PRESELI DOLERITE BLUESTONES: AXE-HEADS, STONEHENGE MONOLITHS, AND OUTCROP SOURCES
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
OLWEN WILLIAMS-THORPE, M.C. JONES, P.J. POTTS AND P.C. WEBB PRESELI DOLERITE BLUESTONES: AXE-HEADS, STONEHENGE MONOLITHS, AND OUTCROP SOURCES Summary. Chemical compositions and magnetic susceptibility data were compared for 12 dolerite bluestone implements including axes, axe-hammers and battle-axes, 11 Stonehenge monoliths (chemical data only), and potential source outcrops in Preseli, South Wales. Most of the studied artefacts are of spotted dolerite, a small number being unspotted dolerite. Bivariate graphs, discriminant analysis and t-tests were used singly and in combination to show, respectively, that the implements found at sites in England are mainly similar to Stonehenge monoliths, while the implements found in Wales have a variety of compositions and are much less similar to Stonehenge monoliths. The dichotomy between English and Welsh dolerite bluestone implements could be explained by exploitation of different Preseli outcrops or erratic assemblages derived from them. A small number of spotted dolerite implements have previously been shown to have chemical compositions atypical of and marginal to Preseli, suggesting the possibility of a source of spotted dolerite outside Preseli. Previously published analytical data in combination with the new implement/outcrop comparisons presented in this paper support derivation of the majority of analysed Stonehenge monoliths at one particular outcrop within the group of four identified by Thorpe et al. 15 years ago. Analysis of all the extant bluestone monoliths at Stonehenge (now possible using non-destructive methods) would allow progress in identifying monolith outcrop sources, and in understanding the links with the bluestone axe trade. introduction and background Archaeological ‘bluestone’ studies have traditionally focused on the stones used for the Stonehenge monoliths, but more recently attention has turned to the polished axe-heads and other implements manufactured from bluestone (Williams-Thorpe et al. 1999; Jones and Williams-Thorpe 2001; Williams-Thorpe et al. 2004). In this paper, we bring together information on the sources of dolerite bluestone implements, and of the Stonehenge monoliths, in order to comment on the relationship between these two groups of artefacts. In particular, we examine the question of whether the monoliths and the implements originated at the same sources within Preseli, and what this tells us about the links between the procurement of bluestone for Stonehenge, and the bluestone employed in England and Wales for implements. OXFORD JOURNAL OF ARCHAEOLOGY 25(1) 29–46 2006 © 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd., 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street Malden, MA 02148, USA. 29
PRESELI DOLERITE BLUESTONES Recent non-destructive geochemical and magnetic studies of spotted dolerite implements (including axes, axe-hammers, battle-axes and maces) showed that fewer than a dozen can be reliably assigned to the Preseli source area (Williams-Thorpe et al. 2004). That work resulted in a revised, and much reduced, distribution of these implements (which are also referred to as ‘Group XIII’, following the terminology of the British Implement Petrology Committee (IPC); Clough and Cummins (1988)). Within the dataset for the implements confirmed as Preseli by Williams-Thorpe et al. (op. cit.), there appeared to be small variations in chemical and magnetic characteristics, suggesting that not all originated at the same outcrop source(s) within Preseli. This observation prompted us to investigate whether these small differences were statistically significant, and to assess the feasibility of provenance determinations at outcrop level. In particular, we wished to investigate the relationship between outcrop origins and archaeological aspects such as find- location and morphology of the implements. Eleven (out of 27 remaining) Stonehenge monoliths of dolerite (nine spotted, two unspotted) were analysed by Thorpe et al. (1991). They were all confirmed as Preseli in origin, and most were linked to a fairly general source area of eastern Preseli (four outcrops, including Carnmenyn which is the outcrop traditionally regarded as the Stonehenge bluestone source). table 1 Chemical compositions and magnetic susceptibility of Preseli dolerite implements, and chemical compositions of selected Stonehenge monoliths Reference SH33 Ha41 Wi118 Wi109 Wi108 SH44* Wi302* Find location Stonehenge Bankes Stockton Stonehenge Stonehenge Stonehenge Wilsford Heath, Earthworks south Bournemouth barrow 54 Artefact type monolith axe axe miscellaneous miscellaneous monolith battle-axe artefact artefact fragment fragment K2O wt % 0.18 0.54 0.42 0.31 0.36 0.42 0.72 CaO 11.30 9.96 11.11 10.49 12.83 9.06 13.29 TiO2 1.15 0.96 0.74 1.01 0.71 1.30 1.29 MnO 0.16 0.10 0.11 0.15 0.13 0.19 0.12 Fe2O3 9.27 9.28 9.05 9.52 9.94 11.80 8.88 Ba ppm 120 100 173 122 144 225 207 Nb 6 4 1 3 2 7 4 Rb 10 5 2 −1 2 12 8 Sr 225 215 199 239 173 235 239 V 210 159 220 203 220 240 257 Y 21 13 16 14 17 23 24 Zr 67 59 57 55 61 71 73 PXRF n 6 5 5 5 5 Mean magnetic 0.52 0.44 0.47 0.43 0.61 susceptibility (10−3 SI) s.d. 0.04 0.04 0.04 0.03 0.05 Mag sus. n 10 6 10 6 10 Notes: Data from Thorpe et al. (1991) (monoliths), and Jones and Williams-Thorpe 2001 and Williams-Thorpe et al. 2004 (implements; Pb omitted (large s.d.) and V added to support groupings (new data, rsd 20–50 %)); analytical details in those papers. Major and minor elements are given as weight % (iron as Fe2O3 total iron), trace elements as ppm. Negative values are retained, so means can be zero. The implements are grouped according to their chemical compositions (cf. text discussions). All artefacts and monoliths are spotted dolerite except those marked*, which are unspotted dolerite with Preseli compositions. Artefact types after Thorpe et al. op. cit. and Williams-Thorpe et al. op. cit. table 4. Abbreviations: An Anglesey, Car Carmarthenshire, Dev Devon, Do Dorset, Ha Hampshire, M Monmouthshire, Me Merionethshire, Ra Radnorshire, Wi Wiltshire; SH Stonehenge. OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors 30 Journal compilation © 2006 Blackwell Publishing Ltd.
