This chapter details the analysis of a part of the Belderrig quartz assemblage. Section 9.2 introduces the excavations at Belderrig, Co. Mayo and details the Trench Two component of the excavation which is the focus of the present analysis. Section 9.3 describes the geological and palaeoenvironmental context of the area pertinent to the Mesolithic and Neolithic periods. Section 9.4 concerns the analysis of the quartz from Trench Two, beginning with the condition and sources of the quartz in Section 9.4.1; Section 9.4.2 discusses the cores and debitage for all contexts, with the C.203 group covered in Section 9.4.3, C.206 in Section 9.4.4, C.202 in Section 9.4.5, and C.205 and C.215 in Section 9.4.6. Section 9.4.7 outlines the non-quartz component of the assemblage. Section 9.5 concludes the chapter with an overview and discussion of the analysis.
As outlined in Chapter 3, north Mayo has received a wealth of archaeological attention, beginning with the early twentieth century discovery of Neolithic and Bronze Age field systems. A local schoolteacher, Patrick Caulfield, was a key figure in the recognition of the sub-blanket bog walls as being prehistoric, and his early discoveries were elaborated on by his son, Seamas Caulfield, who eventually became an archaeologist at UCD and continued surveying and mapping the field systems (e.g. Caulfield 1983; Caulfield et al. 1998).
The quartz-dominated lithic scatter (Irish Grid Reference F992415) at Belderrig harbour was first noted by Patrick and Seamas Caulfield – the scatter was eroding from the cliff close-by to Seamas Caulfield’s Belderrig home; over the years they collected material as it was eroding out of the cliff face. At the point where the archaeological material – which included organic material such as hazelnuts and fish bone – was being eroded was a section of the cliff which was experiencing substantial erosion, caused by changes in the up-slope drainage regime (Warren 2005b). The Mesolithic character of the quartz lithic scatter was noted by Woodman et al. (1999) and Costa et al. (2005), and Woodman also took organic samples from the erosion scar (Warren 2005b).
Warren inspected the site in 2002 with Caulfield and began fieldwork in 2004 consisting of test pit survey, monitoring of erosion scars coupled with geophysical and topographic survey (Warren 2005b). From this, the 2005 season involved the excavation of a further six test pits, the complete excavation of a c. 4m x 3m trench (Trench Two) located on the surmised concentration of lithics (and organic material identified in the cliff face) close to the erosion scar, a further 1m x 0.5m trench (Trench ‘Cliff’) on the cliff edge itself targeting a visible exposure of fish bone, and the initiation of the excavation of a 30m x 2m trench (Trench One) running upslope starting a few metres from the erosion scar, and attempting to connect the concentration of stone tools with sub-bog structural features including field walls (Warren 2005b). In 2006 another 1m x 0.5m trench (Trench Three) was excavated on the cliff close to Trench Three and Trench ‘Cliff’; excavations continued in Trench One in 2006-2008 with a substantial annex to Trench One also excavated (Plan 1).
Trench One, Trench Two, Trench Three, six test pits and along the cliff edge have shown that a part of the Mesolithic activity at the locale consisted of the construction of stony layers/platforms (Plan 1 and Plan 2), with Mesolithic dates returned for activity in and below these stony layers. Trench One and its annex uncovered Neolithic structural activity in the form of field walls, cairns, and a horseshoe-shaped structure (Warren 2009b). The date range of the excavations shows intermittent activity for over two thousand years, from the Later Mesolithic to the end of the Neolithic (Appendix A- 50).Top of Page
The quartz assemblage from Trench Two (hereafter TR2) has been selected for the following analysis. Graeme Warren suggested that the TR2 assemblage would be a suitable assemblage as it provided a completed trench with excellent spatial and control over the data, including presumably in situ material. The trench was excavated in 50 cm squares, identified through an alphanumerical system (A1, A2, etc.); all spoil was dry sieved through a 5mm mesh (Warren 2005b). Initially, an attempt to record the 3D location and angle of dip of all ≥30 mm laminar artefacts where the angle of rest in the soil could confidently be ascertained was made. However, problems with the recognition of quartz in the field, the non-laminar character of many artefacts and the highly compacted stony layers means that this data is of little use (Warren 2009a pers. comm.) and is excluded from this study, which uses the grid square as the primary spatial information.
There are some difficulties with the TR2 assemblage which should be noted. The first difficulty occurred in 2007, while TR2 artefacts were being washed by student volunteers in the Belderrig Research Centre – the error occurred because students took out and washed lithics from a number of contexts and squares without properly noting where they had derived from – it is thought, however, that the contexts themselves were not mixed up, and the problem concerns the inter-context spatial information. This error was caught quickly, and it was decided that I would take over the job of washing the rest of the artefacts myself at a later stage. A note from Warren concerning this mix up is in Appendix A- 52.
Appendix A- 52 shows that all squares from C.200, the sod layer, were mixed up; for C.202 a third of the rows of squares were mixed up; for C.205 five squares from one row were mixed up. Because of this, it was decided to complete a preliminary catalogue of all the material, and subsequently exclude the mixed up material from full analysis. Consequently, for C.202 a full analysis was completed on 50% of the rows (which provided over 900 artefacts) hence excluding the two mixed up rows and also excluding Row 3, which left the alternate Rows 2, 4, and 6 for full analysis. For C.205, the problem was negligible as only nine artefacts were involved in the mix-up, and were simply excluded from the analysis.
A second difficulty arises from issues concerning contexts; Warren has provided the following comments on the Trench Two contexts.
"The interpretation of Trench Two, which was completely excavated in 2005, is greatly facilitated by the ongoing interpretation of the much larger Trench One, which was also more extensively sampled in order to understand site formation. Notably, in retrospect, some ambiguous features and layers in Trench Two, especially those below the stony layers, were not accorded the attention that they may have deserved, and are not well represented in plans/archives. Inevitably given the reliance on student volunteers minor inconsistencies in recording and recognition of very subtle archaeological features and distinctions are present, and the ongoing programme of post-excavation work is engaging with these problems" (Warren 2009a pers. comm.).
In terms of the present analysis, Warren suggests that for the main layers these problems should not negatively affect the analysis.
TR2 is located at the point of the substantial erosion scar, in-between the cliff and the bank, with the eroded pathway running over the trench (Figure 9-1). Quartz artefacts were visible on the surface of TR2 before excavation, with the thin layer of turf/peat having been eroded away along the thoroughfare. A full account of the stratigraphic sequence for TR2 appears in Warren (2005b). A brief description is provided here, based on the 2005 Belderrig Report (Warren 2005b) and from Warren (2009a pers. comm.). The contexts in TR2 proved to be complex, with the main artefact-bearing layers’ boundaries proving to be difficult to define from each other. A micromorphology analysis was conducted on the cliff face in order to interpret the context of context formation (Guttman 2005) and this will be supplemented by further analysis of micromorphological samples from Trench One (Warren 2009a pers. comm.).
While Trench One and the test pit ‘X’ on the east side of the bank from TR2 had significant depths of peat, TR2 was characterised by comparatively thin peats – as is the entire cliff edge running southwards towards the harbour. This is provisionally interpreted as being a signature of cut-away bog, which respects the bank to the east, forming a post-medieval trackway (Warren 2005b).
C.200: The sod, C.200, which contained numerous artefacts, overlay thin lenses of sand (C.201) and marls (C.207) – both of which were “assumedly sub-peat waterborne or other erosional features” (Warren 2005b) – and peat (C.208).
C.202: A root-penetrated brown clay mineral soil with high stone content, which “was very artefact rich, but has clearly suffered from erosion and is present in variable depths” (Warren 2005b).
C.206: Under C.202 was C.206, a metalled layer which was “much more formal than anything else we have seen in Trench One, and respects the field bank” to the east (Warren 2009a pers. comm.). C.206 was also artefact-rich, and may be a modification of the upper levels of the context below (C.203) – this modification may be related to the surmised post medieval trackway; Warren (2009a pers. comm.) notes that a “possibility is that the true horizon…[of C.206] was missed in this trench, possibly because of [this] later disturbance”.
C.203: C.203, which merges/underlies C.206 “is a compact dark brown clay sand with many stone inclusions, especially at its top: it contains abundant artefacts, many of which appear to be in situ [sic]. Artefacts are found throughout C.203, and were often compacted within the stone inclusions” (Warren 2005b); “[i]nitially interpreted as a sealed land surface, we might also see the upper parts of…[C.203] as part of the stony layers” (Warren 2009a pers. comm.). The micromorphology analysis suggested that C.203 represents a pre-peat, very biologically active woodland topsoil; because of its shallow nature, the context had probably been either compressed, truncated or subject to erosion (Guttman 2005).
C.205: C.205 “is often described as co-eval with 203, but with much less charcoal staining (it appears to be slightly different in date, and is sometimes described as below 203 in the records)” (Warren 2009a pers. comm.).
C.211: Located beneath C.203 and described as a residual peaty layer. In places it was very thin, while where thicker it has been interpreted as being “sometimes in possible features such as irregular stake-holes or hollows or scoops…it is possible that these deposits of 211 were the fill of features, such as pits and irregular hollows, similar to those excavated in Ferriter's Cove…It is also possible that the four small irregular shaped holes found in F6 are stake-holes, although field interpretations suggested that they were stone holes…They were irregular in plan and three of the four were vertical in profile. They ranged in surface diameter between 10-12cm with a max depth of 8cm. These small holes were located to the east of the large irregular pit-like depression” (Warren 2005b).
C.215: Appeared in just one 0.5m² square, underneath C.203, and was described as a sandy peat lens.
C.214: Sealed subsoil C.209 in places – it is interpreted as a possibly heavily leached organic soil horizon.
Figure 9-1 Top: facing east. People congregating on top of erosion scar; Trench Two to immediate right; foreground showing psammite bedrock on top of domed metadolerite intrusions; base of cliff is shattered psammite bedrock in form of large tabular boulders. Bottom left: facing south. Pre-excavation Trench Two in middle ground in-between two volunteers with bare soil and rocks visible. Bottom right: facing north. Mid-excavation Trench Two with eroded pathway leading up to it along cliff and bank to right; Trench One beyond to right
Table 9-1 highlights that the three main artefact-bearing contexts are C.202, C.206, and C.203; all three are dominated by vein quartz artefacts with minimal amounts of rock crystal and non-quartz artefacts. C.203 is interpreted as pre-peat woodland topsoil, which is artefact-rich and subsequently compressed, truncated, or eroded. This context has hazelnut shells dated to c. 4300-4500 cal BC (Appendix A- 50). C.203 has extensive, if irregular stone in its upper layers, where there is an interface with C206, which is the formal stony layer. Correspondent dates from the stony layer in Trench Three place activity in this context at c. 4000-4200 cal BC (Appendix A- 50).
|Context||Vein quartz||Rock crystal||Non-quartz|
|202 (50% sample)||907||13||3|
Table 9-1 BDG TR2 artefact quantities; non-quartz includes seven hammerstones
The metalled layer, C.206, may have been modified in the recent past as part of the cutaway and track outlined above, but may also be in situ. If C.206 has been disturbed then the upper layers of C.203 are also likely disturbed and C.202 also results from later disturbance. Alternatively, if C.206 is unmodified, then by comparison with Trench One, C.202 layer is also in situ and should post-date the main construction of the platforms (Warren 2009a pers. comm.). C.214 is the only context with a substantial proportion of non-quartz artefacts. 29% of the 28 artefacts were chert and flint flakes; C.214 is interpreted as a leached soil horizon and lay under C.203, the woodland soil. For the following analysis, both C.211 and C.214 were grouped with C.203.Top of Page
Along with the hazelnuts mentioned above in relation to the dating, TR 2 produced a limited amount of fragmented fish bone. In order to widen the scope for the analysis of the faunal record, a small section of the immediate cliff edge was excavated, where the fish bone was noted in the exposed cliff face. The analysis suggested that the fish bone was highly fragmentary and poorly preserved; the two identified species were conger eel and ballan wrasse, both of which represent shallow water, rocky shore catches (Parks 2006).Top of Page
The following description of the north Mayo geology is abbreviated from Long et al. (1992). A simplified geology of the north coast of Mayo (Figure 9-2), running from west to east and forward chronologically, begins with the Precambrian Erris Complex which has two main components, the oldest of which is comprised of c. 1900 million year-old granitic rocks metamorphosed to gneiss c. 1000 million years ago and the younger being 950-650 million year old greywackes metamorphosed to schists.
