Petrography

Petrographic method Each bracer was petrographically described using a x10 hand lens/low power binocular microscope and without knowledge of their geochemistry. Individual identifiers are denoted in parenthesis (see Table 5). Particular attention was paid to breaks/fractures in the artefact as they provided 'fresh', unpolished surfaces and the true colour of the rock. All lithological features, including mean grain size, presence of clasts, megacrysts, fossils, veining, bedding, laminae and foliation planes were noted and measured. The colour of the polished and any broken, natural surface was recorded and standardised using the Geological Society of America's rock-color chart. A lithological identification for each bracer was made based upon these macroscopic characteristics. In addition the same characteristic were used to group the bracers into lithologically coherent groups. Post-depositional carbonate, silica and limonite coatings on the bracers were noted as was the presence of secondary copper minerals formed by the alteration of copper/bronze.

Green bracers Microscopic examination demonstrated that the five arched bracers (i.e. of concavo-convex section) appear to form a petrographically coherent group comprising examples from Barnack, Cambridgeshire (3),Ferrybridge, Yorkshire (21), Driffield, Yorkshire (8), Tring, Hertfordshire (16) and Hemp Knoll, Wiltshire (2). Four of them are of arched Type C. These are greenish-grey (5G 4/1 5G 6/1), fine-grained, metavolcaniclastics characterised by a sub-conchoidal to sub-hackly fracture, laminae, small rock clasts and oxidised pyrite porphyroblasts. The bracer from Calne, Wiltshire shares some macroscopic features with these and may be an outlying member of the same group or has been made from a very similar lithology.Petrographiccomparison with the polished stone axes from the English Lake District defined as Group VI by the Implement Petrology Group (Keiller et al. 1941), and rocks collected from Langdale suggests that both they and the arched bracers share a common lithological origin.

The dark greenish colour of the bracers has some resemblance to dark jade but there is a significant difference in the polish between the two materials. Despite the presence of layering in the rocks they are essentially structurally isotropic (unlike the lithologies of the grey bracers that have a defining planar foliation) and hence, like jade, could be shaped in the round. Their ability to be shaped is an important physical property and one which appears to characterise the morphology of these bracers making them distinctive from other types. Other examples made from a similar darkish green rock are housed in the National Museum of Scotland and have been the subject of preliminary study by one of the authors (AW). Like several of the English examples, these also exhibit small red-brown oxidised pyrite porphyroblasts. Interestingly, no Group VI examples are known from Ireland where only three of the 96 bracers listed by Harbison are described as being 'dark green' but are of steatite and porphyritic igneous rock (1976: 31-33, Appendix C).

Grey bracers The majority of the flat bracers can also be seen to form a loosely defined group, although sharing different petrographical features. They are fine-grained, most are within the colour range bluish-grey (5B 7/1) to greenish-grey (5GY 6/1), show a pronounced planar foliation with flat surfaces parallel to this main foliation and contain green to brown-green mottling and/or the presence of up to 1-2mm diameter chlorite porphyroblasts and/or chlorite/green amphibole veining. These rocks were identified as being thermally metamorphosed metasediments and more especially spotted slates/hornfels. A core group of 11 items consists of bracers from Bishops Cannings, Wiltshire (17); Brandon, Suffolk (6); Cotswold Community, Gloucestershire (25); Gravelly Guy, Oxfordshire (24); Wellington, Herefordshire (19); Sewell, Bedfordshire (7); Aldbourne, Wiltshire (9); Sutton Veny, Wiltshire (15); Sturry, Kent (1); Sitttingbourne, Kent (5), and West Stafford, Dorset (18). The bracers from Roundway, Wiltshire (13) and Thomas Hardye School, Dorset (20) lack the presence of chlorite spots and amphibole but probably belong to the group.

Hornfelses are metamorphosed, fine-grained mudstones and siltstones and hence are macroscopically quite variable; therefore a grouping with a broad colour variation is consistent with this interpretation. Hornfels is not a common rock-type within England and Wales but exposures are known surrounding the exposed granite outcrops of Devon and Cornwall where Devonian-Lower Carboniferous sediments have been baked by the Cornubian batholith.

Black bracers Two fine-grained, black bracers, from the Amesbury Archer burial, Wiltshire (23) and from Stonehenge, Wiltshire (26) are identified as metamudstones/slates. Their grain-size difference suggests they are from different rocks. Dark metamudstones/slates are common in Britain and hence provenancing is difficult. However, the nearest suitable rocks to the find spots include regionally metamorphosed Lower Carboniferous rocks of Devon and a number of Palaeozoic rocks from South Wales. Other areas would include much of central Wales and the English Lake District. Despite an initial visual resemblance and quite high iron content (below) the Stonehenge bracer is unlikely to be a porcellanite from Northern Ireland, well known as the important Implement Petrology Group IX and a likely source for some of the black Irish bracers (Harbison 1976: 31-33, Appendix C).

Other bracers The red Amesbury Archer bracer (22) is probably made from a banded mudstone and may be less metamorphosed than other flat bracer lithologies. Red mudstones are abundant and widespread but the nearest outcrops to the find spot are in the Devonian and Permo-Triassic sequences of Cornubia and Lower Palaeozoic rocks of South Wales. It may have been an attempt to 'imitate' the red jaspers known to crop out in Galway and Mayo and which provide a conspicuous proportion of the Irish bracers (see Harbison 1976: Appendix C). Within the bracers from Ireland described in Harbinson's work, and including the two Irish bracers analysed during the present project at the British Museum (11 & 12), 28 per cent were of jasper, and a further 17 per cent were made from other red or pink rocks (such as micaceous or silicified siltstones), apparently in an attempt to replicate the red hue of the jasper examples. The red Amesbury Archer bracer shows no such Irish petrographic association. A solitary Scottish example is described as being of 'reddish hue' (Ashmore 1989: 70) and awaits petrographic description.

