An experimental approach to distinguishing different stone artefact transport patterns from debitage assemblages

Journal of Archaeological Science 65 (2016) 44e56

Contents lists available at ScienceDirect

Journal of Archaeological Science journal homepage: http://www.elsevier.com/locate/jas

An experimental approach to distinguishing different stone artefact transport patterns from debitage assemblages Kane Ditchfield M257, Archaeology, The University of Western Australia, 35 Stirling Highway, Crawley, Perth, 6009, Australia

a r t i c l e i n f o

a b s t r a c t

Article history: Received 19 August 2015 Received in revised form 14 October 2015 Accepted 28 October 2015 Available online xxx

This paper experimentally demonstrates the ability of a set of indices to distinguish between different stone artefact transport patterns represented in debitage assemblages. Stone artefacts were transported extensively in the past and this is an important component of technological organisation. However, most stone artefacts occur as part of debitage assemblages. From these assemblages, where mostly nontransported artefacts remain, it can be challenging to identify what artefacts, if any, were transported in anticipation of future use. A series of indices; the cortex ratio, volume ratio, flake to core ratio, noncortical to cortical flake ratio and flake/core diminution tests are presented to meet this challenge. These are tested on an experimental assemblage where three different transport scenarios are simulated. Results suggest that the indices are sensitive to artefact transport and are capable of empirically distinguishing between the three transport scenarios, even when raw material form varies. The results also indicate that artefact transport is capable of exerting a significant influence on stone artefact assemblage formation. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Stone artefact transport Technological organisation Assemblage formation Experimental archaeology Cortex ratio Volume ratio

1. Introduction In archaeology, the importance of stone artefact transport as a behavioural strategy is noted as far back as the Oldowan (Potts, 1991; Toth, 1985, 1987). The ability to transport, and ‘curate’, different toolkits in anticipation of future use(s) is an important component of how past technology was organised in relation to localised socio-economic, environmental and functional contexts, especially among highly mobile hunteregatherers (Andrefsky, 2009; Bamforth, 1986, 2003; Goodyear, 1989; Kelly, 1988; Kuhn, 1992, 1994; Meltzer, 1989; Nelson, 1991; Odess and Rasic, 2007; Shott, 1986, 1996; Torrence, 2001). Along with raw material constraints (e.g. Andrefsky, 1994; Brantingham et al., 2000; Elston, 1990), variation in these contexts placed different constraints on hunteregatherer technology and thus different transport patterns were advantageous for different contexts. For example, large, thin flakes might be transported to ensure a sharp edge is available when required and, because of their size, large flakes generally had longer use-lives than smaller flakes (Andrews et al., 2015; Close, 1996; Dibble, 1997; Douglass et al., 2008; Eren, 2013; Key and Lycett, 2014; Lin et al., 2013; Morrow, 1996; Odess and Rasic,

E-mail address: kane.ditchfi[email protected]. http://dx.doi.org/10.1016/j.jas.2015.10.012 0305-4403/© 2015 Elsevier Ltd. All rights reserved.

2007; Roth and Dibble, 1998; Terradillos-Bernal and Rodriguez, 2012). Due to their durability and use-life potential, thick flakes may also represent a good transport solution under conditions of high mobility (Eren and Andrews, 2013). Transporting cores, or coretools, may also represent an attractive option because flakes can be created as required until the core is exhausted (Bamforth, 2003; Braun et al., 2008a; Close, 1996; Kelly, 1988; Kelly and Todd, 1988; Nelson, 1991:73e76; Phillipps and Holdaway, in press). However, cores can be large and heavy items so their transport can come at some cost (Beck et al., 2002). To avoid the transport cost, cores can be ‘prepared’ at their source location by removing unnecessary exterior weight and/or they can be stockpiled in strategic locations for use over multiple occupations (Close, 1996; Kuhn, 1992, 1995). Numerous small flakes, as opposed to a smaller number of large flakes, may also be a desirable transport option (Kuhn, 1994) especially where microliths are an essential part of the transported toolkit (e.g. Hiscock et al., 2011). In Australia, the ethnographic literature indicates that a wide variety of artefacts might be used and transported depending on localised context (Holdaway and Douglass, 2012). In addition, factors such as the design, flexibility, function and maintainability of artefacts influenced past transport decisions (Bleed, 1986; Goodyear, 1989; Kuhn, 1992; Nelson, 1991). However, these different systemic stone artefact transport patterns will clearly leave a variety of different archaeological signatures