After these are the Dalradian Supergroup, composed of metamorphosed sedimentary rock – quartzite and psammitic rock (both of which are metamorphosed sandstone) – with subordinate volcanic and intrusive rocks which includes vein quartz. The psammite at Belderrig is part of this Supergroup; the metamorphosing of the Dalradian began from c. 750 million years ago. C. 400 million years ago the Caledonian Igneous Suite began to form, with small intrusions on the north coast at Belderrig in the form of dolerite, which subsequently metamorphosed to metadolerite. C. 360 million years ago the Carboniferous Period began with the sandstones, limestones (with consequent cherts), shales, siltstones and mudstones and so forth subsequently occurring, with limestone to hence dominate as 60% of present-day Irish bedrock.
From this simplified geological description, one can see that vein quartz occurs in the immediate vicinity of Belderrig in relation to the psammite and the metadolerite, as well as further to the west and southwest, with the extensive bog coverage today masking possible sources of vein quartz away from the coast. The largest noted quartz vein is located on Achill Island – which is dominated by quartzites and psammites – southwest of Belderrig where a vein estimated at about 85,000 metric tons of high purity quartz has been mined, primarily for architectural chippings (Long et al. 1992, 34). To the east of Belderrig, the series of carboniferous rocks dominate, producing the sandstones, mudstones, siltstones, and chert-bearing limestones.
Figure 9-2 North Mayo bedrock geology; data adapted from GSI (2005a; 2005b)Top of Page
The quaternary deposits in north Mayo were principally laid down during the last glacial retreat, with Long et al. (1992, 31) noting that the movement of the ice in north Mayo “was one of the most complex anywhere in Ireland”. There is currently no data concerning the Mayo coast Holocene sea levels (see Brooks and Edwards 2006). The consequent post-glacial mineral soil is comprised of glacial till – clays, silts, muds, sands and gravels – as well as medium and large glacial erratics. Today, the landscape of Belderrig is practically bereft of trees, with heather-covered blanket bog giving way to rough pasture on blanket bog.
During the Mesolithic the region was forested, consisting of full, mixed woodland on mineral soil, dominated by pine along with oak, elm, birch, and hazel as canopy trees (O'Connell and Molloy 2001) (Appendix A- 51). The forest would also have been interspersed with blanket bog; peat from a low ridge beside this project’s source of Rose Cottage quartz in the Belderrig valley dates to 4358-4767 cal BC; in west Mayo a pine on 80cm of peat dates to 6124- 6597 cal BC with the peat underneath the pine dating to 7517- 8206 cal BC (Caulfield et al. 1998, 637-8). The pollen diagram in Appendix A- 51 from the Céide fields, a few km to the east of Belderrig, shows signs of pre-Neolithic basin peat and woodland firing, with the authors commenting that it is not possible to ascertain whether this was a deliberate or natural forest fire(s) (O'Connell and Molloy 2001, 103-4). A well-defined Neolithic landnam began a century after the initiation of the Elm decline, with the latter taken to be c. 3890 cal BC; the landnam was characterised as “substantial-to-widespread woodland clearance and intensive pastoral farming” which continued for c. six centuries followed by a lull in farming evidence for a further c. six centuries, after which low level activity once again is apparent (O'Connell and Molloy 2001, 116).Top of Page
The following analysis primarily concerns the vein quartz component of the TR2 assemblage, therefore where the term ‘artefact’ is used this implies vein quartz unless stated that other materials are included. All of the quartz from TR2 appears in a fresh condition, with sharp edges. Pieces with sub-rounded and rounded edges – generally smaller fragments – were regarded as natural quartz and excluded from further analysis. One difficulty noted with the analysis of the vein quartz artefacts was ascribing a time period to some of the fractures. Initially, a distinction was made regarding the staining on the artefacts, whereby areas that were heavily stained were deemed to be old surfaces. During the accidental breakage of a number of artefacts during washing and measuring, it became apparent, however, that macrofractures in the artefacts allowed fluid to penetrate thereby leaving a residue and creating stains on the interior of the artefacts. When these artefacts broke (inevitably) along the macrofractures, the fresh breaks were therefore stained, giving the appearance of older breaks. Consequently, the use of staining to identify old breaks is irrelevant. Indeed, a possibility may be that fresh breaks, or breaks created at some stage after deposition in soil may in fact be identified by heavily stained surfaces, as the residue deposited by the fluid in the macrofractures may stain the quartz more thoroughly than a surface open to the soil.
Following from this, a related issue concerns to what extent the development of fractures in the artefacts has occurred since deposition. During knapping, numerous incipient fractures in the quartz were noted, and on returning to the experimental assemblage on a number of occasions over the course of the analysis of the assemblages, many smaller, and some larger, artefacts had fragmented further even though they were individually bagged and handled carefully. Therefore, it is likely that such post-depositional fragmentation occurred with the archaeological material even without extensive or moderate trampling/disturbance and so forth. It is of course impossible to quantify this without further experimentation with trampling experimental assemblages in various degrees of burial in soil.
Only 11 artefacts (0.2% of total) were noted as burnt (n = 1) or possibly burnt (n = 10). All were debitage and from all contexts except C.206; five were from C.203 and came from three adjoining squares; the two from C.202 were also from adjoining squares, while the two from C.211 were from the same square; C.214 and C.215 had one each. The burning experiment discussed in Chapter 8 showed that changes in hue are apparent in burnt quartz, with these changes apparent on larger pieces and where the original hue of the quartz is known. Therefore, while a small number of burnt artefacts may have been missed, it appears that a very low proportion of the assemblage was burnt. Indeed, of the 11, 10 were described as possibly burnt and therefore may represent an over-count of actual burnt artefacts.
5275 vein quartz and rock crystal artefacts were analysed from the five context groups (Table 9-2). A small number of artefacts were categorised as either indeterminate or as possible artefacts. The vein quartz was categorised into source materials where possible. The source categories used were quartz-beach, quartz-psammite, and quartz-metadolerite – uncategorised quartz was labelled as quartz. Only 6% (n=222) of the ≥10mm vein quartz artefacts could be identified definitively to a source; 3.3% to quartz-beach, 2.6% to quartz-psammite, and 0.2% to quartz-metadolerite. For the cores, 10.9% were identified as quartz-beach, 9.1% as quartz-psammite, with the remainder as generic quartz and no quartz-metadolerite cores identified. 30 pieces of rock crystal debitage were also in the assemblage, over half of which was debris, with no crystal cores. It is uncertain where the rock crystal was sourced – today, numerous small rock crystals can be found in the spoil heap associated with a copper mine across the bay.
Table 9-2 BDG vein quartz and rock crystal artefacts analysed
The preponderance of non-cortical flakes, and the low frequency of artefacts retaining parent rock bedding planes, precluded assigning most of the quartz to a source. Figure 9-3 compares the Belderrig (hereafter BDG) flakes to the Experiment flakes for cortex proportions. While just 10% of the BDG flakes were cortical, cortical flakes accounted for 42% of the Experiment flakes; 1% of the BDG flakes had 100% cortex, while 5% of the Experiment flakes had 100% cortex. Figure 9-4 compares the cortex proportions on the BDG and Experiment complete platform cores; 39% of the BDG cores and 76% of the Experiment cores retained some cortex. This low proportion of cortical artefacts suggests that the initial knapping of cobbles/blocks began away from the excavated trench area, with cores being brought to that area once decortified to some extent, as well as having bedding planes removed. The smaller proportion of cobbles (meaning beach cobbles) suggests the quarrying of quartz veins; these therefore are covered by less cortex and/or bedding planes to begin with, creating a lower proportion of cortical artefacts.
Appendix A- 53 provides the cortex proportions for artefacts by context. Analysis was conducted using Multinomial Logistic Regression, using the C.203 group as the reference category. The C.203 group had more cortical artefacts than the other two main artefact-bearing contexts (C.202 and C.206) with the difference significant (χ² = 59.364; df 12; p = 0.000). C.205 also had more cortical artefacts than C.202 and C.206; when analysed excluding C.203 (using C.215 as the reference category), the difference between these three contexts was significant, but weaker (χ² = 10.857; df 3; p = 0.013). This statistical analysis must be interpreted with an appreciation of probable intermixing of artefacts within the context boundaries mentioned previously.
Figure 9-3 BDG and Experiment flakes: cortex
Figure 9-4 BDG and Experiment complete platform cores: cortex
The visual characteristics of the >10mm materials were noted for grain, opacity, hue, appearance, as well as inclusions, bedding, staining, and heat alteration. Overall, 83% of the debitage was smooth-grained, with sugar-grained at just 2%; there were no sugar-grained cores, suggesting a clear preference for smooth-grained material. Table 9-3 describes the debitage, for the category groups that accounted for >1% of the total; these six category groups cover 82% of the total. The largest group consisted of semi-translucent, smooth-grained, cloudy glass appearance, with a cloudy hue, followed by a group consisting of semi-translucent, smooth-grained, and vitreous, with a metallic grey hue. Almost all the rest were slight variations on the six main groupings. Indeed, the devised categories are not separated sharply, but rather are gradations of each other. The clear preference throughout the contexts was for smooth-grained quartz, but C.215 had the smallest proportion of smooth-grained material, followed by the C.203 group (Table 9-4).
A small number (six in total including one core) were of distinctly different quartz types, and unlike any quartz noted around Belderrig. It may well be that these, as well as other quartz in the assemblage, may represent a use and/or deposition of non-local quartz in the excavated area. These six artefacts of different quartz came from contexts 203, 205, and 206; the two from C.206 are flake conjoins.
|Hue||Appearance||Grain||Opacity||Total||% total debitage|
Table 9-3 BDG >10mm debitage: Appearance categories that accounted for >1% of total
Table 9-4 BDG >10mm debitage: grain by context
It appears, therefore, that quartz sourced from metadolerite played a minor role in the material repertoire, with no quartz-metadolerite cores present. However, the question remains as to what extent the 94% of the debitage and 80% of cores that were uncategorised were sourced from metadolerite. It is also possible that the metadolerite sources that are today right beside the cliff (Figure 9-1) may have been covered by land in early prehistory, with subsequent erosion exposing them on the foreshore. Interestingly, the outcrops available today in the immediate locality are dominated by the smooth-grained metadolerite and the sugar-grained psammite. However, there was little evidence for the metadolerite quartz veins being utilised even though the assemblage is dominated by smooth-grained quartz. The psammite quartz veins today are predominantly sugar-grained, but the quartz-psammite identified in the assemblage is equally split between smooth-, sugar/smooth-, and sugar-grained quartz. This suggests that veins of differently grained quartz were available and utilised in prehistory, and/or that the use of the metadolerite quartz veins is masked by the lack of cortical and bedding plane signatures on the artefacts.Top of Page
Overall, cores accounted for just 1% of the assemblage, with a further three artefacts categorised as core/debitage. Table 9-5 and Appendix A- 54 provide a breakdown of core types; including possible cores, 80% are platform cores (Appendix B- 14, Appendix B- 15, Appendix B- 16, Appendix B- 17), 13% are bipolar (Appendix B- 18), with remaining being conical pieces (Appendix B- 18) and radially split pieces; 71% are complete cores. Out of the definite platform cores, multiplatform types dominate at 70% of the assemblage. Five of the platform cores had impact points suggesting inelastic support, i.e. a platform-on-anvil technique – all five were from C.203. A core of distinctly different quartz came from C.203; it was a small dual, opposed core with no apparent similar debitage material noted. The only other example of a dual, opposed core from C.203 contained a bedding plane of indeterminate parent rock. The third and final dual, opposed core in the assemblage came from C.215; it was quartz-psammite, and was over five times heavier than the smaller example from C.203, and almost three times heavier than the larger C.203 example.
|Possible conical piece||Fragment||1||-||-||-||-||1||1.6|
|Grand Total||Grand Total||14||6||37||3||3||61||100.0|
Table 9-5 BDG Core types; including possible cores
Looking at the vein quartz debitage (Table 9-6), 31% of the assemblage is <10mm debitage, and flakes account for 40% of the total; 5% of the flakes are complete. (As noted in Chapter 5, due to the difficulties in identifying sequential breaks and sequential fragments consistently in the archaeological material, these categories were excluded from analysis.) Only three bipolar flakes were identified. For the analysis of the archaeological assemblages, for the categorising of flakes the definitive bipolar flakes were labelled ‘bipolar’ while platform flakes and flakes in general were labelled ‘platform’, and no third category of flake type was used. While this may deflate the proportion of bipolar debitage, the fact that just three bipolar flakes were noted in the assemblage signals strongly that the lack of bipolar debitage is more real than apparent.