The Tytherington, Wiltshire (14), bracer is a cross-laminated metasediment with some colour and tactile affinities with the main grey group; however, the presence of many void spaces in the rock precludes it from membership. The Mildenhall, Cambridgeshire (4) example is an olive - grey metasediment perhaps with garnet. Petrographically, it is unlike any other bracer. The last five bracers discussed have little lithological coherence with any other bracers or themselves.

Geochemical analysis

PXRF method Selected lithic artefacts were analysed using a Spectrace TN9000 portable X-ray fluorescence (PXRF) spectrometer and the compositional data is listed in Table 5. The methodology largely followed procedures described in earlier publications (Potts et al. 1997a, b; Williams-Thorpe et al. 1999). Measurements were made with the PXRF analyser in 'lab stand' configuration and samples were excited sequentially using Cd-109, Fe-55 and Am-241 sources for count times of 100, 60 and 20 s respectively. X-ray spectra were quantified using the instrument manufacturer's algorithms, which are based on region-of-interest peak integration and a fundamental parameter matrix correction. As most samples generally had an elongated, often rectangular, wafer shape, two measurements were made, one on each side of the wafer where possible, and the average taken as representative of the bulk composition of the sample. Corrections were made for (a) surface topography (that is deviations from an ideal flat analysis surface) using procedures described by Potts et al. (1997a); (b) calibration bias, based on the average deviation from parity between recommended and analysed values of approximately 25 reference materials analysed as compressed powder pellets. In general, this correction amounted to ≤±10 per cent. As the samples were generally fine grained, an average of two measurements was sufficient to account for any mineralogical effects (Potts et al. 1997b). The comparability of analytical data derived during the entire period of the project was monitored by analysing in duplicate two control samples (powder pellets of a microgranite and a dolerite), at the start of each operating session. Consistency of the data was considered to be satisfactory.

All 26 bracers together with some samples of possible source materials were analysed and their compositions assessed using standard principal component (PC) analysis on the correlation matrix (e.g. Baxter 2003) using SPSS Version 11 (Table 6). The analysis was based on contents of elements K, Ca, Ti, Mn, Fe, Sr, Zr, Rb and Ba, which were those measured with greatest precision by PXRF and mostly at concentrations well above detection limits. Any 'missing data' (below detection) were 'imputed' by the minimum value for that element.

The first two principal components account for 69.5 per cent of the variation and the results are plotted in Figure 5. This reveals a group of 12 bracers with broadly similar compositional characteristics: low levels of K, Rb and Ba (some undetectable); moderately low Ti, Mn and Fe; and moderate levels of Ca, Sr and Zr. This group comprises the bracers from Cotswold Community (25), Roundway (13), Wellington Quarry (19), Bishops Cannings (17), Sewell (7), Brandon (6), Gravelly Guy (24), Sturry (1), Thomas Hardye School (20), West Stafford (18), Sittingbourne (5), and Sutton Veny (15). This group matches the grey bracers (spotted slates) minus Aldbourne (9) but plus the two 'probable' spotted slate pieces.

Most of the other artefact compositions appear to be less closely related, but some pairs are similar and there is a loose group of 6 bracers, which tends to be characterised by moderate levels of K, Ca, Ti, Fe, Sr, Zr and Ba, and moderately low Mn and Rb. This group comprises Ferrybridge (21), Tring (16), Hemp Knoll (2), Driffield (8), Calne (10) and Barnack (3) bracers. These match the six green bracers described above. Furthermore, in terms of PC analysis, the comparative samples of Langdale volcanic tuff (two analyses of source rock; four analyses of IPC Group VI axes: Ixer et al. 2004, with additional data pers. comm. Williams-Thorpe) match more closely with the group of six bracers than most others. In a direct comparison of their compositions with this group, the two rock samples have similar levels of most major elements, K, Ca, Ti and Mn, but slightly higher Fe and have similar levels of traces Sr, Zr, Rb and Ba. Comparative data for four Group VI Langdale axes (Ixer et al. 2004) were similarly matched except for relatively low levels of Ti which may have been a problem in those previous PXRF analyses of the stone axes.

Some of the other bracers have compositional features in common, but do not form coherent groups. Some bracers are distinguished by unusually high levels of particular elements: for example, Fe in the Stonehenge, Wiltshire bracer (26), Zr in the Mildenhall, Cambridgeshire (4), bracer, K and Rb in the Tytherington, Wiltshire bracer (14), or low levels: such as Fe in the Amesbury Archer, Wiltshire (23) black bracer.

A plot of Sr/Ca against Mn/Fe has also been constructed (Figure 6). This plot was chosen as Sr substitutes for Ca, and Mn for Fe in rock forming minerals, and changes in these ratios may signify different mineral hosts and potentially different modes of origin for the rocks. The group of 12 are again distinct, but sample 9 (Aldbourne) also plots in the same place. This group precisely matches the set of grey bracers including the possible spotted slates. The remainder of the bracers are more scattered, but the other group of six recognised by PC analysis, i.e. sample nos 2, 3, 8, 10, 16 and 21 are clustered. The IPC Group VI axes and one of the two Langdale 'source' samples also plot within that cluster, suggesting a significant degree of comparability.