OLWEN WILLIAMS-THORPE, M.C. JONES, P.J. POTTS AND P.C. WEBB The chemical database given by Thorpe et al. (op. cit.) allows us to comment on the likelihood of these 11 monoliths, and Preseli dolerite implements, coming from the same outcrop sources, or from different outcrop sources. This information will help us to interpret the putative links between the selection of raw material for axes and other implements, and for Stonehenge monoliths (cf. e.g. Atkinson 1979; Green 1997). To support the present paper, we will draw on our recent in situ characterization of outcrops of the Preseli spotted dolerite area using portable X-ray fluorescence (PXRF) (Jones et al. 2005) to extend the available information on the chemical composition of potential source outcrops. the axes and other implements Twelve implements that were linked to a Preseli source by Williams-Thorpe et al. (2004) are considered in the present paper, and compositions of these implements are listed in Table 1 (together with other analytical data, which are discussed later). Eight artefacts in Table 1 are of Preseli spotted dolerite, and two (Wi302 and An10) are of Preseli unspotted dolerite (strictly, therefore, members of IPC Group XXIIIb rather than XIII; cf. Clough and Cummins Do41 Dev1 Car23 Ra4 M8 Me8 An10* SH42 Maiden Castle High Peak, Trelech a’r Clap yr Arian, St Brides, Bwlch-Gwyn Llanfaethlu Stonehenge Sidmouth Bettws Llansantffraed- Netherwent Farm, near Cwmdeuddwr Arthog reworked axe mace battle-axe battle-axe axe axe-hammer axe-hammer monolith 0.72 0.62 1.76 0.75 0.99 0.99 0.36 0.15 11.61 9.67 8.07 9.68 9.68 9.31 5.91 11.53 1.12 0.56 1.12 1.07 0.73 0.70 0.96 1.37 0.17 0.13 0.14 0.12 0.15 9.04 9.49 9.06 9.43 10.94 8.75 8.54 9.17 222 129 115 98 132 192 23 123 6 4 4 4 0 −6 −5 7 3 −2 10 −3 −5 32 −1 8 225 210 230 256 329 518 246 235 249 267 231 245 213 22 14 14 19 11 10 60 31 72 50 64 81 72 94 98 101 5 5 5 5 5 1 1 0.43 0.44 0.40 0.59 0.44 0.52 0.54 0.03 0.02 0.05 0.05 0.02 0.05 0.02 6 5 21 11 5 6 5 OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd. 31
PRESELI DOLERITE BLUESTONES 1988). M8 and Me8 have spotted dolerite mineralogy, but chemical compositions that are marginal to Preseli (cf. Williams-Thorpe et al. op. cit., 372). The dating evidence and styles of the implements were noted by Williams-Thorpe et al. (2004). In summary, they include Neolithic as well as Bronze Age types, and rare round barrow and Beaker associations. Three battle-axes – Car23, Ra4 and Wi302 – stand out as more developed in style, and are the most carefully finished and highly polished of the pieces studied. The artefacts were found variously in southern England and in Wales, and the find locations are shown on Figure 1. the preseli source outcrops, and characterizing data available Spotted dolerites crop out within a small (c.5 × 2 km) area of eastern Preseli (Fig. 2). They form one part of the much larger south-west Wales Lower Palaeozoic volcanic and intrusive complex (Bevins et al. 1989), within which spotted dolerite rocks are reported only within the area shown on Figure 2 (Evans 1945; cf. Thorpe et al. 1991). Previous analytical provenancing studies of spotted dolerite artefacts have relied mainly on comparative source data drawn from conventional laboratory X-ray fluorescence analyses of rock samples (Bevins et al. 1989; Thorpe et al. 1991; summary in Jones and Williams-Thorpe 2001). These data represent very good coverage of the Preseli spotted dolerite area as a whole (they include 45 samples from 14 outcrops), but several of the outcrops are represented by the compositions of only one or two rock samples each. Jones et al. (2005) used a new PXRF survey of Preseli dolerite outcrops to show that the geochemical variance (distinguished from analytical and sampling variances) within a single outcrop often exceeds the variance derived from the limited sample analysis that is traditionally used to characterize sources for geological work. Jones et al. also considered some of the Preseli outcrops, in particular Carnmenyn, Carnbreseb and Cerrigmarchogion, to be geochemically heterogeneous for some elements. Bearing in mind this intra-outcrop variation, for our investigation of outcrop-specific provenancing of artefacts we will use the PXRF database reported by Jones et al., which provides more detail for individual outcrops. The dataset comprises 122 PXRF measurements made at 55 locations dispersed over eight outcrops representing good geographical coverage of the Preseli spotted dolerite area (Fig. 2). In addition to the geochemical data, a recent survey of magnetic suceptibility of Preseli outcrops, reported in summary by Williams-Thorpe et al. (2004), offers useful characterizing data. The magnetic susceptibility dataset comprises 853 measurements dispersed over 15 outcrops or discrete exposures. Mean measurements for these 15 areas are given in Table 2 and their geographical distribution is shown in Figure 2 above. While Williams-Thorpe et al. considered the magnetic susceptibility in terms of a range for the whole of the Preseli spotted dolerites (op. cit., 366), in the present paper we shall investigate these data in terms of individual outcrops in order to identify any distinctions that might be useful in provenancing artefacts. OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors 32 Journal compilation © 2006 Blackwell Publishing Ltd.