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

where it can be difficult to distinguish what signature belongs to what transport pattern. Addressing this issue represents an important goal towards understanding past technological organisation and the contextual constraints operating on past transport decisions. Some transport patterns can be distinguished through application of techniques such as geochemical sourcing (e.g. Boulanger et al., 2015; Braun et al., 2008b; ten Bruggencate et al., 2015; Nash et al., 2013; Shackley, 2011), refitting (e.g. Close, 2000;
pez-Ortega et al., 2011) and stone Delagnes and Roche, 2005; Lo tool retouch indices (e.g. Andrefsky, 2008; Clarkson, 2002a, 2002b) to stone artefacts that were demonstrably transported in the past. However, demonstrably transported stone artefacts are not common in the archaeological record where most stone artefacts occur as part of debitage assemblages (here defined as including all components of flaked stone discarded at a given site (or other analytical unit) including cores, flakes, flaked pieces and retouched/ used tools). Debitage assemblages commonly consist of thousands of stone artefacts manufactured from a limited range of localised raw material(s) where the desired items were either used expediently (Nelson, 1991:64) or transported away (e.g. Douglass et al., 2008). As such, all that remains in debitage assemblages (with the possible exception of non-local raw materials from other transport events) are those artefacts which were not transported from their place of production. Although the applications of new methods are beginning to show the extent to which debitage assemblage formation was influenced by artefact transport in the past (e.g. Douglass et al., 2008; Holdaway et al., 2008), beyond establishing that artefact transport did occur in the past (and in the absence of demonstrably transported artefacts), it remains difficult to distinguish exactly what artefact transport patterns were responsible for past debitage assemblage formation, especially under conditions of varying raw material forms. To help address these issues, this paper aims to use a novel combination of the cortex ratio, volume ratio, flake to core ratio, non-cortical to cortical flake ratio and a flake and core size diminution to provide quantitative criteria capable of distinguishing between different stone artefact transport patterns. These indices will be rigorously tested using an experimental assemblage where different artefact transport scenarios will be simulated and the indices applied to each simulation. Each index measures different aspects of stone artefact assemblage composition (see below) meaning that, as assemblage composition is altered by different artefact transport scenarios, each index should respond differently where the results may be characteristic of different transport patterns. Further, the influence of different stone artefact transport patterns on assemblage formation can be closely tracked. 2. Background Extensive archaeological application of the cortex and volume ratios have established both as robust measures for artefact transport behaviour (Ditchfield et al., 2014; Douglass et al., 2008; Douglass, 2010; Holdaway et al., 2008, 2010; Phillipps, 2012; Phillipps and Holdaway, in press). While controlling for raw material shape and size, the cortex ratio determines whether all the cortical products of reduction are present in a stone artefact assemblage where cortex is the weathered surface of a rock. It does this by comparing, in the form of a ratio, the observed amount of cortical surface area with the expected cortical surface area (Douglass et al., 2008). The expected cortical surface area is the amount which should remain in the assemblage under expedient conditions. If no artefact transport occurred (i.e. the assemblage was expediently produced) then the observed amount of cortical surface area will not differ from the expected amount. If artefact

45

transport did occur, the observed amount of cortical surface area will differ from the expected amount. However, the cortex ratio can be limited in applicability because it mostly requires assemblages to be produced from fully cortical nodules. Dibble et al. (2005) provide the experimental proof of concept while two separate studies in western New South Wales, Australia (Douglass et al., 2008; Douglass, 2010; Holdaway et al., 2008) provide the initial archaeological applications of the cortex ratio method. In both archaeological studies, observed cortex ratios are significantly below one due to large cortical flake transport. A host of further archaeological work, experimental studies and computer simulations (Ditchfield et al., 2014; Douglass, 2010; Douglass and Holdaway, 2011; Douglass et al., 2008; Lin et al., 2010; Parker, 2011), have determined this is the case while further applications in Egypt (Holdaway et al., 2010; Phillipps, 2012; Phillipps and Holdaway, in press), Middle Palaeolithic France (Lin et al., 2015), and the southern Cook Islands (Ditchfield et al., 2014) have reinforced the archaeological applicability of this methodology. The volume ratio is similar to the cortex ratio, except that it uses assemblage volume instead of cortex (see below for measurements). This can be advantageous because the volume ratio does not always require assemblages to be produced from fully cortical nodules. Compared to the cortex ratio, however, the volume ratio has seen only three applications (Ditchfield, 2011; Ditchfield et al., 2014; Phillipps, 2012; Phillipps and Holdaway, in press) where it was successfully developed to track or check for the transport of cores (or core-tools) in the Fayum, Egypt (Phillipps, 2012; Phillipps and Holdaway, in press), south-western Tasmania (Ditchfield, 2011) and on Moturakau, Aitutaki (Ditchfield et al., 2014). The other ratios and analytical techniques (the flake to core ratio, non-cortical to cortical flake ratio and the flake and core size diminution tests) have seen wide application in stone artefact analysis where they are commonly used to measure reduction intensity and occupation duration (e.g. Henry, 1989; Holdaway et al., 2004; Roth and Dibble, 1998; Shiner, 2006, 2008; Shiner et al., 2007). Compared to the cortex and volume ratios, these indices are more simplistic and not as extensively explored but their inclusion here will help further investigate the applicability of these indices and their relationship with artefact transport. For example, because each measure is essentially based on the frequency of complete flakes or cores (see below), it can be expected that, as these frequencies change in correspondence with artefact transport, the ratios will respond accordingly. 3. Materials: an experimental application To test whether the set of indices is capable of distinguishing between different artefact transport patterns, an experimental approach was selected to simulate different stone artefact transport scenarios using an experimentally produced debitage assemblage. Three transport scenarios were selected for experimental simulation: 1. The transport of flakes from an assemblage produced from fully cortical nodules. 2. The transport of flakes from an assemblage produced from partially cortical nodules. 3. The transport of cores from an assemblage produced from fully cortical nodules. As each of these transport simulations is carried out, the proposed suite of indices can be applied to the experimental debitage assemblage at set increments to quantitatively track assemblage compositional changes caused by the artefact transport simulations. Once the simulations are carried out, results can be compared