Table 9-6 BDG Quartz debitage
Appendix B- 19, Appendix B- 20, Appendix B- 21, and Appendix B- 22 are examples of complete and flake fragments, both retouched and non-retouched. Figure 9-5 provides the max. length of the assemblage’s complete and fragmented flakes and complete flakes only; over half of the flakes are <20mm, with a mean length of 21.7mm. (This accounts for length only; some flake fragments have a greater width than length.) For the complete flakes only, a third of the flakes are <20mm in length, with a mean length of 27.8mm. All but two of the 107 complete flakes are <60mm in length. For both flakes and debris, 80% of the assemblage is <20mm and 8% is >30mm in length.
Figure 9-5 BDG flakes: max. length histogram for complete and fragments. Inset: complete flakes only
Appendix A- 55 provides the flake attributes of bulb, platform occurrence and type, termination, and curvature for the appropriate flake categories. 33% of the ≥10mm debitage had complete or partial platforms, which is approximately the same proportion as the Experiment dataset (see Table 6-6). Overall, just 1.8% of complete and proximal flakes had bulbs, with this ranging from 0% to 3.6% between contexts. 38.6% had complete platforms with an inter-context range of 27-50%; 42% were platform fragments, while 19.4% had collapsed platforms – the latter ranged from 6.3-45.9% between contexts. A third of the ≥20mm length flakes were straight, ranging in contexts between 24.1-42.9%. 82% of the distal present flakes had feather terminations, 7.2% irregular, 6.2% plunging, 3.6% step, with remainder being fragmented or retouched/wear marks.
The cores and debitage do not provide any evidence of platform preparation such as faceting or abrasion in the chaîne opératoire from any contexts. However, during the experimental knapping the cores were tapped during knapping in order to loosen incipient fractures created from previous strikes, and the cores were sometimes trimmed in order to create suitable striking angles – the core tapping produced debitage that was indistinguishable from otherwise ‘normal’ flakes and the core trimming was not apparent as obvious trimming and also appeared as ‘normal’ flakes and flake scars. While no platform preparation was noted, on one single platform core it was clear that the striking platform was set up on a veinlet in the quartz, presumably to make use of the existent fracture line in the block (Appendix B- 14).
A small number of diagnostic types were identified. The definition of diagnostic types in the assemblage includes types such butt trimmed flakes and artefacts exhibiting retouch/wear marks. The term retouch/wear mark is used in order to include both possible types of modification in the identification without differentiating them. The shorthand of ‘retouch’ is used in the text. Appendix A- 56 lists, by context, the vein quartz diagnostic types, their delineation, and their angle. The majority came from the C.203 group, including the one example of a butt trimmed flake and one borer; a point/notched piece came from C.202. Along with the 30 retouched artefacts were 23 with possible retouch. For the former, the majority had convex or rectilinear retouch, with the majority of both categories having an abrupt retouch angle; for the latter, almost half were convex, with abrupt or semi-abrupt retouch. While these 23 possible retouched artefacts are probably actual retouch/wear marks, a conservative approach was taken in this regard in order to avoid an inflation of the retouched numbers as is a danger with vein quartz. A more liberal approach could include them as retouched. It is unclear how to interpret the general sense of size clustering for the possible retouched seen in Figure 9-6; whether this is fortuitous or related to difficulty in identification of retouch on smaller artefacts is uncertain – while the possible retouched are generally smaller than the retouched, they are not all the smallest, and some are also larger.
Appendix A- 57 compares the means for the various metrics of the diagnostic types and the non-retouched debitage, showing that the diagnostic artefacts are larger in all dimensions and weight, with the possible retouched artefacts falling in between the retouched and non-retouched groups. The diagnostic types were analysed with GLM. Using the metrics’ log transformations, the difference in mean size and weight between the retouched and non-retouched artefacts was significant (Appendix A- 58); the Bonferroni post hoc test indicates that the difference between the retouched and non-retouched was significant for all metrics; comparing the retouched and possible retouched, there was no significant difference in the means, while comparing the possible retouched and the non-retouched the only significant difference was between the mean length of the artefacts – the possible retouched artefacts’ means fall in between the other two groups (Appendix A- 59).
Figure 9-6 BDG vein quartz diagnostic type: length/width ratio
Figure 9-6 shows the general clustering of the possible retouched artefacts. The analysis was run a second time, excluding artefacts <20mm in length (Appendix A- 60). Again, the difference between groups was significant, with the diagnostic types heavier and larger in all dimensions (Appendix A- 61). The Bonferroni post hoc test indicates that the difference between the retouched and the possible retouched/non-retouched was significant for all metrics, and no significant difference between the possible retouched and non-retouched for all metrics (Appendix A- 62). A third test was run, using ANOVA, with the retouched and possible retouched grouped together, examining the ≥20mm length artefacts. Compared to the non-retouched artefacts, the difference with this group’s means was significant for all metrics except for thickness, with the non-retouched artefacts being on average smaller and lighter (Appendix A- 63).
As noted in Section 9.2.2, no clear horizons were detected between the three main artefact-bearing contexts, and it is possible that C.206 and C.202 were disturbed contexts, related to a post medieval trackway. It is difficult to determine from the assemblage as to whether this is the case; the three main artefact-bearing contexts’ artefacts do not appear dissimilar, and what exact effect a disturbance of cutting away the bog and forming a trackway would have on the assemblage in terms of fragment size is a moot point – presumably this would create a fragmented layer on top, but conversely add larger fragments to form a trackway. From an examination of the fragment sizes, it does not appear that the upper context, C.202, has a significantly greater degree of fragmentation, or smaller fragments than the lower contexts. Figure 9-7 presents the proportions of fragment groups for the three main artefact-bearing contexts, with the fragment groups arranged with the groups with greater mean length at the top and mean length decreasing per group going down the chart.
Figure 9-7 BDG C.202, C.206, and C.203 group debitage fragment groups; fragment groups with greater mean length towards top
This graph shows that while C.202 has a greater proportion of debris, and especially <10mm debris, which may imply a fragmenting into indistinguishable fragments, the clearest difference between the three contexts is with the high proportion of smaller fragments from the middle context, C.206. While a certain degree of intermixing may have occurred during excavation, thus distorting the actual size pattern, from the data available it does not suggest that the upper level, C.202, is fragmented into smaller fragments to a greater degree – while C.202 did have proportionally more debris than C.206, and more ≥20mm debris, it also had proportionally more complete and lateral fragments than C.206 – these lateral fragments are less likely to occur from post-depositional breakage than from knapping breaks and the retention of proportionally more complete flakes does not suggest more disturbance for C.202 than for C.206.Top of Page
The C.203 group (which groups C.203, 211, and 214) accounts for almost half of the vein quartz assemblage, over a third of the rock crystal assemblage, and over half of the non-quartz assemblage including four out of the seven hammerstones (Table 9-7 and Figure 9-8). The four hammerstones had impact marks, but none had evidence of sustained or heavy usage. The artefacts were concentrated in the centre and towards the northeast of the trench, forming an area of 3.5m² out of the 12.25m².
Table 9-7 Context 203 group artefacts by material
In order to ascertain any spatial differential in the material, the trench was divided into three areas based on the total quartz artefact count – the concentration (28% of the squares), adjacent (21% of the squares), and periphery. 55% of the assemblage came from the concentration and 22% came from the adjacent squares; therefore 23% came from the 51% of the squares forming the periphery. Figure 9-8 highlights that just under half of the cores was found in squares outside of the concentration and also the heavier debitage was for the most part also found outside the concentration; both the cores and heavier debitage were mostly found in the adjacent area. Analysing with ANOVA, the mean weight for flakes was significantly different between the three areas (p = 0.007), with the mean flake weight for the adjacent area over twice that of flakes from the periphery and 1.4 times the weight of flakes from the concentration (Table 9-8).
Table 9-8 BDG C.203 group flakes: mean weight by area
While the periphery generally had above average quantities of debris compared to both the other areas, the adjacent area had a substantially above average proportion of <5≥1mm debris. In terms of diagnostic types, there are a butt trimmed flake (Appendix B- 22), a borer (Appendix B- 21), 17 retouched (Appendix B- 20), and 14 possible retouched artefacts. An above average proportion of these came from the concentration; excluding the possible retouched artefacts, 79% (n=19) came from the concentration, with just one from the periphery.
While no extensive refit/conjoin programme was undertaken during the analysis, possible refits and conjoins were looked out for and noted when present; therefore, the amounts of refit/conjoins noted will invariably be less than those possible within the assemblages. In terms of refits and conjoins, there were eight sets, with one considered a fresh break. One core/flake refit pairing was noted from Square A4. Five conjoin pairings were noted, with the respective conjoins coming from the same squares – A1, E4, E5, and H1. One conjoin pair came from different squares – D3 and E1; this pairing was of a distal and mesial fragment. It is difficult to identify if the flake conjoins originate from breakages during knapping or from ‘old’ post-depositional fracturing.
Figure 9-8 BDG C.203 group. Top left: non-quartz count. Top right: average debitage weight per square for debitage >2g. Bottom left: count of all quartz and rock crystal artefacts . Bottom right: count of debris <10=5mm with cores indicated by count. Top left square = A1, bottom right = H6Top of Page
This context group contained 60% of the assemblage’s quartz cores, and was dominated by multiplatform cores, with single platform cores at 21% (Appendix A- 54 and Appendix A- 65). Along with the 33 cores were three possible cores and one possible core fragment. There were two bipolar core fragments, one conical piece, and one radially split core. 15% of the cores were quartz-beach, 9% were quartz-psammite, with the remainder being uncategorised quartz. The cores were on blocks, cobbles, as well as three core-on-flakes (Appendix A- 65). Five cores – three multiplatform and two single platform – have distal impact marks, suggesting a platform-on-anvil technique was used. For the complete cores, the cores on beach cobbles were generally the largest, as were the multiplatform cores (Figure 9-9 and Appendix A- 66; see also Figure 9-16). As mentioned in the previous section, two of the three dual, opposed cores from the assemblage came from C.203. Both were small cores – their mean weight made them the smallest core type –and one had bedding from an indeterminate source, while the other was a cobble of a distinctly different material than the rest of the assemblage with no apparent resultant debitage.
Figure 9-9 BDG C.203 group complete coresTop of Page
29% of the vein quartz debitage is <10mm (Table 9-6); out of the three main artefact-bearing contexts, the C.203 group contained the smallest proportion of <10mm debitage, while containing the greatest proportion of cores. Two of the assemblage’s three bipolar flakes came from the C.203 group. Appendix A- 67 and Table 9-9 provide the fragment types. 3.8% of the flakes were complete, with proximal flakes accounting for the largest proportion after debris. Both the proportions of debris and mesial fragments are considerably less than the experimental datasets – in Section 126.96.36.199 it was noted that mesial fragments were more likely to be interpreted as debris. The low frequency of both debris and mesial fragments in the C.203 group may suggest a number of possibilities:
1. A greater expertise of the Belderrig knappers with the material may have resulted in less fragmented debitage.
2. Different method/techniques used created less mesial/debris fragments.
3. The vein quartz used may fracture slightly differently.
4. The C.203 group does not represent a ‘complete’ knapping floor debitage assemblage, as the 20% sample of the experimental assemblage did – certain fragment types may have been brought in and others taken away from the excavated area.
|(P) Fragment||BDG 203 group||H.H.D.||S.H.D.||Total||BDG 203 group||H.H.D.||S.H.D.||Total|
Table 9-9 (P) fragment types and debris, excluding <10mm debris: BDG and Experiment direct percussion
When all of the ≥5mm debitage is included in the tally, however, the proportions of flakes between the datasets becomes closer, with a greater proportion of <10≥5mm debitage in the C.203 group (Table 9-10). The clearest difference between the datasets is the low proportion of <20≥10mm slivers in the C.203 group, with almost three times less compared to the experimental assemblage. It is unclear how to interpret this lack of thin debris (sliver = debris <3mm max. thickness). In terms of the four possibilities listed above, all may be valid, but the similarities between the smallest size grade may go against the first two possibilities – greater expertise or differing method/technique – to some extent. A possible interpretation is that some of the slivers have been subsequently fragmented, thus adding them to the smaller size grade of slivers – this could account for some of the difference including the slight increase in the smallest sliver size grade in the C.203 group, but not all of it. Following the fourth possibility listed, it is also possible that these slivers were taken away from the excavated area, along with other thin mesial fragments for subsequent use.
