The rock crystal assemblage

During excavations at Dorstone Hill between 2011 and 2019, a total of 2686 pieces of chipped stone were recovered. Flint made up the majority of the assemblage (86.8 per cent); rock crystal comprised 12.5 per cent of the assemblage with a total of 337 pieces. There are also 18 examples of polished axes and axe flakes (0.7 per cent of the assemblage), both of flint and stone, which were probably acquired as finished objects. Petrological analysis is on-going; however, the flint axes are made of a type of flint usually associated with flint from primary geological contexts (e.g. mines), whie the others of volcanic rock probably originated in north Wales and Cumbria. Due to local geology, all raw materials used in the chipped-stone assemblage had to be imported to the site; the closest potential sources for flint is c. 20 km away (Elliot Reference Elliot2019).

Potential sources of rock crystals large enough to produce the artefacts at Dorstone (the largest removal being 34 mm long) are much further away than those for most of the potential flint sources. Geological specimens measuring in excess of 5cm in length have been recovered from Ogof Grisial (Crystal Cave) at St David's Head (Pembrokeshire) and multiple sites in Snowdonia, in north Wales (100 miles and 80 miles away from Dorstone Hill respectively, as the crow flies), with larger specimens up to 10 cm in length being recovered from Snowdonia. Both sources represent surface or close to the surface sources, which could feasibly have been accessible during the Neolithic (Cotterell Reference Cotterell2011; forthcoming; Griffiths Reference Griffiths2009; National Museum Wales n.d.). The South Wales coalfields around Merthyr Tydfil (c. 30 miles from Dorstone Hill) may be another potential source; however, known specimens are smaller (a maximum of 4 cm) and have been recovered recently as a result of industrial mining at depth (National Museum Wales n.d.), and may not have been a source easily accesible in the Neolithic. High-quality rock crystal has also been recovered from mines surrounding Tintagel, Cornwall (125 miles from Dorstone Hill) (Cotterell Reference Cotterellforthcoming), although these are subject to the same caveats as for the Merthyr samples. It is also possible the rock crystal originated from Ireland, Scotland or continental Europe (see introduction).

The techno-typological analysis of rock crystal artefacts is challenging, as the fracture characteristics of rock crystal are not identical to those of flint and chert (cf. Fernández-Marchena et al. Reference Fernández-Marchena, Ramón Rabuñal, Mateo-Lomba, Lombao, Hernando, Cueva-Temprana and Cazalla2020; Tardy et al. Reference Tardy, Vosges and Varoutsikos2016). Most pertinently, the anisotropic nature of rock crystal means that the energy from a blow may not travel evenly through the crystal (Fernández-Marchena & Ollé Reference Fernández-Marchena and Ollé2016; Rodríguez-Rellán Reference Rodríguez-Rellán2016), and the unique shape of the parent crystal also affects the reduction process. Therefore, while our approach is based on established recording systems for other types of lithic materials, some terminology has been adapted for the specific features of the rock crystal assemblage. The technological categorization that follows should be regarded as preliminary, pending experimental replication to understand better how rock crystal behaves when knapped or pressure-flaked.

For the purpose of this study, we recorded attributes of the crystal, included clarity, texture, inclusions and evidence for heat alteration, the techno-typological category of each artefact (according to conventional definitions with the addition of whole and partial crystal), the presence of visible modifications (e.g. edge retouch), as well as completeness and metric data. For the larger pieces we also recorded technological attributes including bulb type, striking platform type, platform edge treatment, termination type, dorsal scarring pattern, longitudinal profile and the presence of the natural edge of the crystal (where applicable). Those pieces exhibiting one or more crystal edges offered the potential to locate the specimen within the parent crystal. Where possible, specimens were recorded as ‘body’ faces along the length of the crystal (characterized by growth-lines orientated horizontally to the crystals longitudinal axis), ‘point’ faces at the tip, (lacking growth-lines), or ‘base’ faces (characterized by more complex formations where the crystal joins other crystals and/or the vein quartz substrate).