46

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

to observe whether the set of indices produce characteristic quantitative markers that can distinguish the simulated transport patterns from one another. This being the case, they may serve as characteristic markers in archaeological assemblages too. This approach is designed to test of the ability of the indices to distinguish different artefact transport patterns. The aim is not to exhaustively test every possible transport scenario. These simulations were selected for two main reasons. First, to test whether the combined indices are capable of distinguishing between flake and core transport patterns created in debitage assemblages. Secondly, to do the same for debitage assemblages produced from different raw material forms or transported core states. This is particularly important for expanding the future applicability of the cortex ratio which has largely required assemblages to be produced from fully cortical nodules. As Ditchfield et al. (2014) note, nodule form can heavily affect the cortex ratio's ability to accurately reflect transport behaviour. Nodules can occur naturally in a partially cortical state or be transported to a site following de-cortification at another location. Simulation 2 can account for both possibilities (while simulations 1 and 3 can only be considered as starting locales for artefact transport).

3.1. The experimental assemblage For the simulations, an experimental dataset (Tables 1 and 2), knapped in reduction stages from silcrete cobbles from western New South Wales by Matthew Douglass and Sam Lin (the artefact producers and data collectors), was used (Douglass, 2010:131e133; Lin et al., 2015). For this assemblage a reduction stage is defined as a reduction episode where a series of flakes were removed from the core. There was no required number of flakes for each stage, while the decision to end one stage and begin the next was arbitrary. When a reduction stage was completed the core and flakes produced during that stage were measured and attributed to that stage. To produce a varied dataset, the number of stages was varied for each nodule (see Table 2). The use of staged reduction and data collection is particularly advantageous for this paper because it allows the experimental assemblage to be reconstructed to a stage of reduction which reflects either a cortical or partially cortical assemblage from which artefact transport can be simulated (see below for detail on simulation constructions). The experimental assemblage was produced from cobbles ranging in size from 49.1 cm3 to 602.9 cm3 which were reduced by free-hand percussion in co-ordination with a general Australian model (Flenniken and White, 1985) to resemble assemblages located in western New South Wales produced from the same material (Douglass, 2010:111). The goal was not to produce distinctive core or tool types (Douglass, 2010:131). This experimental assemblage totals 617 artefacts including 29 cores measuring over 25 mm in maximum dimension. Artefacts below this value were not measured. Although some artefacts below this

Table 2 Summary of the experimental assemblage organised by artefact type and reduction stage. This table only includes the cores as they are from the final stages of reduction. Artefact type

Reduction stage

Angular Fragment Broken Flake Complete Flake Complete Split Core Distal Flake Medial Flake Proximal Flake Proximal Split Total

1

2

3

8 3 122 13 2 18 4 14 1 185

11 2 120 14 7 25 2 7

9 7 92 14 17 23 1 8 5 176

188

4

5

6

7

21

8

6

1 6

2 2

1

2

1

3

3

4

12

11

12

25

8

Total

2

30 8 379 41 29 72 7 39 6 617

4 1 1

8

value may retain cortical surface area (e.g. resharpening flakes), other experimental analyses indicate that the exclusion of <25 mm flakes has no effect on the methods or results (e.g. Dibble et al., 2005; Douglass, 2010). In total the experimental assemblage has 7331.17 cm3 of volume and 6052.36 cm2 of cortical surface area. It should also be noted that, on average, this experimental assemblage was minimally reduced where reduction was not continued on some cores past the preliminary stages. Unfortunately, it was not possible to access the original material to either further reduction or take additional measurements. More information regarding the experimental assemblage, including more detail on morphometric attributes and some images (of the flakes and cores), can be found in Douglass (2010) as well as Lin et al. (2010). For each simulation, the appropriate assemblage was prepared from the dataset and the transport simulation carried out by removing the required artefact types from the assemblage in set increments. The suite of indices was applied at each increment. For simulation 1 (the transport of flakes from an assemblage produced from fully cortical nodules) all reduction stages were used (i.e. the whole experimental assemblage, 617 artefacts) where 10% of the assemblage was removed per increment (62 artefacts per increment). The number of cores (n ¼ 29) was held constant. Artefacts were ordered by surface area from largest to smallest meaning the largest artefacts were removed first. Ordering flakes by surface area assumes that the largest flakes were intended for transport (see above) and combines the effects of removing large cortical and non-cortical flakes. For simulation 2 (the transport of flakes from an assemblage produced from partially cortical nodules) all artefacts from the first stage of reduction (Table 2) were removed from the assemblage. The core measurements from the end of the first (and beginning of the second) stage of reduction were treated as the original nodules prior to reduction. These were partially cortical. This experimental dataset totalled 432 artefacts with 27 cores but was still highly cortical; at approximately 85%. This has some implications explored in greater detail below. Like the first simulation, artefacts were ordered based on surface area and removed in 10%

Table 1 Basic attributes for the experimental assemblage. All standard deviations are ±1. Weight is in grams (g) while surface area and platform area in millimetres squared (mm2). Artefact class Angular Fragment Broken Flake Complete Flake Complete Split Core Distal Flake Medial Flake Proximal Flake Proximal Split

n

Surface area (m)

