Home     PhD Contents     PhD Chap 1     PhD Chap 2     PhD Chap 3     PhD Chap 4     PhD Chap 5     PhD Chap 6     PhD Chap 7     PhD Chap 8     PhD Chap 9     PhD Chap 10     PhD Chap 11     PhD Bibliography     PhD Appendices     PhD Plans    [PhD PDF]

6 Experimental knapping experiments

[Please note that part of this chapter has been published in the Journal of Archaeological Science]

6.1 Introduction

This chapter presents the experimental knapping experiments. Section 6.2 outlines the selection and collection of the raw materials used, and the results from the thin sectioning of the material. Section 6.3 outlines the experiments, with Section 6.3.1 discussing the experiments’ organisation and recording; Section 6.3.2 describing the measurements, equipment, and sampling strategy used for the resultant assemblage; and definitions of the debitage categories discussed in Section 6.3.3. Section 6.4 presents the analysis of the experimental assemblage, with debitage covered in Section 6.4.1, looking at the debitage fragmentation rate, break and fragment types, complete and proximal flakes, debitage types, and lastly curvature, cortex, and flake termination. The cores are dealt with in Section 6.4.2. Section 6.5 concludes the chapter with a discussion and overview of the experiments and results.

6.2 Collecting the materials and assessing the quartz

As outlined in Section 1.2, this series of experiments has focused on vein quartz. The quartz used for the experiments was collected from Belderrig, Co. Mayo (Figure 6-1 and Figure 6-2). The psammite quartz (hereafter P.Q.) was quarried from the quartz vein in the psammite bedrock cliff face just north of the excavation site, and also picked up from the base of the cliff where blocks of the vein had been eroded out (Figure 6-3). The metadolerite quartz (hereafter M.Q.) and the Rose Cottage quartz (hereafter R.Q.) are both quartz veins associated with the metadolerite outcrops, with the M.Q. coming from the intertidal zone (Figure 6-4) and the R.Q. from a vein exposed from a cutaway bog 1km inland from the coast. The beach (hereafter B.Q.) cobbles used were a mix of quartz derived from the metadolerite and the psammite and were collected, along with the psammite hammerstones, from the stretch of shore by the river mouth. The psammite used for hammerstones is an impure quartzite (metamorphosed sandstone), and is therefore a ‘hard’ stone.

Over 100kg of quartz was collected and about 30kg was used to familiarise myself with knapping the material. Approximately 70kg of quartz was subsequently used for the 40 quartz knapping events. These 40 events produced an assemblage of over 14,000 ≥5mm experimental artefacts. The impactors used were a series of 14 psammite hammerstones, ranging from 124-627g, and a deer antler (Appendix A- 9). The chert used was a festooned chert collected from an outcrop at Lough Derraveragh (Figure 6-1) in the vicinity of a Mesolithic quarry (Little 2005; O'Sullivan et al. 2007). This material was selected as festooned chert was also found at the Belderrig site during excavations and the site is a known Mesolithic quarry.

Samples from the three different veins used and a beach cobble have been analysed macroscopically and in thin section by Dr Julian Menuge, School of Geological Sciences, UCD. The thin sections (Figure 6-5) show that the crystal size was 1-5mm making all of them coarse-grained raw materials and were variable in character in terms of crystal orientation and fracture development, but with all of them being of massive habit. The samples contained multiple macro- and micro-fractures, some of which led to the subsequent development of veinlets of quartz within them; a veinlet is visible in the P.Q. thin section as a dark line. The R.Q. quartz contained many micro-fractures of uncertain age (Menuge 2007 pers. comm.), which appear in the thin section as thin white streaks.

When examining the quartz macroscopically, the large crystal size is not always apparent – while the quartz is defined as coarse-grained, the majority of it lacks the appearance of a sugary texture. Instead, the quartz appears as smooth-textured even though the thin sections clearly show the large individual crystal sizes, and the massive habit of the quartz. The material could easily be taken for finer-grained quartz, or indeed what Ballin has described as ‘milky quartz’. It is difficult to distinguish the grain/crystal boundaries from macro-fractures (Menuge 2007 pers. comm.). In some cases of the smooth-textured quartz, however, subtle lines of lamellar grain boundaries can be seen macroscopically, and take the appearance of a 3-D contour map of a hill or the terraces of rice paddies, with the grain boundaries lined up akin to the contour lines.

Quartz lithic source mayo

Figure 6-1 Raw material sources

quartz and hammerstones

Figure 6-2 Quartz. Planes of weakness can be seen as flat surfaces on some blocks. Bottom right: psammite hammerstones

psammite bedrock quartz vein

Figure 6-3 Facing northeast. Psammite outcrop. Inset: close up of quartz vein located at top of ranging rod in main picture

Metadolerite bedrock quartz vein

Figure 6-4 Facing north. Metadolerite outcrop at low tide. Ranging rod at quartz vein. Inset: quartz vein. Psammite outcrops on cliff at right of picture

quartz thin section

Figure 6-5 Quartz thin sections. Upper left: B.Q. Upper right: P.Q. Lower left: M.Q. Lower right: R.Q. Scale = 1mm (1000µm)

quartz texture

Figure 6-6 Quartz texture. Left: M.Q.. Right: P.Q.

Figure 6-6 provides a comparison of the M.Q. and P.Q. using a backlight to view the grain texture; the P.Q. has a sugar texture, while the M.Q. has a smooth texture with multiple macrofractures, but both have similarly sized crystal grains. None of the M.Q. or R.Q. used had a sugar texture, while the P.Q. ranged from sugar to smooth, with the majority being sugar; a third of the B.Q. used had a sugar texture. As noted in Chapter 4 and above, researchers have categorised vein quartz in a number of ways, and it is difficult to place the Belderrig quartz into their categories, apart from de Lombera Hermida’s (2009), which divides quartz into four categories depending on absence or presence of grain and plane. Using his categories, all the Belderrig quartz used in the experiments are grain/plane – or S/S as he abbreviates it. Compared to Ballin’s (2008) descriptions of quartz, the Belderrig material cannot easily fit. Ballin made a distinction between massive, milky quartz – which he described as the predominant quartz used in Scottish prehistory – and fine- and coarse-grained quartz. However, the Belderrig quartz is both massive, milky quartz and clearly coarse-grained, even though the graininess is often invisible, thus taking the appearance of a lack of grain. The thin sections, therefore, have highlighted the dangers of using macroscopic identification to interpret crystal/grain size and therefore through thin section it is possible that, for example, some of Ballin’s ‘milky quartz’ would be found to be in fact coarse-grained and of massive habit.

Top of Page

6.3 The experiments

6.3.1 Recording the knapping events

The experiments were structured by defining each knapping of a block or cobble as an event (Appendix B- 1). Hence, the experiments consisted of 40 events plus one extra for the chert event; Appendix A- 10 lists the weights of the raw material used for the different techniques – bipolar, hard hammer direct (H.H.D.) and soft hammer direct (S.H.D.). An event began with a block or cobble and this was knapped until no more usable flakes could be removed; for some events an event also used a secondary core called a child core that had been produced during the event – this child core was derived from either the block or cobble splitting or a removal of a large flake during the knapping event. The debitage was recorded as to what core – the initial or child core – it was removed from. Of the 32 direct percussion events, 34% produced split cores and an additional 28% produced large flakes; some of these split cores and large flakes were subsequently used as child cores mentioned above. While it is noted in the literature (e.g. Callahan 1987; Knutsson 1988a) that cores were worked by both platform percussion and bipolar, this type of combination of reduction strategies was not done in these experiments.

The aim of the knapping was to reduce the core until no more flakes could be removed – this was due to either the cores’ small size or lack of good angles for further flaking. For the direct percussion the aim was to produce medium-sized flakes (defined here as flakes 30-90mm in length); for the bipolar percussion the aim was to produce small flakes (defined here as flakes 10-29.9mm in length). On some occasions the quartz proved difficult to knap and the cores were left quite big; in these instances – as long as the event had already produced enough flakes – I did not attempt to reduce every core to a small size.

The knapping was conducted in a large room, with a floor space of c. 40m² covered with two drop cloths. The initial recording strategy was to collect and bag each strike’s debitage of ≥10mm together in order to observe the fracture properties per strike and to allow for rapid conjoining to assess the fracture patterns. The bags were numbered, allowing the sequence of strikes to be noted and for sampling; each fragment in a bag was also given a number. (Using Appendix B- 2 as an example, the ‘32’ signifies that this was the 32nd strike with four fragments in the bag, numbered 32-1 to 32-4). An event’s technical procedures such as core trimming or core tapping (to loosen incipient fractures common with quartz knapping) were also bagged per technical procedure.

