7 Copper Alloys and Typology

7.1 Introduction

Early work on archaeological copper alloys has concentrated almost entirely on the alloys themselves (and this was the approach used earlier in this study). More recently (i.e. since the early1970s), the compositions of copper alloys have been related to the typology of the objects concerned. One of the primary aims of the present research has been to obtain a representative sample of all copper alloys used in Iron Age and Roman northern Britain. As a result many samples have been obtained from generic artefacts such as wire or sheet, and so there is relatively little scope for an in-depth examination of alloy composition in terms of typology. Nevertheless, where a substantial proportion of typologically distinct objects has been analysed the results are considered (especially where previous research already provides some information).

As part of the sampling procedure, artefacts were assigned to functional categories: Personal, Household, Military, Transport, Metalworking Waste, and Uncertain. These categories were useful in monitoring the selection of samples and ensuring representativeness. The wider usefulness of the categories is questionable, however, as many items occupy intermediary or uncertain positions within such a classification system. In addition, a classification based solely on function must rely on modern rather than ancient perceptions of function and fails to take into account possible symbolic meanings. The proportions of alloys used for the categories are shown in Figure 44. There is some variation in the proportions of the alloys used for different categories of artefact. The proportion of brass is highest for Transport and Military items and lowest for Metalworking Waste. The variation, however, is relatively slight, especially when this is compared to the variation seen in alloys over time and on different sorts of sites.

Fig.44 Alloys used for different categories of artefact
(Actual number of analyses given in brackets)

7.2 Brooches

A variety of different types of Iron Age and Roman bow brooches are known. Two detailed typological examinations of these brooches have been carried out by Don Mackreth and by Mark Hull, but neither has been fully published (Hull's typology and catalogue of Iron Age brooches has been published - Hull & Hawkes 1987). General typological divisions are well-known (based mostly on Collingwood 1930b) but a knowledge of the detailed sub-division of types is largely restricted to a small group of specialists.

Iron Age brooches are plainly different from Roman ones: they are rarer than Roman ones, there are fewer types and most Iron Age brooches are actually made of iron rather than copper alloy. Straightforward comparisons should, therefore, be avoided. There is no chronological overlap in types between the early Iron Age and the Roman period. Late Iron Age types do overlap, however. One-piece brooches derived from La Tène III types (e.g. Colchester A and Nauheim derivative) and early hinged brooches (e.g. Rosette) span the 20 or 30 years around the Roman Conquest. The Roman period sees an expansion in the number of different types of brooches produced (Colchester B, Polden Hill, Dolphin, Trumpet, Headstud, etc. - see Figures 45-52). These brooches are common on Roman sites occupied in the first two centuries AD. The later Roman period sees a dramatic decline in the deposition of bow brooches which may or may not reflect a decline in the use of such brooches. Knee brooches belong to the 2nd and 3rd centuries while Crossbow brooches are dated to the 3rd and 4th centuries.

The range of alloy compositions of Iron Age and Roman brooches has been dealt with in detail by Bayley (1992). The alloys used were often specific to the types of brooches, and distinct alloy compositions also pointed out flaws in the typology used (Hull's). A re-assessment of Hull's typology in the light of Bayley's analyses is forthcoming (Bayley & Butcher forthcoming). All the results of the brooch analyses carried out for this research agree with those carried out by Bayley (1992) except where noted. In this section most attention will be devoted to those brooch types which are found predominantly in northern Britain (Trumpet, Headstud, and Dragonesque brooches). The numbers of these analysed by Bayley were small as they are rare in southern Britain.

All of the Early Iron Age brooches were made of tin bronze. This alloy was typical of the period. There are no differences in the alloy used for different types of Iron Age bow brooches, except perhaps in the case of Birdlip brooches which have a relatively high tin content (12-14%) compared to other brooches (8-11%).

La Tène III brooches have their origins in the pre-Roman period but similar types (e.g. Nauheim derivative - see Figure 45) continue in production after the Roman conquest of Gaul (and possibly after the Roman conquest of Britain). These brooches can be split into two groups on the basis of their composition: some are made of tin bronze with no zinc present, while others are made of brass with 20% or more zinc. The use of either brass or bronze (but not gunmetal and always with low lead) for these brooches is also found in Bayley's (1992) results. The first group of brooches has an alloy composition similar to most Iron Age objects (tin bronzes with little or no zinc). A simple chronological separation of such brooches into pre-Roman bronze and Roman brass is unlikely, however, as the bronze used to make these brooches does not have the same impurity pattern (regular presence of arsenic) that is found in almost all Iron Age alloys.