Surface area (s)

Weight (m)

Weight (s)

Platform area (m)

Platform area (s)

30 13 379 42 29 72 7 39 6

767.17 349.54 1372.09 998.21 16437.66 644.82 558.86 821.46 477.33

888.04 151.91 1308.19 1145.90 9575.96 465.51 231.25 596.22 219.04

10.27 1.85 16.27 7.50 382.01 5.17 3.43 6.05 4.00

19.63 1.21 26.57 9.49 328.70 9.18 2.07 7.79 4.29

0.00 0.00 99.63 0.00 0.00 0.00 0.00 95.21 0.00

0.00 0.00 139.35 0.00 0.00 0.00 0.00 129.51 0.00

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

47

increments (43 artefacts). For simulation 3 (the transport of cores from an assemblage produced from fully cortical nodules) the entire experimental assemblage (617 artefacts) was used where all cores were removed from the assemblage in batches of two ordered by reduction set until none remained (15 removal increments in total). The methods applied at each increment are calculated as follows.

to derive a two-dimensional surface area. For cores, three dimensional surface area is calculated by entering the maximum length, width and thickness semi-axes into an equation for the surface area of an ellipsoid:

4. Methods

where a, b and c are the respective semi-axes for maximum length, width and thickness, and p is 1.6075 (Thomsen, 2004). The surface area of each artefact is multiplied by the midpoint of its ordinal cortex category (i.e. 0, 0.25, 0.75 and 1) to derive cortical surface area. These are summed to arrive at the total observed cortical surface area. The volume for each artefact is also calculated by dividing its weight by material density. These values are summed to determine the total amount of observed volume. It is possible that some of these metric attributes could be more precisely measured using laser scanning and photographic morphometric techniques (e.g. Bretzke and Conard, 2012; Cardillo, 2010; Grosman et al., 2008; McPherron and Dibble, 1999; Picin et al., 2014; Shott, 2014). Although these techniques may increase measurement precision, their use will also increase the time taken to generate the measurements from archaeological debitage assemblages (commonly consisting of thousands of artefacts, all of which require measurements for the cortex ratio). Significantly, Lin et al. (2010, 2015) have twice assessed precision for the mechanical and ordinal measurements used to generate observed cortical surface area through comparison against measurements derived from 3D laser scans. In the first study (Lin et al., 2010), using the four ordinal cortex categories, they find the mechanical and ordinal measurements to be accurate although observed cortical surface area tends to over-estimate scanned values by as much as 10% due

4.1. Cortex ratio The cortex ratio requires the observed and expected values for cortical surface area to be calculated which, following previous applications (Dibble et al., 2005; Douglass et al., 2008), requires the following attributes for all stone artefacts in an assemblage: the percentage of cortical surface area, weight (g), and maximum dimensions (mm) for length, width, and thickness (see Holdaway and Stern, 2004:138e140). Percentage of cortical surface area is visually recorded using four ordinal cortex categories; 0%, 1e49%, 50e99%, and 100%. These broad categories are used to reduce inter-observer error (Gnaden and Holdaway, 2000) and to promote comparability with previous studies (e.g. Douglass et al., 2008). Material density is also required to provide an accurate measure for the volume of each artefact. Density is calculated by a measure of material displacement in water (Berman, 1939). The calculation method for the cortex ratio (in combination with the volume ratio) is graphically illustrated in Fig. 1. 4.1.1. Cortex ratio: calculating observed cortical surface area Observed cortical surface area is obtained by, first, multiplying maximum length by maximum width for all artefacts except cores

1  apbp þ apcp þ bpcp p S ¼ 4p 3

Fig. 1. A schematic diagram illustrating the calculation procedure for both the cortex and volume ratios. A ‘ŧ’ marks where expected assemblage volume can replace observed assemblage volume to recalculate the cortex ratio.

48

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

to an over-approximation of artefact surface area. In the second study (Lin et al., 2015), when using seven cortex categories (0%, 1e9%, 10e39%, 40e59%, 60e89%, 90e99% and 100%), observed cortical surface area was under-estimated by 7.4% when compared to scanned values. These two studies suggest that the metric and ordinal measures used to derive observed cortical surface in this paper are accurate enough so that laser scanning and photographic morphometric techniques are not necessary. Further, estimations of assemblage volume were proportionally increased by 10% for simulations 1 and 2 (multiplication by 1.10) to account for the 10% over-estimation in the observed surface area measurements identified by Lin et al. (2010:700), who used the same experimental assemblage as this paper. This was not carried out for simulation 3 because, when cores are transported, proportionally more volume is removed than surface area cancelling out the effect of an overapproximation in surface area (see below). Recent research (Morales et al., 2015) has also demonstrated that water displacement can be an imprecise method of deriving density for stone artefacts due to impurities. Where stone artefacts have few impurities, this may not be the case (e.g. Ditchfield et al., 2014). The original materials were not available to test the precision of the water displacement method for the experimental assemblage but Douglass (2010) makes no mention of any issues with this measure. Nevertheless, in light of recent research (Morales et al., 2015), it is possible that density imprecision may produce some effect on the cortex ratio. 4.1.2. Cortex ratio: calculating expected cortical surface area Calculating the expected cortical surface area requires an estimate of the original nodule size and shape followed by a measure of the number of cores used to produce the assemblage (Dibble et al., 2005, Fig. 1). As Douglass (2010) and Phillipps (2012) suggest, an ideal approach for estimating original nodule size is to identify the raw material source location and collect raw material size (and shape) data at that location. Another approach is provided by Douglass (2010) who, based on work by Braun (2006), developed multiple regression equations to predict the amount of mass lost from core reduction based upon core attributes (see Douglass, 2010:146e180). In this case, the amount of mass lost can be used to recalculate original nodule size. However, if raw material source data or the required core attributes are not attainable, another approach is to simply divide observed assemblage volume by the number of cores in the assemblage for an estimate of original nodule volume (Dibble et al., 2005:557; Douglass et al., 2008:519). This is the approach taken in this study. It should be noted that this assumes one core is produced per nodule. When this assumption is not possible, such as in heavily reduced assemblages, alternative approaches can be taken (Lin et al., 2015). The value for original nodule volume is entered into a geometric surface area equation to derive original nodule surface area while approximating original nodule shape. This is dependent on the selection of an appropriate geometric model which most closely approximates the average shape of nodules selected in the past as well as the geometric relationship between volume and surface area represented in the assemblage. Dibble et al. (2005:549) provide shape formulas for a cube, sphere and right cylinder. These equations are as follows:

