5 Iron Age

5.1 Introduction

This section will describe the results of the analyses of Iron Age copper alloys from northern Britain. The section is divided into four sub-sections, covering the origins of Iron Age copper metallurgy, the Iron Age proper, 'Celtic' metalwork, and finally the end of the Iron Age and the beginning of the Roman period. The archaeological distinctions between the three chronological phases (and one cultural phase) are at times a little vague but the analytical results do show important changes in metal use through the Iron Age. The earliest Iron Age covers evidence from the late Bronze Age-early Iron Age transition period. The late Iron Age covers the period during the 1st century AD when Roman contacts and influence are suggested by the presence of continental/Roman imports. As the Roman conquest of Britain was a long drawn out, and not entirely successful, exercise, the late Iron Age is here taken to include objects and sites which were in use after the start of the Roman conquest, but in spheres relatively uncontrolled by Rome (i.e. settlements remote from Roman towns and forts). The comparison of 'Celtic' metalwork with the Iron Age results reveals that most of the former was produced in the late Iron Age or after the Roman conquest.

5.2 Late Bronze Age/Earliest Iron Age

Considerable attention has been devoted in previous archaeometallurgical research to the examination of the early phases of copper alloy use (e.g. Junghans et al. 1960; 1968; 1974). This section examines the origins of Iron Age copper metallurgy and its relationship with late Bronze Age metallurgy (Brown & Blin-Stoyle 1959; Northover 1982a). The basis of prehistoric chronologies is the Three Age system which assumes the primacy of technology as a means of dating. There is, however, no certainty that changes in technology will always occur in step with changes in social organisation. It is now widely agreed that the distinction between the late Neolithic and the early Bronze Age is somewhat artificial. While technology does change, the forms of social organisation seem to continue. It is suggested here that these two sites belong to a transitional period rather than to the Bronze Age or the Iron Age.

In order to examine the origins of Iron Age metallurgy in northern Britain, a small number of samples were taken from sites dated to the transitional, late Bronze Age/early Iron Age, period. The two sites available for analysis were Staple Howe (Brewster 1963) and Scarborough Castle (Smith 1927), both in North Yorkshire. Only 8 samples were taken so these are shown in Table 5.1 (rather than in charts).

XRFIDSiteObject CuSn PbZn FeNi As
2029Staple HoweRazor90.93 8.650.4ndndndnd
2030Staple HoweChisel90.85 7.082ndnd0.05nd
2031Staple HoweRazor85.01 9.795.12ndnd0.08nd
2121Scarborough Armlet99.72ndndnd0.07nd0.2
2124Scarborough Gouge80.4713.654.87nd0.480.070.45
2127Scarborough Axe92.544.452.91ndnd0.09nd
2128Scarborough Casting Jet93.94 1.963.85ndnd0.080.17
2130Scarborough Axe68.588.2922.64nd0.05nd0.33
Table 5.1. Late Bronze Age/Early Iron Age results (nd = not detected)

The collection of bronze objects from Scarborough was recovered during the excavation of the late Roman signal station. They were found lying on an old land surface adjacent to pits containing early Iron Age pottery (Smith 1927: 179) and so may not be closely associated with the pottery. Stylistically, the bronze objects belong to the late Bronze Age, while the pottery has been dated to the early Iron Age (Rutter 1959; Cunliffe 1993: 67-8) and the late Bronze Age (Barrett 1980). The reliability of the 'context' of the bronze objects is compromised by XRFID 2122and 2123 the two heavy (?harness) rings found with the other bronze objects. XRFID 2122 and 2123 both contain high levels of zinc. The mixed alloy (containing zinc, tin and lead) used for these objects is typical of this type of object in the Roman period (see Sections 6 and 7). These two rings probably relate to the use of the hill in the Roman period as a signal station (Frere & St Joseph 1983: 82-3). The excavation account (Smith 1927) does not make clear how far below the Roman levels the bronzes were found.

There is only slightly more consensus over the dating of the artefacts from Staple Howe (e.g. Megaw & Simpson 1979, deal with the site in both late Bronze Age and Iron Age chapters). The pottery from the site is dated to the early Iron Age, the copper alloy objects find closest parallels with typologically Halstatt objects (usually dated to the late Bronze Age in Britain), and the radiocarbon dates support either an early Iron Age or late Bronze Age date.

Most of the objects (Table 5.1) are tin bronzes (with the addition of varying amounts of lead). Late Bronze Age artefacts are usually made from leaded bronze (Brown & Blin-Stoyle 1959; Northover 1982a) while Iron Age objects only rarely have high levels of lead (i.e. >5%). Zinc was not detected in any of the samples and is rarely found in late Bronze Age or Iron Age artefacts. The one pure copper artefact (XRFID 2121) is an armlet with overlapping terminals. The object was probably wrought rather than cast and the low alloying element content would be well-suited to this method of manufacture. Arsenic was present as an impurity in half of the samples. This is broadly analogous to the results obtained from Iron Age artefacts (discussed later). Only one of the objects had a substantial proportion of iron present. Low iron levels are a feature of Bronze Age alloys, but iron is a regular impurity in Roman and later alloys. Craddock & Meeks (1987) suggest that the change in iron levels reflect a change in smelting methods. Early prehistoric copper may have been produced by a low temperature non-slagging smelting method, while Iron Age and Roman copper was probably produced by a tap-slagging method. This conclusion is supported by the work of Pollard et al. (1991) which shows that the presence of arsenic in early copper indicates a low temperature (and slag-free) smelting method.