To an extent this collecting per strike was achieved but many smaller pieces, i.e. <20≥10mm, were not collected per strike because it proved too time consuming to collect each fragment. Consequently a lot of pieces were left on the drop cloth and bulk bagged at the end of each event. For the bipolar technique’s eight events the recording was different because the process of bipolar knapping invariably produces a series of child cores and debitage; for these events the cores and debitage were not bagged per strike but instead were bulk bagged per event.

Top of Page

6.3.2 Sampling strategy

The experimental assemblage produced over 14,000 ≥5mm artefacts and for full analysis the assemblage was sampled at a rate of at least 20%; all the sampling was conducted using the SPSS 15.0 (2006b) random number generator. The cores from the direct percussion technique events were not sampled but instead all were analysed. The direct percussion component was sampled by the individually bagged strike piles and not by the amount of debitage in each pile, i.e. if a knapping event had 10 strikes, two strike piles were sampled no matter how many pieces of debitage were contained within each strike pile. In order to avoid a bias in earlier or later phase strikes being analysed, the strikes were divided into three arbitrary phases – a beginning, middle, and end – and then sampled at 20% for each phase. A second consideration was that in one event a large piece of debitage or a core split – the child cores discussed above – was sometimes returned to and used in the same event; in these cases, the same sampling strategy of dividing the piles into three phases and sampling at least 20% was followed for the child core as well, and it was recorded that a core and debitage derived from the child core. For the strike bags that arose from the technical procedures or tapping or trimming, these were excluded from the sampling of the debitage in order to focus on the attributes of the flakes produced alone.

Because the amount of strikes would not always be conveniently divisible, a minimum sampling rate of 20% was applied (Table 6-1). The bulk collected debitage (i.e. that could not be assigned to an individual strike) was first divided into groups of ≥20mm, <20≥10mm, <10≥5mm, <5≥1mm, and <1mm and all were ≥20% sampled. The <5mm and <1mm debitage were defined by using 5mm and 1mm calibrated sieves. (Of course, the use of a 5mm sieve entailed that some pieces that were greater in length than 5mm would pass through it, as long as their width or thickness were <5mm.)

Sample group Sample rate Mean (%) Sample rate range (%)
Strike pile bags 21.10 20-25
Bulk Debitage≥20mm 29.70 20-100
Bulk Debitage <20mm 20.53 20-23.5
Bipolar 20.25 20-20.4
Bipolar <10mm 21.86 20-23.1

Table 6-1 Sample rates and ranges

In order to avoid a size bias in the sampling of the <20 mm debitage, they were divided into 10 piles (also in a manner to avoid size bias) and using the SPSS random number generator two of these groups were selected for analysis giving the 20% sample rate. For the bulk collected ≥20mm debitage these were then recorded individually at the 20% sample rate as were the <20≥10mm flakes; any <20≥10mm debris was bulk recorded (Table 6-2). Because of the small amount of bulk bagged ≥20mm pieces per event, Table 6-1 shows that a greater amount of these were invariably sampled – for two events this was 100% sampling as only one piece per event was there. For the bipolar knapping events, all ≥10mm cores and debitage were laid out in rows of 20 and random sampled at a rate of at least 20%. The same recording procedure was then followed as outlined in Table 6-2.

Size Category Debitage recording
≥20mm Debitage Full recording
<20≥10mm Debris Counted and bulk recorded
<10≥5mm Debitage Counted and bulk recorded
<5≥1mm Debitage Weighed
<1mm Debitage Weighed

Table 6-2 Debitage recording

Top of Page

6.3.3 Debitage categories

For the experimental assemblage the fragments were classified as detailed in Figure 6-7. This allowed a given artefact to be classified into a myriad of permutations of fragment plus an additional class for cases where a flake fractured sequentially – a sequential fracture occurred when a fragment was removed from either the dorsal or ventral face of the debitage. This sequential fracture sometimes occurred when a fracture had been initiated from a previous strike and subsequently the second strike removed the main flake and the previous incipient flake. Also, the use of the bipolar technique in a sense produces sequential fragments because a single strike will result in a number of flakes. The category of sequential was used to classify both a fragment and a break. A minor amount (< 1%) of the breaks were difficult to categorise and were labelled ‘?’. 26 permutations of fragment classes were noted and subsequently grouped into five fragment groups – proximal, mesial, distal, lateral, and sequential (Appendix A- 11 and Appendix A- 12). For example, if a flake was a distal right missing this became a proximal as it retained a platform; if a flake was a proximal right missing it became a proximal as it also retained a (partial) platform; a distal missing became a proximal and vice versa.

Five categories were used for debitage breaks – siret, uneven, clean, crystal plane, and sequential; the uneven category includes languette breaks (Figure 6-7). Crystal plane breaks are where breaks form along crystal planes creating a clean, smooth break surface. Because of the multiple fracture patterns of the quartz, two longitudinal and two transverse breaks were noted per artefact – diagonal breaks were treated as longitudinal. While some researchers (e.g. Odell 1989, 192; Redman 1998, 38) define a flake with a step termination as a flake fragment, these have been included here as complete flakes (11.4% (n=19) of complete flakes had step terminations).

flake attributes

Figure 6-7 Debitage fragment classes and breaks. Example for fragments: an artefact may be a 'distal right' if the only piece is from the distal right quadrant, while it may be a 'distal right missing' if the missing fragment is the distal right quadrant

Theoretically, none of the ≥10mm platform debitage produced during the experiments should be classified as debris because with the aid of conjoining of the ≥10mm debitage bagged per strike, all of the artefacts can be shown to be the result of a certain manufacture technique. In practice, however, this did not occur for three reasons:
1. As mentioned in Section 6.3.1, much of the ≥10mm debitage was not collected per strike and no attempts were made to place the bulk collected material into their respective strike piles.
2. Not all of the pieces that were collected in a single strike pile were able to be conjoined and remained ‘floating fragments’ which could either be classified as a flake fragment if they had diagnostic attributes or, as was more usually the case, classified as debris.
3. For some of the successfully conjoined flake fragments, they nevertheless appeared as debris when analysed without their conjoin siblings. In other words, only when conjoined was it recognisable that they were indeed flake fragments.

The non-conjoinable ‘floating fragments’, and the ‘conjoinable debris’ fragments, therefore, needed two entries in the database to record what was known about them and what was perceived about them. Therefore, for the category of debitage class, the artefacts’ class was recorded in two columns – the first column called ‘(A) Class’ recorded its ‘actual class’ while the second column called ‘(P) Class’ recorded its ‘perceived class’; this was also done for the artefacts’ categories of type and fragment (Appendix A- 11 and Appendix A- 12). In the following analysis, both the (A) and (P) categories are used for comparative analysis.

The <10mm debitage was not classified as flake fragments, instead these pieces were classified as debris. While this categorisation will inevitably miss some small flake fragments, this system attempted to strike a balance between a thorough analysis and limited time period with which to analyse an assemblage; these small fragments are the hardest – and most time consuming – to identify flake attributes on, therefore the time spent on classifying these small pieces can often be of limited value.

However, as noted in Section 5.6, the <10≥5mm debitage (and all debris) was subdivided into debris and slivers: ≥3mm thick were debris, while <3mm thick were slivers; the 3mm size was chosen after experimentation showed that this size appeared to create a meaningful division. The use of the term ‘chunk’ has been used by various analysts to describe what is otherwise termed debris (see Section 5.4). As Ballin (2000, 10) pointed out, using the term chunk in this way will inevitably include pieces that are not sensu stricto chunky, with a chunk generally being defined as a thick piece (Chunk 2009). While in agreement with Ballin that such a term should be reserved for pieces that are ‘chunky’ – i.e. thick – the consequence of this is that one needs to quantify the thickness of a piece. Therefore, at least two dimensions must be measured for each artefact, with a certain ratio of width/thickness, if not length/width/thickness, designated as a cut-off point for determining what ‘thick’ is, and the thinner pieces called something else other than ’chunk’ – because there must be some point at which a piece is not ‘thick’ any more. While this analysis set out to be thorough, a balance had to be struck with the input of time and the resultant outcome. Consequently, it was decided that pieces <10mm would not be measured for length/width/thickness, but instead would be categorised as debris once they were determined to be <10mm using a piece of paper with a series of measured circles allowing for rapid size sorting; for <20≥10mm pieces that were categorised as debris, these were also not measured for length/width/thickness. Therefore, these pieces were categorised in terms of a minimum and maximum dimension which allowed for rapid measuring but entailed that a width/thickness ratio could not be computed for each artefact and consequently could not be designated a certain ratio of thickness with which to define as a chunk. However, the subdivision of the artefacts into slivers and non-slivers could be done rapidly by setting a callipers to 3mm, with those that passed through the gap defined as slivers, no matter what the width/thickness ratio was.