Cube : Sphere : Right Cylinder :

S ¼ 6V 2=3  2=3 3V S ¼ 4p 4p  2=3 V S ¼ 4p p

To test the accuracy of each of these equations, Dibble et al. (2005) calculated the cortex ratio for their experimental assemblage, produced from nodules of different shapes and sizes, using each of these geometric equations. Each geometric model returned a value close to one indicating that the cortex ratio, despite variation in raw material shape, is capable of accounting for almost all the products of reduction using different geometric models. The sphere model proved to be the most accurate. Finally, once a geometric model is selected, the original nodule surface area is multiplied by the number of cores (as a proxy for the original number of nodules) to arrive at the expected amount of cortical surface area that should be present if all the products of reduction remain. Division between the observed and expected cortical surface area produces the cortex ratio. 4.2. Volume ratio The volume ratio follows a similar procedure (requiring both an observed and expected statistic) and allows for similar inferences, as the cortex ratio. Like the cortex ratio, its calculation procedure is also graphically illustrated in Fig. 1. Observed assemblage volume is obtained by simply summing the volume of all artefacts in the assemblage (much the same as calculating observed assemblage cortical surface area for the cortex ratio). Previously, to derive expected assemblage volume for a volume ratio, applications have concentrated on accurately calculating original nodule size, shape and frequency specifically for core transport (Ditchfield et al., 2014; Phillipps, 2012; Phillipps and Holdaway, in press). However, because the volume ratio is also applied to flake transport in this paper (where the cores remain in the assemblage), an approach for calculating expected volume that is applicable to all transport scenarios is required. This is a simple approach based on the fact that all stone artefacts possess volume and surface area meaning that, when artefacts are removed from, or added to, an assemblage, the assemblage must respectively lose, or gain, both surface area and volume. The cortex ratio quantifies the addition or loss of assemblage surface area and, because surface area must be lost from, or added to, an assemblage in conjunction with volume, the cortex ratio can be taken as an indicator of how much volume is lost or gained. As such, expected volume can be calculated by dividing or multiplying observed assemblage volume by the cortex ratio. In situations where the cortex ratio is below one observed volume can be divided by the cortex ratio to derive expected assemblage volume while multiplication can be used if the cortex ratio is above one. Essentially, these calculations estimate expected assemblage volume under situations of artefact transport induced assemblage compositional change as reflected by a cortex ratio different to one. The only assumption here is that the cortex ratio adequately reflects this compositional change (see above for supporting literature). The volume ratio can then be calculated through division of observed assemblage volume by the predicted original assemblage volume (i.e. expected volume). As with the cortex ratio, values below one indicate that volume is under-represented whereas the opposite is true if values are above one. Should this prove to be an accurate approach, the calculated volume ratios should closely match the actual volume ratios produced during each experimental simulation. On a separate note, because the volume ratio calculation allows expected volume to be calculated, a small modification to the cortex ratio calculation procedure can be made. In calculating original nodule size as part of estimating expected cortical surface area for the cortex ratio, a division between observed assemblage volume and the number of cores is used (as above, also see Fig. 1). This assumes that observed assemblage volume is a reasonable