The above analyses show similarities between Scarborough and Staple Howe and represent an apparently uniform picture. The results as a whole also show a range of similarities with both late Bronze Age results previously published and with the Iron Age results discussed below. The use of a leaded tin bronze is almost universal in the late Bronze Age, while arsenic is a common impurity in Iron Age alloys. This reinforces the proposition that the Bronze Age/Iron Age transition should be seen as a period in its own right.

5.3 Iron Age copper alloys

5.3.1 Introduction

This section considers the Iron Age itself, as distinct from the transitional period with the Bronze Age which precedes it and the later transition to the Roman period. In southern Britain this period is often referred to as the Middle Iron Age. The Iron Age objects were selected in the first instance from 'Arras' culture burials (Stead 1979). These showed the widespread use of tin bronze with arsenic as a common impurity (although, unlike earliest Iron Age artefacts, there is little or no lead). In order to give the results a wider relevance, objects from Iron Age settlement sites were also selected. These also showed the widespread use of tin bronze with arsenic as a common impurity. Dating settlement occupation and associated finds to the pre-Roman Iron Age can be difficult. Many of the sites examined continued to be occupied into the Roman period, e.g. Thorpe Thewles phase III covers most of the Iron Age but seems to end in the early Roman period (Heslop 1987). Where there was room for doubt, finds were classified as late Iron Age and are dealt with in the next section.

The present analysis of a range of 'Celtic' metalwork (cf. Macgregor 1976) shows that some of this is made of the same 'Iron Age' alloy and these objects are included here (e.g. Piggott's class II swords [Piggott 1950]). A few other finds are included in the Iron Age analyses (e.g. a mini terret from Piercebridge which is typologically identical to those from 'Arras culture' burials, and has the same chemical composition).

The find spots and types of artefacts analysed are shown in Table 5.2. The analytical results from a range of 'Arras culture' burials (Arras, Burton Fleming, Cowlam, Danes Graves, Kirkburn, Rudston, Sawdon, Wetwang/Garton Slack) are collected together here. It can be seen that there are more samples from the 'Arras' burials than all the other sites put together. In addition the types of finds found differ. While over half of the objects found in burials are items of horse harness, these are almost never found on settlement sites. The occasional horse harness fittings found on settlements (e.g. Huckhoe [Jobey 1959]) usually date to the late Iron Age or Roman period. The evidence from Weelsby Avenue consists mainly of debris from metalworking but this is rare at other sites.

BroochArmletSheet WireHorse HarnessSwords MiscDebris Total
'Arras' burials910--37211-69
Stanwick, Tofts4------26
Dragonby4 -1---4-9
Weelsby Ave1-34---1220
Stray finds----132-6
Total1810 44 385 1914 112
Table 5.2. Distribution of analysed Iron Age finds in northern Britain

5.3.2 Alloying in the Iron Age: Tin

Previous workers (especially Craddock and Northover) have established that Iron Age copper alloys are almost exclusively of tin bronze. Levels of lead and zinc (common as alloying elements in the Roman period) are very low in Iron Age alloys (see below). Figure 16 shows the distribution of tin content in Iron Age alloys analysed for this article.

[Tin content of Iron Age copper alloys]
Fig.16 Tin content of Iron Age copper alloys

The overall distribution is nearly normal, peaking around 11%, and may suggest the widespread use of a 'standard' alloy type. This overall distribution is similar to that found in previously published Iron Age copper alloys analyses (Barnes 1985, Hunsbury, Northants; Cowell 1990, Camerton hoard; Craddock 1986. various stray finds from Britain and Ireland; Northover 1984a, Danebury, Hampshire; Northover 1987, Hengistbury Head, Dorset; Northover 1991a, Maiden Castle, Dorset; Northover 1991b, Danebury, Hampshire). A few of the Hengistbury Head results are unreliable (they were obtained from corroded samples) and so are not used here). The overall distribution of tin contents shown in Figure 16 is nearly normal but is not exactly symmetrical about the mean value. A closer examination of the results shows that cast objects (Figure 17) tend to have higher tin levels (mean = 11.2%) than wrought ones (Figure 18: mean = 8.8%).

