Journal of Archaeological Science 82 (2017) 50e61 Contents lists available at ScienceDirect Journal of Archaeological Science journal homepage: http://www.elsevier.com/locate/jas From commodity to singularity: The production of crossbow brooches and the rise of the Late Roman military elite Vince Van Thienen a, *, Sylvia Lycke b, c a Ghent University, Department of Archaeology, Historical Archaeology Research Group, Sint-Pietersnieuwstraat 35, UFO, B-9000, Ghent, Belgium Ghent University, Department of Archaeology, Archaeometry Research Group, Sint-Pietersnieuwstraat 35, UFO, B-9000, Ghent, Belgium c Ghent University, Department of Analytical Chemistry, Raman Spectroscopy Research Group, Krijgslaan 281 S12, B-9000, Ghent, Belgium b a r t i c l e i n f o a b s t r a c t Article history: Received 16 April 2016 Received in revised form 14 April 2017 Accepted 25 April 2017 Available online 8 May 2017 The diachronic development of the crossbow brooch was examined for northern Gaul during the Late Roman period (3rd to 5th century) in order to extrapolate changes in production organisation from variation in the copper-alloy composition and shape of the artefacts. A combined method of surface analysis by handheld XRF and dimensional analyses by focussing on variation or similarity was applied and interpreted in a typological framework to confirm and enhance the traditional production model of the crossbow brooch. This led to new insights into the processes of regionality and state-control in Roman metal production in a provincial context. From a simple military commodity to an elite symbol of power and authority, these changes in production and consumption reflect sociocultural changes in the Late Roman West. © 2017 Elsevier Ltd. All rights reserved. Keywords: Copper-alloy Production hXRF Dimensionality Crossbow brooch Variation 1. Introduction The relationship between the production of objects and social organisation is vital for archaeologists working to understand the link between artefacts and past societies (e.g. Costin, 2001; Peacock, 1982; Underhill, 2003). Traditionally, the distribution of stylistic features of objects such as lithic tools, ceramic pots or metal dress accessories are used to investigate production organisation. Recently, more studies focus on compositional analysis to explore production processes. Too often, these two approaches are used separately, whereas the chaîne op
eratoire paradigm demonstrates that both aspects are expressions of social practice (Sellet, 1993; Tite, 1999). A third aspect, represented by the object biography and life history concepts (Dannehl, 2009; Kopytoff, 1986), stresses the importance of incorporating the sociocultural context in order to understand the changes in both artefact and society. Despite the success of these approaches, the production and compositional variation related to social change remains understudied for many Roman brooches, such as the crossbow brooch. * Corresponding author. E-mail addresses: Vince.vanthienen@ugent.be (V. Van Thienen), Sylvia.lycke@ UGent.be (S. Lycke). http://dx.doi.org/10.1016/j.jas.2017.04.005 0305-4403/© 2017 Elsevier Ltd. All rights reserved. This specific brooch type is closely associated with the Late Roman world (3rd e 5th century) and is well documented in archaeological and art-historical studies. Despite its rich research history, not much progress has been made recently in understanding the link between changes in the production organisation and the transformations of the Late Roman society. Hypotheses on state control and regionality have been formed based on the styledistribution patterns (Swift, 2000), but still remain unchallenged by other methods. This study explores the changes in the crossbow brooches' physical properties e composition and dimensionality e over time and changing social context to gain new insights in the production organisation and its reflection of the growing influence of Late Roman military elite. 1.1. The Roman crossbow brooch The crossbow brooch is a Roman bow brooch. Its main functional attributes are the bow and arms (or crossbar) containing the hinge mechanism and pin attachment, which is set in the foot. Its most distinctive features are the knobs (or terminals) at the end of the arms and at the front of the bow (Figs. 1 and 7). Production techniques include casting in moulds, the working of solid metals, or assembling various parts made by either casting or working. V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 51 Fig. 1. Selection of Low Countries crossbow brooches in the various types as presented in Table 1: 1. Type 0, GRMT 74.A.2, Tongeren; 2. Type 1, KAM 09, Nijmegen; 3. Type 2i, RAM burial 41, Oudenburg; 4. Type 2ii, RAM burial 59, Oudenburg; 5. Type 2iii, GRMT 74.A.35, Tongeren; 6. Type 3/4, RAM burial 72, Oudenburg; 7. Type 5ii, RAM burial 152, Oudenburg; 8. Type 6i, KAM 54, Nijmegen; 9. Type 6ii, RAM burial 111, Oudenburg. They are often decorated by motifs and patterns along the foot and bow and contain decorative features on the arms and base of the bow. These decorations were either present in the mould or made by cutting, stamping and soldering. Some brooches can also exhibit gilding, silvering and niello as added decorative features (Bayley and Butcher, 2004; Dandridge, 2000; Swift, 2000). Interest in the crossbow brooch started in the first half of the 20th century and a wide variety of scholars created many different typologies (for an overview see Swift, 2000; Van Thienen, 2016). The typology applied here is based on the model of Keller (1971), consisting of six successive types, with the adjustments made by € ttel (1988) and Swift (2000) (Table 1). In order to incorporate Pro the direct predecessor of the crossbow brooch, this study in€ ttel-Swift model’ (Fig. 1.1). troduces a ‘type 0’ into this ‘Keller-Pro Although, despite the longstanding research history, some issues concerning chronology and provenance have to be noted. The dating of the (sub)types is limited by the uncertain lifespan of an individual brooch between production and deposition or recycling. Also, a large number are stray finds or have poor contextual information. Additionally, there is a depositional bias towards burials favouring crossbow brooches from the mid-4th century onward. These burial gifts were taken out of circulation, and the recycling flow, and are more likely found by archaeologists than their predecessors. Initially, these brooches were considered to be military artefacts exported from a central state-run workshop (fabrica) in