Top of Page

6.4 Analysis

6.4.1 Debitage Fragmentation rate

Two blocks/cobbles of each quartz source material were used for each technique and support variable, with the one chert block used for H.H.D. (Appendix A- 10). The 20% sample of the quartz component of the assemblage produced 2760 ≥5mm debitage; a quarter was in the 5-10mm range, a third was debris >10mm, with the rest being flakes; no count was taken of the <5mm debitage. The total weight of the sampled debitage was 11.2kg, with the <10mm debitage accounting for 7.8% of the total weight and the majority of this being <5≥1mm (Table 6-3).

(P) Class Debitage Flake Debris ≥10mm Debris <10≥5mm Debris <5≥1mm Debris <1mm Total
Count 1139 935 686 N/A N/A 2760
% count 41.3% 33.8% 24.9% N/A N/A 100.0%
Weight (g) 9727.5 627.4 183.7 477.6 221.9 11238.1
Weight (g) 9727.5 627.4 183.7 477.6 221.9 11238.1

Table 6-3 20% quartz sample – (P) Class debitage count and weight

Figure 6-8 compares the average ≥10mm fragment per strike rate for H.H.D. and S.H.D., both including and excluding complete flakes. (This was tallied by dividing the quantity of ≥10mm debitage by the number of strikes; excluding the complete flakes was tallied by dividing the debitage by the amount of strikes, minus the complete flakes and minus the strikes that produced a complete flake.) The chert fragment rate was just 1.2 fragments per strike with the overall quartz average at 5.4 fragments per strike. Excluding the complete flakes, the average for chert is 2 per strike and 6.6 per strike for the quartz; the quartz fragment rate excluding complete flakes ranged from 3.6 to 14 fragments per strike. This chart shows clearly the significantly different fragmentation patterning between the chert and quartz. In general the P.Q. (the grainier, more sugary-textured) produced the least amount of fragments and the S.H.D. Inelastic generally produced less fragments across the material source range except for the P.Q. where the H.H.D. Inelastic produced less.

quartz fragment rate

Figure 6-8 Direct percussion. =10mm debitage per strike including and excluding complete flakes. Chert knapped with H.H.D. Elastic only

Top of Page Break and fragment types

In the last section we saw that for the direct percussion component of quartz the ≥10mm fragment rate per strike ranged from 3.6 to 14 fragments. Up to four breaks were noted per debitage, two longitudinal and two transverse, with diagonal breaks categorised as longitudinal. The breaks were noted as a primary and secondary longitudinal break and primary and secondary transverse flake; the primary and secondary flake was assigned by size, with the greater break being the primary. While clean breaks and crystal plane breaks were initially recorded separately, for analysis these were grouped together because few crystal plane breaks were noted, and these were inevitably clean breaks. For the direct percussion’s primary break (n=1174), siret breaks accounted for 21.6%, sequential for 34%, uneven for 38.4%, clean or crystal plane for 2.5%, and uncertain breaks for 3.5%. If the secondary break on the same axis was different from the primary, this was recorded – i.e. if a siret break also had an uneven longitudinal break. 20% (n=228) has different secondary breaks. For the siret with a different secondary break (n=140), 59.3% of the secondary breaks were uneven, 37.9% were sequential and 2.9% were crystal plane.

Appendix A- 13 lists the primary longitudinal and transverse breaks, with the sequential breaks grouped with uneven breaks. The bipolar technique is clearly dominant in terms of complete flakes and transverse only uneven breaks, and with lesser proportions of longitudinal or transverse clean breaks than with direct percussion. Figure 6-9 provides the break types (excluding ‘?/?’, i.e. uncertain types) with sequential, uneven, and clean breaks grouped. The significant difference with the bipolar debitage is clear, with a/breaks (transverse-only breaks) dominating followed by siret/a breaks (siret/no transverse break). S.H.D. Inelastic produced a greater proportion of a/breaks (transverse only breaks) than the other direct percussion, with hard hammer and bipolar producing more siret breaks than soft hammer. Because the break categories and grouped break categories proved unwieldy for statistical analysis, the breaks were subsequently categorised into siret and non-siret breaks to ascertain the effect of the source materials and technique/supports on the occurrence of siret breaks (Figure 6-10).

The first analysis of the siret breaks, using GZLM, examined the techniques and source materials with S.H.D. and R.Q. as the reference categories. The difference in siret proportions was significant for technique, with bipolar and H.H.D. having on average almost twice the proportion of siret breaks than S.H.D.; neither the source nor the interaction of source and technique were significant (Table 6-4 and Appendix A- 14). While the B.Q. produced a much lower proportion of siret breaks for bipolar, it conversely produced a much higher proportion for S.H.D. than the other materials.

Tests of Model Effects
χ² df p
(Intercept) 184.091 1 0.000
Technique 23.988 2 0.000
Source 0.881 3 0.830
Technique * Source 2.565 6 0.861

Table 6-4 GZLM. Dependent Variable: Siret/non-siret. Model: (Intercept), Technique,
Source, Technique*Source. Reference categories: S.H.D. and R.Q.

quartz flake breaks

Figure 6-9 Break type proportions (excluding ?/?) for technique/support

siret break proportions

Figure 6-10 Siret break proportions. Technique/support and source

The second analysis included the support variable, with S.H.D. Inelastic and R.Q. as the reference categories. While technique/support was significant, and source was not significant, in the proportions of siret breaks, the interaction of the two variables proved to be significant (Table 6-5 and Appendix A- 15). Appendix A- 15 provides the parameter estimates for the variables; compared to S.H.D. Inelastic, the proportions of siret breaks from both bipolar and H.H.D. Elastic were significantly different, but not from S.H.D. Elastic or from H.H.D. Inelastic. In terms of the source materials’ interactions with technique/support, compared to R.Q., the B.Q.’s siret proportions proved significantly different for H.H.D. Elastic and S.H.D. Elastic, but a weaker significance for the latter. For H.H.D. Inelastic, a significant difference was between the M.Q. and R.Q.

Tests of Model Effects
χ² df p
(Intercept) 244.403 1 0.000
Technique/Support 34.075 0.000
Source 2.293 3 0.514
Technique/Support * Source 36.115 12 0.000

Table 6-5 GZLM. Dependent Variable: Siret/non-siret. Model: (Intercept), Technique/Support,
Source, Technique/Support*Source. Reference categories: S.H.D. Inelastic and R.Q.

The proportions of siret breaks, therefore, occur more generally with hard hammer and bipolar than with soft hammer. However, the support used has an effect; when an anvil is used the occurrence of siret breaks is diminished for both hard hammer and soft hammer. The effect of the materials is less clear-cut, with the B.Q. both increasing and decreasing the likelihood of siret breaks.

In terms of the fragment types, the clearest difference between the techniques was the lateral fragments which are related to siret breaks – the majority of lateral fragments result from a siret break; occasionally, a lateral fragment will be formed by a fracture away from the impact point. S.H.D. produced less lateral fragments than the H.H.D. and bipolar (Appendix A- 16). The bipolar technique produced the greatest proportion of complete flakes at 26.9% of the debitage; for H.H.D. and S.H.D., complete flakes ranged from 4.1% to 8.2% of the debitage. Appendix A- 16 and Appendix A- 17 compare the two categories given for each artefact in terms of their (A) Class and (P) Class as outlined in Section 6.3.3. For H.H.D. and S.H.D. 70% of the (A) Class sequential flakes were categorised as (P) Class debris, as were 36% of the mesial flakes and 16% of the distal flakes; less than 1% of the proximal or lateral fragments were considered (P) Class debris. These few flakes with partial platforms that were designated as (P) Class debris were indistinguishable because the partial platform was devoid of impact marks or fissures. As the bipolar debitage were not bagged per strike, it was difficult to designate them as being sequential, hence the apparent lack of sequential fragments; also the bipolar flakes have a high proportion of ‘fragments’ – these were fragments that were probably mesial or sequential fragments that were uncertain, but not initially categorised as debris; for the (P) Class these were subsequently categorised as debris. Therefore, by far the most problematical artefacts for identification are sequential fragments followed by mesial fragments.