[Tin content of cast copper alloys]
Fig.17 Tin content of cast copper alloys

[Tin content of wrought copper alloys]
Fig.18 Tin content of wrought copper alloys

The lower tin content of wrought objects would have made them easier to work (this can also be seen in the comparison between the lead content of cast and wrought objects - see below). This may be seen as evidence for the sophistication of the Iron Age smiths. It shows that they had a good empirical understanding of the properties of the alloys they used (even if their understanding was not based on a knowledge of elements), and they selected their alloys accordingly. It is even possible that they could manipulate the composition of the alloy (perhaps through weighing the copper and tin and mixing according to 'recipes'). The lower tin content of the wrought objects may, however, be accidental rather than deliberate. Wrought objects cannot be satisfactorily made from high tin alloys as they tend to break if worked. Therefore attempts to make wrought objects from high tin bronzes would be more likely to end in failure. Such failed objects may be melted down as scrap and re-used. The lower tin content of wrought objects could be an indirect result of physical metallurgy. It is relatively straightforward to recognise the differences in tin and lead content; it is less easy to know to what extent it was deliberate.

5.3.3 Alloying in the Iron Age: Lead

The lead content of Iron Age alloys (shown in Figure19) is generally low. The mean lead content is 0.9%, and three-quarters of all alloys have less than 1% lead. Such low levels of lead are probably impurities in the metal. This contrasts with the late Bronze Age (Northover 1982a) where almost all objects have several percent or more of lead present. The lower lead levels in Iron Age copper alloys make it unlikely that recycled Bronze Age scrap was a significant source of metal in the Iron Age. It also contrasts with the Roman period (see Section 6) where, although many objects have low levels of lead, a small proportion have quite high levels (up to 40%). Those few Iron Age objects which have moderate levels of lead are usually cast objects (see Table 5.3). While 31% of all cast alloys have 1% or more lead, only19% of wrought alloys have more than 1% lead. A small addition of lead to the copper alloy would reduce the melting point and increase the fluidity of the alloy, and so reduce the chances of producing a flawed casting. The difference in lead content between cast and wrought alloys might indicate the sophisticated understanding of the copper smiths, but might, like the tin contents, be an inevitable outcome - leaded alloys are more likely to break when wrought.

[Lead content of Iron Age copper alloys]
Fig.19 Lead content of Iron Age copper alloys

Less than 1% Lead1% lead or more
Wrought81% 19%
Table 5.3: Proportions of objects which are leaded and unleaded.

5.3.4 Alloying in the Iron Age: Zinc

Copper alloys containing zinc (brasses and gunmetals) are common in the Roman period, but are almost entirely absent from Iron Age copper alloys (see Figure 20). The few exceptions can be divided into two groups: those with high levels of zinc (usually over 15%) where this was a major deliberate addition, and those with low levels of zinc (under 5%) which may be impurities rather than deliberate additions. The former group includes two brooch spring and pin fragments from Dragonby (XRFID 1711 and 1728). The fragments appear to come from simple one-piece La Tène III brooches (e.g. Nauheim derivative). This type of brooch is found in Britain and on the continent and was being produced before the Roman Conquest of Britain. Some Nauheim derivative brooches were made of brass (Bayley 1992). These samples, therefore, may be imports to northern Britain and not relate to traditional copper alloy production. Alternatively, the objects may be intrusive and not relate to the Iron Age contexts in which they were found. At Dragonby, most of the Iron Age objects came from a context (Field Number 3) which also contained a Headstud brooch (XRFID 1731). Such finds are usually dated to the late 1st century and early 2nd century AD. There are no high zinc brasses from secure Iron Age contexts in northern Britain, and it is suggested here that high zinc alloys are not a normal feature of the Iron Age in Britain. The use of zinc alloys in Iron Age and 'Celtic' objects is discussed in the following section on the late Iron Age.

[Zinc content of Iron Age copper alloys]
Fig.20 Zinc content of Iron Age copper alloys

The second group of Iron Age copper alloys which contain low levels of zinc (XRFID 1667, 1729, 1851, 1863, 1871, and 2228) may also be intrusive within the Iron Age contexts in which they were found. Alternatively, the zinc may be an impurity in the alloy. This may either result from the use of scrap containing small amounts of possibly imported brass, or from the use of copper ores which contain zinc (this last possibility is discussed by Northover [in Musson et al. 1992] in relation to Iron Age copper alloys from the Welsh Borders).

5.3.5 Impurities in Iron Age alloys: Introduction

Copper smelting aims to recover copper from copper ores (oxides, carbonates, etc.). Inevitably this is rarely 100% efficient - some copper is lost in the slag. Similarly other metallic elements present in the ore may be reduced with and dissolve into the copper. These other metallic elements are impurities in the smelted metal. The level of such impurities depends on the levels of impurities in the ore (and other materials present during smelting such as the slag and the furnace lining), the smelting conditions, and the degree of metal purification. The interpretation of these metal impurities is fraught with difficulties, in particular, the equation of impurity patterns in a metal object with a specific ore source is difficult. Northover (1982b; 1984b) has attempted to use these metal impurity patterns to determine the source of metal used in the Iron Age. While impurities tell us something about the production processes, our ability to interpret impurity patterns is hampered by a lack of knowledge about the mineral sources and the smelting procedures. The impurities that are seen by Northover as most useful are cobalt and nickel. Craddock & Meeks (1987) have suggested that iron levels reflect something of the smelting procedures used. The analytical results for each of these impurities and their significance is dealt with separately below. Arsenic, however, is seen as the most relevant impurity for this research project as it is regularly found in Iron Age alloys but is almost never found in Roman alloys.