Many of the Hod Hill and other early hinged brooches analysed for this research have compositions comparable to those found in Bayley (1992). Most of these brooches have a relatively high zinc content (15-25%) which is often accompanied by a moderate level of tin (0-5%). The remaining early hinged brooches are made of tin bronze with little or no zinc. In all cases lead is absent (or present at only very low levels). The lack of consistency in the choice of alloy (brass or bronze) for early hinged brooches is also found in Bayley (1992). It is also reflected in the wide typological variation of these brooches (Hull's typology includes 31 different variants of the Hod Hill brooch), and exact parallels for a particular Hod Hill brooch are rare. This lack of exact typological standardisation can also be seen in the military fittings catalogued by Oldenstein (1977). This catalogue is indispensable for finding parallels for newly excavated finds, but exact parallels are extremely rare. This suggests that each object was unique and that production was on a relatively small scale (Allason-Jones 1994).

Only two Langton Down brooches were analysed during this project but these have compositions which are similar to those found in Bayley (1992). They are made of brass (moderately high zinc - rarely over 20%) with low (0-3%) levels of tin and lead present. Only one possible Rosette brooch was analysed. This was made of brass (as are those analysed by Bayley) but has very low tin content and no lead present, unlike those analysed by Bayley.

Fig.45 Principal brooch types: Nauheim Derivative (XRFID 1230)

Fig.46 Principal brooch types: Colchester A (XRFID 1010)

Fig.47 Principal brooch types: Dolphin (XRFID 1605)

Fig.48 Principal brooch types: Trumpet loose headloop (XRFID 1234)

Fig.49 Principal brooch types: Fantail, fixed headloop (XRFID 1635)

Fig.50 Principal brooch types: Headstud, fixed headloop (XRFID 1766)

Fig.51 Principal brooch types: Dragonesque (XRFID 1392)

Fig.52 Principal brooch types: Penannular, type A2 (XRFID 1505)

Eight Colchester A brooches were analysed during this research project (Figure 46). (It was only realised that no Colchester B's had been selected after all analysis had been completed. This may reflect the fact that Colchester B's are relatively rare in northern Britain.) All but one of the Colchester A's had a typical composition (brass with little or no tin or lead). The Colchester A from Elginhaugh (XRFID 1201) was the only one with an atypical composition - it had a relatively high lead content for a one-piece brooch. (It is hardly surprising that the brooch had broken - possibly during manufacture. One other incomplete Colchester brooch was found on the site.)

The Polden Hill and Dolphin brooches (Figure 47 ) are all of bronze or gunmetal (the zinc content is fairly low and never exceeds 10%, while the tin content never falls below 3%) with variable lead content.

The work of Bayley & Butcher (forthcoming) has considerable impact on the typology of Trumpet brooches (Figure 48). These were sub-divided into various types on the basis of variations in the nature of the moulding in the middle of the bow (first by Collingwood 1930b, and later by Hull). Analysis of Trumpet brooch composition by Bayley (1992: 154-6) has shown that the nature of the headloop is a more useful criterion for sub-division. The re-categorisation of Trumpet brooches briefly reported in Bayley (1992: 154) argues for five types:

A 'A standard undecorated Trumpet brooch . . . with sprung pin, loose headloop and fully-rounded waist moulding . . . most of these are brasses or gunmetals' (Bayley 1992: 154).

B 'devolved plain ones with fixed headloops and half-round waist mouldings . . . which are leaded bronzes' (Bayley 1992: 154).

C 'A further group of devolved imitations . . . are all leaded bronzes whether they have loose or fixed headloops. They have lower tin and higher lead contents on average than the Trumpet B leaded bronzes' (Bayley 1992: 154-156).

D 'decorated Trumpets . . . which display a high degree of craftsmanship with a number of distinctive patterns of enamelling . . . Most are unleaded alloys with brass predominating.' (Bayley 1992: 156).

E 'assorted Trumpet-headed brooches . . . mainly have low lead contents with zinc-rich alloys predominating . . . this is not a tight compositional grouping' (Bayley 1992: 156).

This new typology is as yet a little difficult to use as there is no discussion of the criteria required, for example for a brooch to be regarded as 'devolved'. The new Trumpet brooch typology still draws on fairly subjective appreciations of the style of moulding derived from Hull's typology. Nevertheless some agreement with the principles of the new typology can be found in the analyses of Trumpet brooches described in this report (Figure 53).

All of the Trumpet brooches with fixed (cast) headloops are of leaded bronze or leaded gunmetal (the lead content is at least 2% in every case, and the average is 6.6%). There are no obvious sub-divisions in the alloy composition that can be related to the nature of the moulding (whether acanthus or not, and whether all-round or front only). Those with loose (wire) headloops are made of varied alloys (brasses and bronzes, with and without lead). They can, however, be sub-divided into two groups based on the style of the moulding. Those with plain mouldings (all-round and front only) are mostly brasses (zinc 10-20%) with low lead contents (average = 0.8%). Those with acanthus mouldings (all-round and front only) all have higher lead contents (average = 3.8%). There is no obvious sub-division based on whether the moulding is all-round or front only.