Table 9-10 Debitage: BDG 203 group and Experiment dataset (P) debitage
Looking at the mean thickness of the various flake fragment types, there appears to be no difference between the datasets, with the C.203 group similar to the experimental data, especially the S.H.D. component (Appendix A- 68); the greatest difference between the means, albeit slight, is for the mesial fragments which have a greater mean thickness for the C.203 group; when only flakes with a <5mm max. thickness are tallied, the C.203 group’s mean thickness is slightly less than the experimental dataset (Table 9-11). This suggests that while the C.203 group mesial fragments are generally thicker, for the <5mm thickness examples, they are fewer and generally thinner than the experimental data. Consequently, this may imply that a similar range of thin mesial fragments were in fact produced during the knapping, but, like the slivers, thinner mesial fragments may have been selected out and removed from the excavated area for use. Additionally, as noted in the previous section at least one of the cores had no apparent debitage present, and there were very few bipolar debitage noted to go along with the bipolar cores – hence there is a mismatch between cores and debitage.
|BDG 203 group||3.55||32||0.68||0.46||2.9|
Table 9-11 BDG C.203 and Experiment mesial flakes max. thickness <5mm
For the context’s break types, a third of the flakes had no longitudinal break; of these fragments, the majority of the transverse breaks were uneven. Almost half of the flakes had uneven longitudinal breaks, with 17% siret breaks (Appendix A- 69). The C.203 group was compared to the experimental assemblage’s proportions of siret breaks using GZLM. Figure 9-10 and Appendix A- 70 compare the C.203 group debitage with the experimental assemblage, highlighting that the C.203 group’s siret proportions are aligned with the S.H.D. assemblage, falling in between the S.H.D. Elastic and S.H.D. Inelastic datasets.
Figure 9-10 BDG C.203 group and Experiment siret breaks
The first statistical analysis looked at technique, using the C.203 group as the reference category. For the techniques, the difference in siret break proportions was significant (Table 9-12), with the difference between the C.203 group and the S.H.D. assemblage not significant (χ² = 0.288; p = 0.592). The second analysis included the support; the difference in siret proportions was significant (Table 9-13). Appendix A- 71 provides the parameter estimates: compared to S.H.D. Elastic, the difference was not significant; compared to S.H.D. Inelastic, the difference was significant, albeit weaker than the significant differences between H.H.D. Elastic and H.H.D. Inelastic.Tests of Model Effects
Table 9-12 GZLM. Dependent Variable: Siret/non-siret Model:
(Intercept), Dataset/technique. Reference category: C.203 group
Table 9-13 GZLM. Dependent Variable: Siret/non-siret Model:
(Intercept), Dataset/technique/support. Reference category: C.203 group
During the analysis of the archaeological assemblage, it was noted that the identification of siret breaks on non-proximal flake fragments appeared to be less than the experimental dataset. This difference in the datasets appeared to be as a result of the distal or mesial siret break fragments being harder on the archaeological samples to identify without the aid of platforms. Consequently, it was likely that the occurrence of siret breaks would be undercounted in the archaeological assemblage compared to the experimental assemblage, with ‘actual’ siret breaks without platforms being described as non-siret breaks. Table 9-14 highlights that the exclusion of non-proximal fragments was indeed greater for the experimental assemblage. In order to limit this bias, non-proximal fragments were excluded from the second round of analysis on the siret break proportions.
|Siret breaks||C.203 group||H.H.D.||S.H.D.|
|Proximal and non-proximal||151||175||89|
Table 9-14 BDG C.203 group and Experiment siret breaks: all and excluding non-proximal
Appendix A- 72 and Figure 9-11 give the siret break proportions excluding non-proximal fragments, highlighting that the exclusion of non-proximal fragments changes the proportion of siret to non-siret breaks greatly. Nevertheless, for the experimental dataset the difference in siret proportions between technique/supports remained significant (Table 9-15). In the analysis with the archaeological data, the difference in siret break occurrence was significant when analysed for technique alone (Table 9‑16), with the difference between the C.203 group and the S.H.D. assemblage not significant (χ² = 1.356; p = 0.244). For technique/support, the difference was also significant (Table 9-17); Appendix A- 73 provides the parameter estimates: the difference between the C.203 group and S.H.D. Inelastic and was not significant; compared to S.H.D. Elastic the difference was significant, albeit weaker than the significant difference with H.H.D. Elastic and H.H.D Inelastic.
Figure 9-11 BDG C.203 group and Experiment siret breaks: excluding non-proximal fragmentTests of Model Effects
Table 9-15 GZLM. Experimental dataset; excluding non-proximal flakes.
Dependent Variable: Siret/non-siret Model: (Intercept), Technique. Reference category: S.H.D.
Table 9-16 GZLM. Experimental and BDG dataset; excluding non-proximal flakes.
Dependent Variable: Siret/non-siret Model: (Intercept), Dataset/technique. Reference category: C. 203 group
Table 9-17 GZLM. Experimental and Belderrig dataset; excluding non-proximal flakes.
Dependent Variable: Siret/non-siret Model: (Intercept), Dataset/technique/support. Reference category: C. 203 group
Therefore, excluding the non-proximal flakes produced broadly the same results in terms of the difference between the experimental dataset’s soft and hard hammer assemblages and the C.203 group. For the S.H.D. Elastic and Inelastic, however, the differing proportions were reversed. In the first analysis the difference in siret proportions with Elastic was not significant: in the second analysis the difference with Inelastic was not significant. This suggests that in terms of siret proportions the C.203 group is similar to the S.H.D. assemblage, and especially the S.H.D. Inelastic assemblage – as noted in the previous section a quarter of the context group’s platform cores had evidence for inelastic support.
The similarity to the S.H.D. assemblage in terms of the proportion of siret breaks must be seen in light of the list of possibilities mentioned above. The proportions of siret breaks may be altered by the removal or addition of debitage in the area. The expertise, along with the method/technique employed with a hard hammer, may have resulted more closely with what was achieved with a soft hammer during these experiments, i.e. thicker flakes less likely to form a siret break. Appendix A- 68 highlights that the mean thickness for complete flakes was similar for S.H.D. and the C.203 group. However, as noted in Chapter 6, the difference in thickness between H.H.D. and S.H.D. complete flakes was not an appropriate indicator of technique.
Figure 9-12 compares the mean length, width, thickness and platform width and thickness for the complete flakes in the datasets, highlighting that the C.203 flakes were on average shorter and narrower than, but as thick as, the S.H.D. flakes, and had similar platform dimensions. The length/width ratio for complete flakes fell primarily between 2:1 and 1:1 with the mean length of flakes 29mm (Figure 9-13). Figure 9-14 provides the length by fragment group, highlighting that numerous non-complete flakes have a greater length than the average complete flakes, and that the majority of the longest flakes are proximal flakes (these long proximal flakes are distal missing flakes).
The experimental knapping also showed that the soft hammer component was less likely to exhibit the strike’s impact mark on the platform, with 79% of the complete soft hammer flakes having impact marks compared with 96% of the hard hammer direct and bipolar flakes. The C.203 group’s complete flakes had impact marks in 62.9% (n=39) of the cases, which also suggests the use of a soft hammer.
Figure 9-12 BDG C.203 group and Experiment complete flakes: means
Figure 9-13 BDG C.203 group complete flakes: length/width ratio
Figure 9-14 Boxplot. BDG C.203 group: median length for flakes by fragment groupTop of Page
The C.203 group consisted of three contexts – C.203, C.211, and C.214 – with the main artefact-bearing context, C.203, formed on a very biologically active woodland topsoil, which had been compressed, truncated or subjected to erosion. While today the excavated area is on the cliff edge, during the time period under scrutiny the wooded area would presumably been at some remove from the cliff edge, close to the mouth of the small river. C.211 was beneath C.203 and may represent the fill of various features, while C.214 is possibly a heavily leached organic soils horizon sealing the subsoil. As with all the contexts, the clear majority of the quartz artefacts from the C.203 group are undifferentiated by source due to a lack of cortex and bedding planes with which to identify them. This context group, however, produced the significantly greatest proportion of cortical artefacts, albeit just slightly more than C.205. This suggests a pattern of preferential collecting of material by the communities of practice from quarried veins instead of cobbles, and possibly that cobbles were initially knapped away from the excavated area and brought decortified to some extent, if not almost completely. As well as the quartz cobbles, a number of psammite cobbles were collected for use as hammerstones. The cores deposited in the excavated area derived principally from blocks with five from cobbles. Three of the cores were core-on-flakes; whether this reuse of large flakes as cores was a deliberate strategy or a fortuitous use of the material is unclear, but only a limited amount of core-on-flakes was noted throughout the assemblage. However, the strategy of core-on-flakes may have been more prevalent but not so clearly identifiable on the resultant cores because further reduction of the core-on-flakes could remove visible evidence of the original flake. The majority of the cores were multiplatform with a large range in weight and dimensions; the smaller quantity of single platform cores had similar dimension means but a lesser range than the multiplatform. The cores show evidence for both freehand direct percussion and direct percussion with an anvil.
While these platform cores generally appear to match the debitage – including the clear match of the core and flake refit from Square A4 – the two dual, opposed cores present a different pattern of deposition in the excavated area. Just three dual, opposed cores came from the assemblage, and the two from C.203 were small, thin cores one of which was a distinctly different quartz type derived from a cobble with no matching debitage, while the other non-cobble core had an indeterminate bedding plane attached. This suggests that as well as the excavated area representing a knapping floor – defined by the extensive deposition of large and small debitage and the cores – these two cores may represent an importation of cores by the communities of practice which were not further reduced in situ but instead deposited; both were knapped with a different strategy than the others, and on different quartz than available locally. Additionally, there appears to be a low proportion of cores from this context group for the amount of debitage present, which may imply that not all the cores were deposited in situ. Moreover, the non-quartz assemblage from the C.203 group contained cores without matching debitage, again suggesting the deposition of cores in the excavated area but not in situ knapping.
The minor amount of bipolar artefacts does not suggest the pervasive use of a bipolar technology by the communities of practice, but rather, may represent a less structured use of a bipolar technique, especially with just two bipolar flakes and two bipolar core fragments noted. Additionally, the conical piece and the radially split core may represent a less structured use of bipolar knapping, or indeed may be by-products of direct percussion, as noted by Knutsson (1988a). While the conical piece and radially split core were found away from the bipolar cores, the one bipolar core fragment from Square D3 was found with a bipolar flake in the same square, with the other bipolar core fragment from F5 found with the other bipolar flake in the same square; for the latter, both came from C.211 which is interpreted as a residual peaty layer beneath C.203. Whether this represents in situ knapping or the deposition of material knapped elsewhere is unclear, but these two pairs of artefacts clearly represent just a fraction of a bipolar knapping event which will result in numerous cores and debitage – especially considering that the cores are both fragments.
Comparisons to the experimental assemblage suggest that the communities of practice possibly used soft hammers for knapping, and the four psammite hammerstones from the context group only showed evidence for minor use. If psammite, or other stone, impactors had been used more extensively by the communities, the impactors were not deposited in the area. The cores and debitage do not provide any evidence of platform preparation. The minor amount of complete flakes present consisted of short, narrow, thick flakes, and many of the flake fragments were of similar dimensions to the complete flakes.