5.3.6 Impurities in Iron Age alloys: Arsenic

The most striking metal impurity in Iron Age copper alloys is arsenic (see Figure 21). The relatively high arsenic levels found in Iron Age alloys from northern Britain can be paralleled with the analysis of Iron Age alloys from elsewhere in Britain, although arsenic levels in northern Britain are slightly lower. Arsenic as an impurity is therefore a regular feature of Iron Age alloys in Britain. This relatively high arsenic content of all the Iron Age alloys examined contrasts most strongly with those of Roman copper alloys from northern Britain (see Figure 43) - 85% of all Roman copper alloys had less than 0.10% arsenic.

[Arsenic content of Iron Age copper alloys]
Fig.21 Arsenic content of Iron Age copper alloys

5.3.7 Impurities in Iron Age alloys: Iron

Most Iron Age samples analysed contained some iron (see Figure 22). The levels of iron in Iron Age alloys is generally higher than that found in the Scarborough/Staple Howe alloys (see above) or late Bronze Age alloys (Brown & Blin-Stoyle 1959; Northover 1982b). The mean iron content of the Scarborough/Staple Howe samples was 0.07% (and that largely due to a single high result), whereas the mean iron level in Iron Age samples was 0.20%. The iron content of Iron Age alloys shows more similarity with Roman alloys (Figure 23).

[Iron content of Iron Age copper alloys]
Fig.22 Iron content of Iron Age copper alloys

[Iron content of Roman copper alloys]
Fig.23 Iron content of Roman copper alloys

The mean iron content of Roman alloys is 0.24% (only slightly higher than those of the Iron Age). In terms of the iron content, Iron Age alloys can be seen to bear more resemblance to Roman ones than to Bronze Age ones.

5.3.8 Impurities in Iron Age alloys: Cobalt

Cobalt was detected in a little over half of all the Iron Age samples analysed in this paper (Figure 24). This is broadly similar to the range of cobalt levels found in a range of other Iron Age alloys. Overall the cobalt levels in northern Britain, however, are slightly lower than in southern England. While 46% of the samples from northern Britain contained no detectable cobalt, only 34% of the southern samples contained no cobalt.

It is not possible to compare Iron Age cobalt results with Roman ones as cobalt was not determined for Roman samples.

[Cobalt content of Iron Age copper alloys]
Fig.24 Cobalt content of Iron Age copper alloys

5.3.9 Impurities in Iron Age alloys: Nickel

Nickel was detected in approximately a third of all the Iron Age samples analysed (Figure 25). This is somewhat lower than the results from southern England. While 40% of the samples from northern Britain have at least 0.05% nickel, 54% of the southern samples have at least 0.05% nickel. The nickel levels in Iron Age alloys is noticeably higher than that found in Roman alloys (see Figure 42 - only 10% of Roman alloys have 0.05% or more nickel).

[Nickel content of Iron Age copper alloys]
Fig.25 Nickel content of Iron Age copper alloys

5.3.10 Impurities in Iron Age alloys: Conclusions

The above charts show the incidence of metal impurities in the Iron Age alloys of northern Britain analysed for this article. These have been compared with similar results for the Roman alloys from northern Britain and with Iron Age results from southern Britain.

In general, impurity levels in Roman copper alloys are lower than those found in Iron Age alloys. The higher purity of Roman copper alloys might be seen as a reflection of the view that the Roman Empire was an improvement on the cultures it conquered. However, low levels of metal impurities are not detrimental to many of the properties of copper (except electrical properties which are irrelevant for the Roman period). The higher purity of Roman copper would have been achieved by fire-refining. The molten copper would have been exposed to blasts of air which would oxidise metal impurities and they would dissolve in the slag. Invariably some copper is also lost during this process (Merkel 1991). This would suggest that Roman copper production may have been less efficient than Iron Age production.

It might be tempting to see the difference in arsenic content between Iron Age and Roman copper alloys in terms of ore sources used. It might be thought that the Roman period saw the use of new ore sources and perhaps the abandoning of older sources. However, such an approach does not take into consideration the complexities of ore chemistry, smelting processes, and metal use and re-use strategies (as discussed above).