Appendix B- 2 shows a fragmented H.H.D. flake, E1-32, detailing the fragments on the left and the re-conjoined flake on the right. 32-1 is a sequential fragment from the dorsal surface which lay over the two siret break fragments. 32-2, 32-3, and 32-4 are siret breaks, with 32-3 and 32-4 also having uneven breaks. The sequential fracture here occurred presumably from an incipient fracture formed from a previous strike; the sequential fragment lay over the zone of the siret break, but did not split in two, but instead popped off with a siret break snapping it in two. Appendix B- 3 shows a fragmented H.H.D. flake, number E1-45, showing the lateral left fragment’s siret break; the centre and right hand of the image shows the lateral left fragment re-conjoined with the lateral right fragment.

Appendix B- 4 is a fragmented S.H.D. flake, number E1-12-10, detailing the fragments on the right and bottom left, with the re-conjoined flake on the upper left. This fracturing followed a pattern of a pair of sequential breaks of overlapping lateral fragments – E1-12-10-1 contained the impact point on the platform. This pair of overlapping sequential flakes fractured along a similar fracture line and then also broke transversely. The transverse break was an uneven break and the exact conjoin fit was impossible as the transverse conjoin zone fragmented into multiple <10mm fragments, resulting in a gap in the image of the four re-conjoined flakes.

Top of Page Break and fragment types

Complete and proximal flakes are generally seen as flakes which retain the most information for lithic analysis. In Knutsson’s (1988a, 91) quartz experiments, he noted that “irrespective of reduction strategy, diagnostic flakes, that is flakes with a preserved striking area, account for approximately 16-19% of the total number of flakes on a ‘fresh’ quartz knapping floor”. It is unclear whether his ‘total flake’ count includes what is categorised here as debris, and also whether Knutsson means both complete platform and partial platforms when calling it a ‘preserved striking area’ as both could apply.

For this experiment series, 11.7% (n=133) of the flakes were complete and 35% of the ≥10mm debitage had complete or partial platforms, with 12% having complete platforms which somewhat corresponds to Knutsson’s range of 16-19%, albeit quite lower than his range. Unlike Knutsson’s results however, there was a significant difference between the bipolar and direct percussion techniques; as least twice as much bipolar debitage retained a complete or partial platform (Table 6-6). This does not correspond with Knutsson’s research, but he does not state what size of ‘flakes’ he included in the tally and he may have counted bipolar debitage <10mm that was not counted here which may explain the significant discrepancy.

Platform Bipolar
(n-193) (%)
H.H.D. Elastic
(n=401) (%)
H.H.D. Inelastic
(n=604) (%)
S.H.D. Elastic
(n=430) (%)
S.H.D. Inelastic
(n-446) (%)
(n=2074) (%)
Absent 31.1 65.3 67.5 71.6 67.9 64.7
Present, complete 42.5 6.5 9.1 7.4> 13.7 12.3
Present, fragment 26.4 28.2 23.3 20.9 18.4 23

Table 6-6 Absent/present platform on all ≥10mm quartz debitage

Morphologically, the quartz direct percussion platforms can often take the same form as the chert comparison assemblage – having a convex ventral platform edge. Appendix B- 5, Appendix B- 6, and Appendix B- 7 give eight examples of characteristic platforms and one characteristic pseudo-platform. These three figures do not follow typical conventions for lithics (see Martingell and Saville 1988) as they are aimed specifically at certain characteristics, and therefore are presented ‘upside-down’ (or more specifically ‘upside-up’ to how it’s knapped) and angled to show the platform in plan and in perspective to the ventral/dorsal surface.

E1-4-37 (Appendix B- 5, bottom right) is a typical direct percussion platform with a convex platform edge and the impact point on the edge. The impact point is a distinctive whitened area, formed from the multiple fractures at the impact point partially filling with quartz dust and the micro- and macro-fracturing leading to an increased opacity; this distinctive, whitened impact area can also form on the core from which the flake is detached; this can be seen on the E1-4-37’s platform dorsal face, formed from the previous flake removal. (Often, however, the impact point is not on the ventral platform edge, resulting in no whitened area on the core). The fractures at the impact point zone are partially made up of radial fissures. Along with the radial fissures, transverse platform fissures can form just below the impact point; if these fissures fracture completely, step fractures form on the platform. This flake also exhibits edge damage formed during flake formation – here, a fracture is formed running from the number ‘3’ on the image, diagonally up to a few mm below the right side of the platform, to the point where a slight indentation can be seen. This type of edge damage and fracture formed during flake formation is common, and the fracture can lead to the flake fragmenting completely rather than just an incipient fragment as in this example. This type of flake formation edge damage can be difficult to discern from edge damage created during subsequent use or post-deposition. Moreover, this fracture inevitably leads to a plane of weakness whereby the flake will more easily snap during use or post-deposition.

E1-3-9-1 (Appendix B- 6, left) is an example of a characteristic triangular-shaped fracture. The apex of the triangle is at the point of impact; the fracture lines are visible on the platform and the ventral surface, continuing down to the flake’s distal end; a second, subtler, triangular fracture occurs on the dorsal side of the impact point, also radiating from the impact point. In some cases these incipient triangular fractures are completed, whereby the platform fragments, leaving a distinct triangular-shaped platform ventral edge, as in the example E1-29-16 (Appendix B- 5, top left); here the impact point has been crushed, with the distinctive triangular edges angled towards the dorsal face. E1-37-17 (Appendix B- 7, right) provides an example of a morphologically similar triangular platform created using a bipolar technique. While many platforms have the triangular form with a reasonably sharp apex, in other cases the triangle’s apex is more rounded, appearing as a convex fracture; if this type fragments, the fragment inevitably appears as a convex fragment.

E1-6-29 (Appendix B- 6, right) is an example of a sequential break where the platform fractured in two, in effect creating two substantial flakes, with most of the impact point absent due to shattering (a small gap in the conjoined flake, representing the collapsed impact point, is somewhat visible in the image). The sequential break’s fracture line has been darkened in the image for clarity, with the two separated conjoins at the bottom of the image. The aim for this flake removal had been for the flake to detach at the point of impact, resulting in the first of the two flakes, but the flake formation continued along a plane of weakness further in towards the core resulting in the second, even bigger, flake also being detached at the same time. The resultant flake fragments could easily be interpreted as two separate, complete flakes with complete platforms.

The pseudo-platform on E1-15-4 (Appendix B- 5, bottom left) is another example of misleading flake morphology. Here, a distal fragment has edge damage created during flake formation, with the flake fracturing transversely; this formation is similar to that discussed in relation to E1-4-37 above. This fracturing created a distal fragment whose proximal end is an apparent ‘platform’ with an apparent ‘impact point’ on the edge; the ‘impact point’ has a characteristic fracture running to the dorsal face and has radial fissures. This distal fragment could easily be identified as a complete flake with flat platform and impact point on edge.

E1-40-42 (Appendix B- 5, top right) is an example of a soft hammer impact mark. The soft hammer created a small, compact impact mark with less crushing and radial fissures, and with a diffuse bulb. These bulbar ventral surfaces were uncommon with quartz, with more occurring on B.Q. than other materials and also more occurring while using a soft hammer than hard (see below p. 128).

Looking at the bipolar platforms, a characteristic feature is that they tend to be rounded, with the steeper side of the roundedness on the ventral face. A common feature is for the platform angle to be reversed compared to direct percussion debitage – E1-22-21 and E1-37-24 (Appendix B- 7, left and centre) both show the platform angles dipping towards the dorsal face. E1-22-21 is a complete flake with the impact point at the edge of the flake, with the remaining platform appearing flat; E1-37-24 is a siret break. As mentioned, E1-37-17 is a fragmented platform that is morphologically more similar to a direct percussion strike. In addition to the impact points, radial and transverse fissures on platforms, and the edge damage marks noted above, non-proximal radial fissures are a consistent pattern on the quartz. However, these fissures occur regularly on non-knapped quartz and therefore are not by themselves indicative of knapped quartz.