Pannonia or Illyricum (Riha, 1979). Although the general shape is very similar Table 1 € ttelCrossbow brooch typology applied here is based on the models of Keller-Pro Swift after Swift (2000). The introduction of ‘type 0’ signifies the direct 3rd century predecessor. The model of Hull and Hawkes (1987) on the ‘light’ and ‘developed’ crossbow brooches is added as comparison. Representative brooches are depicted in Fig. 1. Type Model 0 1 2 3/4 5 6 1 2i 2ii 2iii 3/4a 3/4b 3/4c 3/4d 5i 5ii 6i 6ii Date (AD) Hull 3rd century Light crossbow brooch 280e320 300e365 Developed crossbow brooch T190 T191A T191B T192 325e355 350e410 330e410 330e410 350e415 390e460 throughout the different types, there are distinct stylistic differences per region, postulating that a series of smaller workshops could have been active in different locations. The interregional comparison made by Swift (2000) demonstrated that the crossbow brooches' evolution is shaped by regional dynamics, chronological overlap and changing production processes. Swift created a 52 V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 narrative of a continuing mainstream trend that achieved a wide distribution alongside parallel smaller divergent groups with regional characters and restricted spread. Workshops locations in northwest Gaul and the Danubian provinces were proposed for the early 4th century (type 1 and 2), with limited distribution clustering along the frontier zone. For the largest part of the 4th century, the Pannonian production (type 3/4) is considered to have exported crossbow brooches across the entire Roman Empire, alongside localised pockets of production, until this central workshop ceased producing at the end of the 4th century. At this point, small workshops in the area west of the Rhine are suggested to have continued production into the 5th century (type 5 and 6), characterised by a distribution away from the frontiers and changes in symbolism resulting in a higher status. Evidence of small scale localised production was found in a workshop in northern Italy (Giumlia-Mair et al., 2007), but more confirmation for Swift's styledistribution model was acquired by the compositional study of British brooches (Section 3.1). 2. Materials and methods The physical properties of Roman brooches are mainly investigated by compositional analyses (e.g. Bayley and Butcher, 2004; Bayley, 1998; Dungworth, 1997; Giumlia-Mair, 2005; Riederer, 1993), although often brooches cannot be damaged for analyses because they are part of invaluable collections. Portable X-ray fluorescence (XRF) provides a mobile, quick and inexpensive solution to analyse large collections non-destructively, which motivated the choice for using a portable device here. However, for copper alloy brooches, this type of surface analysis offers many challenges related to corrosion, patinas, soil contamination, irregular shapes and limited detection of low Z elements (Arai, 2004; Elia, 2013; Mantler and Schreiner, 2000; Milazzo, 2004; Nicholas and Manti, 2014). These issues generated semi-quantitative methods focusing on trends and patterns, rather than exact
n-Torres, 2016; Martino
ncompositional data (Asinelli and Martino Torres et al., 2014; Orfanou and Rehren, 2015). Furthermore, the life history approach to recycled and re-alloyed metal objects, such as brooches, regards compositions as snapshots in the dynamic flow of metal circulation of that time (Pollard et al., 2015). Here, we consider brooches belonging to the same specific type in the crossbow brooch's typology (Table 1) as an assemblage of objects that represents a specific snapshot in its life history from which information on the social context is known (Van Thienen, 2016). Additionally, Dungworth (1997) demonstrated that relationships between typology and alloy composition exist, e.g. the Roman dragonesque and Colchester brooches. Moreover, a typology is constructed based on observed variations that are expressions of the different choices made by craftsmen and their workshops, which reflects access to raw materials, the manufacturing process and the intended consumer population (Caple, 2006). In other words, typology contains information that can help explain compositional variation caused by different manufacturing techniques or production centres, changes in consumer demands, as well as factor in aspects of chronology and regionality. Thus, typology provides a good tool to examine the wide range of potential explanations for the observed variation in analytical studies. A total of 187 brooches were collected from 12 different sites in northern Gaul (the Netherlands and Belgium) (Fig. 2, Table 2) and will be referred to as the Low Countries crossbow brooches, which encompasses all existing variations for its entire life history in this region between the 3rd and 5th century. The method applied in this paper aims to contribute to the crossbow brooch's production model, and by doing so, to offer a way to constructively investigate multidimensional copper alloy artefacts by (1) combining compositional with dimensional information in relation to typology, (2) inserting physical changes in a style-distribution model, and (3) interpreting these results in relation to the changing sociohistorical context of the artefact to gain new insights in the development of production organisation. 2.1. Handheld X-ray fluorescence The compositional analysis was performed by a commercial handheld X-Ray Fluorescence (hXRF) instrument (Olympus InnovX Delta). An Rh-target based X-ray source produces a polychromatic X-ray beam allowing to record elemental information from approximately a 5  5 mm2 sample area by means of a silicon-drift detector. The experiments were conducted in air using a shielded chamber, with 40 kV/79 mA tube voltage/current and 300 s measurement time. Each brooch was measured at multiple different locations to compensate the heterogeneity of the copper alloy and minimise the potential alloying or corrosion effect. On average, three spots were measured per brooch. Exceptionally, two, four or five spots were chosen. When needed, and if permission was granted, the corrosion was removed locally by fine manual cleaning using a scalpel to expose the surface metal. Nevertheless, the (scraped) surface of many of the samples reflect the metal patina rather than the bulk of the object, which will have increased the variability within and between the brooches. Evaluation of the individual XRF spectral data was performed using the AXIL (Analysis of X-rays by Iterative Least Squares) software package that allows a mathematical description of the fluorescence peaks and the spectral background (Vekemans et al., 1994). The net peak intensities were used for further data processing involving a qualitative assessment of the spectral data and a semi-quantitative approach to investigate trends and patterns in the major elements. 