OLWEN WILLIAMS-THORPE, M.C. JONES, P.J. POTTS AND P.C. WEBB Figure 1 Map of part of Britain showing find locations of 12 axes and other implements that have been linked to a Preseli source by previous work, glacial systems (directions and limits, shown respectively as arrowed lines and dotted lines), and the locations of places mentioned in the text. The broken line in southern England encloses four implements that are compositionally similar to each other and to the majority of analysed Stonehenge monoliths. Me8 and M8 (with question marks) are spotted dolerite implements with chemical compositions marginal to Preseli (cf. text). Glacial information is mainly taken from Bowen (1991) and Charlesworth (1957), with additions after Briggs (1994) and John (1984) (small arrows near Preseli); the dot-dash arrowed line represents the controversial proposed further extension of Pliocene ice towards Salisbury Plain as suggested by Kellaway (2002). OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd. 33
PRESELI DOLERITE BLUESTONES 0m 25 Garn Ddu Fach Carngoedog 0.49 0m 25 *Carnbreseb 0.59 Carn Ddafad-Ias 0.49 33 y d d Carn Gwr 0.43 B 0.70 y n A 0.59 Sheepfold East 0.78 M e l i smaller 0.54 e s Sheepfold 0.58 P r Carngyfrwy 0.61 Mynydd-bach Carnbica 0.47 Carnmenyn *Cerrigmarchogion 0.55 main 0.48 m Carnarthur 0.46 250 Carnsiân 0.50 32 11 12 Craig Talfynydd (unspotted) 250 31 m Mynachlog-ddu Area of 30 main map 13 14 15 Figure 2 Sketch map of eastern Preseli in South Wales showing the locations of dolerite outcrops, magnetic susceptibility data (means for each of 15 exposures, cf. text and Table 2), and high Zr outcrops (indicated by *). Outcrops measured by PXRF are named in bold italics. investigating links between implements and with the outcrops In order to compare the implements with each other, and with outcrop sources, a number of different approaches were investigated, with varying success. We will describe first the tests based on geochemical characteristics, and then those using the magnetic data. The geochemical data In earlier work, Jones et al. (2005) identified some distinctions between Preseli outcrops using the PXRF dataset. Using Discriminant Analysis (DA) in the standard SPSS software, Jones et al. (op. cit.) measured the separation in terms of SPSS’s cross-validation estimate (that is, the percentage of data points, treated in turn as unknowns, for which SPSS correctly predicts the OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors 34 Journal compilation © 2006 Blackwell Publishing Ltd.
OLWEN WILLIAMS-THORPE, M.C. JONES, P.J. POTTS AND P.C. WEBB table 2 Magnetic susceptibility data for Preseli spotted dolerite exposures Outcrop Magnetic susceptibility (×10−3 SI) mean n relative s.e. of mean (%) Carnmenyn ‘smaller’ 0.54 119 1.6 Carnmenyn ‘main’ 0.48 33 1.6 Carn Ddafad-las 0.49 69 0.9 Carnbreseb 0.59 73 1.8 Carn Sian 0.50 30 2.6 Carn Arthur 0.46 44 1.4 Cerrigmarchogion 0.55 133 1.4 Carngoedog 0.49 71 1.4 Carnbica 0.47 71 1.6 Carn Gwr 0.43 44 1.4 Carngwfrwy 0.61 52 8.4 Sheepfold 0.58 53 3.4 Sheepfold East 0.78 36 5.0 Outcrop ‘A’ 0.59 12 5.4 Outcrop ‘B’ 0.70 13 5.2 Total n 853 group, or outcrop, membership), obtaining reclassification rates (for various element combinations) of up to 72 per cent. Encouraged by this, we decided to extend the DA to investigate the relationship of our 12 Preseli-related implements with the outcrop sources, and with each other. However, there is an important difference between the treatment of the PXRF data required for Jones et al.’s work, and for the present paper. PXRF outcrop data may be affected by weathering processes that can significantly alter the chemical composition of the (measured) surface layer from that of the bulk rock. However, the 12 implements discussed here were measured on generally unweathered surfaces, and so their PXRF analyses are not similarly affected. Therefore, in order to compare outcrops with implements, the outcrop data must be adjusted to compensate for weathering effects. Weathering correction factors were derived by comparing (mean) PXRF analyses for eight outcrops (total n = 110) with (mean) laboratory XRF analyses (total n = 23) of rock samples from the same outcrops. Regression analysis from SPSS provided estimates of slopes and their standard errors for each element in turn. The correction factors derived from these slopes for the elements used in this paper (and with 95 per cent confidence limits that reflect variations in the behaviour of individual elements and rocks) are as follows: K 0.78 ± 0.17, Ca 2.30 ± 0.72, Fe 1.72 ± 0.36, Ba 1.28 ± 0.23, and Sr 1.47 ± 0.31. Support for these correction factors comes from the work of Potts et al. (2006) who evaluated analogous factors using the ratio of the surface (weathered) concentration of Preseli dolerite samples, to the corresponding (fresh rock) concentration below the altered layer. The corrections derived by Potts et al. (op. cit.) agree well with ours except for Fe and Sr for which our corrections are larger but nevertheless agree with Potts et al. within the 95 per cent confidence limits reported, respectively, by Potts et al. and above. Potts et al.’s factors form part of a detailed study of weathering behaviour in four samples from three outcrops, while the OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd. 35
PRESELI DOLERITE BLUESTONES PXRF/laboratory XRF comparison is based on measurements at a much larger number of outcrops and locations within outcrops. Corrections from the latter dataset may better reflect the variations in weathering behaviour in Preseli, and, therefore, are used in the present paper. (Note that, since all the corrections were derived from Preseli dolerites, they are not directly applicable to other areas or rock types.) In Figures 3(a) and 3(b), then, we use DA to compare the 12 implements (treated as unknowns by the program) with the (weathering corrected) Preseli PXRF outcrop dataset. The choice of elements largely follows that of Jones et al. (2005). Thus, we use Fe, Ba, Sr, Ca and K, but omit Zr (a relatively minor contributor to the discrimination noted by Jones et al.) because it showed some slight but systematic depletion in comparisons with laboratory XRF discussed below. K is added separately, in Figure 3(b), in order to illustrate the effect this addition has on the position of, in particular, Car23, on the plots. A noticeable