The metal impurities in Iron Age alloys have been used by Northover (1982b; 1984b) to determine the ore sources used for the production of copper. Northover's early publications (e.g. 1984a) divided samples into two classes: class I and II. Class I has higher cobalt and lower nickel, while class II has lower cobalt and higher nickel. The labels for these two groups have changed in later publications (e.g. the use of a range of letter codes in Northover [1987]) and the groups have undergone some sub-division. The levels of cobalt and nickel have remained important: Northover sees a chronological trend from metal with cobalt as a principal impurity to metal with nickel as the main impurity (i.e. from class I to class II). This is difficult to reconcile with the results from northern Britain as both nickel and cobalt levels in northern Britain are slightly lower than in southern England. If the northern samples were generally early then the nickel levels would be low but cobalt would be high; if the northern samples were generally later than the southern ones then the reverse would be true. The fact that nickel and cobalt are both lower in northern Britain suggests that the North had a different supply of copper, which may either have come from another source or have been smelted differently.

The high cobalt and nickel content of some of the southern English bronzes has been interpreted by Northover as indicating the use of particular ore sources. While the results of analysis have been published piecemeal over almost a whole decade, the interpretation has changed considerably. When considering the higher cobalt metal, Northover suggested in the first Danebury report (Northover 1984a) that the source was somewhere in south-west England. In the Hengistbury Head report (Northover 1987) the suggested source is 'Alpine'. In the Maiden Castle report (Northover 1991a) the source is again south-western England (and probably the Tamar valley). Finally, in the second Danebury report (Northover 1991b) the suggested source is also south-western England but it is noted that the same metal type is used for vessels from the site at La Tène. Given the problems of interpreting trace elements in terms of ore source (discussed above), such claims need to be carefully examined. Cobalt is chemically similar to iron and may be introduced to the metal from the flux or furnace lining rather than the ore source. Nickel is chemically similar to copper, and so it should be of more use in attempting to determine the ore sources used. Very little nickel was found in objects from 'Arras culture' burials (84% of these objects had <0.05% nickel). Nickel was found in most of the objects from Weelsby Avenue (only 20% of these objects had <0.05% nickel). The copper alloys from northern Britain have lower levels of nickel than the alloys of southern Britain. This may indicate a variety of metal supply networks for the north and south.

5.4 Iron Age alloys - Summary

There is little variation in the composition of Iron Age copper alloys analysed for this thesis. Tin is the only alloy element regularly added and so Iron Age copper alloys can be safely referred to as bronze. The same cannot be said of Roman metalwork which is made from a range of alloys; often containing tin, zinc and lead. Iron Age bronzes usually contain some arsenic (0.1-1.0%) and this pattern is repeated in bronzes from southern England. The northerly extent of this alloy type is uncertain as the North has produced fewer bronze objects. Nevertheless this alloy type (a tin bronze, with arsenic as an impurity but no zinc present) has been found at Broxmouth and Traprain Law. The alloy type may extend even further north - the analysis of two samples from Sculptor's Cave, Covesea on the Moray Firth (reported in Benton 1930-1), shows the use of zinc-free bronze (in one case with a substantial proportion of arsenic).

The uniformity of the tin bronze used in the Iron Age can be seen in a variety of sites (burials, settlements, and stray finds). There was, however, insufficient information regarding the circumstances of discovery to allow detailed comparisons to be made either of these different contexts or the associated artefacts and their alloy composition (cf. Hill 1994).

While tin bronze is the standard alloy of the pre-Roman Iron Age, many of the stray 'Celtic' finds are made of alloys containing at least some zinc (see below). It is suggested here that the presence of zinc in 'Celtic' alloys is of some chronological significance. This is discussed further below.

5.5 'Celtic' metalwork

The collection of metalwork catalogued by Macgregor (1976) has been regarded as 'Celtic' on stylistic grounds. The dating of this material is difficult, however, as it was mostly discovered in the19th century and only limited data are available concerning the context or date of deposition. On stylistic grounds it has been assumed that many 'Celtic' metalwork objects were made in the Iron Age. Some of the finds actually come from Roman forts, and so have been dated to the 1st century AD. It will become clear from the results shown below that most 'Celtic' belongs to the late Iron Age. The two categories are discussed separately, however, as they are defined differently: late Iron Age by context date, 'Celtic' on stylistic criteria.

The analysis of 'Celtic' metalwork may help to shed light on the dating of these objects. A scatter plot showing the zinc and tin contents of 'Celtic' metalwork shows a range of different alloys being used - from bronze to brass (Figure 26). In order to compare these results with those from both late Iron Age and Roman contexts, the results are summarised as a barchart using the classification system discussed earlier (Figure 27). While a substantial proportion of 'Celtic' metalwork is made of bronze, the majority contains some zinc and many items have zinc as the principal alloying ingredient. This is in marked contrast to the Iron Age alloys discussed above which are largely zinc-free, but is broadly similar to the range of alloys used in Roman Britain. The relatively high incidence of brasses in 'Celtic' metalwork finds closest parallels with late Iron Age alloys and with those from farmsteads of the Roman period. The high incidence of brass in such indigenous contexts is surprising as brass is conventionally regarded as a 'Roman' metal. This is discussed further below.