2.2. Metric variation The metric data was compiled by measuring the total and partial dimensions of each brooch (Fig. 7). To investigate the typologically related variation and to estimate the degree of standardisation, first, a multivariate data processing of principal component analysis (PCA) was carried out on the dimensional data. Followed by calculating the standard error on the mean (STdev) for each typological group by dividing the standard deviation (s) of the group by the square root of the number of brooches (n) in that group (STdev ¼ s/√n). Second, the coefficient of variation (CV) was calculated for each attribute per type, by dividing the standard deviation (s) by the group average (x) to explore the varying degree of variation and control throughout the life history of this artefact type. 3. Results 3.1. Composition and typology In the 557 measurements performed on a total of 187 brooches from the Low Countries, 14 elements were detected: Si, K, Ca, Fe, Ni, Cu, Zn, As, Ag, Sn, Sb, Au, Hg and Pb. Their varying presence and intensities are related to the major and minor elements of the copper ore (Cu, Ni, As, Sb, Ag), alloying metals (Zn, Sn, Pb), decorative precious metals (Ag, Au, Hg), corrosion and soil contaminants (Si, K, Ca, Fe), and the hXRF device (characterisation of the hXRF instrument showed a contamination of Ni (De Langhe, 2015)). The first qualitative observation of the spectra revealed that all brooches are copper alloys (Fig. 3), with the exception of one silver alloy brooch. Additionally, a small number of samples contained precious metal decoration in the form of silvering (Ag, n ¼ 1) and V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 53 Fig. 2. The geographical location of sampled sites in the region of northern Gaul corresponding with present day Belgium and the Netherlands. Table 2 List of sites with the number of sampled crossbow brooches. NL NL NL NL BE NL NL BE NL BE BE NL Site Samples Types Alphen aan de Rijn Beneden-Leeuw Beuningen Den Haag Gent Maasdriel Nijmegen Oudenburg Ravenstein Tienen Tongeren Wijk bij Duurstede total 1 1 2 1 1 3 106 50 1 1 14 6 187 0 3/4 1 0 2 3/4 0-1-2-3/4-5-6 0-1-2-3/4-5-6 5 3/4 0-1-2-3/4-5 0-1-2-3/4 gilding e either leaf (Au, n ¼ 14) or mercury gilding (Au-Hg, n ¼ 3). Hereafter, the term ‘copper alloy’ indicates the range of copper alloys without gilding or silvering. The precious metal decoration was identified by visual observation and major peaks of the corresponding elements in the XRF-spectra. A typological overview is presented in Fig. 4. Although comparing XRF-spectra is suitable to distinguish large differences in composition or detect the presence of metal decoration, it is less ideal to investigate the alloy variation indicated by Zn (zinc), Sn (tin) and Pb (lead). The relative proportion of these three major elements (Fig. 5) shows an overall continuous variation mainly influenced by Zn and Pb, less by Sn. As the result of the surface analysis by hXRF, the ternary diagrams can display an underrepresentation of Sn and an overestimation of Pb. In part, this relates to the use of intensities here, but can also reflect a Pbenrichment of the patina layer on some of the brooches. While some reservations towards the data is understandable, the interpretation presented here takes the known limitations of surface analysis into account and focuses on the main trends (Asinelli and
n-Torres, 2016; Orfanou and Rehren, 2015). Martino Some loose grouping can be discerned within each type, but without much consistency that could point to a large main production line or smaller deviating workshops (Section 1.1). Moreover, it is unclear if these brooches had a specific predetermined alloy. Roman brooches are likely formed by the mixing of alloys creating ternary and quaternary alloys, as well as very likely to have been part of the recycling process (Bayley and Butcher, 2004; Dungworth, 1997; Pollard et al., 2015). The mixing of alloys and recycling would explain the observed continuous variation. 54 V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 Fig. 3. XRF spectra from a typical ‘plain’ copper alloy brooch (left) and a gilded copper alloy brooch (right). Fig. 4. Typological overview of the number of ‘plain’, silvered and gilded copper alloy and silver alloy brooches. Furthermore, the crossing of modern alloy boundaries is explained by the different choices made by craftsmen in Antiquity from present day notions. Consequently, a principal component analysis was performed on the net peak intensities of Zn, Sn and Pb to further explore potential patterns in the copper alloys. Prior to this statistical test, the measurements influenced by corrosion or contamination (high levels of Fe, Ca, Si, K) were removed and the remaining measurements were averaged per brooch. Additionally, the silver alloy, silvered and gilded brooches were not taken into account, given their evident difference from the ‘plain’ copper alloys. This left a total of 119 brooches to test the major elements (Zn, Sn, Pb) with PCA. The minor elements (Ni, As, Sb, Ag) were not taken into account due to inconsistent detection and instrumental contamination in the case of Ni. The typological information was added to move beyond a black box approach of the compositional data (Speakman and Shackley, 2013). Chronology or provenance could also be used for this purpose, but were not favourable here due to dating issues, chronological overlap and the uneven distribution of samples (Table 2). Despite poor grouping in the resulting score plot (PC2 versus PC1), the typological division does demonstrate a distinction between either the influence of Zn or the combined influence of Pb and Sn. Overall, type 3/4 deviates most from the other types by the more pronounced influence of Zn that appears to be lacking in all other types (Fig. 6). The cluster of type 3/4 consists of two parts. One that illustrates the difference from the other types and one that behaves similarly from the influence of Pb and Sn. There is no distinction between the two parts of the group. Unfortunately, there are too few samples from types 5 and 6 to make any meaningful statements on their behalf. V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 