Fig.53 Lead and zinc content of Trumpet brooches

Only one possible Aesica brooch (XRFID 1447) was analysed during this research. Only half of the brooch survives and so the identification is uncertain. This brooch is made of brass (13.5% zinc) with a small amount of tin and lead. This is unlike the two Aesica brooches analysed by Bayley which were made of bronze with high lead levels. This difference in alloy composition does not necessarily rule out the identification of XRFID 1447 as an Aesica brooch however, as the Aesica type is not typologically homogeneous and only three brooches of this type have ever been analysed.

The fantail brooches analysed (Figure 49) can be divided into two groups on the basis of their composition. The first group consists of two brooches with discs on the bows. These are both made of brass (zinc 10-15%) with low levels of tin and lead also present. The second group consists of small fantails often with curvilinear enamel decoration on the tail. These are all made of leaded bronze but the tin and lead levels are relatively modest (always less than 10%).

Just as Trumpet brooches can be split into those with fixed and those with loose headloops, the same can be done with Headstud brooches (Figure 50). Most of the Headstud brooches with fixed headloops are made of leaded bronze (Figure 54). There are three exceptions (XRFID 1425, 1791 & 1905) which are made of brass (with little or no tin) but there is no obvious typological basis for separating these brooches from others with fixed headloops. The majority of Headstud brooches with loose headloops are made of brass with low levels of lead.

Fig.54 Zinc and lead content of Headstud brooches

Knee brooches and Crossbow brooches are almost always made of leaded bronze with little or no zinc. Insufficient numbers of either type of brooch were analysed to be sure if compositional variation could be related to typological variation

A limited number of plate brooches have been analysed during this research project. Relatively few attempts have been made to devise typological frameworks for plate brooches as there is considerable variation (Hull lists 81 different types of plate brooch - not including variants). Most plate brooches seem to have been used throughout the Roman period and few can be more precisely dated. Dragonesque brooches are one of the few sorts of brooches that are found almost exclusively in the north of Britain and so were especially selected for analysis. They are discussed more fully below.

Bayley suggests that some early plate brooches were made of brass while most other plate brooches were made of leaded bronze (Bayley & Butcher 1989; Bayley 1992). The results of the present research supports this as the few high zinc plate brooches are of early types (e.g. Cruciform, XRFID 1595, 1876) or from early contexts (e.g. Stanwick, XRFID 1850; Redcliff, XRFID 1014). The remaining plate brooches are of mixed alloys with leaded bronze predominating.

Dragonesque brooches (Figure 51) have been studied in a series of short articles by Bulmer ( 1938) and Feachem (1951; 1968). The Bulmer/Feachem typology divides brooches according to the nature of the enamel decoration (Feachem 1968: 100). Bulmer envisaged a chronological sequence from unenamelled Dragonesque brooches to type iv which gave rise to both type iii and type i. Finally type ii developed out of type iii (see Bulmer 1938: Figures 3 and 4). Dating for Dragonesque brooches suggests that they appeared roughly at the same time as the Roman Conquest and they continued in use until the mid/later 2nd century AD.

Analysis of 16 Dragonesque brooches during this research supplements the single analysis carried out each by Bayley (1992) and Craddock (1975). As a whole, Dragonesque brooches are made from varied alloys (see Figure 55). A distinction can be made using a simplified version of the Bulmer/Feachem typology. In this simpler system there are four types: plain ones (P), those with simple circular decoration (R), those with square decoration (S), and those with a lattice arrangement of decoration (L).

The P and R types generally have modest zinc contents and relatively high tin contents (mean values are 3.3% and 7.9%, respectively), while the S and L types have higher zinc and lower tin contents (mean values of 8.7% and 3.4%, respectively). All that remains of one of the brooches, from Kelco cave (Feachem 38), is the head and so it cannot be assigned to a type with certainty. The lack of enamel on the surviving fragment and the composition (high tin and low zinc) suggest that it belongs to an early P or R type.

The generally modest zinc levels (usually less than 15%) in all Dragonesque brooches suggests a limited use of scrap metal derived from the Roman World (although some was used as only one of the brooches had no zinc present). The trend from low zinc, high tin alloy to moderate zinc, low tin alloy may indicate a gradual increase in the use of Roman scrap metal (especially brass).

One of the Dragonesque brooches which has been analysed, that from Milking Gap, has a very low zinc content. In fact the alloy used for this brooch is very similar to that used for Iron Age artefacts in northern Britain. From this it may be deduced that Dragonesque brooches were first produced before the Roman Conquest (it is possible, however, that this alloy type continued in occasional use after the Conquest). If the Milking Gap brooch is one the earliest surviving examples of this type then the Bulmer/Feachem typological scheme (i.e. P before R) may be incorrect.