[Plot of zinc and tin content of 'Celtic' metalwork]
Fig.26 Plot of zinc and tin content of 'Celtic' metalwork

[Barchart showing the different alloys used for 'Celtic' metalwork]
Fig.27 Barchart showing the different alloys used for 'Celtic' metalwork

The contrast between the composition of Iron Age alloys and those of the late Iron Age and Roman period may be of chronological significance. This may perhaps be illustrated by reference to Piggott's 'Celtic' swords (Piggott 1950). Piggott suggested (on typological grounds) that the class III swords were produced in the Iron Age while the class IV were made somewhat later (roughly AD 50-150). All of the class III swords are made of tin bronze (with little or no zinc or lead, and arsenic as a common impurity), while the class IV swords are made of alloys containing zinc. Many of the class IV swords have zinc as the principal alloying element. Thus, the analysis of the metal composition tends to support Piggott's typology and dating (Figure 28) (although the 'indigenous' alloy type of class III swords could continue to be manufactured after the Roman conquest). The exception to the rule is, however, Pilling Moss (Macgregor 1976: No. 155). This dagger scabbard is usually regarded as belonging to class IV but all of the components are made of tin bronze with little or no zinc present. The division between class III and IV is largely on the basis of the moulding at the tip of the scabbard. The Pilling Moss dagger scabbard has an unusual moulding and it may be argued that it belongs to class III.

[Plot of zinc and tin content of 'Celtic' swords]
Fig.28 Plot of zinc and tin content of 'Celtic' swords

The high incidence of brass in 'Celtic' metalwork is strikingly illustrated by the Melsonby (Stanwick) Hoard (Macgregor 1962; Haselgrove et al. 1990: 11-13). This large hoard consists of a number of fragmentary sets of horse harness and other items. Most of the items in the hoard are made of brass rather than bronze, and even the bronzes usually contain at least some zinc. Dating the hoard is difficult as it was recovered in the19th century and few of the finds can be closely dated on typological grounds. The conventional date of the hoard is the mid 1st century AD (Macregor 1962: 36-7). The analysis of a large number of items from the Melsonby hoard has allowed the reappraisal of the grouping of objects into sets. On the whole the sets suggested by Macregor (1962) and Leeds (1933) on stylistic grounds are strengthened by the analytical results. Sets A, B and C are all brasses while set D is of bronze. A closer examination of the results for sets A, B and C shows that these are, on average, distinguishable from each other. They all have high zinc contents (mostly 16-22%) and low tin contents. The tin contents overlap but Set A has the highest tin content (mean = 1.34% [excluding one outlier XRFID 2545b]), Set B has the next highest (mean = 0.86%), and set C the lowest (mean = 0.38%). Some of the items analysed do not fall into the compositional groups of Sets A to D. In particular, some items shared a gunmetal composition (mostly 10-15% zinc and 2-6% tin). These items (XRFID 2056, 2565, 2568, 2569, 2570, 2572, 2573) are all plain and so have not stood out as a group in their own right. They are shown on Figure 29 as a possible fifth set (?E). This possible fifth set is not the same as that suggested by Spratling (1981), a subdivision of Set A into gilded and ungilded items. Spratling's fifth set is not convincing as there are gilded and ungilded items which are stylistically similar (if not identical). The presence or absence of gilding on the surviving items should not be credited with too much significance. Most of the items in the hoard are broken, and some have been distorted by high temperature (Macgregor 1962: Nos. 65 and 75). The selection of illustrations in Macgregor (1962) is a little deceptive as the badly distorted items are not usually illustrated. The distortion of some of the items by high temperatures has previously been diagnosed as signs of miscasting (Macgregor 1962: 20; Spratling 1981: 14). Temperatures high enough to distort some but not all of the items could be obtained in a funeral pyre. The Melsonby hoard may perhaps be the debris from a funeral pyre similar to Folly Lane, St Albans (Selkirk 1993), It is perhaps no coincidence that one of the items from the Folly Lane burial is stylistically similar to items from set C of the Melsonby hoard.

[Melsonby Hoard: plot of zinc and tin content]
Fig.29 Melsonby Hoard: plot of zinc and tin content

[Melsonby Hoard: enlarged plot of zinc and tin content]
Fig.30 Melsonby Hoard: enlarged plot of zinc and tin content

It was possible to use the overall differences in metal composition between the different sets to confirm some doubtful items, and as a guide when considering those items which had not been assigned to a set (Table 5.4).

XRFID NumberMacgregor NumberMacgregor GroupAssigned Group
2004a & b79? D
Table 5.4 Melsonby hoard: assignment of artefacts to sets

The suggestion that zinc-free alloys are a regular feature of the pre-Roman Iron Age and that alloys containing zinc are largely a phenomenon of the Roman period is not a new one in the context of 'Celtic' metalwork. This issue was discussed by Savory (1964) and Spratling (1966) in relation to the Tal-y-Llyn hoard and by Megaw (1967; 1971; 1973) in relation to 'Wraxall' collars.