The images highlight that the impact point on quartz flakes is generally well defined as a whitened area of multiple fractures, often with radial and transverse fissures. Nevertheless, the impact point is not always clearly visible and can be a more ephemeral marking than the examples provided, and in some cases not visible at all. While the vast majority of the complete hard hammer and bipolar flakes retained a visible impact point, with all materials equal in their proportions, the soft hammer had a significantly lesser proportion (Table 6-7). The lack of a whitened area with some of the soft hammer platforms is a result of a lack of micro-fractures and macro-fractures forming.

Technique Impact point
Present Absent Total Present Absent
Count Count Count % %
Bipolar 51 2 53 96.2 3.8
H.H.D. 44 2 46 95.7 4.3
S.H.D. 26 7 33 78.8 21.2
Total 121 11 132 91.7 8.3

Table 6-7 Impact point presence on complete flakes

For the direct percussion debitage, the partial platforms were recorded as either fragments or collapsed. Platform collapse is defined as the platform collapsing at the point of impact, and platform fragment when the platform fractured away from the point of impact; siret platforms are not treated as platform collapse, as these are splits rather than collapse, although some present as collapsed at the split. There were 173 debitage with complete platforms and 420 with partial platforms (Table 6-8).

Platform fragment H.H.D.
Count Count Count Count Count
Non-siret fragment 34 47 41 39 161
Siret fragment 68 78 45 27 218
Collapsed 9 15 2 15 41
Total 111 140 88 81 420
% % % % %
Non-siret fragment 30.63 33.57 46.59 48.15 38.33
Siret fragment 61.26 55.70 51.14 33.33 51.91
Collapsed 8.10 10.71 2.27 18.52 9.76

Table 6-8 Platform fragment types. Direct percussion

Statistical analysis comparing the proportions of platform collapse for technique/support with source material proved not to be testable; therefore analysis was conducted on technique and source. Using GZLM, with S.H.D. and R.Q. as the reference categories, there was no significant difference for collapse proportions for technique, source, nor for the interaction of technique and source (Appendix A- 18). However, when compared with support (Table 6-9), with Inelastic as the reference category, the difference is collapse proportions was significant (Appendix A- 19), with Elastic support having significantly less platform collapses than Inelastic. The analysis was conducted a second time excluding siret break, i.e. comparing non-siret platform fragments to collapsed platforms. For these, the difference between supports was also significant (Appendix A- 20).

Platform fragment Support
Elastic Inelastic Elastic Inelastic
Count Count % %
Siret fragment 113 105 56.78 47.51
Non-siret fragment 75 86 37.69 38.91
Collapsed 11 30 5.53 13.58
Total 199 221 100.00 100.00

Table 6-9 Platform fragment types. Direct percussion support

Conchoidal fracturing is considered a key fracture type of humanly struck lithics, with the resultant bulbs and compression rings the obvious signature marks. As noted in the previous chapter (Section 5.6), however, bending fractures also play a key role especially when using soft hammers. For vein quartz, a significant part of the fracture initiation stems from fracture planes already in existence in the raw material; this is especially the case in coarse-grained quartz such as that used in these experiments, which contained numerous major and minor fractures.

While compression rings were looked for, just two flakes had visible compression rings. Both were R.Q. and were produced by both soft and hard hammer; all but one of the chert flakes had compression rings. The analysis therefore focused on bulbs. Table 6-10 lists the percentage of complete and proximal quartz flakes that exhibited bulbs. The bipolar flakes are excluded as just one bipolar flake had a bulb. Overall 5% (n=31) had visible bulbs; for the chert, which is not listed in the table, 88% (n=16) had visible bulbs. Analysis was conducted with GZLM, with chert as the reference category. The difference between the bulb presence on chert and quartz was significant (Appendix A- 21). Looking at the quartz alone, while bulbs occurred more frequently with S.H.D. and B.Q., there was no significant difference for the materials, techniques, nor interaction of the two for the proportions of bulb presence (Appendix A- 22).

Source Bulb H.H.D.
(n= 139) (%)
(n=196) (%)
(n=122) (%)
(n=143) (%)
(n=600) (%)
Beach Present 6.3 12.5 10.0 9.4 9.2
Metadolerite Present 6.5 - 11.6 - 4.1
Psammite Present 6.5 - 7.7 - 2.9
Rose Cottage Present 3.3 6.3 4.7 7.9 5.7

Table 6-10 Presence of bulb – complete and proximal quartz flakes

This lack of visible bulbs, however, does not imply that only 5% of the quartz initiation fractures were conchoidal, but rather, that the visible characteristics of conchoidal fracturing are less apparent on the quartz. This was especially the case for compression rings, which were not apparent at all. This suggests that while conchoidal fracturing is an important signature to look for in a quartz assemblage, low numbers are to be expected. While at a micro-scale, the quartz crystals fracture conchoidally, in the aggregation of crystals in the xenomorphic quartz a clear fractal pattern does not occur as with cryptocrystalline quartz. On many artefacts, the debitage fractured away from the point of impact as happens with bending fractures; but often this is the result of an already existent fracture plane or plane of weakness in the material, rather than an example of a bending fracture in cryptocrystalline materials as described by Cotterell and Kamminga (1987).

As noted, the aim for the direct percussion was to produce medium-sized flakes, defined here as flakes with a length range of 30-90mm. Figure 6-11 and Figure 6-12 compare the 26 complete chert and quartz H.H.D. Elastic flakes. The mean length of the chert flakes was 50mm, with the mean length for the quartz flakes similar to the mean chert flakes’ width; on average, the quartz flakes have a length/width ratio of 1:1; while of lesser length and weight than the chert, the quartz flakes were thicker (Appendix A- 23). The chert and quartz flakes were compared using ANOVA. The differences between the chert and quartz flakes’ length, and the ratios of length/width and length/thickness, were significant, while differences for the remaining variables were not significant (Appendix A- 24). Figure 6-13, Figure 6-14, Appendix A- 25, and Appendix A- 26 compare the complete quartz flakes by material and technique/support. Overall, S.H.D., and especially S.H.D. Elastic, produced bigger, thicker flakes with a greater length/width ratio. Few flakes had a length/width ratio of 2:1 or greater. The S.H.D. flakes’ length range is the greatest and overall P.Q. gave the smallest length range (Figure 6-14). That the S.H.D. produced the bigger, thicker flakes goes against the usual perception that using a soft hammer will produce flakes of a diminutive size compared to a hard hammer (e.g. Kooyman 2001, 78-81).

complete hard hammer flakes

Figure 6-11 Means. Chert and quartz complete H.H.D. Elastic flakes

complete hard hammer flakes

Figure 6-12 Chert and quartz complete H.H.D. Elastic flakes. Length/width ratio

complete quartz flakes

Figure 6-13 Length/width ratio. Complete quartz flakes

complete quartz flakes

Figure 6-14 Means. Complete quartz flakes

Analysis was conducted using UNIANOVA. This comparison of the means of the complete quartz flakes – using the log transformation of the metrics because of non-normal distributions – showed no significant difference for length, width, or weight for technique/support or source material (Appendix A- 27). The difference in thickness was significant; post hoc tests using LSD showed a significant difference (p = 0.016) between S.H.D. Elastic and H.H.D. Inelastic, but using Bonferroni (p = 0.095) or Tukey’s HSD (p = 0.073) did not. Therefore, while the more liberal post hoc test proved significant, the more conservative ones did not. In comparing the means of platform width and thickness, all complete and flake fragments with complete platforms were included. For the 174 platform flakes with complete platforms, there was no significant difference for the different source materials or technique/supports for either platform width or thickness (Appendix A- 28). Figure 6-15 highlights that many flake fragments’ length was greater that the mean length of complete flakes. This boxplot[14] provides the grouped fragments whereby ‘proximal’ include distal missing fragments and ‘distal’ include proximal missing fragments – both these fragment groups can invariably be of substantial length, with just a short distal or proximal fragment fractured off. This is especially so for the proximal fragment group. A distinction is apparent with the lateral fragments, with the S.H.D. lateral fragments having a smaller range, and tending to be shorter than the H.H.D.