Fig.55 Zinc and tin content of Dragonesque brooches

The standard work on penannular brooches (Figure 52) is still Fowler (1960). This divides brooches into types according to the nature of the terminals (moulded, bent, etc.). Type A consists of those with slight thickening of the terminals and no other decoration. Fowler places these at the start of her typology and this is supported by the metal composition. The type A brooches analysed for this project were all made of bronze with little or no zinc or lead. This alloy is relatively rare in the Roman period but is typical of the early Iron Age. Two type A brooches are from mid (XRFID 2100) or late Roman (XRFID 2110) contexts at Rudston villa but they may be residual (especially as there was pre-Roman occupation at Rudston).

Type A2 and A4 both have terminals with oblique incisions in the terminal moulding (referred to as 'milled terminal' by Fowler). These brooches were usually made of bronze with only low levels of zinc. The common presence of lead in these brooches would seem to be sensible as they would have been cast rather than wrought.

A3 penannulars consist of those with plain but repeated mouldings on the terminals. Some of the brooches are made of leaded bronze (with low levels of zinc) while others are made of brass (15-20% zinc). There are no obvious typological correlations between the alloy used and either the size of the brooch or the complexity of the mouldings.

Type D penannulars have terminals formed by bending the terminals back on themselves (these are often shaped and are decorated with incised lines, etc.) and type C have terminals formed by bending the terminals back on themselves into spirals. Both C and D penannulars are made from a range of alloys (brass, gunmetal and bronze) but both C and D are wrought and so they usually have lower lead levels than most A3 brooches which have cast terminals. There are no other correlations between typology and metal composition.

During the selection of artefacts for analysis, a type of penannular brooch was identified which does not seem to have been described before. This is a very simple brooch with terminals formed by wrapping sheet metal around the ends of the loop. In all four cases examined, the brooch was incomplete and in fragments and so it is possible that the 'sheet terminal' is actually the remains of the pin wrapped around the body of the penannular loop. In at least one case (XRFID 1646), however, both terminals have survived. All of these brooches are made of tin bronze, occasionally with low levels of zinc or lead.

7.3 Other items of personal adornment

A considerable proportion of the copper alloy artefacts from Roman Britain are objects of personal adornment. This includes the bow, plate and penannular brooches discussed above, and bracelets, rings and other items discussed below. Discussion of the possible relationships between metal composition and typology is inhibited by the lack of explicit typological discussion of these items. Where no explicit typology exists, Crummy's (1983) catalogue of Roman artefacts from Colchester provides a framework. This volume is of limited use, however, as it only deals with material excavatedfrom Colchester.

Bracelets have been subjected to more study than many other personal items (Wheeler & Wheeler 1932; Allason-Jones & Miket 1984; Crummy 1983). Bracelets are probably of most interest as some types are restricted to the late Roman period and so are useful as dating material. The Wheelers' (1932) study was restricted to late Roman decorated strip bracelets and so concentrated on the 'grammar' of the decoration (zig-zag, ring-and-dot, etc.). The catalogues by Allason-Jones & Miket (1984) and Crummy (1983) both cover a wider chronological range of material. The former suffers from a concentration on the method of fastening with little attention paid to the types and organisation of decoration. For the present research, a simplified typology was devised, using Allason-Jones and Miket (1984) for overall form, with late Roman decorated strip bracelets divided into four groups: 'crenellated', 'ring-and-dot', 'notch-decorated' (or 'running wave'), and miscellaneous.

Attempts to relate the composition of late Roman decorated strip bracelets to their typology have not met with success. A range of tertiary alloys (copper with zinc and tin) were used with no obvious correlation between alloy type and the 'grammar of decoration'. All of these decorated strip bracelets, however, have a consistent (if relatively small) addition of lead (mean 2.85%; s.d. = 1.25%). The presence of at least a small amount of lead would suggest that most of these bracelets were cast rather than wrought.

Solid bracelets all have thick, circular or oval sections (3-7mm). Most are made of leaded bronze and few can be closely dated. Various wire bracelets (plain, twisted wire, spiral wire, etc.) have been analysed. These are all united by having a low lead content; this is unsurprising as all wire bracelets were wrought, and wrought alloys generally have low lead contents. The wire bracelets were usually made of brass, copper or bronze, with relatively little use of tertiary or quaternary alloys. Three examples of one bracelet type, with all-round incised groove decoration (cf. Allason-Jones & Miket 1984: catalogue 245), were analysed and all were composed of 'impure copper' (the only alloy element present was tin in the range 1-3%).