feature of these figures is that the implements which were found in England (‘English’ implements) all lie in the upper portion of the plots, while those found in Wales (‘Welsh’ implements) lie in the lower portion. Me8 is a compositionally marginal Preseli spotted dolerite (Williams-Thorpe et al. 2004, 372), whose outlying position on the plots may be partly a function of its unusually high (single) Sr determination (cf. Table 1). M8 is also marginal Preseli composition, though its differences are less apparent on the DA plots. Our initial (by eye) examination of the data had suggested close similarities between certain implements (particularly within the English finds), with putative groupings consisting of: Ha41, Wi118, Wi109 and Wi108; and Wi302 and Do41 (cf. Table 1). The similarities between Wi302 and Do41 are reflected in Figures 3(a) and 3(b), but the distinction of these two from the other English artefacts is not supported by this DA treatment. The assigning of artefacts to specific outcrops is much more difficult. Firstly, the outcrops are closely related geologically (and geographically) and are all fairly similar in composition, so it is unsurprising that many outcrop fields overlap in Figure 3. Secondly, the weathering corrections applied to the PXRF outcrop dataset have large uncertainties (cf. above). While these uncertainties do not affect the relative positions of the outcrops in Figure 3, they can affect the positions of the implements in relation to the outcrops (because the implements do not require the same weathering corrections as the outcrops). In effect, each implement plotted on Figure 3 has an uncertainty on its position relative to the outcrops, an uncertainty that can be represented by very approximately ± 1 unit on the Function 1 scales, and ± 0.5 units on the Function 2 scales. Given these limitations, DA can at present only give the most tentative indications of likely, and unlikely, outcrop sources for implements. However, from Figure 3, it is notable that most of the Carnmenyn and Carn Ddafad-las samples are drawn off by the DA to the left-hand side of the diagrams, away from the majority of the implements. In Figure 3(b) the implement Car23 is placed near to Carnbica and Cerrigmarchogion reflecting higher K content in Car23 and in these outcrops (cf. Table 1; PXRF Preseli data are given in full in Jones et al. 2005, Appendix and cf. table 2). The magnetic susceptibility data Eight hundred and fifty-three measurements of magnetic susceptibility were considered in terms of individual outcrops. The division of the data into outcrops is less easy than might be expected, because some outcrops that are named separately on Ordnance Survey maps are OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors 36 Journal compilation © 2006 Blackwell Publishing Ltd.
OLWEN WILLIAMS-THORPE, M.C. JONES, P.J. POTTS AND P.C. WEBB FeBaSrCa (a) English implements Welsh implements Me8 FeBaSrCaK (b) English implements Welsh implements Me8 Figure 3 Discriminant analysis plots for eight Preseli spotted dolerite outcrops (n = 122) and for 12 implements (means; n = typically 5 per object, one each for An10 and Me8), using concentrations of (a) Fe, Ba, Sr and Ca, and (b) Fe, Ba, Sr, Ca and K. Element concentrations were all determined by PXRF, and the outcrop analyses have been adjusted for weathering effects using corrections described in the text. The implements are named on the figures. The dashed lines (which have no statistical significance) are drawn between implements found in England, and those found in Wales. Car23 is strongly affected by the addition of K (Fig. 3(b)) because this implement has a relatively high concentration of K (cf. discussion in text). OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd. 37
PRESELI DOLERITE BLUESTONES nearly contiguous, while others bear only one name but comprise more than one discrete exposure. For this exercise, we divided the data into 15 areas, each one a discrete exposure on the ground, and all marked on Figure 2 above (note that Carnmenyn was divided into two areas, called by us ‘main’ and ‘smaller’ respectively). The magnetic susceptibility readings for outcrops and for implements are illustrated on Figure 4. Most outcrops have overlapping measurements, but some, such as Carngwr and Carn Ddafad-las, have relatively restricted ranges. The magnetic susceptibility ranges of outcrops are not simply related to the outcrop size. Sheepfold East, for example, is one of the smallest outcrops, yet it has a larger magnetic susceptibility range than Cerrigmarchogion, an outcrop that extends over some 500 m in length. The implements plot between about 0.40 and 0.60 × 10−3 SI on Figure 4. T-tests were used to compare each of the implements to each outcrop in turn, following the procedures established by Williams-Thorpe et al. (1996). Given the overlapping outcrop ranges shown on Figure 4, we did not expect to identify specific outcrop sources, rather to use the tests to investigate similar, or dissimilar, behaviour of the artefacts. The interpretation of t-tests rests heavily on the selected level of significance (called ‘alpha’) at which outcrops are ‘allowed’ to be similar to an artefact. A level of significance might be determined by empirical factors (such as the level at which some already-known provenance is ‘allowed’ by the procedure, cf. Williams-Thorpe et al. 1996), or a commonly used significance level of, say, 95 per cent might be selected (equivalent to alpha of 0.95). In our present case, a much higher alpha of 0.999 was indicated. The key to use of such a high alpha is to note that 3.00 Carnmenyn smaller Carnmenyn main Carn Ddafad-las Carnbreseb Carn Sian Carn Arthur Cerrigmarchogion Carngoedog Carnbica Carn Gwr Carngwfrwy Sheepfold Sheepfold East Outcrop A Outcrop B Implements 2.50 Magnetic susceptibility (S.I.x10-3) 2.00 1.50 1.00 0.50 0.00 Figure 4 Chart showing 853 magnetic susceptibility measurements for 15 spotted dolerite exposures in eastern Preseli, together with average magnetic susceptibilities for each of the 12 implements discussed (data from Table 1). Measurement precision is 5 per cent relative or better (Williams-Thorpe and Thorpe 1993). Note that the plotted positions of the implements overlap on the figure. OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors 38 Journal compilation © 2006 Blackwell Publishing Ltd.
