr 3 1997