Typologies for finger rings have been proposed by Henig (1978) and Crummy (1983); the former, however, includes only those rings which could have held a gem, while the latter lumps all of these different rings into one single group (bezelled). The largest group of rings analysed here, however, consisted of plain rings with no decoration or bezel. While some of these are the right size to be finger rings, this is no guarantee that they are in fact finger rings. Some of these plain rings had large sectional diameters (6-8mm) and are almost certainly not finger rings. They may have been items of horse harness or belt fittings. Others with smaller sectional diameters may have served similar utilitarian functions. Plain rings were made from a range of alloys with mixed (zinc and tin) being the most common, and almost all are leaded. Some of these leaded alloys have extremely high lead contents (20% or more). These highly leaded rings form a substantial proportion of all highly leaded alloys. Out of a total of 1212 Roman alloys, 33 have more than 20% lead, and eight of these 33 are rings.

Spiral finger rings are all made of brass with only low levels of tin or lead. The lack of lead is sensible from a metal working stand point but the choice of brass over bronze seems strange when the three ear-rings analysed were all made of tin bronze (again with little or no lead). This may show that there were 'social rules' governing the wearing of jewellery which included the composition of the metal. It is difficult to know whether the composition of the metal in this case reflects rules governing the colour of the metal (brass being golden, and bronze a more pinkish colour), or a more passive reflection of workshop practice or metal supply.

Henig's (1978) typology for bezelled finger rings is a little difficult to use as the rings are only defined by simple illustrations. Three of the rings analysed here (nine were analysed in total) belong to a type which seems to be transitional between types II and III. There is no apparent correlation between Henig's types and composition, but leaded bronzes are the most numerous.

Button-and-loop fasteners have been studied in detail by Wild (1970) who divided them into types based on the shape and decoration of the button. Most of the button-and-loop fasteners analysed here were made of quaternary alloys with moderate levels of zinc, tin and lead. In few cases did the level of any alloy element rise above 10%. The only fastener with a relatively high zinc content was of class IX which seem to be related to lorica segmentata tie loops (Robinson 1975: fig 184).

7.4 'Household objects'

This section deals with the range of (mainly Roman) artefacts which are believed to have been used primarily in domestic settings, but were not used as items of personal adornment. However, the term needs to be used flexibly as the exact use to which objects were put is unknown (similarities with modern objects is no proof of utilitarian function, let alone symbolic meaning). Toilet items (e.g. probes, spoons, tweezers, etc.) can easily be envisaged as being used in a domestic setting for personal grooming. This does not, however, rule out the use of such items as symbolic markers (perhaps badges of Romanisation) especially when used in social contexts far removed from the Mediterranean world.

The term 'toilet implements' is used here to cover a range of items which are assumed to relate to personal grooming: tweezers, probes, nail cleaners, and scoops. Many of these objects were wrought rather than cast and so a low lead content is not surprising. A low lead content is also found, however, in those toilet implements which were cast.

There has been little systematic study of these items and the only typological distinctions made here are for the toilet scoops (ligulae). These are divided into three groups: those with small round bowls and long handles, those with small round bowls and short handles (made of sheet metal), and those with elongated bowls (Figures 56-8).

Fig.56 Three categories of ligulae analysed in this paper: Long Handle, Small Bowl

Fig.57 Elongated Bowl

Fig.58 Short Handle, Small Round Bowl

The first group is usually made of brass, the second group of bronze, and the last group of gunmetal (Figure 59).

Fig.59 Alloys used for ligulae

Tweezers are usually made from a single strip of sheet metal which is bent back on itself, so it is not surprising that they are rarely made of alloys containing lead. Tweezers are made from both brass and bronze.

Of the numerous types of Roman spoons studied by Riha & Stern (1982), only the plain round-bowled type has been analysed during this research. The seven spoons analysed here are made of a range of alloys but zinc is rarely a major component. Some spoons are wrought and have virtually no lead present (and often low tin content). Others are cast and have high lead content. The cast spoons can be distinguished typologically from the wrought ones by the presence of a rib, extending from the handle, part way along the bowl. Stern's results (Riha & Stern 1982) seem to show the use of very highly leaded bronzes (almost pewter). However, Stern's analyses were all made of the surface of the spoons, which could be misleading as many spoons were corroded and some had surface tinning (nevertheless Stern treated the results quantitatively).

All of the Roman mirrors analysed during this research have been found to be made of speculum (a high tin bronze). A tin content of c.15% or higher can give a silvery-white alloy which can be polished to give a reflective surface. Craddock found that the Roman mirrors that he analysed had tin contents of 18.6% to 22.8% (with one exception of 8.5% (Craddock 1975: 152). All of the mirrors analysed during this project have tin contents of 18% or more. In addition, all of the mirrors had relatively high lead contents (8-26%) and only trace levels of zinc. There is very little overlap between the tin contents of Roman mirrors and other artefacts with high tin contents (see Figure 60). Of the 14 artefacts with 15-20% tin only one was a mirror, and only one of the 12 artefacts with more than 20% tin was not a mirror. (This sample was a piece of casting waste from Rudston villa, and may indicate the production of speculum mirrors in Britain. Limitations of the composition of casting waste are discussed in Chapter 8.)