OLWEN WILLIAMS-THORPE, M.C. JONES, P.J. POTTS AND P.C. WEBB t-tests are tests for equality of means of source and artefact values, while we are really interested in whether artefact values lie within the distribution of source values. T-tests with high alpha are therefore being used as proxies for the latter, determining that an artefact did not come from a source only if their respective means are considerably different. This increase in alpha value ‘allows’ a larger number of outcrops to be treated as the potential source for each artefact, thus realistically reflecting the overlaps between outcrops. For most of the artefacts, the t-test selects a large proportion of the outcrops – six to 13 – effectively allowing most of the source area. However, there are differences between the test responses. Of the English artefacts, six have between seven and 13 selected outcrops; Devon 1 appears to be a little different, with only four selected outcrops. The Welsh axes are also variable: Car23 has only one selected outcrop, and M8, four. The other Welsh artefacts, An10, Ra4, and Me8, have six, seven and 13 outcrops respectively. Magnetic susceptibility data for the artefacts may also be usefully combined with chemical data (Fig. 5), choosing Zr, a frequently used geochemical discriminator (e.g. Rollinson 1993), to represent trace elements. Figure 5, like the DA plots, suggests some differences between the Welsh and the English implements, with three of the former (Ra4, An10 and Me8) having higher Zr, and generally higher magnetic susceptibility, than the latter. investigating links between the implements and the stonehenge monoliths Only 11 (out of 27) Stonehenge dolerite monoliths have yet been analysed for chemical compositions, and there are no magnetic susceptibility data for any of the monoliths. They are 0.65 Wi 302 Magnetic susceptibility (S.I. x 10-3) 0.60 Ra 4 0.55 An 10 Ha 41 0.50 Me 8 Wi 109 0.45 Wi 108 M8 Dev 1 Wi 118 Do 41 0.40 Car 23 0.35 English implements Welsh implements 0.30 40 50 60 70 80 90 100 110 Zr (ppm) Figure 5 Chart showing magnetic susceptibility plotted against Zr concentrations (PXRF analyses) for the studied implements, distinguishing the implements found in England (diamonds) and those found in Wales (squares). The error bar shows a typical standard error of the mean. OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd. 39
PRESELI DOLERITE BLUESTONES therefore not included in the statistical comparisons described above. Nevertheless, some observations concerning the relationship of the monoliths with the Preseli dolerite implements can be made using the limited monolith dataset in Thorpe et al. (1991). Figures 6(a), (b) and (c) show the concentrations of selected chemical elements in the monoliths, and in the axes and other implements included in this work. Note that here we are comparing laboratory (wavelength-dispersive and energy-dispersive) X-ray fluorescence analyses for the monoliths with PXRF analyses (not corrected for weathering) for the implements. Thorpe et al. (1991) separated the 11 monoliths into three compositional groups, and these groups are reflected in the data shown in Figure 6: firstly, a group of eight monoliths that may be conveniently named after the stone that has the identifier ‘SH (Stonehenge) 33’; secondly, SH44 and SH45; and finally, SH42. SH62 (named on Figure 6) was included by Thorpe et al. in the SH33 grouping on the basis of similar concentrations of geochemically immobile elements. Nevertheless, it has some compositional differences (cf. Figures 6(a) and 6(b)), and was also subsequently proved to be distinctive in its opaque mineral assemblage (Ixer 1997). Figure 6 also suggests some compositional similarities between the monoliths and some of the implements, mainly the English implements. Five implements (Wi108, Wi109, Wi118 and Ha41 all found in England, Car23 found in Wales) consistently plot close to the SH33 group. Wi302 and Do41 plot nearer to SH44 and 45. Examination of the full range of analysed elements (Table 1) best supports the parallel of four English implements with the SH33 group. We noted that Zr is slightly and systematically lower in the implements than in the monoliths (cf. Figure 6(c)); this could be a result either of analytical factors (remembering that we are comparing here PXRF of implements with laboratory XRF of monoliths), or slight weathering-related depletion of Zr in the implement surfaces measured. The unweathered appearance of measured surfaces argues for an analytical explanation. Figure 6 shows variability of composition within the Welsh implements, but little consistent similarity of these Welsh implements to monolith groupings. An10 plots near to the monolith SH42 on Figure 6(c) (only), and in respect of this SH42 is included in Table 1 to allow a fuller comparison. Car23 plots within the SH33 group on Figure 6, but other elements differ, and also its separation from the English implements was underlined by the DA. Overall, the Welsh artefacts appear compositionally unlike the monoliths. These observations carry the implication that some of the implements could have originated at the same outcrop sources as some monoliths. And, equally importantly, some of the implements do not appear to be from sources that provided monoliths. summary of findings, and discussion The axes and other implements The investigations described above consistently support compositional groupings within the studied implements, and a dichotomy between implements found in England, and those found in Wales. Such groupings and distinctions are likely to reflect derivation of the raw material from different outcrops or areas of the Preseli spotted dolerite region. It should be noted that the two unspotted implements that have Preseli chemical and magnetic characteristics (An10 and Wi302) could potentially be derived from ‘spotted’ outcrops, since these frequently contain unspotted parts (cf. Williams-Thorpe et al. 2004). OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors 40 Journal compilation © 2006 Blackwell Publishing Ltd.