It is unclear as to what time range the deposition of artefacts the C.203 group represents, and it could be anything from one generation to a plethora. The spatial distribution shows that over half of the artefacts came from just over a quarter of the squares, and over a fifth came from a fifth of the squares which were adjacent to the main concentration. This adjacent area contained most of the cores and heavier debitage, possibly suggesting the working and deposition by the communities of practice of cores and heavy debitage in the adjacent area – which also had a substantially above average proportion of <5≥1mm debris – or a clearance of the larger and heavier artefacts from the concentration squares into the adjacent ones. An above average proportion of the diagnostic types came from the concentration, including the one example from the assemblage of a quartz butt trimmed flake. Apart from the borer and butt trimmed flake, the remaining diagnostic types were all retouched flakes, with the retouched flakes’ mean dimensions significantly larger than the unmodified flakes. It is difficult to discern if the retouched artefacts represent the reuse of non-retouched tools, or if they represent a different kind of tool which necessitated retouch from the outset. An apparent lack of small slivers and mesial fragments may suggest a removal of such artefacts from the excavated area for use elsewhere by the communities of practice. Without the aid of use wear analysis, it is unclear to what extent the non-retouched artefacts were in fact used for varying tasks.Top of Page
C.206 accounts for a third of the vein quartz assemblage, over a quarter of the rock crystal, and over a quarter of the non-quartz assemblage including three of the seven hammerstones (Table 9-18). The artefacts were mainly located in the top half (northeast) of the trench, with a couple of empty squares (Figure 9-15). Unlike the previous context group, there is less of a distinct concentration/adjacent/periphery divide, with a more even spread of material across the top half (northeast), and the lower quarter (southwest) having substantially less – the three platform cores came from low frequency squares in the top right, while the retouched and possible retouched debitage came mainly from the top half.
While C.206 contained less artefacts than the C.203 group, it contained 13 conjoin sets, one of which was considered a fresh break and two of which were possible fresh breaks. All were from the same squares as their respective conjoins and the other 10 sets consisted of eight pairings and two sets of three. One of the three conjoined, from Square B1 consisted of three pieces of debris that conjoined yet were still indeterminate as flakes, presumably because they were mesial or sequential fragments. The second set of conjoined three came from Square C3, with another conjoin set from the same square. The remaining sets came from B2, B3, two from B4, B5, E5, and G1.
Table 9-18 BDG C.206 artefacts by material
Figure 9-15 BDG C.206 distribution. Top left: diagnostic types and non-quartz artefacts. Top right: average debitage weight for debitage >2g. Bottom left: count of quartz and rock crystal artefacts. Bottom right: count of debitage <10=5mm; count of cores, with non-platform cores highlighted. Same square numbering as previousTop of Page
Compared to the C.203 group, C.206 had a very low proportion of cores, with just two complete multiplatform cores, one multiplatform core fragment, two bipolar core fragments, and one conical piece and no single platform cores; cores comprised just 0.4% of the vein quartz component. All the cores were metallic grey, vitreous, smooth-grained, semi-translucent quartz, with one of the multiplatform cores exhibiting a bedding plane of quartz-psammite; the rest had 0% cortex.
Both the complete multiplatform cores were similar in size; Figure 9-16 provides the metrics for these cores, comparing them to the C.203 group – the two cores are amongst the smallest, and the mean weight is almost half of the C.203 group (Appendix A- 74). Looking at the bipolar cores, all were fragments; Appendix A- 76 provides the means for the bipolar cores. The four bipolar core fragments were similar in size to the complete bipolar cores.
Figure 9-16 BDG C.206 and C.203 group complete multiplatform coresTop of Page
29% of the C.206 debitage was <10mm, a similar proportion to the C.203 group (Appendix A- 67); however, as noted there were significantly less cores. The proportion of debris and flakes was also similar between the contexts (Table 9-6). No bipolar flakes were noted. Unlike the C.203 group, there was no apparent lack of mesial fragments, but there was a similar diminutive amount of <20≥10mm slivers compared to the direct percussion experimental datasets (Table 9-10, Table 9-19, and Table 9-20). Again, there were slightly higher proportions of the smaller size grade of slivers, which may explain some of the apparent difference; alternatively, as posited for the C.203 group, some of these slivers may have been removed from the excavated area.
Table 9-19 BDG C.206 fragment types and debris, excluding <10mm debris
Table 9-20 BDG C.206 debitage types
Compared to the C.203 group, C.206 had proportionally less transverse only breaks, longitudinal and transverse clean breaks, and was dominated by Uneven/a breaks (uneven longitudinal and no transverse break) (Appendix A- 69). Overall, the flakes were generally thinner than the C.203 group, except for mesial and lateral fragments (Appendix A- 68). C.206 was compared to the experimental assemblage’s techniques, and techniques/supports for proportions of siret breaks (Appendix A- 70 and Figure 9-17). The proportions were the same as S.H.D. Inelastic, and less than the C.203 group. Using C.206 as the reference group, the difference in siret proportions was significant compared with technique (Table 9-21), with the difference of a weaker significance compared with S.H.D. (χ² = 5.406; p = 0.020) than with H.H.D. (χ² = 57.908; p = 0.000). For technique/support, the difference was also significant (Table 9-22); compared with S.H.D. Inelastic there was no significant difference; compared with S.H.D. Elastic the difference was significant, as it was for H.H.D Elastic and H.H.D. Inelastic (Appendix A- 77).
Figure 9-17 BDG C.206 and Experiment siret breaksTests of Model Effects
Table 9-21 GZLM. Dependent Variable: Siret/non-siret Model:
(Intercept), Dataset/technique. Reference category: C.206
Table 9-22 GZLM. Dependent Variable: Siret/non-siret Model:
(Intercept), Dataset/technique/support. Reference category: C.206
As with the C.203 group, the analysis was run again excluding the non-proximal fragments (Figure 9-18 and Appendix A- 72). For technique alone, the difference was significant (Table 9-23), with the difference of a weaker significance with S.H.D. (χ² = 5.516; p = 0.019) than with H.H.D. (χ² = 50.644; p = 0.000). For technique/support, the difference was also significant (Table 9-24); compared with S.H.D. Inelastic the difference was not significant; compared with S.H.D. Elastic the difference was significant, as it was for H.H.D Elastic and H.H.D. Inelastic (Appendix A- 78).
Similarly to the C.203 group, C.206 is more closely aligned with the soft hammer dataset than with the hard hammer. Unlike the C.203 group, however, the exclusion of the non-platform fragments did not result in a reversal in terms of the support component – C.206 was closely aligned with the S.H.D. Inelastic dataset in both analyses, having the same proportions of siret breaks in each. Therefore, in comparison to the experimental data, C.206 compares with the soft hammer, with the caveats outlined for the C.203 group. The C.206 complete flakes exhibited impact marks to a greater degree than the C.203 group, but at 81.5% of the complete flakes, this is again similar to the experimental soft hammer component of 79%, suggesting again the use of soft hammers in this context.
Figure 9-18 BDG C.206 and Experiment siret breaks: excluding non-proximal flakesTests of Model Effects
Table 9-23 GZLM. Excluding non-proximal flakes. Dependent Variable:
Siret/non-siret. Model: (Intercept), Dataset/technique. Reference category: C.206
Table 9-24 GZLM. Excluding non-proximal flakes. Dependent Variable:
Siret/non-siret. Model: (Intercept), Dataset/technique/support. Reference category: C.206
Just 2.2% of the debitage was complete flakes, compared with 3.8% from the C.203 group. While the mean weight and dimensions of the complete flakes were smaller compared to the C.203 group, C.206 contained less of the smaller size range (Figure 9-19 and Appendix A- 81), and, analysing with ANOVA, the difference in inter-context means was not significant for any of the metrics (Appendix A- 83). The majority of the complete flakes had a length/width ratio close to 1:1 with just one greater than 2:1 (Figure 9-20). Figure 9-21 gives the mean length for the flake fragment groups, highlighting that the proximal group had numerous longer flakes than the complete flakes, and that the mean length of the lateral flakes was slightly greater than the complete flakes.
Figure 9-19 BDG C.206 and C.203 group complete flakes: mean length, width, thickness, and platform width
Figure 9-20 BDG C.206 complete flakes: length/width ratio
Figure 9-21 Boxplot. BDG C.206: median length for flakes by fragment groupTop of Page
The artefacts from C.206 came from within the formal, flat, metalled layer of pebbles and cobbles. The construction of the metalled layer marks this context out as the greatest difference from the C.203 group, as well as the lesser quantity of cortical artefacts, which was statistically significant. Again, this is interpreted as a preferential use of quarried quartz by the communities of practice, with a psammite bedding plane noted on one core and the remaining cores all having no cortex. As with the C.203 group, the presence of just two bipolar core fragments and no bipolar flakes does not suggest a pervasive use of a bipolar technology. The proportion of cores was very low (and low for non-quartz cores as well) and those present were small, and the proportion of larger debitage fragments was also low. This suggests a size grading was at play in this context, possibly with smaller artefacts falling within the matrix of the metalled layer; some may also have formed part of the metalled layer. A couple of squares had no quartz, while six had no <10mm artefacts, with the nearly three quarters of the artefacts from the top half of the excavated area, and just 8% from the bottom quarter. As with the C.203 group, the break patterns are comparable with the experimental dataset’s soft hammer component and use of direct percussion with an anvil, however none of the three platform cores had visible distal impact marks. The proportion of complete flakes with impact marks also suggest the use of soft hammer. Again, like the C.203 group, no core preparation was noted in the chaîne opératoire. C.206 had, proportionally, few diagnostic artefacts, with those identified being retouched flakes.Top of Page
C.202 was sampled at 50% of the squares, and held almost a fifth of the vein quartz artefacts, almost half of the rock crystal, and just 4% of the non-quartz artefacts (Table 9-25). An analysis of the spatial distribution in the trench is less informative due to the 50% gaps. Nevertheless, Figure 9-22 highlights that over half of the squares contained less than 30 artefacts per square and a third had less than 5 debris <10≥5mm, and a fifth had none. Unlike the previous contexts, the majority of the cores came from higher density squares, but similarly, most of the heavier debitage came from low density squares. Three sets of conjoins were noted, one of which was considered a fresh break; both of the other sets were pairings and both came from Square C6.
Table 9-25 BDG C.202 artefacts by material
Figure 9-22 BDG C.202 distribution. Top left: quartz cores and retouch/wear mark artefacts. Top right: average debitage weight per square for debitage >2g. Bottom left: count of all quartz and rock crystal artefacts . Bottom right: count of debris <10=5mm. Same square numbering as previousTop of Page
C.202 had a similar proportion of cores to the C.203 group, consisting of six multiplatform (one fragment), three single platform (one fragment), four bipolar (two fragments), and a possible conical piece; one of the multiplatform cores was a core-on-flake, One single platform core was quartz-beach, while the rest were undifferentiated quartz; nearly all were semi-translucent to semi-opaque, metallic grey, vitreous quartz. The five complete multiplatform cores were similar in size range to the C.203 group (Figure 9-23 and Appendix A- 74). The two complete single platform cores were the largest in mass from the assemblage with one twice the mass of the rest (Appendix A- 75 and Figure 9-24). C.202 held the two complete bipolar cores from the assemblage; their mean weight and dimensions were smaller than the means for the bipolar core fragments from C.202 as well as the other contexts (Appendix A- 76).
Figure 9-23 BDG C.202 and C.203 group complete multiplatform cores. Weight, length, width, and thickness
Figure 9-24 BDG C.202, C.203 group, and C.205 complete single platform cores. Weight, length, width, and thicknessTop of Page
C.202 had a lower proportion of flakes to debris than C.206 and the C.203 group, with 36% being <10mm debris; there were more ≥20mm debris, but less <20≥10mm debris (Table 9-6, Table 9-26, and Table 9-27). As with the previous contexts there is a lack of <20≥10mm slivers, but many <10≥5mm slivers which may suggest a considerable fragmentation of artefacts. But, as with C.206, there does not appear to be a diminutive quantity of mesial fragments compared to the C.203 group. Just one bipolar flake was noted.