The breakdown by category (Table 1) indicates that complete rock crystals were being reduced at Dorstone Hill. A total of 60 pieces retained portions of the natural edge surface of the crystal (see Table 2 and discussion below), which are comparative to pieces retaining cortex in flint assemblages. These, combined with the presence of one complete crystal and two partial crystals, confirm that the material was being introduced to the site as whole crystals, which were then reduced on site. The larger removals (pieces over 10 mm) were divided into the conventional categories of blades, blade-like flakes and flakes, and the smaller removals (under 10 mm) were comprised of chips and flakes. The 31 larger removals represent 8.7 per cent of the total assemblage, or 73.8 per cent if we discount the smallest debitage. These larger removals range from 10 mm to 34 mm in length, and are mainly knapped from high-quality crystal, although some pieces show pronounced juddering/stepping towards the distal termination, which can be caused by areas of coarser crystal, and the anisotropy of the crystal (Tardy et al. Reference Tardy, Vosges and Varoutsikos2016, fig. 17). The smaller removals (chips and fragments), some of which are miniature versions of the larger flakes, are almost certainly the by-products of knapping, confirming reduction was being undertaken locally. The pieces comprising the smaller portion of this category are too small to be deliberately struck pieces, and seem to be a knapping by-product. No ‘cores’ in the traditional sense are present in the assemblage; however, two pieces have only negative scars and exhibit clear signs of multiple removals being taken from two faces in a systematic manner: these are the most convincing parent pieces, and have therefore been classified here as ‘cores’. The larger of these cores (#017, see Figure 5) measures 15×20 mm across one of its flaked surfaces, making it smaller than many of the blades/blade-like flakes/flakes, and both cores appear to at the end of their workability. The ‘bifacial’ nature of the reduction of these pieces could also indicate they were the intended product, especially given the small size and limited utility of the visible removals. Overall, the assemblage indicates the introduction of whole crystals to Dorstone hill, which were reduced, producing larger blades and flakes, smaller debitage and cores in the form of exhausted crystals.

Figure 5. Examples of larger pieces within the Dorstone Hill assemblage, including cores and pieces exhibiting crystal edges.

Table 1. Overall quantities of rock crystal recovered from Dorstone Hill, divided by ‘type’ and area of the site.

Table 2. All pieces exhibiting natural crystal edges, distributed by piece type, number of crystal edges visible for each piece and the position of the edges within the parent crystal structure. Note that all pieces under 10 mm are recorded in the ‘Chips & fragments’ category.

The flint assemblage, like the rock crystal, was largely dominated by small flakes, which, with the presence of cores, indicate flint was reduced on site, but also modified to create ‘formal’ tools (Table 3). After the removal of the smallest flake/flake fragment category, cores make up 1.8 per cent of the flint assemblage, but 4.7 per cent of the rock crystal assemblage, possibly indicating more on-site knapping of the latter. The two assemblages also contrast in the frequency of formal tools; a total of 102 formal tools were recovered from the Dorstone Hill sites, representing 4.4 per cent of the total assemblage, or 10.1 per cent after the removal of the smallest category of debitage (see Tables 3 & 4). In contrast, the rock crystal assemblage contained no formal tools, retouched pieces, or pieces with macroscopic edge-use, and the crystals continued to be reduced after the point at which the removals would have been a usable size. Among the flint, some of the formal tools, notably the axes, almost certainly arrived at the site as finished pieces; however, the scrapers, perforators, fabricators, knives and some of the arrowheads were made on larger flakes of the same dark grey flint that dominates the assemblage. Furthermore, the arrowheads of a different type of flint appear to be accompanied by debitage to suggest they, too, could have been made, or at least finished/repaired, on site. Therefore, while both flint and rock crystal were being brought to Dorstone Hill and reduced there, the rock crystal, transported from much greater distances, was not being used to make any formal tools, nor were the removals being used in any utilitarian way.

Table 3. Overall quantities of chipped stone (all materials) from Dorstone Hill.

Table 4. Comparison of rock crystal and flint and other lithic material assemblages (with smallest debitage category removed).