OLWEN WILLIAMS-THORPE, M.C. JONES, P.J. POTTS AND P.C. WEBB 600 (a) English implements Me 8 500 Welsh implements Stonehenge monoliths 400 Sr (ppm) M8 300 Wi SH 45 Ra 4 Wi An 10 109 SH 302 SH 62 Car 23 42 SH 44 Do 200 Ha 41 Dev 1 41 Wi Wi 118 SH 33 group 108 100 0 0 100 200 300 400 500 600 Ba (ppm) 13 (b) 12 SH 44 M8 11 Fe2O3 (wt.%) Wi 108 SH 45 10 Wi 109 SH 62 Ra 4 Do Dev 1 Wi 118 Ha 41 41 9 Car 23 Wi Me 8 An 10 302 8 SH 33 group 7 6 0 100 200 300 400 500 Ba (ppm) (c) Me 8 500 400 Sr (ppm) M8 300 SH 62 SH 44 SH 45 Ra 4 An 10 Wi 109 Car 23 Dev 1 Ha 41 SH 42 200 Wi 118 Wi 302 Wi 108 Do 41 SH 33 group 100 40 60 80 100 120 Zr (ppm) Figure 6 Graphs of (a) Sr (ppm) vs. Ba (ppm), (b) Fe2O3 (wt %) vs. Ba (ppm) and (c) Sr (ppm) vs. Zr (ppm), showing the studied implements (diamonds and squares) and 11 Stonehenge dolerite monoliths (triangles). Lines are drawn round the monolith groupings proposed by Thorpe et al. (1991) and have no statistical significance. The error bars show typical standard errors of the means for implements, and analytical uncertainties for monoliths (not shown if within symbol size). The elements shown were selected in order to best reflect monolith groupings based on Thorpe et al. (op. cit.) and further information in Ixer (1997), and to illustrate the relationship of the implements with monolith groupings. The data are taken from Thorpe et al. (op. cit.; monoliths) and Williams-Thorpe et al. (2004; implements). OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd. 41
PRESELI DOLERITE BLUESTONES The implements that were found in England (all found in the Stonehenge area or in the south-west, cf. Figure 1) are also those which show the most obvious compositional similarities to Stonehenge monoliths. Three implements that were found at or very near to the site of Stonehenge (Wi108, 109, 118), together with Ha41 (found on the south coast), are most similar to the main monolith group (the SH33 group). Support for this link comes from Ixer (1997), who made the observation (Ixer op. cit., 16) that a particular opaque mineral assemblage characterizes the SH33 Stonehenge monolith group, and also appears to be ‘very significant numerically’ within the Group XIII examples in the South-West England Museums’ implement collection. That collection includes all four of our most SH33-like artefacts – Ha41, Wi118, Wi108 and Wi109. The implements found in Wales not only differ from those found in England, but also lack the compositional similarities to Stonehenge monoliths that we observed in the English implements. A weakness of our observations is that they are based on small numbers of artefacts: that is, the 12 Preseli-related implements; and only 11 of the 27 remaining dolerite Stonehenge monoliths. However, a reasonable interpretation of the available evidence is that Preseli implements found in England are generally derived from the same outcrops as the monoliths, while the implements found in Wales are not. This adds weight to the possibility that the ‘English’ implements might have been manufactured from the same rock assemblage, whether outcrop or erratic material (or waste material from the monoliths themselves), that was exploited to build the bluestone settings at Stonehenge. These implements may then have been distributed, perhaps traded, to other parts of south-west England, within the framework of Beaker period or Early Bronze Age activities. The compositional distinction of the Welsh implements, on the other hand, suggests a separate resource procurement strategy involving exploitation of several different Preseli outcrops, either at primary (outcrop) sources, or at secondary (e.g. glacially or river-transported) sources. The extent of human distribution of the Welsh implements cannot be inferred in the same way as for those found in south-west England, because of the opportunity for glacial dispersion of rocks in Wales. While the main ice flow directions across Preseli were broadly from north-west to south-east (Fig. 1), more local, northward moving, tongues of Devensian ice (Briggs 1994) and also westward movement related to the course of the Gwaun Valley (e.g. John 1984) could have provided mechanisms for more complex natural dispersal of Preseli rocks. Curiously, all three implements that were found at sites north of Preseli (An10, Ra4, Me8) have markedly higher Zr levels (between 81 and 94 ppm) than the implements that were found at sites south of Preseli (Zr between 50 and 73 ppm) (Table 1; the higher Zr is generally correlated with higher magnetic susceptibility (cf. Figure 5) giving us confidence that the differences in Zr are not simply weathering effects). The highest Zr source outcrops (with average Zr of >80 ppm from published laboratory XRF analyses, again correlated with relatively high magnetic susceptibility) are Carnbreseb and Cerrigmarchogion. These outcrops are located, respectively, on the northern and western edges of the Preseli range (Fig. 2), positions from which natural boulder transport northward, as well as westward, is feasible. Exploitation of varied secondary source material dispersed by a number of different mechanisms may offer an explanation for the compositional variation seen within the Welsh artefacts. An alternative explanation for the compositional variations observed within both English and Welsh implements is that all these implements may have been selected randomly from a single, mixed (secondary) assemblage within Preseli. However, this explanation is made OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors 42 Journal compilation © 2006 Blackwell Publishing Ltd.