quartz flakes boxplot

Figure 6-15 Boxplot. Complete, lateral, proximal, and distal flakes

The aim with the bipolar technique was to produce a series of small flakes, taken here as ranging from 10 to 29.9mm in length. As mentioned, the bipolar technique produced a substantially greater proportion of complete flakes than direct percussion. For the complete bipolar flakes, the P.Q. produced smaller, lighter flakes, as it had with the direct percussion, and the length/width ratio was less than with the other materials (Figure 6-16, Figure 6-17 and Appendix A- 29). However, the difference for weight, length, width, or thickness between the source materials was not significant (Appendix A- 30). Overall, a greater number of bipolar flakes had a length/width ratio approaching 2:1 and greater. Similarly to the direct percussion, the B.Q. provided the overall greatest length/width ratio for the bipolar flakes. In terms of the flake curvature of complete flakes (Appendix A- 31), 55.8% of the bipolar flakes were straight. Compared to the platform flakes, this fell in the upper range but below the S.H.D. Inelastic which produced 59.1% straight flakes; S.H.D. Elastic produced 45.5% while H.H.D. Elastic and Inelastic produced between 20-22.6%.

quartz bipolar flake

Figure 6-16 Means. Complete bipolar flakes

quartz bipolar flake ratio

Figure 6-17 Length/width ratio. Complete bipolar flakes

Top of Page Debitage types

Overall, the 20% sample analysed of the H.H.D. and S.H.D. quartz knapping produced 2482 ≥5mm artefacts and 1881 ≥10mm artefacts (Table 6-11). Appendix A- 32 gives a breakdown for (A) Type platform debitage by source and technique/support, highlighting that unlike the chert component, there were no eraillure flakes noted for the quartz material; while the majority of the chert flakes were regular flakes, the majority of the quartz flakes were irregular flakes; the M.Q. produced the greatest proportion of regular flakes (4.2% of the flakes) with the other three source materials producing similar, smaller proportions (average of 2.8% of the flakes). This low proportion of regular quartz flakes highlights the coarse-grained nature of the raw materials. Overall, the B.Q. produced the least proportion of debris. Appendix A- 33 provides a breakdown for (A) Type platform debitage excluding source.

20% sampled H.H.D. and
S.H.D. quartz debitage
Debitage ≥5 mm 2482
Debitage ≥10mm 1881
Flake ≥10mm bulk bagged 117
Debris ≥10mm bulk bagged 636

Table 6-11 Count of 20% sampled debitage

20% sampled H.H.D. and
S.H.D. quartz debitage (P) Class
Count %
Flake ≥10mm 993 52.8
Debris ≥20mm 38 2.0
Debris <20≥10mm 524 27.9
Sliver <20≥10mm 326 17.3
Total 1881 100.0

Table 6-12 20% sampled direct percussion – (P) Class debitage

52.8% of the ≥10 mm debitage was classified as (P) Class flakes (Table 6-12 and Appendix A- 34). This highlights that almost half of the ≥10mm debitage produced during the experiments was, without the aid of conjoining, non-diagnostic fragments. Almost a fifth of the ≥10mm debitage was slivers (debris with a max. thickness of <3mm) and slivers accounted for over a third of the ≥10mm debris. 60% (n=1125) of the ≥10mm debitage was collected in strike piles and 22.4% (n=252) of these fell into the category of conjoinable fragments but perceived as debris, i.e. a part of the (P) Class debris category. This substantial amount, at 22.4%, of conjoinable but otherwise unrecognisable fragments highlights the difficulty in determining a lot of the quartz debitage without the aid of conjoining. Only 2% of the debitage was classified as >20mm debris. 2.6% of the platform debitage was categorised as (P) Type bipolar flakes (Appendix A- 34); this accounted for 4.6% of the (P) Type flakes, highlighting that a small minority of platform flakes can take the appearance of bipolar reduction, especially in the case of H.H.D. Elastic.

The 20% sample of the bipolar debitage produced 280 ≥5mm debitage and 193 ≥10mm debitage (Table 6-13). Table 6-14 looks at bipolar debitage ≥10mm. Just as some of the platform flakes were interpreted as bipolar flakes, some bipolar flakes were interpreted as having attributes of platform flakes. For the bipolar flakes, however, a greater proportion appeared as platform flakes – at 17.1% of the total ≥10mm debitage and 22.6% of the bipolar flakes. Nevertheless, the bipolar technique produced more diagnostic debitage than the direct percussion techniques – just a quarter of the assemblage was classified as debris, compared to almost half of the direct percussion (see Table 6-12).

20% sampled bipolar
debitage (P) Class
Count %
Flake ≥10mm 146 52.1
Debris ≥20mm 3 1.1
Debris <20≥10mm 24 8.6
Sliver <20≥10mm 20 7.1
Debris <10≥5mm 87 31.1
Total 280 100.0

Table 6-13 20% sampled bipolar debitage

20% sampled
bipolar (P) Class
Count %
Irregular bipolar flake 113 58.5
Irregular platform flake 33 17.1
Debris ≥20mm 2 1.0
Sliver ≥20mm 1 0.5
Debris <20>10mm 24 12.4
Sliver <20>10mm 20 10.4
Total 193 100.0

Table 6-14 20% sampled bipolar debitage ≥10mm

During the analysis it was noted that the <20mm debris consisted of numerous thin fragments, therefore it was decided to sub-divide the debris into two categories, the general debris and the slivers, which are <3mm in max. thickness to ascertain any differences between the materials and the technique/supports. Analysis was conducted with GZLM, using S.H.D. Inelastic and R.Q. as reference categories. In terms of the proportions of debris/sliver for <20≥10mm, while bipolar produced a slightly greater proportion of slivers, there was no significant difference between the technique/supports or sources (Table 6-15).

Tests of Model Effects
Source Type III Wald χ² df p
(Intercept) 20.973 1 0.000
Tech/Support 7.383 4 0.117
Source 3.800 3 0.284
Tech/Support * Source 17.365 12 0.136

Table 6-15 GZLM. Dependent Variable: Debitage P Sub-type <20≥10mm Model:
(Intercept), Tech/Support, Source, Tech/Support * Source.
Reference categories: S.H.D. Inelastic and R.Q.

However, the support used had an effect on the proportions. Looking at the direct percussion proportions alone (using Inelastic and R.Q. as reference categories) while the source did not have an influence on the proportions, the differing supports did, as did the interaction of support and source (Table 6-16). The inelastic support produced a greater proportion of slivers overall, but the M.Q. reversed the trend with the largest proportion of slivers produced with elastic support and the smallest proportion produced with inelastic support (Appendix A- 35). Therefore, while the use of an anvil influenced the proportions of slivers, the M.Q. proved to reverse the pattern compared to the other materials.

Tests of Model Effects
Source Type III Wald χ² df p
(Intercept) 40.706 1 0.000
Support 5.766 1 0.016
Source 2.980 3 0.395
Support * Source 8.402 3 0.038

Table 6-16 GZLM. Direct percussion only. Dependent Variable: Debitage P Sub-type <20≥10mm
Model: (Intercept), Support, Source, Support * Source. Reference categories: Inelastic and R.Q.

Top of Page Debitage types

The analysis of flake curvature examined all ≥ 20mm debitage, providing a sample of 613 quartz debitage. While the majority of curved flakes were convex, a small number of direct percussion flakes exhibited a concave curvature (Table 6-17). Appendix A- 36 provides the proportions of curvature with concave and convex flakes grouped. For the quartz debitage, the bipolar debitage generally produced more straight flakes but this varied considerably by source material. Analysis was conducted with GZLM, using S.H.D. Inelastic and R.Q. as reference categories. Overall, none of the variables’ effect on the curvature proportions was significant (Table 6-18). While the chert component is not tallied in the appendix, all of the chert flakes were convex, highlighting a clear distinction on the materials.

Technique Curvature
Concave Convex Total Concave Convex
Count Count Count % %
Bipolar - 22 22 - 100.0
H.H.D. 11 186 197 5.6 94.4
S.H.D. 10 136 146 6.8 93.2
Total 21 344 365 5.8 94.2

Table 6-17 Concave/convex proportions for ≥ 20mm debitage.
For curved/straight proportions see Appendix A- 36

Tests of Model Effects
χ² df p
(Intercept) 8.488 1 0.004
Technique/Support 8.724 4 0.068
Source 1.942 3 0.585
Technique/Support * Source 16.348 12 0.176

Table 6-18 GZLM. Dependent Variable: Curvature grouped. Model: (Intercept), Technique/Support,
Source, Technique/Support * Source. Reference categories: S.H.D. Inelastic and R.Q.