Working rock crystal at Dorstone Hill

The unique six-sided hyaline form of rock crystal affords the opportunity to establish a detailed sequence and directionality of crystal reduction at Dorstone Hill and as such provides the first detailed account of rock crystal reduction in a British or Irish context. A total of 60 specimens exhibited crystal edges (Table 2), with 33 of these specimens exhibiting only one face, and a further 27 exhibiting 2 or more faces. Given the specific morphology of the tip, body and base crystal faces (see above), it is possible to identify the specific faces in 28 of these pieces (Table 2). In a number of pieces, the crystal face morphology indicated the location of the piece more precisely within the parent crystal, allowing the reduction to be characterized further in a number of ways.

Firstly, a total of 14 pieces could be located precisely within the parent crystal (Table 6), which allowed the direction of crystal reduction to be established. When orientated within the crystal, 13 of these pieces had dorsal surfaces facing the crystal tip end, indicating a reduction sequence that began at the tip end, while a single specimen evidenced base-first reduction. The clearest examples of the tip-first reduction are two pieces which used platforms on the pyramidon to take a removal down the point faces (#009 and #054). A further three pieces (#005, #057 and #067) had been struck from platforms within the point area of the crystal, with the removals running down through the point-body boundary (see Figure 6). Finally, the larger of the two cores (#017) exhibits natural edge morphology to suggest it may be from the base of a crystal, indicating this is the exhausted remnant of a crystal reduced from the tip.

Figure 6. Pieces that evidence position within the crystal and direction of removal.

Secondly, 16 pieces, when orientated within the parent crystal, evidenced the direction of their removal (Table 5, Fig. 6). 11 were removed diagonally to the longitudinal axis of the parent crystal, two were removed horizontally and three were removed longitudinally. The vertical removals had at least one natural crystal surface, and appear to have been attempts to remove the outer edge of the crystal; these pieces exhibited a lower quality of crystal, which is most likely to be a product of the anisotropic nature of the material, inhibiting fracture along the longitudinal axis of the crystal (cf. Tardy et al. Reference Tardy, Vosges and Varoutsikos2016).

Table 5. Quantification of pieces with direction of removal from the crystal (relative to the parent crystal longitudinal axis), and direction of crystal reduction identified.

Thirdly, of the 62 specimens exhibiting evidence of previous removals (Table 6), the majority (40) presented dorsal scarring characteristic of systematic reduction. Of these, 28 pieces exhibited evidence of systematic unidirectional reduction (use of the same platform for multiple previous removals), with lower numbers of pieces showing truncated scarring from other directions. Based on the attributes used to determine reduction strategies relating to flint, the nature of the bulbs of force suggests direct percussion, and their diffuse appearance might indicate the use of a soft hammer (Whittaker Reference Whittaker1994, 180, 187ff, fig. 8.8). This contrasts with other assemblages that have undergone detailed technological analysis from Spain and Greece (Rodríguez-Rellán Reference Rodríguez-Rellán2016; Tardy et al. Reference Tardy, Vosges and Varoutsikos2016), where the use of pressure techniques as well as direct percussion in working this material is documented. The striking platforms are narrow, and some have tiny remnants of crystal faces, indicating the use of the natural crystal edge as a platform. Distal terminations tend to be broken, but where surviving they often overshoot, following the inherent properties of the crystal; this tends to corroborate the use of a soft hammer as it is similar to that known to occur on thinning flakes in bifacial reduction (Whittaker Reference Whittaker1994).

Table 6. Character of reduction process, distributed by piece ‘type’. Note that all pieces under 10 mm are recorded in the ‘chips and fragments’ category.