OLWEN WILLIAMS-THORPE, M.C. JONES, P.J. POTTS AND P.C. WEBB unlikely by the compositional correlation between the English artefacts (but not the Welsh artefacts) and Stonehenge monoliths. We were not able to identify correlations between implement style and composition or outcrop source (though shaft-hole implements are more prevalent among the Welsh studied artefacts). The three stylistically developed battle-axes, although extremely alike in form (Car23 and Ra4 are Roe’s Type IIa, and Wi302, Type IIc; Roe 1966, 1979), are of three different compositions and probably, therefore, from three different Preseli outcrops. And conversely, material that could be from the same outcrop was used for axes of different styles – for example, the small reworked axe from Maiden Castle (Do41) is of very similar composition to the Wi302 battle-axe. Williams-Thorpe et al. (2004) noted that some battle-axes of analogous style to Wi302 were made from non-Preseli rocks, for example Whin Sill dolerites (cf. Williams-Thorpe et al. 2003). Comments on bluestone outcrop sources The new chemical (PXRF) data, analysed above using DA, prompt us to consider that the Carnmenyn outcrop, traditionally regarded as the most important source both of monoliths and of axe-heads, may not be the source of any of the implements that we studied. If that is the case, neither is it likely to be the source of the most numerous monolith group, the SH33 group. Thorpe et al. (1991), using the available (laboratory XRF) analytical data on Preseli outcrops, identified the SH33 group with ‘East Preseli, Carnmenyn-Carngyfrwy or Cerrigmarchogion-Carngoedog’. Two monoliths were likened to Carn Ddafad-las, and one to Carnbreseb (Thorpe et al. op. cit., 139–40), but in general Thorpe et al. remained cautious about identifying specific outcrops because of the compositional similarity of many of these outcrops. Nevertheless, their published data do contain some clues that may inform the search for further supporting evidence. It is apparent, for example, that the (very spotted) Carngoedog outcrop contains dolerites that are compositionally closer to the SH33 monoliths than are those of Carnmenyn (cf. Thorpe et al. op. cit., 130–2, 143). Interestingly, in his comparison of opaque mineralogy of monoliths and Preseli outcrops, Ixer (1997) suggested the exclusion of several outcrops and narrowed the likely source of most of the monoliths to either Carngoedog or Carnmenyn. Analysis of all the Stonehenge dolerite monoliths using PXRF (a possibility which we are investigating) would allow progress in this debate. An alternative spotted dolerite source? One further question arises from our work, a question that is both worrying and intriguing. Is there another source of spotted dolerite outside Preseli, a source that contains rocks of similar appearance to those of Preseli, but with a different chemical composition? The uniqueness of Preseli spotted dolerites has, of course, already been considered during the long bluestone debate, with the consensus that eastern Preseli is the only relevant potential source area for monoliths (Thomas 1923, 250; Thorpe et al. 1991, 119). It is, indeed, most unlikely that further outcrops or exposures of spotted dolerite large enough to provide the Stonehenge monoliths have escaped the scrutiny of geologists. Smaller artefacts such as axes and axe-hammers, on the other hand, could feasibly be produced from rock exposures of much more limited extent. OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd. 43
PRESELI DOLERITE BLUESTONES Williams-Thorpe et al. (2004) identified an axe (Sx178; found in Newhaven, Sussex; cf. Figure 1) which has mineralogical characteristics analogous to Preseli spotted dolerite, but a chemical composition clearly unlike the available Preseli data, suggesting the existence of a source not represented within that dataset. Further, M8 and Me8, discussed in the present paper, are spotted dolerites but have marginal Preseli compositions. These anomalous data suggest that a reconsideration of alternative spotted dolerite sources is now appropriate. Scores of rock outcrops and areas have been discussed as potential bluestone sources (cf. summary in Thorpe et al. op. cit., 120 table 5), or are described as ‘spotted’ in geological literature. In many cases the spots are mineralogical features, such as crystals, that are not the same as the metamorphic spots in Preseli (compare, for example, Elsden (1905, 594) and Cox (1915, 311) both describing rocks in north Pembrokeshire). One case, however, merits further investigation. This is the ‘insignificant outcrop of similar rock [similar, that is, to Preseli spotted dolerite; present authors’ insertion] in the Cader Idris district’ (Thomas 1923, 250, quoting a personal communication from Professor A.H. Cox, University of Wales). While this outcrop was ‘disregarded as a possible source’ of monoliths (Thomas, op. cit.; presumably because of its small size), it may be relevant to axe/implement provenance. Cader Idris lies within the Aran igneous province (Dunkley 1979) which together with the Rhobell complex (Kokelaar 1986) constitutes the Ordovician volcanics and intrusions of north-central Wales (general locations are on Figure 1 above). Dolerites within both the Aran and Rhobell provinces have undergone similar metamorphic alteration to those of Preseli (cf. Dunkley op. cit., 600; Kokelaar op. cit., 892) and thus might resemble, mineralogically, those of Preseli. In this context, it may be more than coincidence that one of the spotted axes with rather anomalous chemistry (Me8) was found at Arthog, on the north-western margins of Cader Idris. conclusions The bluestone spotted dolerite implement group (Group XIII) does not comprise a single compositional assemblage, but, rather, a mixed grouping that probably originated at several outcrops mainly in Preseli and perhaps at another (unidentified) spotted dolerite source. Spotted dolerite axes and other implements that have been found in England are mainly similar in chemical composition to analysed Stonehenge spotted dolerite monoliths. Most of the spotted dolerite artefacts found at Stonehenge and in the south and south- west of England could have been manufactured from the same assemblage of outcrop rocks or erratics that was used for the Stonehenge monoliths. Spotted dolerite implements found in Wales have a variety of chemical signatures and show little or no similarity to analysed Stonehenge monoliths. The information on outcrop level provenancing implies that resource procurement strategies for the spotted dolerite used in England were different from those used in Wales. Geochemical evidence points, overall, to Carngoedog as the most similar outcrop to the largest group of analysed Stonehenge monoliths, and to the geochemically analogous (English) implements. An alternative source of spotted dolerite rocks, perhaps within the Aran or Rhobell complexes of north-central Wales, may have been exploited to provide material for spotted dolerite implements that have compositions different from, or only marginal to, Preseli dolerites. OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors 44 Journal compilation © 2006 Blackwell Publishing Ltd.