Table 9-26 BDG C.202 fragment types and debris, excluding <10mm debris
Table 9-27 BDG C.202 debitage types
C.202 had substantially less clean breaks and more siret breaks than the previous contexts, with Uneven/a breaks dominating (Appendix A- 69). The siret proportions was analysed with GZLM. Compared to the experimental data’s techniques for siret proportions, the difference was significant (Table 9-28), with no significant difference with S.H.D. (χ² = 1.601; p = 0.206). For technique/support, the difference was also significant (Table 9-29), with no significant difference with S.H.D. Elastic; for S.H.D. Inelastic the difference was significant as it was for H.H.D. Elastic and H.H.D. Inelastic (Appendix A- 79).Tests of Model Effects>
Table 9-28 GZLM. Dependent Variable: Siret/non-siret Model:
(Intercept), Dataset/technique. Reference category: C.202
Table 9-29 GZLM. Dependent Variable: Siret/non-siret Model:
(Intercept), Dataset/technique/support. Reference category: C.202
When the non-platform flakes are excluded, the pattern remains the same, but with a weaker significance. Compared to technique the difference was significant (Table 9-30), with no significant difference with S.H.D. (χ² = 0.166; p = 0.684). For technique/support, the difference was also significant (Table 9-31), with no significant difference with S.H.D. Elastic; for S.H.D. Inelastic the difference was significant, albeit weak; for H.H.D. Elastic the difference was significant and also for H.H.D. Inelastic (Appendix A- 80). C.202, therefore, has a higher proportion of siret breaks, and while aligned with S.H.D. and S.H.D. Elastic, the significance is lesser than with the previous contexts. 75% of C.202’s complete flakes exhibited impact marks, which is a similar proportion to the experimental soft hammer component.Tests of Model Effects
Table 9-30 GZLM. Excluding non-proximal flakes. Dependent Variable:
Siret/non-siret Model: (Intercept), Dataset/technique. Reference category: C.202
Table 9-31 GZLM. Excluding non-proximal flakes. Dependent Variable:
Siret/non-siret Model: (Intercept), Dataset/technique/support. Reference category: C.202
2.8% of the debitage was complete flakes, compared with 2.2% from C.206 and 3.8% from the C.203 group. The complete flakes were generally smaller, and thinner, but their mean width and platform width was greater than the previous contexts (Appendix A- 81 and Figure 9-25). The majority of the flakes had a length/width ratio of below 1:1 (Figure 9-26). The boxplot for mean length by fragment group (Figure 9-27) highlights that the mean length for the complete flakes was less than the proximal and lateral, and, apart from the distal fragments, the complete fragments were generally smaller than the others including mesial fragments – this boxplot provides a different patterning than the boxplots for the previous contexts, where the complete flakes were in the upper range for length (see Figure 9-14 and Figure 9-21). Figure 9-28 compares the contexts’ complete and fragment for length, highlighting the difference with C.202 compared to the others, while also highlighting that for width (Figure 9-29), the difference between C.202 and the others is not apparent. Comparing the metrics’ means, using GLM, the differences between contexts was not significant, except for the length/width ratio and width/thickness ratio (Appendix A- 82 and Appendix A- 84). The Bonferroni post hoc test shows the difference between C.202 and the C.203 group for these two ratios was significant, but not significant between C.202 and C.206 or between C.206 and the C.203 group (Appendix A- 85).
Figure 9-25 BDG complete quartz flakes. Mean length, width, thickness, platform width and thickness
Figure 9-26 BDG complete quartz flakes. Length/width ratio
Figure 9-27 Boxplot. BDG C.202: Length for flakes by fragment group
Figure 9-28 [top] Boxplot. BDG C.202, C.206, and C.203 group for complete/fragments. Length
Figure 9-29 [bottom] Boxplot. BDG C.202, C.206, and C.203 group for complete/fragments. WidthTop of Page
C.202 contained, proportionally, a high quantity of rock crystal and a low quantity of non-quartz artefacts. Unlike C.206, there does not appear to be a greater degree of fragmentation of artefacts compared to the C.203 group. The lack of cortical artefacts again suggests the preferential utilisation of quarried material over beach pebble and cobbles by the communities of practice. As with the previous contexts, the chaîne opératoires did not include the preparing of cores such as faceting or abrasion. Compared to the Experiment dataset, the C.202 debitage is aligned with soft hammer, free hand direct percussion. While C.202 contained a similar proportion of cores to the C.203 group, it contained a greater proportion of bipolar cores, including the two examples of complete bipolar cores and the only non-quartz (chert) bipolar core. For all materials, bipolar cores accounted for 38% (n = 5) of the C.202 cores. However, there was only one bipolar flake identified. While the greater proportion of bipolar cores may suggest a Neolithic component in this upper context, the lack of bipolar debitage does not suggest in situ knapping, and the remaining cores and debitage were not significantly different from the lower contexts, therefore there is little to distinguish possible differences in the technological repertoires (for clearer differences in chaîne opératoires see Thornhill assemblage, Section 10.4). Indeed, considering that a handful of bipolar cores were also found in lower contexts it is difficult to pronounce the presence of bipolar artefacts as signifying a Neolithic presence only. Nevertheless, the higher proportion of bipolar cores does point to a change in depositional practices at least, possibly with material deposited in the excavated area during the Neolithic onto a previous collection of quartz artefacts.Top of Page
Artefacts from C.205 came from 15 of the 49 squares, with nearly half of the quartz artefacts from just one square, which included two of the three cores and both retouched flakes from the context (Figure 9-30). Proportionally, this context contained by far the most indeterminate/possible artefacts, at 4.8% compared to the average per context of 0.5% (Table 9-2). Two sets of conjoin pairs were noted, from Square A3 and E1, with the respective conjoins from the same squares.
Table 9-32 BDG C.205 and C.215 artefacts by material
Three cores, all complete, came from C.205; one single platform and two multiplatform, one of which was a core-on-flake. While the single platform core had some cortex, none of the cores were identifiable to a source. The two complete multiplatform cores are amongst the smallest of their type from the assemblage (Appendix A- 74) while the single platform was also amongst the smallest of its type (Figure 9-24 and Appendix A- 75).
C.205 had the lowest proportion of debris of all the contexts, with half of the debitage being flakes (Table 9-6). It had the lowest proportion of siret breaks and transverse only breaks and the highest proportion of longitudinal/transverse breaks, i.e. multiple breaks per artefact (Appendix A- 69). C.205 had the lowest proportion of complete platforms and the highest proportion of platform collapse at over double the average. As mentioned previously (pp. 196-7), along with C.203, C.205 had a significantly higher proportion of cortical debitage than the other contexts. Overall, the fragments were the smallest for all dimensions after C.206 apart from their mean thickness which was the thickest after C.202; while the larger dimension fragments (distal missing, proximal missing, and lateral fragments) were on average the smallest in the assemblage, the usually smaller distal and mesial fragments were on average amongst the largest from the assemblage (Appendix A- 64).
C.215 was confined to one square but with a greater quantity of artefacts than C.205. The core from C.215 was a large dual, opposed core on quartz-psammite, and was one of the largest cores from the assemblage; the core had a different appearance to nearly all the debitage from the context, and while C.215 came from just one square, no artefact conjoins or refits was noted. C.215 had the highest proportion of debris of the contexts, especially ≥20mm debris (Table 9-6). It was the only context with no clean longitudinal or transverse breaks, and had the highest proportion of siret breaks but also the highest proportion of complete platforms resulting in the highest proportion of transverse only breaks (Appendix A- 69). The fragments were generally above average for weight and dimensions apart from thickness (Appendix A- 64); the one complete flake was the fourth largest from the assemblage, and larger than all the contexts apart from the C.203 group (Appendix A- 81).
Figure 9-30 BDG C.205 spatial distribution, with C.215's square marked
C.215 was below C.203 and limited to one square and consisted of a sandy peat lens, and may be interpreted as a discrete deposition. The one core from the context was one of the largest from the assemblage, and the debitage were generally larger and heavier than other contexts and contained the greatest proportion of debris. The core was quartz-psammite, and differed from almost all of the debitage in the context. Compared to the other contexts, very few of the debitage had collapsed platforms and a greater proportion contained complete platforms, but the context contained a very low proportion of complete flakes. The sole complete flake was one of the largest from the assemblage. The five siltstone/sandstone flake fragments may have derived from the same core. C.205 is described as either co-eval or below C.203, and artefacts came from about a third of the squares, with nearly half of the artefacts from one square. The three cores were all small examples, and one was a core-on-flake. C.205 had the highest proportion of flakes, and also the highest proportion of collapsed platforms, with the debitage.Top of Page
While the non-quartz component of the assemblage was not analysed fully, with no measurements of the artefacts taken, a brief overview is provided (Table 9-33). Some of the siltstone artefacts are highly eroded, leaving open the possibility that others may have been missed during the excavation as they would be indistinguishable from natural pieces. None of the non-quartz cores and debitage appear to match each other, but the five siltstone/sandstone flake fragments from C.215 may have come from the same block. Chert accounted for nearly half of the non-quartz component, and the majority of the debitage were flake fragments. The retouched artefacts were two flint artefacts from C.203 and one retouched rock crystal flake fragment from C.206 (Appendix B- 20). The flints were a retouched flake fragment and a retouched flake fragment that is possibly fragment of a distally trimmed flake; the fragmented nature of the artefact makes it difficult to designate it definitively. Another possible retouched rock crystal flake also came from C.206. A small fragment of a pebble of indeterminate material from C.206 was tentatively regarded as a possible axe fragment.
Table 9-33 Non-quartz artefacts by material and context
The C.203 group contained a quartzite multiplatform core and a quartzite flake, but were not similar quartzites; the remaining cores from the context group were one single platform chert core, one chert pebble with a single flake scar, one possible chert core and one possible jasper core. There were 16 chert flake fragments, one complete chert flake and, along with two siltstone flake fragments. There were seven flint flake fragments, including the two retouched artefacts mentioned above. The rock crystal consisted of two flake fragments, two <20≥10mm debris, one <20≥10mm sliver, one <10≥5mm debris, and one <10≥5mm sliver. Three debitage were of indeterminate material – one complete flake, one complete blade, and one flake fragment.
C.206 contained two possible cores – one chert block contained one visible flake scar, while a thin, tabular slab of psammite had numerous flake removals and may represent a core. There was one psammite flake fragment, three chert flake fragments, and one chert blade fragment. The flint consisted of one flake fragment and two <20≥10mm debris and the rock crystal were seven flake fragments, one >20mm debris, and one <10≥5mm sliver. The remaining consisted of one complete siltstone blade, one quartzite flake fragment, one basalt flake fragment, and two flake fragments of indeterminate material. C.202 contained a complete chert bipolar core and two chert flake fragments, one of which contained a triangular-shaped quartz inclusion, resembling a quartz flake. There were four flake fragments, six <20≥10mm debris, and three <10≥5mm debris. C.205 contained four chert flake fragments and one flint <10≥5mm debris.
There were seven psammite hammerstones in the assemblage, four from the C.203 group and three from C.206. All were oblong pebbles or cobbles and none had evidence for extensive use. The smallest weighed 31g, the four complete examples ranged from 98g to 149g, and the two fragments were the heaviest, weighing 177g and 259g. While the psammite hammerstones showed limited evidence for use, the fact that all were oblong suggests that the communities of practice purposively chose oblong-shaped cobbles for use. It is of course not proved that these seven were in fact used as hammerstones, but the impact marks on them are consistent with use as hammerstones. They may however have been used for other, unidentified purposes. Beyond these hammerstones, there is very little evidence for the use of psammite, which as mentioned is a quartzite. From a functional perspective it may be that the psammite is too friable for use as flaked stone tools or, alternatively, it was avoided for cultural reasons. Considering the abundance of psammite in the area, the absence of psammite artefacts is clearly significant and not related to degradation; while probably too friable for stone tools, it is nevertheless still composed of quartz and therefore would not degrade much but rather the edges would damage easily through use.
The assemblage from TR2 selected for the present analysis is part of a larger assemblage excavated on a hillside overlooking the mouth of a river at Belderrig Bay, North Mayo. TR2 is located at the point of a substantial erosion scar, in-between the cliff and the bank, with the eroded pathway running over the excavated area. Quartz artefacts were visible on the surface of TR2 before excavation, with the thin layer of turf/peat having been eroded away along the thoroughfare. The date range of the excavations shows intermittent activity for over two thousand years, from the Later Mesolithic to the end of the Neolithic.