Savory's publication of the Tal-y-Llyn hoard (1964) suggested that the items dated to the 4th or 3rd centuries BC. It was noted that some of the objects were made of a zinc-copper alloy and suggested that such an alloy could be made by reducing local copper ores which occurred with zinc ores. In a response, Spratling (1966) noted that the hoard included a Roman-type lock plate. Spratling also argued (citing Tylecote [1962]) that copper alloys containing high levels of zinc could not be accidentally produced due to the volatility of zinc. The publication of the analyses of some of the Tal-y-Llyn objects in Savory (1971 Appendix 1) revealed that some had zinc contents in excess of 15%. Such high levels of zinc are unlikely to be the results of smelting a copper ore rich in zinc ore (the volatility of zinc is discussed further in Section 7). Analyses of other Welsh 'Celtic' objects (Cerrig-y-drudion, Llyn Cerrig Bach, and Tre'r Ceiri - reported in Savory [1971 Appendix 1]) suggest that tin bronze (rather than brass) was the standard copper alloy of the Welsh Iron Age. As high zinc alloys are unknown before the last quarter of the 1st century BC the Tal-y-Llyn hoard brasses are unlikely to have been made before the end of the 1st century BC.

In discussions of 'Wraxall' type collars, Megaw (1967; 1971; 1973) used analytical data to support a 1st century AD date. Most of the collars analysed contained at least some zinc (see also Beswick et al. 1990). The massive armlets found mainly in Scotland (Macgregor 1976: Nos. 231-50) are another type of 'Celtic' metalwork which has been analysed (Tate et al. nd). Most contain at least some zinc and so should be dated to the 1st century AD or later (this is in agreement with the typological dating of the objects [Macgregor 1976]). Other Scottish material from the Roman Iron Age is also made of alloys often containing some zinc (Fraser Hunter personal communication).

The first widespread production of brass in Europe occurred towards the end of the 1st century BC (see Section 8). The first Roman brass coins provide a terminus post quem for brass artefacts in northern Europe. It is suggested here that 'Celtic' copper alloys containing substantial proportions of zinc were produced from the beginning of the 1st century AD onwards. It is assumed that the chief source of brass in northern Europe outside the Roman Empire was the Empire itself. Alloys with minor levels of zinc (less than 5%) could, however, derive from the smelting of mixed copper-zinc ores and so could be dated much earlier. The presence of 'Celtic' items in the 1st century AD (and later) made of brass is less cause for surprise. Many 'Celtic' hoards also contain Roman military equipment (e.g. Camerton [Jackson 1990]; Fremington Hagg [Webster 1971]; Seven Sisters [Davies & Spratling 1976]; Santon [Smith 1908-09; Spratling 1975]). It is likely that 'Celtic' brass was obtained from the Roman world. It is apparent that many 'Celtic' brasses have fairly low zinc levels (often around 15%) while the theoretical maximum content for Roman brass made by the cementation process is around 28% (Werner 1970; Craddock 1978). The maximum zinc content of 'Celtic' brass analysed for this project was 23% (the highest value for Roman brass from the same area was 26%). The lower maximum zinc content of 'Celtic' brass could easily have arisen due to the loss of volatile zinc during the remelting of Roman scrap. 'Celtic' brass should not, however, be viewed as inferior to Roman brass as the average zinc content of 'Celtic' brass (19.4%) was slightly higher than that for Roman brass (18.8%). It would seem that Roman brass was equally, if not more, likely to derive from scrap as 'Celtic' brass.

5.6 Late Iron Age copper alloys

The category late Iron Age is used to indicate the period of (perhaps intensive) contact between Britain and Rome before the Conquest (Haselgrove 1989). As discussed above, it is discussed separately from 'Celtic' metalwork as the two categories are defined differently. The late Iron Age does not have exact chronological boundaries as the Roman Conquest was not instantaneous and many areas of Britain would have seen little or no evidence of Roman control. Under these circumstances, many aspects of indigenous life (including copper metallurgy) may have continued after the conquest of Britain. For the purposes of this discussion, the late Iron Age in northern Britain is assumed to start in the 1st century BC and end in the 1st or 2nd century AD. For those areas outside the Roman empire (such as most of Scotland), there is no Roman period per se to separate the Iron Age and the early Christian era. It is usual to label the Roman period the Roman Iron Age.

Many of the indigenous rural sites examined have Iron Age and Roman phases, but do not have any significant stratigraphical or ceramic changes to indicate the 'moment' of transition. The exact dating of mid to late 1st century contexts at these sites is almost impossible. As a result, some of the samples collected from late Iron Age sites may post-date the Conquest.

Most of the objects from transitional phases on rural sites are classified as late Iron Age to distinguish them from the Iron Age (with little or no contact/influence from Rome) which was discussed above. Those finds from Iron Age contexts which contained high levels of zinc are clearly 'intrusive' and so are dealt with in this section. The samples from late Iron Age sites are mostly from brooches and, unlike the Iron Age, horse harness fittings are almost unknown (Table 5.5).