The evidence indicates a systematic and established process of crystal reduction at Dorstone Hill; the sequence began with whole crystals, which were reduced from the tip using the sub-90° angle of the pyramidon as a suitable platform. Once past the point region of the crystal, the body of the crystal was then reduced systematically by taking multiple removals primarily diagonally across the longitudinal axis of the crystal, with a smaller number of pieces being taken horizontally or vertically, the latter to remove the natural crystal edges. The full reduction sequence ended (in at least some cases), with an ‘exhausted’ crystal that took the form of a bifacially worked ‘core’ from the base of the crystal. The predominance of diagonal tip-first removals suggests that the knappers were familiar with the properties of the rock crystal, starting reduction from the isotropic pyramidon, before following the diagonal cleavage planes in the anisotropic body of the crystal (cf. Rodríguez Rellán 2016, fig. 1; Tardy et al. Reference Tardy, Vosges and Varoutsikos2016, fig. 15). This tip-first approach is generally similar to the method suggested by Fernández-Marchena and Ollé (Reference Fernández-Marchena and Ollé2016), where the crystal is ‘beheaded’ to create a platform from which bladelets would be struck by direct percussion using the facets of the crystal edge as guides; however, the Dorstone material differs in that it demonstrates reduction, as opposed to removal, of the tip, and diagonal reduction, as opposed to more longitudinal reduction for bladelets. The systematic reduction sequence of rock crystal contrasts with the reduction of flint at the site, which is more ad hoc in character, using both hard and soft hammer, including changes in orientation of the core during reduction. This more opportunistic approach may be the result of the small size of nodules being worked at the site; however, this too draws contrasts with the rock crystal, which was as small, if not smaller, yet worked in a systematic way.

This systematic approach to working the material at Dorstone Hill is further reinforced by the distribution of the rock crystal across the multi-monument complex. Rock crystal was recovered across all three mounds, and the causewayed enclosure (Table 7), however, these are not all contemporary. The structural phases of the mounds have produced dates indicating construction in the thirty-ninth or thirty-eighth century bc (Ray & Thomas Reference Ray, Thomas, Barclay, Field and Leary2020, 109), and a new radiocarbon measurement (Table 8) from a post-mound pit on the southern edge of the western mound, sealed by eroding mound material, indicated the mounds were most likely constructed by the thirty-eighth or mid thirty-seventh century bc. This pit was notable for containing 267 rock crystal specimens from either an in situ knapping event, or a deliberate deposition of knapping debris probably from a single event (see discussion below). The causewayed enclosure has returned dates of 3770–3640 cal. bc for the earliest phase, and 3630–3530 cal. bc for the latest phase of activity (Overton et al. Reference Overton, Ray, Thomas and Last2022, 48). Rock crystal was recovered from secure contexts in the pre-mound, mound and post-mound phases (Table 7), and from primary and recut ditches in the causewayed enclosure, indicating a prolonged engagement with this material, potentially spanning the thirty-ninth to thirty-sixth centuries cal. bc.

Table 7. Distribution of all rock crystal specimens in primary contexts within the Eastern, Central and Western mounds, divided by construction phase.

Table 8. Radiocarbon measurement from western mound pit [518]. Calibrated using OxCal 4.4 IntCal 20 curve (Reimer et al. Reference Reimer, Austin and Bard2020).

Given this temporal spread, the character of the assemblages are notably similar; flakes are the most frequent removal type, followed by blade-like flakes across the mounds and enclosure (Table 1). The only notable difference is the presence of whole crystals, partial crystals and ‘chunks’ at the causewayed enclosure, but not the mounds. These pieces, either complete crystals full of flaws, or small ancillary crystals probably removed from larger crystals prior to working, suggest materials may have been ‘tested’ at the enclosure prior to working, whereas ‘pre-tested’ crystals were introduced to the mounds. However, the presence of cores and large and small removals at both mounds and enclosure indicate this material was actively knapped at all four monuments. More significantly, when the reduction sequence is examined in more detail (Table 9), the tip-first systematic reduction, primarily focusing on diagonal removals, is present at all three mounds and the enclosure. This helps to confirm that this was an established and specific way of working this material, which persisted at Dorstone Hill for generations, and possibly centuries.

Table 9. Summary of technological process within each monument sub-assemblage.

Taken all together, the technological analysis indicates a very specific and distinctive set of practices related to rock crystal at Dorstone Hill. While all knappable material had to be introduced to the site, the rock crystal was a rare exotic, transported over substantial distances. At the site both rock crystal and flint were reduced; however, in contrast to the ad hoc flint reduction, working rock crystal followed a systematic process, informed by knowledge of its material qualities. Furthermore, this approach to rock crystal persisted for many generations, an established way of working this material, which was transmitted across generations as a set of skills, knowledges and practices. The distances involved in moving this material to Dorstone, the skill and knowledge bound into this established and systematic tradition of working and the maintenance of this practice over generations are all the more striking given the fact that this material, in direct contrast to flint, appears not to have been used to make any artefacts or tools. It was the material, and the practice of working it, which was significant; the intention to combine the material, the site, and a particular way of working it, but with no ‘final product’, indicate it was the moment of working itself that was the intended outcome. This can be explored further by examining how the resultant pieces from these events were treated through deposition.