While technically quartz does not form a cortex, the term cortex is used here to describe the exterior surface of quartz which becomes altered due to weathering, natural abrading, and so forth. Figure 6-18 provides the proportions of cortex for quartz debitage, showing a clear correlation with source material and cortex. The B.Q. had the greatest proportion of cortical debitage, while the quarried R.Q. had the least. While the M.Q. and P.Q. were also quarried, the original blocks were of a smaller size than the R.Q., and therefore had a greater proportion of cortex; the M.Q. was sourced from the inter-tidal zone, which therefore produced a more distinctive cortex than the P.Q. Looking at the B.Q. alone – which contained the greater proportion of cortex – for all techniques, complete flakes were less likely to occur at the initial and latter stages of knapping, i.e. flakes with either 100% or 0% cortex; especially, few 100% cortex complete flakes (Appendix A- 37).


Figure 6-18 Cortex by source. Quartz debitage

The flake terminations were dominated by feather terminations (Figure 6-19). Unsurprisingly, no bipolar flakes had plunging terminations, which otherwise occurred frequently for the other technique/supports, especially with S.H.D. Inelastic with a subsequent reduction in feather terminations; this however, changed with the source material. While no step terminations occurred for H.H.D. on B.Q., they occurred frequently with S.H.D. on the same material. Overall no clear patterns could be discerned for technique/supports and materials on the proportions of terminations – apart from the lack of bipolar plunging flakes and the lack of chert flakes with irregular terminations.

flake terminations

Figure 6-19 Flake terminations. Chert knapped with H.H.D. Elastic only

Top of Page

6.4.2 Cores

All of the direct percussion cores were analysed and the bipolar cores were sampled at 20%. The 32 direct percussion quartz events produced an assemblage of 49 complete cores and 13 core fragments (Appendix A- 38). The 13 platform core fragments were derived from five cores – three cores fragmented into two pieces, one fragmented into three, and one fragmented into four pieces. Therefore, overall 10% of the platform cores fragmented, which gave an assemblage consisting of 21% of core fragments. The S.H.D. Inelastic events did not produce a fragmented core; neither did the B.Q. or P.Q. materials.

The term ‘core fragment’ occurs regularly in lithic analysis but it is hard to find this term defined; it would appear that a ‘core fragment’ is a self-evident entity – a core, yet not a complete core. A moot point is whether a ‘core fragment’ should simply be considered as debitage. A core can be described as a “block of raw material from which flakes...have been struck” (Inizan et al. 1999, 137) and debitage can be described as “all removals resulting from the knapping of a core” (Inizan et al. 1999, 138). So, in a sense all debitage are fragments of cores – but there is nevertheless a distinction between a ‘fragment of a core’ and a ‘core fragment’. In this series of experiments of platform percussion, the core fragmentation occurred post knapping, i.e. not in direct relationship to a strike. Of course, the fracturing was a consequence of a series of earlier strikes which had produced incipient fractures along pre-existing planes of weakness/fracture lines. Therefore, these core fragments were not designated as debitage. Without this knowledge of the timing of the fracturing, it would be easy in some cases to designate some core fragments as debitage, thus leading to an under-recording of cores in an assemblage.

The H.H.D. Inelastic core (Appendix B- 8) is an example of core fragmentation – fragments E1-24-81-2, 81-3, and 81-4 were held in my hand while fragment 81-1 was also held in my hand with its distal surface resting on the anvil; the last strike occurred on the opposite end of the proximal surface from where the fragmentation subsequently occurred, approximately where the number ‘81’ appears in the image. Once the last flake was struck, the core was manoeuvred to set up for the next strike at which point the core fragmented in my hand into one large fragment and three smaller fragments. Appendix B- 8 shows on the top left the plan of the re-conjoined core; bottom left shows the profile of 81-1, with the three images on the right, from the top, shows the sequence of the re-conjoining of the three fragments onto 81-1.

The platform core types produced were primarily multiplatform cores – 54 complete cores and core fragments – with four each of single platform and dual, opposed complete and core fragments (Appendix A- 38). Appendix A- 38 highlights that the core fragment types do not match – this is an inevitable result of the visible flake scars on the fragments showing a different patterning than how the whole core had been knapped.

The mean initial weight of the block/cobble was 2.2kg, with the beach cobbles generally the smallest used. The average resultant core weight for the complete cores is 183.3g (Figure 6-20 and Appendix A- 39). While the S.H.D. cores were generally larger, as were the S.H.D. debitage, the greater variability related to the materials, with the B.Q. and M.Q. resulting in below average core sizes. Besides size, the distinction between core types related to the impact marks visible on the Inelastic support groups, formed by the impact with the anvil – all the Inelastic cores had distal impact marks apart from the M.Q. core which fragmented into four pieces, none of which had visible distal impact marks.

The 20% sample of the bipolar cores amounted to 64 complete cores and 41 core fragments (Appendix A- 40). For the complete cores, the average masses were similar across the materials apart from the M.Q. which were thinner and lighter (Figure 6-21). Of the 64 complete bipolar cores, all but one had distal impact marks. Knutsson (1988a, 90-1) noted that the conical piece was a diagnostic quartz artefact, which he stated does “not necessarily bear a close relationship to different reduction stages or strategies...[but] can occur at any time...irrespective of the applied method of reduction”; he described the conical pieces as
"hav[ing] one point of impact formed by several converging, negative and positive fracture facets. The point of impact may be crushed. Opposite this point of impact is a flat surface created by an already existing fracture in the piece of raw material or natural plane. These pieces are produced when a detaching blow crushes the 'flake' and at a point where the fracture front meets a flat fracture face that runs perpendicular to the line of the detaching blow".

While it appears that Knutsson found instances of conical pieces occurring with platform reduction, no conical pieces were produced with the direct percussion technique and method of flake removal used during these experiments, at least not out of the 2000 artefacts that formed the 20% sample of platform reduction. In these experiments all of the conical pieces were the result of bipolar reduction, whereby, as described by Knutsson above, the bipolar core fractured along a fracture plane perpendicular to the impact point, creating a conical piece (Appendix B- 9). Consequently, they have been categorised as bipolar core fragments, with the awareness that this differs from Knutsson’s designation. Nine conical pieces were noted out of the 20% sample, accounting for 8.6% of the total sample of cores and 22% of the core fragments (Appendix A- 40).

platform cores

Figure 6-20 Complete platform cores - initial weight and core weight. Reference line at 183g indicates average core weight

quartz bipolar cores

Figure 6-20 Complete bipolar cores - mean length, width, thickness, weight

Top of Page

6.5 Overview and discussion

The aim of the experimental knapping of vein quartz from Belderrig was to produce an assemblage of artefacts using various techniques and supports with which to investigate the fracture mechanics of the materials, and to use the experimental assemblage to compare with archaeological assemblages. Over 14,000 ≥5mm artefacts resulted from the 40 knapping events of 70kg of quartz from three quartz veins and beach cobbles. A series of psammite hammerstones and a deer antler were used as impactors. One block of chert was knapped to use as a baseline comparison.

The thin sectioning of the quartz has shown that the Belderrig quartz sources, while variable in terms of crystal orientation and fracture development, are all coarse-grained quartz of massive habit, with multiple macro- and micro-fractures. While all the quartz is coarse-grained, the minority of it appears as sugar-grained/textured, with the majority appearing as smooth-grained and it can be difficult to distinguish the grain/crystal boundaries from macro-fractures.

Two blocks/cobbles of each source – B.Q., P.Q., M.Q., and R.Q. – were knapped for each technique/support with the direct percussion debitage bagged per strike and the bipolar artefacts bulk bagged after each event. The assemblage was then random sampled at a rate of at least 20%, providing a quartz assemblage of over 2700 ≥5mm debitage for full analysis, weighing 10.5kg along with 700g of <5mm debitage.

The per strike fragmentation rate for ≥10mm debitage of the direct percussion quartz was dramatically different to the chert component. While the chert fragment rate was 1.2 per strike, the quartz was 4.5 times greater at 5.4 fragments per strike. In general the P.Q. – which is the grainier, more sugary-textured quartz – produced the least amount of fragments per strike and the soft hammer inelastic also produced the least amount. The significant difference between the chert and quartz is clear, as are the implications for analyses of archaeological assemblages – if the differences in the fracture mechanics of the various materials used are not taken into account, misleading interpretations will inevitably result; if an assemblage which consisted of equal knapping of chert and quartz cores is tabulated, it will inevitably appear to be dominated by quartz debitage due to the significantly greater fragmentation rate, and a tabulation of complete flakes will make the chert component appear to dominate. Moreover, the relative ease of ‘reading’ the chert component can lead to the chert appearing as a more carefully crafted component of an assemblage, with the quartz knapping appearing as unstructured and, as Lindgren (1998) put it, without shape. However, the shapes of quartz knapped during this series of experiments are relatively predictable.