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

indicator for original assemblage volume. While research suggests this may be the case (e.g. Douglass et al., 2008; Douglass, 2010), as stated above, it is also true that under conditions of artefact transport (where the cortex ratio differs from one) volume must be removed from, or added to, an assemblage in conjunction with surface area. Under these conditions, observed assemblage volume may not equate original assemblage volume meaning that an estimate for expected assemblage volume (in place of observed assemblage volume) may produce improved results. Here, the estimate for expected assemblage volume, produced for the volume ratio, can be used in place of observed assemblage volume to calculate the cortex ratio. Cortex ratios which use this modification will be calculated separately in the experimental analysis and referred to as ‘modified cortex ratios’. The main assumption in doing this is that the estimate for expected assemblage volume is a reasonable indicator for original assemblage volume. If this is true then it should be substantiated in the simulation results below. The remainder of the cortex ratio calculation (using this new estimate) is no different to the calculation procedure outlined above where there is a clear understanding of how the variables and calculations interact (Dibble et al., 2005; Douglass, 2010). 4.3. Flake to core ratios, non-cortical to cortical flake ratios and size diminution tests The calculation of these measures (following Holdaway et al., 2004) is comparatively simple. The flake to core ratio is calculated by dividing the minimum number of flakes by the number of cores. The minimum number of flakes only includes those flakes with remnants of platforms (complete, proximal and longitudinally broken flakes) to account for fragmentation (Hiscock, 2002). The non-cortical to cortical flake ratio is calculated by dividing the number of non-cortical complete flakes by the number of cortical complete flakes. A flake and core diminution test is based on the fact that, as cores are reduced, those flakes removed earlier in the process should be both larger and have more cortex, whereas those removed later should be smaller and less cortical (Holdaway et al., 2004:50). The test is similar for cores where those which are reduced more should be smaller and have less cortex. This test is conducted by dividing the assemblage into three ordinal cortex categories, 0%, 1e50% and 51e100%, and calculating the average size of complete flakes (average maximum length) and cores (average volume) for each of these categories (Holdaway et al., 2004:50). If flaking occurred in situ, the results should show a progressive diminution in flake and core size with decreasing proportions of cortex. However, if certain artefacts are removed or added to an assemblage, this progression in reduction ‘stages’ may be disrupted. For example, if large flakes are removed, the flakes in the 51e100% cortical class may be under-represented and be smaller or similar in size compared to those in the 1e50% class. Note that the cortex categories used for this test are different from those used to calculate the cortex ratio. For the diminution tests, the 51e99% and 100% category were combined because, in some instances, especially when assemblages are produced from partially cortical nodules (see simulation 2), very few flakes may occur in the 100% cortex category. 4.4. Final methodological considerations It is important to note that the set of indices is applicable to assemblages of varying technological production and can quantitatively reflect artefact transport from most assemblages regardless of technological difference or knapper goals (see Ditchfield et al., 2014; Douglass et al., 2008; Holdaway et al., 2004; Phillipps and

49

Holdaway, in press). This means that many archaeological assemblages are not required to be significantly similar to this experimental assemblage for this set of indices to be applicable. For example, as shown by research in drastically different technological situations, regardless of how an assemblage is reduced, prior to artefact transport, a volume ratio of one will still be produced which then changes as artefact transport occurs (Ditchfield et al., 2014; Phillipps and Holdaway, in press). The same is true for the cortex ratio when applied to assemblages reduced from cortical nodules (Dibble et al., 2005; Douglass, 2010). Nevertheless, to promote inter-assemblage comparability, basic morphological attribute measurements are presented for the experimental assemblage in Tables 1 and 2 By using the real nodule volume and cortical surface area recorded for each experimental nodule prior its reduction, it is also possible to calculate the actual cortex and volume ratios. To test their accuracy at each simulated removal increment, the actual cortex and volume ratios can be compared to those derived from the experimental assemblage (calculated as above). These are statistically compared with a Wilcoxon Signed Ranks Test. This nonparametric test represents a good choice because the data are ordered by removal increment and the number of increments is no more than 15, meaning sample size is small (see VanPool and Leonard, 2011:263e267). Following this, the overall simulation results can be compared to establish whether the set of indices are capable of differentiating between the different transport patterns. Finally, some methodological modifications are made below but, these are an important result of the simulations and so are treated as part of the ‘results’ section.

5. Results 5.1. Simulation 1: the transport of flakes away from an assemblage produced from cortical nodules The results (Tables 3 and 4) indicate that all indices are sensitive to artefact transport in a situation where flakes are transported from an assemblage produced from fully cortical nodules. As flakes are removed from the assemblage, both the volume and cortex ratio drop in close accordance with one another. This reflects the removal of both surface area and volume from the assemblage. When compared with the actual ratios calculated from known values, the modified cortex and volume ratios are equally accurate where the difference between calculated values and actual values is not statistically significant (Wilcoxin Z ¼ 1.309, p ¼ 0.191 and Wilcoxin Z ¼ 1.120, p ¼ 0.263 respectively; Table 3). It's worth noting that the (unmodified) cortex ratio significantly differs from

Table 3 Results of the experimental simulation testing the ability of the cortex ratio (CR), volume ratio (VR), flake to core ratio (F:C) and non-cortical to cortical flake ratio (NCF:CF) to respond to artefact transport in a situation where flakes are preferentially removed from an assemblage produced from cortical nodules. Removal 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

CR

Modified CR

Actual CR

VR

Actual VR

F:C

NCF:CF

1.08 0.94 0.85 0.78 0.73 0.69 0.66 0.64 0.62 0.60 e

1.07 0.85 0.71 0.62 0.55 0.51 0.47 0.44 0.42 0.40 e

1.06 0.80 0.67 0.59 0.53 0.50 0.47 0.45 0.44 0.42 e

0.98 0.86 0.77 0.71 0.66 0.63 0.60 0.58 0.56 0.55 e

0.98 0.78 0.70 0.66 0.63 0.62 0.60 0.60 0.59 0.59 e

15.22 13.50 11.53 9.69 7.93 6.43 5.09 3.38 2.03 0.64 e

0.16 0.18 0.21 0.24 0.27 0.29 0.32 0.45 0.63 0.88 e

50

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

Table 4 A flake surface area diminution test for all complete flakes at each removal increment in a situation where flakes are removed from an assemblage produced from cortical nodules. The count and average surface area of complete flakes is given for three cortex categories; 0%, 1e50% and 51e100%. All surface areas are in millimetres squared (mm2). Removal