Deposition of rock crystal at Dorstone Hill

Of the 337 rock crystal specimens recovered from Dorstone Hill, the majority (318) were retrieved from within or very close to the three long mounds, while 18 came from the Causewayed Enclosure and a single specimen came from beneath an Iron Age/Roman earthwork (Table 1). In the Causewayed Enclosure, rock crystal was recovered from upper ditch fills, which could not be confidently identified as in situ, and most likely represented silting events which incorporated material from surface middens within the enclosure (cf. Overton et al. Reference Overton, Ray, Thomas and Last2022). Within the mounds, 96.54 per cent of the rock crystal recovered was recovered from primary contexts (Table 10), including pits dug along the southern edge and into the body of the Western and Central mounds, the ditch surrounding the mortuary enclosure, and within deposits making up the mound. The remainder were recovered from secondary contexts (1.26 per cent) and topsoil or disturbed contexts (2.20 per cent) relating to the bulldozing of the mounds in the twentieth century. In contrast, 49.03 per cent of the flint and other lithic artefacts were recovered from primary contexts in the mounds, including fills of pits, post-holes and deposits making up the mound structure, with 6.14 per cent from secondary contexts and 44.83 per cent topsoil or disturbed contexts. The higher proportion of flint and other lithic materials recovered from disturbed contexts is probably the result of more of these materials being deposited into the upstanding mound layers, which were later bulldozed.

Table 10. Frequency of rock crystal and other lithic material from primary, secondary and disturbed contexts in the mounds.

The distribution of rock crystal and lithics within primary contexts exhibits a number of important patterns relating to the specific depositional contexts of these materials. Both materials were recovered from structural post-holes, pits dug after the initial construction of the Western and Central earthen mounds ([518], [136] and [528]), pits dug into the top of the Western and Central mounds ([557] and [110]), the ditch surrounding the mortuary structure, and deposits forming the body of the earthen mounds (see Figure 4). Additionally, flint was recovered from buried soil beneath the mound, tree throws and a number of stone-lined cists cut into the mounds. The distribution of this material over the different features shows that rock crystal was overwhelmingly deposited within pits, whereas the majority of flint was deposited relatively equally across pits and mound deposits (Table 11: Overall frequency). Furthermore, when the flint assemblage is split into flakes and formal tools, over 60 per cent of formal tools are recovered from pits. This indicates that, while flakes were present across the site, formal tools were being intentionally selected for deposition in pits, a pattern echoed in the deposition of rock crystal (albeit even more strongly with the latter). It must be noted that while in situ contexts were sieved on site, using a 0.8 mm dry sieve, sediment samples from pit [518] were wet-sieved through a 0.05 mm mesh, which recovered large numbers of very small rock crystal and flint fragments (267 rock crystal and 175 flint). This has inflated the frequency of flint and rock crystal flakes/debitage from pits; with the [518] material removed, 60 per cent of the flint flakes are within mound deposits and 46 per cent of rock crystal is from the mortuary structure ditch, with a further 27 per cent from pits. However, even though the lack of wet-sieving across the site has led to a possible under-representation of micro-debitage across the site, it does not affect the recovery of much larger formal tools, indicating the focus of formal tools within pits is a genuine pattern, and not a product of recovery bias.

Table 11. Distribution and frequency of rock crystal and other lithic material from primary contexts in the mounds.