OLWEN WILLIAMS-THORPE, M.C. JONES, P.J. POTTS AND P.C. WEBB The complete picture of dolerite bluestone implement provenance and its relationship with Stonehenge monoliths will be understood much better in the light of analysis of all the Stonehenge dolerite monoliths. Clarification of the possibility of alternative spotted dolerite source(s) is also needed. Acknowledgements We are grateful to Sid Howells of the Countryside Council for Wales for advice and permissions for work in Preseli (with particular reference to the magnetic susceptibility survey for which full results are given for the first time here). The paper stems from earlier studies of dolerite implements and outcrops, funded by The Leverhulme Trust (grant no. F/00269/C). We thank John Taylor (Figs. 1, 3–6) and Andrew Lloyd (Fig. 2) for cartography. (OWT, PJP, PCW) Department of Earth Sciences (MCJ) Department of Statistics The Open University Milton Keynes MK7 6AA, UK references atkinson, r.j.c. 1979: Stonehenge: archaeology and interpretation (3rd ed.) (Harmondsworth). bevins, r.e., lees, g.j. and roach, r.a. 1989: Ordovician intrusions of the Strumble Head-Mynydd Preseli region, Wales: lateral extensions of the Fishguard Volcanic Complex. Journal of the Geological Society of London 146, 113–23. bowen, d.q. 1991: Time and space in the glacial sediment systems of the British Isles. In Ehlers, J., Gibbard, P.L. and Rose, J. (eds.), Glacial Deposits in Great Britain and Ireland (Rotterdam), 3–11. briggs, s. 1994: The Bronze Age. In Davis, J.L. and Kirby, D.P. (eds.), Cardiganshire County History Volume I (Cardiff, Cardiganshire Antiquarian Society and Royal Commission on the Ancient and Historical Monuments of Wales), 169. charlesworth, j.k. 1957: The Quaternary Era (London). clough, t.h.mck. and cummins, w.a. (eds.) 1988: Stone Axe Studies Volume 2 (London, CBA Research Report No. 67). cox, a.h. 1915: The geology of the district between Abereiddy and Abercastle (Pembrokeshire). Quarterly Journal of the Geological Society of London 71, 273–340. dunkley, p.n. 1979: Ordovician volcanicity of the SE Harlech Dome. In Harris, A.L., Holland, C.H. and Leake, B.E. (eds.), The Caledonides of the British Isles – reviewed (London, Special Publication of the Geological Society of London 8), 597–601. elsden, j.v. 1905: On the igneous rocks occurring between St. David’s Head and Strumble Head (Pembrokeshire). Quarterly Journal of the Geological Society of London 61, 579–607. evans, w.d. 1945: The geology of the Prescelly Hills, north Pembrokeshire. Quarterly Journal of the Geological Society of London 101, 89–110. green, c.p. 1997: Stonehenge: geology and prehistory. Proceedings of the Geologists’ Association 108, 1–10. ixer, r.a. 1997: Detailed provenancing of the Stonehenge dolerites using reflected light petrography. In Sinclair, A., Slater, E. and Gowlett, J. (eds.), Archaeological Sciences 1995 (Oxford, Oxbow Monograph 64), 11–7. OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd. 45
PRESELI DOLERITE BLUESTONES john, b.s. 1984: Pembrokeshire (Newport). jones, m.c. and williams-thorpe, o. 2001: An illustration of the use of an atypicality index in provenancing British stone axes. Archaeometry 43, 1–18. jones, m.c., williams-thorpe, o., potts, p.j. and webb, p.c. 2005: Using field-portable XRF to assess geochemical variations within and between dolerite outcrops of Preseli, South Wales. Geostandards and Geoanalytical Research 29, 251–69. kellaway, g.a. 2002: Glacial and tectonic factors in the emplacement of the bluestones of Salisbury Plain. In Chapman, M. and Holland, E. (eds.), The Survey of Bath and District No. 17 (Bath), 57–71. kokelaar, p. 1986: Petrology and geochemistry of the Rhobell volcanic complex: amphibole-dominated fractionation at an early Ordovician arc volcano in North Wales. Journal of Petrology 27, 887–914. potts, p.j., bernardini, f., jones, m.c., williams-thorpe, o. and webb, p.c. 2006: Effects of weathering on in situ portable X-ray fluorescence analyses of geological outcrops: dolerite and rhyolite outcrops from the Preseli Mountains, South Wales. X-Ray Spectrometry 35. roe, f.e.s. 1966: The battle-axe series in Britain. Proceedings of the Prehistoric Society 32, 199–245. roe, f.e.s. 1979: Typology of implements with shaft-holes. In Clough, T.H.McK. and Cummins, W.A. (eds.), Stone Axe Studies (London, CBA Research Report 23), 23–47. rollinson, h. 1993: Using geochemical data (Harlow). thomas, h.h. 1923: The sources of the stones of Stonehenge. The Antiquaries Journal 3, 239–60. thorpe, r.s., williams-thorpe, o., jenkins, d.g. and watson, j.s. with contributions by ixer, r.a. and thomas, r.g. 1991: The geological sources and transport of the bluestones of Stonehenge, Wiltshire, UK. Proceedings of the Prehistoric Society 57, 103–57. williams-thorpe, o. and thorpe, r.s. 1993: Magnetic susceptibility used in non-destructive provenancing of Roman granite columns. Archaeometry 35, 185–95. williams-thorpe, o., jones, m.c., tindle, a.g. and thorpe, r.s. 1996: Magnetic susceptibility variations at Mons Claudianus and in Roman columns: a method of provenancing to within a single quarry. Archaeometry 38, 15–41. williams-thorpe, o., potts, p.j. and webb, p.c. 1999: Field-portable non-destructive analysis of lithic archaeological samples by X-ray fluorescence instrumentation using a mercury iodide detector: comparison with wavelength-dispersive XRF and a case study in British stone axe provenancing. Journal of Archaeological Science 26, 215–37. williams-thorpe, o., webb, p.c. and jones, m.c. 2003: Non-destructive geochemical and magnetic characterisation of Group XVIII dolerite stone axes and shaft-hole implements from England. Journal of Archaeological Science 30, 1237–67. williams-thorpe, o., potts, p.j. and jones, m.c. 2004: Non-destructive provenancing of bluestone axe-heads in Britain. Antiquity 78, 359–79. OXFORD JOURNAL OF ARCHAEOLOGY © 2006 The Authors 46 Journal compilation © 2006 Blackwell Publishing Ltd.
You can also read