55 Fig. 5. Typological overview of the correlation between the averaged X-Ray intensities per brooch of the major alloying elements Zn (zinc), Sn (tin) and Pb (lead). Fig. 6. Score plot and loading plot of PC 1 and 2 from the PCA on the X-Ray intensities of Zn, Sn and Pb of a total of 119 brooches. The score plot is separated per type. This dual trend aligns with the distinction into two main compositional groups based on the evidence of Britain, where the group characterised as ‘leaded bronze’ indicated an origin in Britain and the ‘brass/gunmetal group’ pointed to a Continental origin, possibly the Danubian region (Pannonia) (Bayley, 1992; Swift, 2000). An exception to this were type 6 brooches belonging to the leaded bronze group, but containing stylistic features from the continental West. Arguably, the distinction visible in the scatterplot (Fig. 6) coincides with these observations: type 3/4 (partially) matches the ‘brass/gunmetal group’ and the ‘leaded bronze group’ corresponds with all types. Moreover, Swift noted the link between the brass/gunmetal group and the stylistic features from the Danubian area. This could suggest that this group represents products from one or more official fabricae, as is indicated by the 56 V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 use of brass in state facilities producing military equipment and the notion of Roman brass as a state-controlled metal (Bayley, 1990; Dungworth, 1997). Consequently, the implication arises that the leaded bronze group is not necessarily associated with British products, but with products from non-state e i.e. local or regional e workshops. 3.2. Dimensionality and variation to support controlled production In order to further explore the production changes, the brooch dimensions have been investigated in search of potential requirements, limitations or freedoms regarding its shape and size. The total dimensions include the length, width and height of the brooch and the separate features consist of the lengths of the arms, bow and foot as well as the maximal knob diameter (Fig. 7). Due to fragmentation, of the original set of 187 brooches, 126 complete brooches were available for PCA and 152 brooches could be used to calculate CV. PCA revealed one main group, comprising of types 2 through 6, and one or two distinct group for types 0 and 1 (Fig. 8). These separate sets of dimensions indicate a change in requirements in the brooch shape. Despite poor clustering, a shift is clearly visible. To explore the validity of this shift, the standard error on the mean was calculated (Fig. 9). Mainly the values of PC1 strengthen both the division and shift between types, although the connection between types 5 and 6 is remarkable, as well as their overlap in error bars of PC1 with type 3/4. The significance of this will be discussed further (Section 3.3). An additional approach focused on examining the changes in variation and control in the production by means of comparing the coefficient of variation (CV) for the dimensional data. Since the adoption of psychological concepts on the limitations of human perceptual capacities in cultural transmission studies (Eerkens, 2000), CV has been increasingly applied in artefact studies to investigate variation, control and change over time, both for dimensional and stylistic properties (de Voogt et al., 2013; Eerkens and Bettinger, 2001; Eerkens and Lipo, 2005; Lassen and Williams,
n-Torres et al., 2014; Underhill, 2003). In this study, 2015; Martino CV is used to investigate two objectives: the first is to explore the degree of variation or control between the types, revealing information on differences per type. The second is to identify the degree of variation/control on the different features, to assess the impact of the stylistic changes through time. The dimensional variation is interpreted by using the ‘Weber fraction’ (5% for production) and the ‘random uniform line’ (RUL) (57.7%) as indicators for the degree of variation/control (Eerkens and Bettinger, 2001). It has to be noted that the use of moulds influences the interpretation of the CV assessment. The use of moulds is a form of automated production that removes human error in the same batch of objects. Here, only type 0 brooches from Oudenburg qualify for reflecting the use of the same mould. All other brooches are unique individuals in both shape and decoration, decreasing the chance of shared mould use. Additionally, the manual adjustments and decorations to the different brooch-parts increases the chance for human error in the copying and imitating of styles, warranting the use of CV to investigate the variations in restrictions or freedoms between types and workshops. In general, the average CV values per type and per feature (Fig. 10, Table 3) are both well above the Weber fraction and do not approach the RUL, indicating an overall strong degree of control. Furthermore, the varying degree of variation/control per type demonstrates a gradual increase in CV for the first three types, whereas the last three types have noticeable lower values. The average values do not fit the general notion of standardisation (Eerkens and Bettinger, 2001), however, certain factors that increase CV values have to be taken into account. First, the multidimensionality of the crossbow brooch makes it a complex object with more chance for error. Second, this brooch has simultaneously a functional and a stylistic nature, which is affected by both design tolerance and constraint (Caple, 2006; Horsfall, 1987). Additionally, multiple workshops actively produced these brooches at the same time, resulting in several craftsmen with different skillsets working together or separately on these brooches. So, in all, the average CV values between 10 and 15% are chosen here to be understood as a high means of control that could be construed as standardisation. To assess the impact of the changes in shape and style on the manufacture, the variations for each dimension are compared. In average, the total dimensions have lower values than the separate features, except for the bow length (Fig. 10). Despite the higher variation present in the different features, the total dimensions show a tendency towards standardisation (10e15%), indicating a manner of compensation: i.e. if one feature is disproportional, mainly for stylistic reasons, other features are adjusted to fit the required overall shape. This supports the impression that the production of crossbow brooches was subjected to strong regulations, likely related to its symbolic