The deposition of both rock crystal and formal tools made of flint and other lithic material is particularly associated with mortuary contexts, both in the form of graves and more broadly in reference to the long mounds. In pit [136], a cremation grave, five large flakes of rock crystal, including the largest pieces in the assemblage (#004) and a flake from the crystal tip (#009) (see Figure 5) were deposited alongside cremated bone, beneath a charred plank. The crystal and bone were accompanied by two flint leaf-shaped arrowheads, a transverse arrowhead, one flake from a greenstone polished axe, one scraper, one serrated flake (both flint) and 53 sherds of prehistoric pottery. In pit [518], another deposit of cremated bone was accompanied by a flint perforator, three rock crystal flakes, and 268 pieces of rock crystal and 175 pieces of flint micro-debitage (recovered through wet sieving). Further cremation deposits in pit [110] and within the mortuary structure ditch were also accompanied by rock crystal and a few flint flakes. Pits without cremated bone also contained notable material; pit [528] on the south edge of the Western mound contained a single rock crystal flake and six flint flakes, pit [557], cut through the Western mound into its burnt core, contained the largest rock crystal ‘core’, a flint scraper and three burnt chips, and pit [102], cut through the Eastern mound over the top of the mortuary structure, contained two polished axes and a bifacially flaked knife and debitage. These pits show the intentional selection of specific materials and artefacts, either as grave goods when associated with cremated bone, or as deposits venerating the mounds. Notably, rock crystal was selected as suitable for these deposits alongside other artefacts, even though it was not used or modified into tools. This further supports the interpretation that the material produced from knapping rock crystal was seen as significant in its own right and was perhaps the intended product from working (in contrast to flint, where the products were transformed into tools).

A number of these pits are also materially tied to moments of transformation. Multiple pits contained cremated remains, the product of dramatic and transformative processes; in the case of the mortuary chamber ditch, the rich charcoal and bone deposit was laid down directly before the chamber and ditch were backfilled and may be made up of the cremated remains of bodies taken out of the chamber before it was decommissioned. In pit [102], the knife was deposited alongside debitage which appears to be related to its production, suggesting this piece was made specifically for deposition. Pit [136] contained the fragments from breaking ceramic vessels and flaking of a polished stone axe. In pit [518], the micro-debitage of flint and rock crystal was the result of either knapping directly into the pit, or the deliberate collection and deposition of knapping debris, probably from a single event (cf. Kadowaki et al. Reference Kadowaki, Nagai, Nishiaki, Nieuwenhuyse, Bernbeck, Akkermans and Rogash2013, 154). This was deposited alongside large quantities of charcoal from a range of taxa; Corylus avellana (hazel), cf. Frangula alnus (alder buckthorn), Maloideae (apple, pear, hawthorn, service tree, rowan, whitebeam), Populus/Salix (poplar/willow), Quercus sp. (oak) and Sambucus nigra (elder). Notably, there was little oak and much of the charcoal derived from roundwood fragments of small diameter. This type of fuelwood would provide a relatively fast heat, typically used for domestic fires, but would not easily provide the high, sustained heat required for human cremation. The general paucity of oak is unusual, given that oak–hazel woodland was prevalent in the Dorstone area in the Neolithic and there is little evidence for significant woodland clearance until the Early Bronze Age (Garwood Reference Garwood and Watt2011). The charcoal from [518] came mostly from scrub or hedgerow taxa, which could have been sourced from the immediate environs of the site. Poplar, willow and alder buckthorn, however, favour wet ground habitats and are more likely to have come from the valley bottom. This combination of a range of species from a range of environs again suggests the intentional selection of materials; in later periods, certain woods have special associations: hazel and elder, for instance, have traditionally been used in England to ward off evil (Watts Reference Watts2007), and it is not inconceivable that the selection of wood for fuel in the Neolithic was neither purely accidental nor solely functional, but may have had symbolic meaning. Smoke and fire are associated with cleansing and purification rituals in many societies, with both the act of burning and the production of the ashes being significant (Chadwick Reference Chadwick2012; Kaliff Reference Kaliff and Insoll2011).

Therefore, rock crystal was one of a suite of materials being intentionally selected for deposition in pits and negative features across the mounds; these materials can be understood as votive or venerative ‘gifts’, but they were also significant as a result of the transformative events they were involved in before deposition. However, rock crystal should not be understood as interchangeable or synonymous with other selected materials and artefacts. In total, 95 per cent of rock crystal pieces were deposited in association with cremated bone deposits, in comparison to 35 per cent of flint flakes and 39 per cent of formal tools. This clearly demonstrates that rock crystal, over other materials, was being intentionally selected and deposited with cremated bone deposits, citing a specific link between this material, the transformative moments of working it, and the cremated remains of the dead.