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

0%

1e50%

51e100%

Table 6 A flake surface area diminution test for all complete flakes at each removal increment in a situation where flakes are removed from an assemblage produced from partially cortical nodules. The count and average surface area of complete flakes is given for three cortex categories; 0%, 1e50% and 51e100%. All surface areas are in millimetres squared (mm2). Removal

Count

Average

Count

Average

Count

Average

51 50 48 44 40 34 29 25 19 7 e

668.67 612.04 564.75 489.45 431.23 373.12 331.93 310.28 277.05 239.86 e

172 160 138 115 92 73 57 34 20 7 e

1144.88 946.79 749.57 621.55 515.40 441.82 393.12 328.94 280.85 219.43 e

156 120 93 72 54 43 33 21 10 1 e

1852.56 1111.69 819.23 644.97 507.26 444.16 392.45 345.38 288.80 220.00 e

the actual cortex ratio (Wilcoxin Z ¼ 2.807, p ¼ 0.005) but, by using the recalculated original nodule volume (as specified above), the ratio becomes accurate. Actual original nodule volume is 258.26 cm3 while the average predicted value is 254.13 cm3. The removal of flakes is represented well by the flake to core ratio which shows a substantial decrease throughout the removal sequence resulting from the loss of flakes relative to the number of cores (Table 3). In contrast, the non-cortical to cortical flake ratio increases throughout the removal sequence (Table 3). This reflects the loss of the cortical flakes. As flakes are removed from the assemblage, the flake diminution test (Table 4) shows a progressive decrease in average complete flake surface area for all cortex categories. However, this decrease occurs at different rates for each cortex category. The 0% and 1e50% categories decrease slowly while the 51e100% category shows a dramatic decrease as the first 30% of flakes are removed from the assemblage whereby the 1e50% and 51e100% categories becomes roughly equivalent. The average surface area of those flakes in the 0% category is mostly no less than 100 mm2 below the higher two categories. 5.2. Simulation 2: the transport of flakes away from an assemblage produced from partially cortical nodules The results for this simulation (Tables 5 and 6) again suggest that the applied indices are sensitive to artefact transport patterns, even when flakes are removed from an assemblage created from partially cortical nodules. Both the cortex and volume ratios decrease as flakes are removed reflecting the loss of surface area and volume. However, at the 0% removal increment, when nothing

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

0%

1e50%

51e100%

Count

Average

Count

Average

Count

Average

45 44 41 36 32 26 23 17 13 1 e

681.56 617.50 541.56 453.19 403.56 341.92 321.83 284.00 265.15 168.00 e

141 124 102 83 67 53 35 24 11 4 e

1187.75 881.33 667.54 540.39 458.55 403.32 344.46 300.79 244.09 196.00 e

71 54 43 30 26 20 16 20 7 0 e

1598.23 935.87 712.88 505.30 448.23 385.70 356.50 306.80 279.57 0.00 e

has been removed from the assemblage, the cortex ratio returns a value less than 1 (compare Tables 3 and 5), obviously reflecting the partially cortical nature of the nodules and not artefact transport. As a result, the modified cortex ratio and the volume ratio (Table 5; ‘Initial Modified CR’ and ‘Initial VR’) produce values which are too low and significantly different from the actual values (Wilcoxin Z ¼ 2.825, p ¼ 0.005 and Wilcoxin Z ¼ 2.812, p ¼ 0.005 respectively; Table 5). Here, ‘Initial Modified Cortex Ratio’ refers to the cortex ratio following modifications outlined above (Section 4.2) while the ‘Modified Cortex Ratio’ refers to the cortex ratio with additional calculations proposed here (below). The same is true for the ‘Initial Volume Ratio’ as compared to the ‘Volume Ratio’. In this circumstance we know that the nodule size calculated utilising the cortex ratio method will be too small and that nodule size calculated for the volume ratio is too large so, a more reliable value can be calculated following:

NS ¼ ½ðCNRS  VRNSÞ=2 þ CRNS where NS is nodule size, CRNS is the cortex ratio nodule size and VRNS is the volume ratio nodule size. By subtracting CNRS from VRNS the difference between the two nodule sizes is obtained which means, when divided by two, half the difference is obtained. Adding this to CNRS produces a nodule size estimate that is halfway between CNRS (the underestimate) and VRNS (the overestimate). This nodule size can be used to re-calculate the cortex and volume ratios (Table 5; ‘Modified CR’ and ‘Modified VR’). When these are compared to the actual cortex and volume ratio values (Table 5; ‘Actual CR’ and ‘Actual VR’) there are no significant differences (Wilcoxin Z ¼ 1.000, p ¼ 0.317 and Wilcoxin Z ¼ 1.069, p ¼ 0.285