meaning and enforced by the military, state or social convention. Compared to the 53%e74% CV average for the bow width from Bronze Age brooches from the Eerkens and Bettinger study (2001), the crossbow brooch values are extremely low. This high level of conformity is related to its need to be recognisable as a symbol of Roman authority (Section 4). Variability is also a good factor to distinguish between local/ regional products and controlled state products. Arguably, a higher variation corresponds with local/regional production and lower variation with fabrica products. This assumption seems promising, supported by all types, except for types 5 and 6. If regionality is indeed expressed by variation, we would expect to see higher CV values in these two types. The variation remains low, equal to the values of the state-controlled type 3/4. Clearly, simply connecting variation with regionality as opposed to central production is thus inaccurate. However, it can be argued that although the workshops producing types 5 and 6 in the area west of the Rhine had a regional distribution, they were in fact set up or transformed into official state workshops after the main workshop(s) in Pannonia had ceased production. 3.3. Contributions to the style-distribution production model Fig. 7. Crossbow brooch (type 3/4) from the 4th century burial site at Oudenburg (Belgium) with indications of the terminology used in the metric analysis. The evidence gathered from the Low Countries crossbow brooches largely confirms the existing production model and adds new information to this dynamic narrative. The type 0 brooches were added to Swift's style-distribution model that mainly focussed on the 4th century types. The compositional results relate V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 57 Fig. 8. Score plot and loading plot of PC 1 and 2 from the PCA performed on the dimensional data from a total of 126 brooches. A distinction was made based on type. Fig. 9. Standard error on the mean for PC1 and 2 on the dimensional data from the typological groups (STdev ¼ s/√n). these 3rd century brooches to the Pb-Sn group (Fig. 6), some very similar (Fig. 5), with a high variation in dimensionality (Fig. 10, Table 3). This makes them very distinct from the 4th century types, arguing for local productions with limited distribution. The findings from the types 1 and 2 brooches largely correspond with the regional character from Swift's model, expressed in the wide alloy variety (Fig. 5). Although, for type 2 brooches it demonstrates a higher Zn contribution (Fig. 6), suggesting a larger number originating from the Danubian area or produced in a state controlled environment. The degree of variation also points to regionality, of which the type 1 dimensions divert most from the main requirements (Fig. 9), although type 2 has the largest overall variation (Fig. 10). This corresponds with Swift's observations that type 2 is highly differentiated with some regional subtypes and others already very similar to the highly controlled type 3/4 that dominated the bulk of the 4th century production (Fig. 1). The contribution of Zn from the PCA for type 3/4 increases immensely for these brooches from the Low Countries, although, there is also a significant influence of Pb (Figs. 5 and 6). The degree of variation declines strongly towards a standardised appearance and the dimensional requirements have become more restricted (Figs. 9 and 10). All of which points to a state-run production from one or more central workshops, submitted to a high degree of control. Further investigation can contribute to the general debate of a state monopoly of Roman brass as posed by Dungworth (1997). In any case, the dimensional restrictions support the notion of a controlled production. However, this does not necessarily mean that there is only one state-controlled workshop that produced and distributed type 3/4 brooches. It is possible that zinc ore or brasses were supplied to multiple fabricae that were allowed to produce official military or state equipment. In contrast, type 5 and 6 brooches again demonstrate less 58 V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 4. Discussion Fig. 10. Average CV per type (top) and per feature (bottom), with indication of the Weber fraction (5%) and the random uniform line (RUL, 57.7%) as constants. influence from Zn and more from Pb-Sn (Fig. 6), arguing a regional character. When the compositional values of both types are considered together in the ternary plots (Fig. 5), distinct groups for these two types can be suggested, potentially linked to multiple workshops. More brooches will have to be analysed to confirm this. The degree of control and dimensional requirements remain similar as to type 3/4 (Figs. 9 and 10). This could indicate a change in resources for the fabricae or in the dimensional expression of regionality. The transformation in symbolism of the crossbow brooches for the late 4th e early 5th century must be considered, implying that compositional results should be interpreted in a different framework from the late 4th century onwards. We believe that a regional character for these brooches is correct, but that the concept of a central state-run production can no longer be considered incompatible with regionality. An important final step is to link these interpretations on production to the attested changes in the sociocultural contexts in the crossbow brooch's life history (Swift, 2000; Van Thienen, 2016). During the second half of the 3rd century, this artefact arose as a new type of bow brooch originating as a soldier's uniform attribute in the Danubian area. Not much is known about this first phase (type 0 to 1), other than that archaeological evidence places it predominantly on military frontier sites. The lack of iconographical and historical evidence suggests that its owners belonged to a lower social military class (Van Thienen, 2016). The compositional and dimensional evidence gathered here indicates a localised production developing into a regional distribution with some freedoms and significantly different requirements than the later types. These earliest samples demonstrate the characteristics of a high output craft production: everyday use, low manufacture cost and skill, minimal decoration and randomly discarded due to loss, damage and wear (Caple, 2006). Along the transition from the 3rd to the 4th century, a first change in the symbolic value of the crossbow brooch occurred. The objects show an increased variation in shape and style and in number