The three main artefact-bearing contexts from TR2 are C.202, C.206, and C.203; all three are dominated by vein quartz artefacts with minimal amounts of rock crystal and non-quartz artefacts. C.203 is interpreted as a very biologically active, pre-peat woodland topsoil, dated to c. 4300-4500 cal BC, that was artefact-rich and subsequently compressed, truncated, or eroded. For this analysis, C.211 and C.214 were grouped with C.203, with the former two contexts probably representing earlier activity. C.214 is the only context with a substantial proportion of non-quartz artefacts – 29% of the 28 artefacts was chert and flint flakes along with one blade of indeterminate material. C.215, also below C.203, appeared in just one square, contained 136 artefacts, and is interpreted as a discreet deposit of material; this context’s artefacts stood out in terms of size, and the one core from the assemblage was large and different from the debitage in the context. The upper levels of C.203 were hard to distinguish from the lower levels of C.206, which is the formal stony layer dated to c. 4000-4200 cal BC (for discussions on other Mesolithic platforms see Driscoll 2009a). C.202 may represent a later disturbance, with artefacts within it representing the upper level of C.206, which was disturbed with the consequence that the horizon of C.206 was indistinguishable. From the analysis of the vein quartz artefacts’ fragmentation and size, it does not appear that C.202 contains more fragmented artefacts than C.206, but rather C.206 contained a greater proportion of smaller fragments than C.202 or the C.203 group; C.202 contained more complete and lateral fragments than C.206, suggesting a lack of disturbance in C.202. While C.202 had a greater proportion of bipolar cores, suggesting a Neolithic component in the assemblage, the direct percussion artefacts do not appear to be significantly different from the earlier contexts in terms of the chaîne opératoires, such as noted in the Neolithic assemblage from Thornhill (see Section 10.4). Furthermore, while a considerable proportion of the cores from C.202 were bipolar, there is little evidence of in situ knapping, with just one bipolar flake noted. This may suggest that during the Neolithic material was deposited onto a previous collection of quartz artefacts.
Exploring the chaîne opératoires at play in the excavated area necessitates a clear understanding of the fracture mechanics of the quartz. The archaeological material from Belderrig for the most part matches the experimental dataset’s direct percussion component in terms of fracturing. However, direct comparisons are unlikely given that the experimental knapping produced a complete, preserved knapping floor without further post-knapping breakages or removal of artefacts occurring. Furthermore, due to the difficulties in identifying sequential breaks and sequential fragments consistently in the archaeological material, these categories were excluded therefore no direct comparisons with these categories were possible. As well as this complication, the experimental assemblage set out with the intention of creating a preserved knapping assemblage, which, understandably, would not have occurred during prehistory with consequent differing chaîne opératoires. Nevertheless, the experimental assemblage provides a framework for understanding the fracturing of the quartz, and for examining similarities and variances within the archaeological material.
5275 vein quartz and rock crystal artefacts were analysed from the five context groups. All of the quartz from TR2 appears in a fresh condition, with sharp edges. Pieces with sub-rounded and rounded edges – generally smaller fragments – were regarded as natural quartz and excluded from further analysis. One difficulty noted with the analysis of the vein quartz artefacts was ascribing a time period to some of the fractures. Following from this, a related issue concerns to what degree the artefacts have fragmented since deposition. During knapping, numerous incipient fractures in the quartz were noted, and on returning to the experimental assemblage on a number of occasions over the course of the analysis of the assemblages, many smaller, and some larger, artefacts had fragmented further even though they were individually bagged and handled carefully. Therefore, it is likely that some post-depositional fragmentation occurred with the archaeological material even without extensive or moderate trampling/disturbance and so forth. It is of course impossible to quantify this without further experimentation with trampling experimental assemblages in various degrees of burial in soil.
The vein quartz was categorised into source materials where possible. However, the preponderance of non-cortical flakes, and the low frequency of artefacts retaining parent rock bedding planes, precluded assigning most of the quartz to a source. Only 6% (n=222) of the ≥10mm vein quartz artefacts could be identified definitively to a source; 3.3% to quartz-beach, 2.6% to quartz-psammite, and 0.2% to quartz-metadolerite. For the cores, 10.9% were identified as quartz-beach, 9.1% as quartz-psammite, with the remainder as generic quartz and no quartz-metadolerite cores identified. This low proportion of cortical artefacts suggests that the chaîne opératoires involved the initial knapping of cobbles/blocks away from the excavated trench area, with cores being brought to that area once decortified to some extent, as well as having bedding planes removed. The overall low proportion of cobbles also suggests the quarrying of vein quartz; which would result in less cortex to begin with, thus creating a lower proportion of cortical artefacts.
Analysis of the proportion of cortical artefacts showed that, while all contexts had low proportions of cortical artefacts, the C.203 group and C.205 had significantly more cortical artefacts than the other two main artefact-bearing contexts (C.202 and C.206). The significantly higher proportion of cortical artefacts in the C.203 group and C.205 suggests a difference in the chaîne opératoires of the communities of practice over time. This difference may relate to a lesser degree of initial knapping taking place away from the excavated area during these earlier contexts and/or with more beach cobbles used in these contexts, and also a lesser degree of decortification of quarried blocks. For the quarried blocks, an alternative explanation is that quarried veins, which had been previously opened and consequently decortified during the C.203 group and C.205, were subsequently quarried further during later contexts with less cortex on the material as it had been decortified already. This interpretation would suggest that the same veins were re-visited for use over the generations.
Overall, 83% of the debitage was smooth-grained, with sugar-grained at just 2%; there were no sugar-grained cores, suggesting a clear preference for smooth-grained material by the communities of practice. The largest group of materials consisted of semi-translucent, smooth-grained, with a cloudy glass appearance and cloudy hue, followed by a group consisting of semi-translucent, smooth-grained, a vitreous appearance and a metallic grey hue. As noted in Section 6.5, the grainier, more sugary-textured quartz produced the least amount of fragments per strike during the experimental knapping. If this fragmentation pattern also held for the prehistoric communities of practice at Belderrig, it may imply that a higher fragmentation rate of materials was not an attribute to be avoided, and/or that characteristics of the material such as the smooth-grained appearance were more sought after than the apparently functional attributes of lesser fragmentation rates. Experimental work on the use of the differing quartzes would shed light as to which types of graininess is suitable for differing tasks in term of edge sharpness and edge maintenance/damage resistance.
It appears that quartz sourced from metadolerite played a minor role in the material repertoire, with no quartz-metadolerite cores present. However, the question remains as to what extent the 94% of the debitage and 80% of cores that were uncategorised to a source were in fact sourced from metadolerite. Interestingly, the outcrops available today in the locality are dominated by the smooth-grained metadolerite and the sugar-grained psammite. However, there was little evidence for the metadolerite quartz veins being utilised even though the assemblage is dominated by smooth-grained quartz. The psammite quartz veins accessible today are predominantly sugar-grained, but the quartz-psammite identified in the assemblage is equally split between smooth-, sugar/smooth-, and sugar-grained quartz. This suggests that veins of differently grained quartz were available and utilised in prehistory, and/or that the use of the metadolerite quartz veins is masked by the lack of cortical and bedding plane signatures on the artefacts. It is also possible that the metadolerite sources that are today right beside the cliff may have been covered by land in early prehistory, with subsequent erosion exposing them on the foreshore.
There is no evidence for platform preparation such as faceting or abrasion in the chaîne opératoires, but rather the cores were primarily knapped directly using direct percussion, using both freehand and anvil – five of the platform cores had impact points suggesting inelastic support, i.e. a platform-on-anvil technique – all five were from C.203. Cores accounted for just 1% of the assemblage, with a further three artefacts categorised as core/debitage. 80% were platform cores, 13% were bipolar, with remaining being conical pieces and radially split pieces; 71% were complete cores. Out of the platform cores, multiplatform types dominate at 70% of the assemblage.
None of the cores resemble classic Later Mesolithic Larnian uniplane cores, as defined by Woodman and Johnson’s (1996) research on flint cobble knapping, where cobbles were opened by removing a cortical flake from the top with the flake scar then used as the striking platform, generally knapping less than 50% of the core face and producing a uniplane surface. These uniplane cores were seen as wasteful by-products:
"the classic uniplane core is not so much an integral feature during the reduction of a core but rather a frequent by-product which is a limitation to the further effective utilisation of the core. It has also been shown that frequently very few useful flakes and blades are produced from each core, with the result that large quantities of by-products are left after the production of a relatively small number of usable flakes" (Woodman and Johnson 1996, 221).
Similar Mesolithic core forms have also been noted away from the flint-rich north, on materials such tuff and chert at Lough Allen (Driscoll 2006, 200), chert in the midlands (Warren et al. 2009), and greenstone at Ferriter’s Cove (Woodman et al. 1999). While the latter were also formed on cobbles, the former were formed on blocks, highlighting that the removal of a cortical cap was negated. Warren et al. (2009) suggest that this form of core will occur where large pieces of raw material were available. However, while large blocks were available at Belderrig, this is not the case. The examples of single platform cores were usually knapped at up to 90% of the platform circumference and the single platform cores were generally subsequently rotated, resulting in dual, opposed cores or rotated again resulting in multiplatform cores. Therefore, there does not appear to be a ‘wasteful’ (as Woodman and Johnson put it) approach to the cores, with flakes continuing to be removed beyond the typical uniplane core; however, conversely, it does not seem that the communities were attempting to conserve the material as such. It is uncertain what the reasons for this are; it might be that the short, thick flakes produced did not need such uniplane faces to be maintained. The absence of uniplane cores is not, however, problematical from a typological perspective – the Ferriter’s Cove excavations highlighted that a low amount of uniplane cores may be expected on Later Mesolithic sites, with the uniplane core a by-product of a certain method of working and not a strict template to be followed (Woodman et al. 1999). However, it is not the case that Belderrig quartz uniplane cores would be unrecognisable, as suggested by Woodman et al. (1999, 76) – since a variety of core types were noted, there is no reason why uniplane cores would not be.
31% of the assemblage is <10mm debitage, and flakes account for 40% of the total; 5% of the flakes are complete. Over half of the flakes are <20mm, with a mean length of 21.7mm. For the complete flakes only, a third of the flakes are <20mm in length, with a mean length of 27.8mm. All but two of the 107 complete flakes are <60mm in length. For both flakes and debris, 80% of the assemblage is <20mm and 8% is >30mm in length. 33% of the ≥10mm debitage had complete or partial platforms, which is approximately the same proportion as the Experiment dataset. Overall, just 1.8% of complete and proximal flakes had bulbs. 38.6% had complete platforms; 42% were platform fragments, while 19.4% had collapsed platforms. A third of the ≥20mm length flakes were straight. Only three bipolar flakes were identified. While a proportion of ‘bipolar flakes’ may have been missed for the reasons outlined in Chapter 6, this tiny proportion of bipolar flakes suggests that it is in fact an actual pattern. Indeed, as noted in Chapter 6, some platform flakes can appear as bipolar flakes as well, thus adding an extra complication to the identification of reduction techniques.
Of the flakes with terminations present, 82% had feather terminations, 7.2% irregular, 6.2% plunging, 3.6% step, with remainder being fragmented or retouched/wear marks. The proportion of step terminations appears small, with numerous step terminations noted on the cores. This raises the possibility that step terminations were under-recorded and other step terminations were interpreted as the distal end of flake fragments, because it can be difficult to determine if it is in fact a termination or a later break in quartz. As noted in Section 6.3.3 some researchers treat flakes with step terminations as being flake fragments rather than a termination of a complete flake in the first place (e.g. Odell 1989, 192; Redman 1998, 38).
The minor amount of complete flakes present consisted of short, wide, thick flakes, and many of the flake fragments were of similar dimensions, if not larger, than the complete flakes. Comparing the metrics’ means of the complete flakes, the differences between contexts was not significant, except for differences between C.202 and the C.203 group for the length/width ratio and width/thickness ratio; the C.203 group had a greater length/width ratio and a lesser width/thickness ratio compared to C.202. The C.203 group had an apparent lack of slivers and mesial fragments compared to the Experiment dataset, raising the possibility that these were removed for use elsewhere. C.206 also had a lack of slivers, but not to such an extent as the C.203 group and no apparent lack of mesial fragments. For C.202, the lack of ≥10mm slivers may relate to a greater fragmentation of them, creating more <10≥5mm slivers and no apparent diminutive quantity of mesial fragments.