Fig.60 Tin content of Roman mirrors and other objects (with tin 15% and over)

The large sheet metal vessels of prehistory known as cauldrons have been studied by Hawkes (1951). Cauldrons first appear in late Bronze Age hoards (e.g. Heathery Burn) and these early vessels are characterised by having suspension rings attached to the body by staples. Simpler vessels with no suspension loops are also known and these are conventionally dated to the Iron Age. These are divided into two types: Santon and Globular. The former (also known as 'projecting-belly') are often made of at least two pieces of metal, while the latter are hemispherical and are made of a single piece of metal. Many cauldrons are only roughly provenanced and are without any archaeological context. This makes their dating difficult. Some cauldrons have been found in hoards which also contain Roman items (Piggott 1952-3; Hawkes 1951: 181) and may have been made during the Roman occupation.

The Santon vessels analysed were all made of tin bronze with no zinc or lead - typical of Iron Age alloys. These vessels, however, are made from a tin bronze which has little or no arsenic present as an impurity (unlike most Iron Age bronzes). It is not possible (from the analysis of the metal) to be certain of the dating of this type of cauldron. They may be Iron Age or Roman in date.

The Globular cauldrons analysed were mostly made of tin bronze but one (Lochmaben XRFID 1950) also contained 1.2% zinc. The presence of some zinc is typical of many Roman alloys, and uncharacteristic of the Iron Age. Nevertheless the presence of small amounts of zinc in pre-Roman alloys (from the use of mixed copper and zinc ores, or the use of some scrap brass imported from the Mediterranean world) cannot be dismissed. A vessel from Welton Wold (XRFID 1492) which came from a 4th century context was made from sheet bronze in a similar way to many cauldrons.

Both Santon and Globular cauldrons occasionally have repairs made to slight tears by the addition of a small patch. Some of these patches are not riveted on but consist of a piece of sheet metal bent back on itself so that two 'wings' can be pushed through the tear. The patch is then smoothed down. These artefacts can be seen in situ in some Scottish cauldrons (Abercairney, Elvanfoot, Kyleakin, and Carlingwark Loch: Macgregor 1976: catalogue nos. 300, 303, 306, and 309). Macgregor (1976) refers to these as 'paper-clip repairs', and Cool (1990) refers to them as 'diamond clips'. All of those analysed here were made of copper or bronze. Zinc and lead are virtually absent. The composition of the 'diamond clips' closely matches that of many cauldrons. It is curious that while the cauldrons are mostly found in Scotland (with a few in southern England, see Macgregor 1976: 170-171; map 21), the 'diamond clips' have been found at a number of sites throughout Britain: Vindolanda (Birley forthcoming), Piercebridge (Fitzpatrick forthcoming b), Dalton Parlours (Cool 1990: 89; fig 72: 51-7), Brough-on-Humber (Wacher 1969: 89; fig 38: 26), Gadebridge Park (Neal 1974: 137; fig 59: 104-6), and Whitton (Jarrett & Wrathmell 1981: 187-8; fig 74: 99, 100).

While most pre-Roman vessels are made from beaten sheet metal, Roman vessels were usually cast (although some may be spun). The large-scale production of paterae and other vessels is assumed to have taken place in Campania in Italy in the late Republican period (Fredericksen 1959). Production seems to have shifted north (ultimately into Gaul) during the Principate (Willers 1907). The study of vessels has been greatly helped by the publication of two Dutch monographs on the subject: a catalogue and typological discussion of the vessels in Nijmegen Museum (denBoesterd 1956), and the spectrographic analyses of these vessels (denBoesterd & Hoekstra 1965). Most of these vessels date to the earlier part of the Empire. A discussion of the typology and metal composition of some later continental vessels can be found in Lindberg (1973).

The present research has included the analysis of 45 vessels or vessel fittings. All of the paterae are dated to the 1st or 2nd centuries and confirm the earlier finds of denBoesterd & Hoekstra (1966), that all of these vessels are made of leaded bronze with little or no zinc. This composition is also found in miscellaneous vessels from the early Empire and many of those which were not closely dated.

A substantial number of 3rd century vessels were made available following the analysis of finds from the cemetery at Brougham (Wilson 1968: 179). Vessels formed a significant proportion of the grave goods and 17 vessels were analysed. Unfortunately the burial rite at Brougham was cremation and all of the vessels are fragmentary. Four of these vessels had wedge-shaped rims which were decorated (usually with lines). This decoration closely resembles the Hemmoor vessel illustrated in Lindberg (1973: fig 17). The four wedge-rimmed vessels are all brasses which is the same as most Hemmoor vessels analysed by denBoesterd & Hoekstra (1965) and Lindberg (1973). The use of brass for a specifically mid to late Roman artefact type confirms Craddock's (1975: 221-4) suggestion that brass production continued into the late Roman period.