of retrieved artefacts (types 1 and 2). Additionally, this brooch type starts to be depicted in iconographic sources, displaying connections with military and Pannonian attributes (Fig. 11). Furthermore, luxurious examples bearing inscriptions praising emperors are found from this period (Deppert-Lippitz, 2000). The art-historical evidence uncovers a dual message connected to the crossbow brooch at this point in its life: on the one hand, it is associated with anonymous members of the military (Fig. 11) and on the other hand it is connected to public figures (Fig. 12). The distinction is not simply a division between military and civilian, because many high-end civilians had military backgrounds or official ties to the military establishment. Furthermore, in the 4th century the official civilian dress was highly influenced by the military. This duality is also visible in the nature of its production. The compositional data for northern Gaul, supported by the data from Britain (Bayley, 1992; Swift, 2000), shows almost an equal number of brooches associated with regionality (mainly influenced by Pb-Sn) as with official production (mainly influenced by Zn) (Figs. 5 and 6). The increased amount of variation suggests limited control and no fixed requirements (Fig. 9). This evidence argues for a largely continued regional production with a smaller amount of imports from state-controlled fabricae. In a monetised and commercialised society, it is normal behaviour for a commodity, such as a brooch, to show increased variation as the result of its growing popularity (Kopytoff, 1986). Around the first quarter of the 4th century, however, this Table 3 Values for the coefficient of variation (CV) and average dimension in mm (x) for each feature per type (CV ¼ s/x). Type 0 x (mm) CV (%) Type 1 x (mm) CV (%) Type 2 x (mm) CV (%) Type 3/4 x (mm) CV (%) Type 5 x (mm) CV (%) Type 6 x (mm) CV (%) Av. CV per feature Length Width Height Arm length Bow length Foot length Knob diameter Av. CV per type 63.00 10.66 68.33 8.67 74.43 14.78 78.43 11.91 80.44 11.43 82.50 13.48 11.82 36.05 18.97 46.27 16.41 50.03 17.42 50.15 13.21 52.55 13.92 56.00 7.89 14.64 25.90 14.55 26.73 12.81 26.63 16.47 26.95 9.22 29.10 9.65 29.20 11.96 12.44 15.55 16.36 14.54 15.44 13.00 18.84 12.87 18.19 15.30 21.36 16.20 15.98 17.70 35.61 12.11 35.41 10.31 34.15 16.99 32.92 10.60 34.60 7.73 34.20 8.11 10.98 22.75 14.58 27.71 18.11 35.67 17.45 40.16 18.16 41.33 16.85 42.50 22.93 18.01 4.84 24.11 7.88 18.75 11.32 21.07 12.13 12.36 14.10 15.75 13.10 7.82 16.64 15.91 14.36 17.57 13.38 13.81 12.60 V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 Fig. 11. Sculpture from a funeral monument at Tilva ros (Bor, Serbia), dated ca. 280-320 AD, depicting two anonymous men with crossbow brooches on a military cloak and a Pannonian hat (after Petkovi
c, 2010). Fig. 12. Detail of the lid of the Projecta Casket, picturing Secundus as a Roman official (courtesy of the British Museum, number 1866, 1229.1 AN493408001). development was intervened to preserve the military authority associated with the crossbow brooch. The crossbow brooch became a strict military object with a high degree of uniformity (Figs. 9 and 10), produced on a large scale in one or more state-run fabricae. This process of singularisation is generally associated with the type 3/4, although the start of a more controlled product already existed in type 2 (Swift's subtype 2ii, Fig. 1.4), fitting with the duality in the previous developments. Despite the increased number of brooches, there is no corresponding increase in iconographic evidence. The depictions are few and still associated with anonymous military identities, supporting the reclaiming of this brooch for the military establishment. Resulting from the dual social significance, a second process of singularisation started in the second half of the 4th century (types 5 and 6). For northern Gaul, this stage is characterised by an immense drop in the number of brooches. In the iconographic material, we 59 see the gradual disappearance of the anonymous figures with brooches (Fig. 11), replaced by personal (often portable) objects with recognisable officials (Fig. 12), such as consuls and members of the senatorial class (Van Thienen, 2016). All of this is in line with the bespoke nature of the crossbow brooch at this point in its life history: unique and highly valued hand-crafted objects, commissioned by a select wealthy group of patrons, representing the taste of the elite, demonstrating wealth and social position (Caple, 2006). Again, it is too simple to classify this development as a civilian elite take-over. Rather, this evolution signifies the rise of members of the military elite to the highest positions in the administrative and political circles of the Late Roman Empire. The excessive decorative nature attests for this, also expressed in the increased use of gilding (Fig. 4). The official nature of this elite class is expressed by taking an object of authority that had become synonymous with the military establishment and redefining it as a symbol for state authority. This process reflects the vast military influence in the Late Roman society and state organisation. This second singularisation is simultaneously a response on the brooch's restriction to the military ranks as well as a claim to the elite's military roots, which is expressed in the standardised dimensional requirements (Fig. 10) in the local/regional products. This duality indicates the capacity and skill to make high-end official state products in local/regional workshops by the end of the 4th century. A final note has to be made on the presence of brass or Zn-rich alloys in crossbow brooches from the 4th and possibly 5th century. Generally, it is considered that there is a ‘zinc decline’ towards the Late Roman period. Studies by Pollard et al. (2015), Bayley and Butcher (2004) and Dungworth (1996, 1997) have shown that the ‘zinc decline’ is not so much a decline in Zn as it is in ‘pure’ brass objects. Gunmetal objects become more frequent in the Late Roman period, indicating that the mixing of alloys became a more common practice than creating new objects from ‘pure’ Zn ore. It is considered that this reflects a change in alloying policy in which recycling plays a more central role (Dungworth, 1996; Pollard et al., 2015). The compositional results gathered from the crossbow brooches is, in our opinion, in line with this view. The continuous variation in the composition of type 3/4 corresponds more with the mixing of alloys than with the