In order to ascertain the differences in what was actually being produced during the knapping, and how the assemblage would be perceived without this knowledge, the debitage types, classes and fragments were categorised twice – as the A Type, A Class, and A Fragment, and then as the P Type, P Class, and P Fragment. For the A Class, debris accounted for 30% (n=628) of the =10mm debitage, while for the P Class the debris jumped to 45% (n=934). This increase consisted mainly of sequential and mesial flakes, highlighting that these two fragment types are the most difficult to identify with certainty and will therefore be under-recognised in assemblages. The fact that almost half of the experimental debitage was categorised as debris could make one despondent with attempting to analyse assemblages produced on coarse-grained vein quartz such as that from Belderrig, and view these ‘shapeless’ fragments as no more than gravel. Nevertheless, while this 45% of debris is substantial, and in many ways disquieting, the remaining 55% is more than amenable to analysis, and clear patterns of debitage formation and breakage can be discerned. As noted, sequential fragments are amongst the hardest to identify in the assemblage, as were sequential breaks. Because of this difficulty in identifying these breaks and fragments consistently, these categories were excluded during the analysis of the archaeological material (see Chapters 9 and 10).

The analysis of the complete and platform flakes has shown that a number of characteristic platform attributes are identifiable, albeit with a number of complicating factors such as edge damage that forms and which can result in pseudo-platforms forming on non-proximal flakes. For all ≥10mm debitage, 35% had complete or partial platforms, with 12% having complete platforms. The impact point on quartz is typified by a whitened area, formed by micro- and macro-fractures which increase the opacity in the area and also partially fill with quartz dust; the whitened impact point is found on the detached flake’s platform and often on the core if the impact point was at the edge of the platform. However, the impact point may be more ephemeral in many cases and in some case not visible at all, as with a sizeable minority of the soft hammer flakes. Along with the impact points, radial and transverse fissures form on the platform, which can result in full fractures if they develop significantly. As well as these fissures, more substantial fractures can develop, with a signature fracture being the triangular fracture with the triangle’s apex forming at the impact point; another triangular fracture can often form radiating towards the dorsal face of the platform as well. The triangle’s apex is often less acute, resulting in more rounded and appearing as a convex fracture. If these fractures develop fully, a triangular-shaped platform fragment is formed, or a more convex shape if the apex was less acute. Another characteristic platform is formed during a sequential break, where the flake breaks into a number of flakes, which some or all can take the appearance of complete flakes, with complete platforms and flake terminations. Another complicating fragment is where a pseudo-platform is created by a transverse break, formed by edge damage during flake formation – these ‘platforms’ appear to have an impact point and radial fissures, and can appear to be complete flakes. The bipolar flake platforms are generally characterised by a rounded platform with the steep side on the ventral face of the flake, with the platform angle reversed compared with direct percussion platforms. As with the direct percussion platforms, bipolar platforms can also fracture in a triangular fashion, leading to triangular-shaped platform fragments. For both bipolar and direct percussion, a less frequent occurrence was bulbs, with compression rings only noted on two direct percussion flakes. Just 5% of the flakes had bulbs, and no pattern was discerned for the occurrence according to technique/support or source, apart from bipolar which had practically none.

The distinction between bipolar flakes and platform flakes was generally clear. Nevertheless, a sizable proportion of bipolar flakes appeared as platform flakes. 17.1% of the ≥10mm bipolar debitage and 22.6% of the bipolar flakes appeared to be platform flakes, while 2.6% of the ≥10mm platform debitage and 4.6% of the platform flakes appeared as bipolar flakes. This highlights that it is more likely that bipolar flakes will be underestimated in an assemblage compared to platform flakes.

Compared to the direct percussion, the bipolar component had a significantly greater proportion of complete flakes at a third of the flakes, with the direct percussion producing less than 8%. This may be related to size, with less surface area available for fragmentation. However, the different recording strategies may have exaggerated the proportion of apparently complete bipolar flakes – as the bipolar debitage was not bagged per strike, it was difficult to identify sequential breaks and fragments; consequently, some of the bipolar flakes that were assigned as complete may in fact be flake fragments with indistinguishable sequential breaks; the surface of the quartz makes it difficult to assess the completeness of flakes compared to flint or chert. The bipolar had substantially more transverse only breaks than the direct percussion, which had both longitudinal and transverse breaks on most artefacts.

The analysis of the proportions of siret breaks showed that the soft hammer produced significantly less siret breaks than the hard hammer or bipolar regardless of the quartz source, and that the use of the platform-on-anvil reduced the occurrence of siret breaks for both soft and hard hammer, but that the B.Q. both increased and decreased the likelihood depending on the technique/support. Therefore, the proportion of siret breaks is useful predictor of the technique/support used. Additionally, while 96% of the hard hammer direct and bipolar flakes’ platforms exhibited impact marks, only 79% of the soft hammer flakes did. Overall, the various materials produced relatively few clean breaks, highlighting the coarse-grained nature of the material and also few regular flakes, with the clear majority having irregular edges.

The complete bipolar flakes were, unsurprisingly, smaller (as was the aim), and with a greater proportion of length/width ratio approaching 2:1 and greater. For the direct percussion, while the S.H.D. Elastic generally produced bigger, thicker complete flakes with a greater length/width ratio, the differences between the technique/supports and source materials for all the metrics were not significant, and therefore not useful as predictors for identifying technique/supports. Comparing the H.H.D. Elastic quartz and chert flakes, the metrics that were significantly different were length and the length/width and length/thickness ratios. Because of the fragmentation rate, many of the non-complete quartz flakes were of a greater size the complete flakes, especially the distal missing, proximal missing, and lateral fragments. Consequently, analysis of just the complete flakes is not necessarily representative of the size range of flakes, but it is difficult to analyse these together.

The analysis of the cores has shown that the process of core fragmentation complicates the interpretation of assemblages. Overall, 10% of the platform cores fragmented, resulting in a platform core assemblage of 21% core fragments – many of these core fragments are indistinguishable as core fragments, possibly resulting in an under-recognition of cores in archaeological assemblages and a consequent over-count of debitage. Neither the B.Q. nor the P.Q. – both the sugar-textured quartz – produced core fragments; nor did the soft hammer elastic knapping events. For the platform cores knapped on inelastic supports, all but one core fragment set had indicative distal impact marks signifying the use of an anvil, as did the bipolar cores which also had proximal impact marks. The bipolar core assemblage produced a greater proportion of core fragments, but the bipolar component is overall easier to identify as bipolar cores and bipolar core fragments (however, see below Chapter 7). One difference noted to Knutsson’s experimental dataset was the occurrence of conical pieces, which he states as occurring from both bipolar and direct percussion knapping. In this series of experiments, all the conical pieces resulted from bipolar reduction, and are treated, cautiously, as bipolar core fragments.

Overall, the four quartz source materials did not result in significantly different outcomes for most attributes recorded with just a couple of exceptions – such as with the interaction with technique/support in siret proportions where the B.Q. produced less siret breaks using bipolar compared to other materials, and more siret breaks using S.H.D compared to other materials; for slivers the M.Q. reversed the proportion of slivers produced with inelastic support compared to other materials. The technique/supports generally had a greater influence in the differences, where significant differences were discerned such as with the siret proportions and platform collapse. Therefore, for some attributes the materials used confounded the otherwise clearer pattern of the technique/support results, highlighting that the idiosyncrasies of the individual blocks/cobbles of quartz will affect the composition of vein quartz assemblages in unpredictable ways, but the materials only altered the patterns in interaction with techniques and/or supports. The clearest difference in materials, however, was between the chert and the quartz, which produced significantly different results for most of the attributes recorded such as the differences between the chert and quartz flakes’ length; the ratios of length/width and length/thickness; curvature; presence of bulb, compression rings, and eraillure flakes; the regularity of flakes; and the difference in fragmentation rate of the debitage per strike and consequently the types of flake breaks produced.

[14]The following boxplots provide the median in the box with the minor outliers signified by circles and major outliers signified by stars.[return].

Home     PhD Contents     PhD Chap 1     PhD Chap 2     PhD Chap 3     PhD Chap 4     PhD Chap 5     PhD Chap 6     PhD Chap 7     PhD Chap 8     PhD Chap 9     PhD Chap 10     PhD Chap 11     PhD Bibliography     PhD Appendices     PhD Plans    [PhD PDF]

Top of Page