The other 3rd century vessels have rims similar to early Roman vessels (straight or turned over). Like 1st century vessels, many of these may have been formed or finished on a lathe. These later Roman vessels, however, are usually made of bronze with little or no lead, in contrast to those from the early Empire which are usually leaded. The single 4th century vessel analysed was a large sheet bowl from Welton Wold (XRFID 1492). The method of manufacture is similar to that of the cauldrons discussed above, and the vessel may have been made some time before the 4th century. The alloy used in this case was a slightly leaded bronze with no zinc (or arsenic) present.

7.5 Military fittings

Roman military equipment is one of the most intensively studied areas of Roman material culture (reflecting traditional interests in the period - see Bishop & Coulston 1993, for a recent review of the subject). Military equipment from archaeological sites has often been used to date the period of occupation at a site, and identify the units stationed there (i.e. legionary or auxiliary). The highly distinctive fittings of the lorica segmentata (the segmented cuirass, Bishop & Coulston 1993: 85-90) have attracted considerable attention as they are identified as items used by Roman legionaries during the first two centuries AD (see also Maxfield 1986, for discussions of the use of military equipment in this way). Many other military items are used over much greater periods of time and cannot be associated with just one particular type of unit. Despite the considerable attention paid to lorica segmentata fittings, it is only recently (Bayley 1985b; Bishop 1989b) that they have been shown to be made of brass (rather than bronze). Bishop & Coulston (1993: 191) suggest that two different sorts of brass were used in the manufacture of lorica segmentata: one containing c.20% zinc used for the sheet fittings (hinges, buckle plate, hinges, etc.), and one containing 10-15% zinc used for rivets.

Eight samples of lorica segmentata were analysed during this research; all but one of these was made of brass. The average zinc content was c. 20%, in line with Bishop & Coulston (1993: 191). One lorica segmentata fitting (XRFID 1082) was, however, made of a bronze with 5.7% tin. This item may have been an ad hoc repair or replacement. Overall, the uniformity of the composition of lorica segmentata fittings suggests that most were made to a strict recipe. This may indicate that they were produced centrally by the state. Alternatively, they may have been produced on a more local scale but to a recipe known and adhered to on a much larger scale (Welbourn 1985). This scale can only be known when fittings from other parts of the Empire are analysed, but typological uniformity hints at metallurgical uniformity.

The rivets and other fittings associated with lorica segmentata fittings have a more varied composition. Some are made of a brass with zinc content similar to the actual sheet fittings, while others have much lower zinc contents (in some cases approaching 0% zinc). Nevertheless, other alloy elements (tin or lead) rarely seem to be deliberate additions. Some rivets etc. do seem to conform to the lower zinc alloy suggested by Bishop & Coulston (1993: 191). The lower zinc content of rivets can be explained, as it would have made the rivets softer and so easier to work. Alternatively (or additionally), the lower zinc content would have made the rivets more coppery and so redder in colour. It is surmised that a colour contrast may thus have been produced, which was then selected on aesthetic grounds. The colour difference that such changes in composition ought to produce was confirmed visually during the sampling of one lorica fitting and its rivet (XRFID 1805a & b). The corroded metal appeared uniformly green over the sheet metal and the rivets, but the removal of corrosion products by air abrasion showed a dramatic colour contrast between the two parts. Nevertheless, in two out of three cases where a lorica segmentata fitting and its rivet could be analysed, the rivet contained the same level of zinc as the fitting.

Some other military fittings of the early Roman Empire ('armour tin pins' and some scale armour) are consistently not made of brass (contra Bishop & Coulston 1993: 191). Other alloys used include bronze and impure copper (usually with a small amount (1-3%) of tin, but very little else). This does not affect whether or not the army's brass was produced under an official monopoly but it does show that a range of alloys were used to manufacture military equipment.

Scale armour (lorica squamata) has been found in contexts from the 1st to the 4th century; the eight samples analysed here, however, are mostly from early Roman contexts. All the armour scales are made of alloys which do not contain lead. This is logical as the scales are wrought from sheet metal. The levels of tin and zinc in scale armour varies: some are made of brass, one of bronze, and two are made of impure copper. The one armour scale made of bronze is also the only one from a late Roman context (XRFID 2169). This change over time from brass (or copper) to bronze fits the wider compositional changes in Roman metalwork.