creation of new objects from ‘pure’ raw materials. The assumed control of the state or military over Zn might have been served as a ‘secondary raw material’. In general, the military equipment was owned by the state, and before the widespread inhumation rite, it can be assumed that this equipment returned to the fabricae or military supply depots to be redistributed. It can be argued that it is possible that the controlled workshops were the only ones in the 4th century that had easy access to brass objects to recycle into new objects. This would also provide an explanation for the few brooches that appear to be ‘pure’ brass. These can be the result of the mixing of high Zn-objects that come across as pure brass Late Roman brooches. The amount of type 5e6 brooches are not sufficient enough to support or deny this hypothesis, nor are the results produced by the hXRF comparable with other studies to that extent. Moreover, the situation becomes much more complex in which the military elites are controlling large parts of the Western Empire and the military is increasingly made up of Germanic auxiliaries and ‘barbarian mercenaries’ (Esmonde Cleary, 2013; Halsall, 2007). Furthermore, there is an increasing decentralisation and growing autonomy in the 5th century of the Western provinces, such as Gaul and Britannia. All of this indicates that the explanation for the disappearance of brass objects, if they can indeed be seen as a monopoly of the state, could be linked to changes in the political and administrative organization in the late 4th and 5th century rather than the result of too much recycling. The idea that the Late Roman state is significantly different from the early Roman Empire is a notion that should be 60 V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 considered more thoroughly in the discussion on changes in copper alloys in Roman history. To summarise, the results delivered by the Low Countries crossbow brooches argue that in the 3rd century a military item linked to military identity was produced in multiple local workshops, made by (low) skilled craftsmen responding on a regional need in the military ranks. Over the course of half a century, this type of brooch and its associated social class had gained enough influence to have become a popular commodity around the turn of the 4th century, creating an ambiguous symbol of social identities. The regional workshops began to produce on a larger scale for both military needs as well as social demand. Together with the major military and administrative reforms under the Tetrarchy, the state reacted on this uncontrolled growth by turning the crossbow brooch into a standardised military object with precise requirements (shape) and certain freedoms (decoration). They were produced and exported on a large scale from state-run fabrica(e), to supply military needs and to guarantee uniformity. In the second half of the 4th century, a second singularisation had been initiated by the military elite class with official functions in the administrative and political circles. The crossbow brooch had become a powerful symbol, making it the perfect choice to turn into an elite object and redefine the brooch as an embodiment of the Roman state. These elite objects were commissioned in one of the multiple small official workshops and custom made for their wealthy and influential owners. For northern Gaul, the life history stops here at the beginning of the 5th century with the withdrawal of the Roman presence from the region. The developments of the crossbow brooch, however, continue until the 6th century in the remainders of the Roman Empire. 5. Conclusion To conclude, we can state that despite certain limitations in nondestructive surface analyses, good results and new insights can be obtained to investigate the relation between composition and production for an extended time period in a complex setting. Additionally, the use of metric variation through shape and dimensional change to explore differences in production organisation displayed much promise. Moreover, by investigating the complete (regional) life history of a symbolic artefact such as the crossbow brooch, this study argues a reconstruction of the relation between regional and state production in the changing Roman society. Specifically, linking the brooch typology with varying degrees of variation and control, expressed in its composition and dimensionality, provided information on the changing nature of a specific craft production. This, in its turn, contributed to the traditional style-distribution production model and delivered aspects for further research in the production of the crossbow brooch. By comparing these new results with attested sociohistorical transformations, new insights were acquired in the complex dynamics of the singularisation of commodities and the interacting processes of regionality and state involvement in the metal production in the Late Roman Empire. Acknowledgements This study is part of the research project “Decline and fall? Social and cultural dynamics in the Low Countries in the Late Roman Empire (AD 270-450)”, supported by the Netherlands Organisation for Scientific Research (NWO) and the Research Foundation e Flanders (FWO) (GA04612N). We wish to thank the following museums and institutions for allowing us to perform analyses on their collections: Agentschap Onroerend Erfgoed, Romeins Archeologisch Museum Oudenburg, Gallo-Romeins Museum Tongeren, RAAKVLAK, Rijksmuseum van Oudheden Leiden, Archeologische Dienst Nijmegen and museum G.M. KAM Nijmegen. We would also like to thank Bart Vekemans and Jolien Van Pevenage for their assistance in processing the results, as well as Wim De Clerq, Peter Vandenabeele, Stijn Heeren and Kaatje Delanghe for proofreading the article. A final word of gratitude goes out to the editors and the reviewers for their useful comments. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.jas.2017.04.005. References Arai, T., 2004. Analytical precision and accuracy in X-ray fluorescence analysis. Rigaku J. 21, 26e38.
n-Torres, M., 2016. Copper-alloy use in a Tyrrhenian medieval Asinelli, M.G., Martino town: the case of Leopoli-Cencelle (Italy). J. Archaeol. Sci. Rep. 7, 597e608. Bayley, J., 1990. The production of brass in antiquity with particular reference to Roman Britain. In: Craddock, P.T. (Ed.), 2000 years of Zinc and Brass. British Museum Publications Ltd., London, pp. 7e28. Bayley, J., 1992. Non-ferrous Metalworking in England: Late Iron Age to Early Medieval. University of London, London.