Comparisons to the experimental assemblage suggest that for the three main artefact-bearing contexts, the communities of practice possibly used soft hammers, and for the C.203 group and C.206 used anvils, for direct percussion knapping. The similarity to the S.H.D. assemblage in terms of the proportion of siret breaks must be interpreted in the light that the proportions of siret breaks may have been altered by the removal or addition of debitage in the area. Alternatively, the expertise of the communities of practice, along with the method/technique employed with a hard hammer, may have resulted more closely with what was achieved with a soft hammer during these experiments, i.e. thicker flakes less likely to form a siret break. The mean thickness for complete flakes was similar for S.H.D. and the BDG artefacts. However, as noted in Chapter 6, the difference in thickness between H.H.D. and S.H.D. complete flakes was not an appropriate indicator of technique; for Redman’s (1998) research, however, (mentioned previously, see Section 5.6) on soft and hard hammer differences, thickness was deemed to be a predictor, albeit weak, of impactor. Seven psammite impactors identified had signs of limited use. If psammite, or other stone, impactors had been used more extensively by the communities, the impactors were not deposited in the area. The experimental knapping also pointed to less soft hammer flakes exhibiting impact marks compared to the hard hammer flakes, and the three main artefact-bearing contexts all had similarly low proportions of visible impact marks on the complete flakes, again possibly suggesting the use of soft hammers.
The indications of the possible use of soft hammer raises the question of what soft impactor was used during the Mesolithic, as no deer were present in Ireland (Woodman et al. 1997) thus excluding the possibility of antler such as used during the experimental knapping. The Later Mesolithic is usually discussed in terms of hard hammer, freehand percussion (e.g. Woodman and Johnson 1996; Woodman et al. 1999), with Woodman et al. (1999, 71) citing the use of hard hammer as signified by “the variation in platform size, the number of prominent bulbs of percussion, and the broad flake scars on many of the cores”. For the experimental quartz dataset, while none of these factors were statistically significantly different, the soft hammer flakes had greater variations in platform sizes, had more occurrences of bulbs (albeit in low numbers for both techniques and none were prominent), and had broader flakes than the hard hammer flakes. In terms of the Irish Early Mesolithic, Costa et al. (2001) argued that the analysis of an assemblage from Donegal suggested that both soft stone hammers and hard stone hammers were used for creating differing artefacts. The authors, however, do not mention organic impactors such as antler, and do not suggest why they argue for soft stone instead of an organic soft hammer; they thus avoid the issue concerning deer in Ireland which has been raised elsewhere by Woodman and McCarthy (2003). While the lack of deer in Ireland rules out the possibility of the use of antler during the Mesolithic at Belderrig, hard woods would have been available, or indeed, various ‘soft’ stones such as suggested by Costa et al. (2001) in relation to the Early Mesolithic.
As well as the possible use of soft hammers identified, another difference from other Later Mesolithic assemblages is the identification of the use by the communities of practice of a platform-on-anvil technique. There has been little discussion of the platform-on-anvil technique in Irish lithic research. Nelis (2004, 866) identified platform-on-anvil cores in Neolithic assemblages from Northern Ireland. Evidence from Ferriter’s Cove (see Woodman et al. 1999, 32) and Corralanna (see Warren et al. 2009, 15), however, may point to the use of a platform-on-anvil technique in the Later Mesolithic. At Ferriter’s Cove Woodman et al. (1999, 32) noted for the flint cores “areas of severe damage are visible on a number of distal ends of single- and dual-alternate platform cores…Although this implies the use of an anvil support, there is little typological or refitting evidence for the use of the bipolar technique”. While in a lithic scatter from Corralanna Warren et al. (2009, 15) noted the discovery of a large stone with pitting, interpreted this as an anvil but commented that “the absence of any evidence of bipolar techniques in the assemblage suggests that this pitting results from activities unrelated to stone-working tasks”. Both of these analyses looked for evidence of bipolar cores specifically rather than for evidence of platform-on-anvil knapping. This may suggest that this type of technique was more prevalent than has been appreciated.
Along with the thousands of non-retouched flakes, many of which can be considered as tools (as discussed in Section 4.3), a small number of diagnostic types were identified. The majority came from the C.203 group, including the one example of a butt trimmed flake (Appendix B- 22) and one borer (Appendix B- 21); a point/notched piece came from C.202; besides these, the rest were predominantly retouched flakes while one with abrupt retouch on its distal end could be considered scraper-like (Appendix B- 21, top left). None of the cores were retouched, while one retouched artefact was categorised as a core/debitage. As well as the 30 retouched artefacts, there were 23 with possible retouch. For the former, the majority had convex or rectilinear retouch, with the majority of both categories having an abrupt retouch angle; for the latter, almost half were convex, with abrupt or semi-abrupt retouch. While these 23 possible retouched artefacts are probably actual retouch/wear marks, a conservative approach was taken in this regard in order to avoid an inflation of the retouched numbers as is a danger with vein quartz (see Lindgren 1998). A more liberal approach could include them as retouched.
Just one ‘classic’ Later Mesolithic type – a butt trimmed flake – was noted. Other flakes are morphologically similar to butt trimmed types, minus the retouch – the example in Appendix B- 22 (top right) is not retouched but the lateral edges have extensive damage, but it is unclear whether this damage represents damage during use or from post-depositional processes. The Ferriter’s Cove site also had a low proportion of butt trimmed forms, with Woodman et al. (1999, 76) noting that the assemblage had a “sufficient number of the diagnostic tools but it also has an interesting and diverse range of other artefacts”. It is unclear if the phrase ‘sufficient number’ means sufficient for the communities who formed the assemblage, or sufficient for the analysis of the site by archaeologists. Be that as it may, the Ferriter’s Cove flaked stone assemblage consisted of 0.5% (n=36) “retouched tools” (Woodman et al. 1999, 29) while the TR2 quartz assemblage consisted of 0.6% retouched artefacts and if the possible retouched artefacts are included this increases to 1%. (The non-quartz component consisted of 3.2% (n=2) retouched artefacts.)
Compared to the non-retouched artefacts, the diagnostic artefacts were significantly larger in all dimensions and weight, with the possible retouched artefacts falling in between the retouched and non-retouched groups. In order to avoid a multitude of smaller non-retouched artefacts skewing the analysis, a second analysis was conducted comparing only >20mm retouched and non-retouched debitage; this second analysis confirmed that the diagnostic types were on average larger in all dimensions. While there is therefore a clear distinction in terms of size between the retouched and non-retouched artefacts, it is difficult to discern if the retouched artefacts represent the reuse of non-retouched tools, or if they represent a different kind of tool which necessitated retouch, and larger size of artefact, from the outset.
In terms of the use of the non-retouched debitage, the apparent lack of slivers, and in some contexts of mesial fragments, was suggested as possible evidence for the removal of such artefacts from the area for use. Considering the communities of practice’s location at a coastal site, a possible interpretation for these small fragments is the use in fish processing. As outlined in Section 4.3, Flenniken (1981) noted that small, non-retouched fragments of vein quartz were hafted for use as fish processing tools, with these erstwhile non-diagnostic pieces of quartz noted as functional tools because they were found hafted in waterlogged contexts; while there, the communities used bipolar reduction to specifically produce such artefacts, at Belderrig a possible interpretation is that such fragments were a useful by-product of the communities vein quartz chaîne opératoire, and a use of the fracture characteristics of the quartz.
The analysis of the quartz artefacts did not clearly identify any ritualised or symbolic use or deposition of the material. However, it could be argued that, conversely, the analysis did not identify clearly any ‘mundane’ or utilitarian use or deposition of the material – while diagnostic types were noted, it is unclear if their use or retouch was indeed ‘simply’ functional or utilitarian or used during ritualised practice. As noted in Chapter 4, the quartz artefacts can be interpreted as having held differing roles and differing significance depending on the context and on the observer or participant. Consequently, it is not a question of an either/or of the sacred and profane, but rather when, where, and by whom. It is the assumption of the present analysis, and most if not all lithic analyses, that the knapping of the quartz by the communities of practice was to produce stone tools. Nevertheless, while this may be a safe assumption, the question remains as to why there, why then, and by whom. In terms of the when, we have a rough idea (in the fifth and fourth millennium BC), in terms of the where, we have a partial answer (a part of the chaîne opératoire occurred at Belderrig), and by whom we are in the dark.
Quartz has been argued as an especially power-full material (e.g. Taçon 1991), an argument that cannot be ignored in interpretations as to why the Mesolithic and Neolithic communities chose the Belderrig hillside by the river mouth at the ocean for recurrent visits. It is unclear to what extent the quartz sources themselves were an attractor – little is known about other quartz sources inland from Belderrig (see above, Section 9.3.1). Moreover, the reasons need not have remained unaltered; the initial visits may have been for different reasons, which subsequently set in train a sequence of visits with the original motives fogged by time but nevertheless the locale became a focus for activities – a part of the traditions of the communities. These traditions were not necessarily rigid, but as Gosden (1994) has put it, they were a dynamic traditionalism, open to elaboration and change through time. A part of this dynamic traditionalism may have been the re-visiting of veins of quartz quarried in previous years or previous generations.
Looking at the chaîne opératoires at play from the perspective of a ritualised taskscape (for taskscape see Ingold 2000) places the question of what the ‘tools’ were used for, and what the excavated area represents, in another dimension: in viewing the quartz assemblage as neither strictly utilitarian nor strictly symbolic, one can consider the material as on a continuum of ritualised practice. Indeed, in viewing the chaîne opératoire as a sequence rather than a more rigid chain compliments the taskscape perspective, which calls for a recognition of the sociality of technology and the interactivities of life – a sequence of inter-related activities, and an enaction of technological practices that encompassed the entirety of communities’ world and world-view. The quartz used by the communities was not a material found ‘out there’ in the landscape, and subsequently exploited and discarded, but the quartz itself was part of the taskscape amongst the other parts such as the humans, plants, and animals, and the river and ocean at Belderrig that the communities of practice visited and re-visited over the millennia or so of indentified activity. Elsewhere, I have discussed the utilisation of raw materials, and local/non-local issues of the use and deposition of materials, during the Mesolithic and Neolithic in the west of Ireland (see Driscoll 2006, 2009b); the convergence of the local and non-local materials brings different nodes of the landscape together through their use in the taskscape – the stone is not just extracted from its source to then become a reified commodity, but probably brought with it the personality or qualities of its source (see Bradley 2000) or at least the stories told of the source.
The clearest representation of symbolic or ritualised practice can be seen in contrasts – a small number of cores have been interpreted as being non-local quartz; these cores had no matching debitage, suggesting an importation of cores for deposition at the locale, which had a relatively lower quantity of cores that matched the debitage present. As well as these non-local quartz cores, the chaîne opératoires included the deposition of non-local, non-quartz cores and debitage; these cores and debitage were a mismatch of materials, suggesting no in situ knapping, but rather the deposition of non-local material. One of the non-quartz artefacts deposited was a chert flake with a triangular – almost flake-shaped – quartz inclusion (Appendix B- 23), thus marrying the local and non-local materials.
This mismatch of materials – both quartz and non-quartz – could be interpreted as the excavated area representing a ‘dump’, a midden of lithics and organic material, with the lithics being the remaining material culture. However, for the C.203 group at least, the analysis of the spatial distribution of material has suggested a patterning of lithic deposition into concentration/adjacent/periphery, which goes against the excavated area being a random dumping of material; instead it is interpreted that the C.203 group is a more structured deposit, with the adjacent area that contained the cores and heavier debitage either the area where knapping occurred or a clearance of larger pieces from the knapping in the concentration area into the adjacent area. The deposition of non-local quartz and non-quartz, therefore, was not necessarily a dumping of material – a midden sensu stricto (for discussions on middens and refuse see Needham and Spence 1997) – but rather can be seen in line with the structured depositing of the local quartz, of activities taking place in the context of a midden sensu lato, where a midden can be seen as an accumulation of material culture, an accumulation of artefacts and ecofacts, without modern categorisations of these as rubbish or waste.
Because C.200 was tallied during the preliminary catalogue, this count is approximate as the catalogue for C.200 may contain some natural pieces.[return]
Approximate total; see footnote above.[return]
5710±90 BP, UCD-C46; OxCal v. 4 Bronk Ramsey (2008); IntCal04 Atmospheric curve Reimer et al. (2004). [return]
7530±100 BP, UCD-C48; OxCal v. 4 Bronk Ramsey (2008); IntCal04 Atmospheric curve Reimer et al. (2004).[return]br /> 8660±130 BP, UCD-C54; OxCal v. 4 Bronk Ramsey (2008); IntCal04 Atmospheric curve Reimer et al. (2004).[return]
From parameter estimate[return]
From parameter estimate[return]
While this 0.5% is calculated from Table 3.1 in Woodman et al. (1999, 29), different quantities are provided in later tables (Woodman et al. 1999, 76-77).[return]