Bishop & Coulston (1993: 191) suggest that Roman cavalry fittings of the early Empire were usually made from leaded brass (primarily on the basis of analyses by Craddock and Lambert in Jenkins 1985). Many of the cavalry fittings (especially junction rings and loops) analysed here are made of brass which also contains small amounts of tin and/or lead, but there does not seem to be the same consistent addition of lead to brass as seen in the Xanten fittings (Jenkins 1985). XRFID 1277 is a bifid pendant of a type typical of the 1st century AD but it comes from a 3rd century context. There are typological parallels with pendants from the Xanten hoard but the composition is most unlike those from Xanten. The Catterick pendant (XRFID 1277) is a leaded bronze with only 2.2% zinc (not untypical of alloys in the 3rd century), whereas most bifid pendants from Xanten are leaded brass. The Catterick pendant may show the late manufacture of bifid types (in contemporary alloys) or the use of a variety of alloys for cavalry fittings in the 1st century. Production of cavalry fittings at Alesia has recently been discussed by Rabeisen (1990). The production of these items within a Roman town suggests that they were produced privately rather than by the state.

The study of mid-Roman military fittings is largely based on the work of Oldenstein (1977) who catalogued all of the 2nd and 3rd century (supposedly) auxiliary equipment from forts in Upper Germany. The catalogue contains a large number of cast belt fittings (often in openwork), few of which are identical. Allason-Jones (1994) has suggested that searches for detailed typological parallels amongst this equipment may always fail because each object was unique. The uniqueness of these fittings is in stark contrast with the limited number of forms of lorica segmentata. Production of late 2nd and 3rd century fittings was probably at the scale of the individual artefact and so each was different. The casting of most mid-Roman military fittings contrasts with the early Roman equipment, which is often wrought. The analysis of mid-Roman fittings shows most of them were made of quaternary alloys. These have only moderate levels of zinc (rarely over 10%) but few are zinc-free. Tin is the most consistently added alloy element. In addition, almost all of these fittings contain considerable quantities of lead (usually over 5%). This composition is in marked contrast to that used for most 1st century military fittings where lead was rarely used. The mixed nature of mid-Roman military alloys could be the result of recycling scrap of varied composition but the overall increase in the lead content must result from the use of metal (whether scrap or fresh) not previously used for most military fittings. The fact that the mixed alloy is ideally suited to the method of manufacture (casting) suggests that such alloys may have been deliberately aimed for, rather than the passive result of recycling.

Fourth century Roman military fittings are relatively rare. This may reflect the fact that the army was smaller (James1984) and may have been less well supplied (and so much less equipment would be abandoned). Some of the most useful finds of Roman military equipment in the early Empire are the result of expansion and military conquest. As units were moving to new locations on a regular basis they may have abandoned some of their scrap metal at each old fort (Bishop 1989a; Allason-Jones & Bishop 1988). The relative paucity of identified 4th century military fittings may also reflect the ways in which the late Roman army was becoming less different from the civilian world (EsmondeCleary 1989). The standard work on these late Roman military fittings is still Hawkes & Dunning (1961), which catalogues chip-carved belt plates (cast from bone or wooden originals), zoomorphic buckles, sheet metal strap ends, and other miscellaneous finds. Chip-carved belt plates are almost unknown in northern Britain so samples for this study were taken from zoomorphic buckles and sheet strap ends. In addition, one belt plate (from Catterick, XRFID 2025) with simple ring-and-dot decoration was analysed. The zoomorphic buckles are made of leaded bronze with little zinc. The strap ends are made of bronze with only low levels of zinc or lead (but rarely absent). The Catterick belt plate is made of a quaternary alloy with moderate zinc content, but the tongue for the buckle is made of leaded brass (with 20% zinc). The tongue is very simple and may be a replacement (and so does not necessarily relate to 'official' military production). The overall low zinc content of late Roman military equipment reflects a general decline in zinc content of copper alloys by the 4th century.

7.6 Conclusions

The above discussion of the relationships between copper alloy compositions and typology illustrates that many items were made to more-or-less strict recipes. The use of recipes is not, however, universal as many objects were made from a range of different alloys. Many of those objects which lack distinctive typologies are also made from a range of different alloys. This suggests that in some cases production was on an ad hoc basis.

This chapter and previous work (e.g. Craddock 1975; Bayley 1992) have identified the use of distinct alloy types for the production of some items. The use of recipes may reflect a number of different metallurgical, economic or social constraints. Roman mirrors are made of speculum (tin over 20%) as this is the alloy which produces a silvery-white surface. Complex cast objects are usually made of leaded alloys as these produce a more fluid molten alloy at a lower temperature and so aid the casting process. Many distinctive alloys may reflect traditions where it is difficult to separate economic and social constraints. The use of lead for some recipes may reflect the abundance of this metal rather than the metallurgical need for it. A leaded alloy may then continue to be used because it is traditional to do so, even if lead becomes more scarce.

© Internet Archaeology
Last updated: Thu Apr 3 1997