Bayley, J., 1998. The composition of roman brooches. In: Nicolini, G., Diedonne
taux antiques: travail et restauration, Montagnac, Glad, N. (Eds.), Les me pp. 113e119. Bayley, J., Butcher, S., 2004. Roman Brooches in Britain: a Technological and Typological Study Based on the Richborough Collection. The Society of Antiquaries, London. Caple, C., 2006. Objects: Reluctant Witnesses to the Past, Routledge, London/New York. Costin, C.L., 2001. Craft Production Systems, Archaeology at the Millennium. Springer, pp. 273e327. Dandridge, P., 2000. Idiomatic and mainstream: the technical vocabulary of a late roman crossbow fibula. Metrop. Mus. J. 35, 71e86. Dannehl, K., 2009. Object biographies. From production to consumption. In: Harvey, K. (Ed.), History and Material Culture: a Student's Guide to Approaching Alternative Sources, pp. 123e138. Routledge, London. De Langhe, K., 2015. Beyond the Beam: Evaluation and Application of Handheld Xray Fluorescence in Archaeology. Ghent University, Ghent, Belgium (PhD dissertation). http://hdl.handle.net/1854/LU-6715561. de Voogt, A., Dunn-Vaturi, A.-E., Eerkens, J.W., 2013. Cultural transmission in the ancient Near East: twenty squares and fifty-eight holes. J. Archaeol. Sci. 40, 1715e1730. Deppert-Lippitz, B., 2000. A late antique crossbow fibula in the metropolitan museum of art. Metrop. Mus. J. 35, 39e70. Dungworth, D., 1996. Caley's' zinc decline' reconsidered. Numis. Chron. 1966 (156), 228e234. Dungworth, D., 1997. Roman copper alloys: analysis of artefacts from northern Britain. J. Archaeol. Sci. 24, 901e910. Eerkens, J.W., 2000. Practice makes within 5% of perfect: visual perception, motor skills, and memory in artifact variation1. Curr. Anthropol. 41, 663e668. Eerkens, J.W., Bettinger, R.L., 2001. Techniques for assessing standardization in artifact assemblages: can we scale material variability? Am. Antiq. 493e504. Eerkens, J.W., Lipo, C.P., 2005. Cultural transmission, copying errors, and the generation of variation in material culture and the archaeological record. J. Anthropol. Archaeol. 24, 316e334. Elia, A., 2013. Application of Electrochemical Methods for the Study and Protection of Heritage Copper Alloys. Ghent University, Faculty of Sciences, Ghent, Belgium. Esmonde Cleary, S., 2013. The Roman West, AD 200e500: an Archaeological Study. Cambridge University Press. Giumlia-Mair, A., 2005. Copper and copper alloys in the Southeastern Alps: an overview. Archaeometry 47, 275e292. Giumlia-Mair, A., De Cecco, C., Vitri, S., 2007. Fibulae production at socchieve (Udine, Italy) in late antiquity. In: Proceeedings of the 2nd International Conference “Archaeometallurgy in Europe 2007”, Aquileia. Halsall, G., 2007. Barbarian Migrations and the Roman West, 376e568. Cambridge University Press, Cambridge. Horsfall, G.A., 1987. Design theory and grinding stones. In: Hayden, B. (Ed.), Lithic Studies Among the Contemporary Highland Maya. University of Arizona Press, pp. 332e377. Hull, M.R., Hawkes, C.F.C., 1987. Corpus of Ancient Brooches in Britain: Pre-Roman Bow Brooches, Oxford. €tro €mische Grabfunde in Südbayern, C.H.Beck'sche VerlagsKeller, E., 1971. Die Spa buchhandlung, München. Kopytoff, I., 1986. The cultural biography of things: commodization as process. In: Appadurai, A. (Ed.), The Social Life of Things. Commodities in Cultural Perspective. Cambridge University Press, Cambridge, pp. 64e91. V. Van Thienen, S. Lycke / Journal of Archaeological Science 82 (2017) 50e61 Lassen, R., Williams, T., 2015. Variation in flintknapping skill among Folsom-era projectile point types: a quantitative approach. J. Archaeol. Sci. Rep. 4, 164e173. Mantler, M., Schreiner, M., 2000. X-ray fluorescence spectrometry in art and archaeology. X-Ray Spectrom. 29, 3e17.
n-Torres, M., Li, X.J., Bevan, A., Xia, Y., Zhao, K., Rehren, T., 2014. Forty Martino thousand arms for a single emperor: from chemical data to the labor organization behind the bronze arrows of the Terracotta Army. J. Archaeol. Method Theory 1e29. Milazzo, M., 2004. Radiation applications in art and archaeometry X-ray fluorescence applications to archaeometry. Nucl. Instrum. Methods Phys. Res. B 213, 683e692. Nicholas, M., Manti, P., 2014. Testing the Applicability of Handheld Portable XRF to the Characterisation of Archaeological Copper Alloys. Orfanou, V., Rehren, T., 2015. A (not so) dangerous method: pXRF vs. EPMA-WDS analyses of copper-based artefacts. Archaeol. Anthropol. Sci. 7, 387e397. Peacock, D.P.S., 1982. Pottery in the Roman World: an Ethnoarchaeological Approach. Petkovi
c, S., 2010. Crossbow fibulae from gamzigrad (romuliana). Starinar 111e136. Pollard, A.M., Bray, P., Gosden, C., Wilson, A., Hamerow, H., 2015. Characterising copper-based metals in Britain in the first millennium AD: a preliminary quantification of metal flow and recycling. Antiquity 89, 697e713. €ttel, P., 1988. Zur chronologie der Zwiebelknopffibeln. Jahrb. Des. Ro €mischPro Germanischen Zentralmuseums Mainz 35, 347e372. 61 €mischer Fibeln aus Kempten, Die ro €mischen Riederer, J., 1993. Metallanalysen ro Fibeln von Kempten/Cambodunum, Cambodunumforschungen, V. Mater. Bayer. Vorgesch. A 63, 45e52. € mischen Fibeln aus Augst und Kaiseraugst, Amt für Museen und Riha, E., 1979. Die ro €ologie des Kantons Basel-Landschaft, Augst. Archa
ratoire; the concept and its applications. Lithic Technol. Sellet, F., 1993. Chaîne ope 18, 106e112. Speakman, R.J., Shackley, M.S., 2013. Silo science and portable XRF in archaeology: a response to Frahm. J. Archaeol. Sci. 40, 1435e1443. Swift, E., 2000. Regionality in dress accessories in the late Roman West, editions monique mergoil. Montagnac 13e18. Tite, M.S., 1999. Pottery production, distribution, and consumptiondthe contribution of the physical sciences. J. Archaeol. Method Theory 6, 181e233. Underhill, A.P., 2003. Investigating variation in organization of ceramic production: an ethnoarchaeological study in Guizhou, China. J. Archaeol. Method Theory 10, 203e275. Van Thienen, V., 2016. A symbol of Late Roman authority revisited: a sociohistorical understanding of the crossbow brooch. In: Roymans, N., Heeren, S., De Clercq, W. (Eds.), Social Dynamics in the Northwest Frontiers of the Late Roman Empire. Beyond Transformation or Decline, Amsterdam. Vekemans, B., Janssens, K., Vincze, L., Adams, F., Van Espen, P., 1994. Analysis of Xray spectra by iterative least squares (AXIL): new developments. X-Ray Spectrom. 23, 278e285.