BroochesToilet ImplementsHorse HarnessSheetDroplet MiscellaneousTotal
Dragonby5 1----6
Redcliff12 3-2-320
Thorpe Thewles---1214
Dod Law2 ----13
Table 5.5. Distribution of analysed late Iron Age finds

The late Iron Age artefacts considered here are those which could be so assigned on stratigraphical grounds and so are distinct from the 'Celtic' objects already considered. The alloys used are similar, however, and it is possible that both 'Celtic' and late Iron Age cover broadly the same material.

There is a higher proportion of brass in indigenous contexts (late Iron Age, 'Celtic' and Roman period rural settlements) than in ordinary Roman metalwork. This is surprising, given that brass is often regarded as a 'Roman' metal. It was not invented by the Romans but they were certainly the first to produce brass in Europe on a large scale. This is all the more striking when it is realised that Roman brass production began in the 1st century BC just as the Iron Age in Europe was ending. The speed with which brass disseminated through Europe indicates the complexity of exchange networks at this time.

5.7 Summary

The analysis of a range of Iron Age alloys (from the earliest possible objects to those which are contemporary with the Roman occupation of most of Britain) has made it possible to describe and explain the copper metallurgy during this period. There are relatively few differences between the metals of the Bronze Age and the Iron Age and the change from Bronze Age to Iron Age should be seen as a transition. The only two substantial differences between late Bronze Age alloys and Iron Age ones are the higher iron content and lower lead content of Iron Age alloys. The higher iron content probably relates to the use of a new smelting procedure making more use of free-running slags. The lower lead contents are curious as there is no metallurgical reason why Iron Age alloys should have low lead contents (most are castings). The lower lead levels may reflect changes in the mining and supply of metals or wider social and economic changes in later prehistoric Britain.

The copper alloys of the Iron Age proper are all tin bronze (occasionally with a little lead). Zinc is almost never present in these alloys but arsenic is a frequent impurity. While previous analyses of Roman copper alloys have shown that they are actually a range of different alloys often containing zinc, tin and lead, Iron Age copper alloys can safely be referred to as bronze. The Iron Age alloys of northern Britain are similar to those previously published results for (mainly) southern England. The British Iron Age shows a great uniformity in its alloying tradition and implies wide exchange of materials or knowledge throughout Britain at this time.

The use of a single alloy type (bronze) ends some time towards the end of the Iron Age in Britain. Many items of 'Celtic' metalwork from the late Iron Age are made of alloys containing at least some zinc. This zinc probably came from imported Roman brass. Firm dating for the start of this change is not available and the change was almost certainly not instantaneous throughout the whole country. A succession of changes occur in the late Iron Age, which begin in the 2nd century AD and continue through to the Roman Conquest and after. 'Celtic' coinage appears in Britain in the last half of the 2nd century BC, and later coins are produced in Britain. Amphorae are imported from c.100 BC onwards, and fine pottery, glassware and metalwork from the last half of the 1st century BC onwards. Roman (or Gallo-Roman) imports do not appear in northern Britain until the 1st century AD. There does not seem to be a single historical event which marks either the beginning or the end of the late Iron Age in Britain. The 'Celtic' metalwork made of brass would seem to belong to the late Iron Age. An even more precise date can be assigned in most cases as brass was not produced on any scale in Europe prior to the Augustan coin reforms of 23 BC. This provides a terminus post quem for the production of most brass in Europe. It is likely that there was a delay before brass appeared in northern Britain. It is suggested here that all brass in northern Britain was produced from the beginning of the 1st century AD onwards.

Brass is a relatively common alloy in 'Celtic' and late Iron Age metalwork - even more common than in many Roman contexts. In addition the quality of 'Celtic' brass (as measured by the mean zinc content) is not inferior to Roman brass. 'Celtic' metalworkers probably obtained their brass from the Roman world but this brass would not necessarily have been looted or stolen. They could obtain ingot quality brass, perhaps gifts as part of a treaty between the Roman Empire and its northern neighbours (Braund 1984). It must also be considered that non-Roman metalworkers may have learned the cementation process and begun their own production of brass.

There are many changes in copper alloy metal composition through the course of the Iron Age. These changes are not always in step with the chronological horizons of conventional archaeology. The earliest Iron Age shows little change from the late Bronze Age, but considerable change does occur at the start of the middle Iron Age and during the late Iron Age. This undermines a traditional notion of time, characterised by Collingwood as being neatly divided up into distinct eras:

each with peculiar characteristics of its own, and each marked off from the one before it by an event which in the technical language of this kind of historiography is called epoch-making. (Collingwood 1993: 50)

The mis-match between historical and archaeometallurgical evidence is not unique. A similar phenomenon can be seen in the production of Medieval pins (Caple 1991).

All barcharts used in this paper were produced using the Excel computer spreadsheet programme. The labels for the individual bars are given as single numbers (in the case of Figure 16 as 0, 1,.2,.3, etc). Each bar actually represents a range of compositions (in the case of Figure 16, 0-0.999, 1-1.999, 2-2.999, 3-3.999, etc.).


© Internet Archaeology
Last updated: Thu Apr 3 1997