Table 5 Results of the experimental simulation testing the ability of the cortex ratio (CR), volume ratio (VR), flake to core ratio (F:C) and non-cortical to cortical flake ratio (NCF:CF) to respond to artefact transport in a situation where flakes are preferentially removed from an assemblage produced from partially cortical nodules. Removal 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

CR

Initial Modified CR

Modified CR

Actual CR

Initial VR

Modified VR

Actual VR

F:C

NCF:CF

0.85 0.77 0.72 0.69 0.66 0.64 0.63 0.62 0.61 0.60 e

0.72 0.61 0.55 0.50 0.47 0.45 0.43 0.42 0.41 0.40 e

0.78 0.68 0.62 0.58 0.55 0.52 0.51 0.49 0.49 0.47 e

0.84 0.68 0.61 0.57 0.54 0.52 0.50 0.49 0.48 0.47 e

0.77 0.70 0.66 0.62 0.60 0.58 0.57 0.56 0.55 0.54 e

0.87 0.82 0.79 0.77 0.75 0.74 0.73 0.72 0.71 0.70 e

0.97 0.83 0.78 0.75 0.73 0.72 0.72 0.71 0.71 0.70 e

11.06 9.72 8.28 6.81 5.70 4.59 3.39 2.30 1.37 0.22 e

0.21 0.25 0.28 0.32 0.34 0.36 0.45 0.50 0.72 0.25 e

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

respectively). Both the flake to core and non-cortical to cortical flake ratios both display decreases for the same reasons as simulation 1. However, in comparison, the flake to core ratios display much lower values than simulation 1 while the non-cortical to cortical flake ratio shows slightly higher values. These reflect the lower number of flakes per core in this assemblage as well as a proportionally higher number of non-cortical flakes per number of cortical flakes. The flake diminution test (Table 6) shows a similar pattern to simulation 1 where, once 30% of flakes are transported, the average surface area of flakes in each of the cortex categories becomes approximately similar (except that the average surface area of the flakes in the 0% category are generally slightly smaller than the 1e50% and 51e100% categories). 5.3. Simulation three: the removal of cores from an assemblage produced from cortical nodules These results (Table 7) suggest that, in a situation where cores are preferentially removed from an assemblage, the indices remain sensitive to artefact transport. However, the cortex ratio behaves in a different way from the first two simulations. As cores are removed, both cortical surface area and volume are lost creating the expectation that the cortex ratio should drop in accordance with the volume ratio. Against expectations, however, it rises where both the cortex ratio and initially modified cortex ratio (Table 7; ‘CR’ and ‘Initial Modified CR’) become significantly different from actual values (Wilcoxin Z ¼ 3.408, p ¼ 0.001 and Wilcoxin Z ¼ 3.153, p ¼ 0.002 respectively). ‘Initial Modified Cortex Ratio’ and ‘Modified Cortex Ratio’ follow the same definitions as outlined above. The reason for the rise in the cortex ratio is that as cores are reduced, their cortical surface area ‘transfers’ into the assemblage in the form of flakes whilst the cores themselves still retain a large portion of (often non-cortical) volume. As a result, as these cores are removed from the assemblage, an increasing proportion of cortex will remain relative to the remaining volume. Despite this trend for cortex, the volume ratio is accurate, showing no significant difference from actual values (Table 7; Wilcoxin Z ¼ 0.228, p ¼ 0.819). These two results, a cortex ratio below above one and a volume ratio below one, could be taken as a signature of core transport. However, one approach towards redressing the unexpectedly high cortex ratio is offered here. The main mathematical reason the cortex ratio does not drop is because the number of cores remaining in the assemblage is no

51

longer an accurate proxy for the original number of nodules responsible for assemblage production (a point also made by Phillipps and Holdaway, in press). If the cortex ratio is to be improved to approximate actual values, the original number of nodules must be estimated. Since the volume ratio is accurate, it can now be used as a measure for under-representation (like the cortex ratio, with the same assumptions, for the first two simulations). By dividing the number of cores which remain by the volume ratio (as a measure of their under-representation), an indication of original number of nodules is obtained. Dividing this value by the remaining (observed) assemblage volume provides an improved estimate of original nodule size. Finally, and assuming this process does indeed provide a more accurate indication of nodule size, dividing predicted assemblage volume by the new nodule size estimate will further improve the indication for the original number of nodules. These new values (nodule size and number of nodules) can be used to rectify the cortex ratio in a situation where cores are transported. The use of these values to calculate the cortex ratio (‘Modified CR’) are shown in Table 7 where the modified cortex ratio is not significantly different from the actual values (Wilcoxin Z ¼ 0.682, p ¼ 0.496). There are potential issues of circularity in this calculation procedure, and, this is discussed further below. The flake to core ratio, compared to the simulations where flakes are removed, shows a different pattern. Instead of decreasing, it rises as cores are removed from the assemblage (Table 7). Because flakes are not transported, the non-cortical to cortical flake ratio will show no change in this simulation and, as such, is not displayed in Table 7. It maintains a constant value of 0.16 throughout. Because cores are not removed in any particular order, the size diminution (Table 8) tests show relatively constant size averages in the 1e50% and the 51e100% cortex categories. In most instances, those in the 51e100% category are larger than those in the 1e50% category. 6. Discussion 6.1. Distinguishing the artefact transport patterns The controlled nature of these simulations allowed the formation and structural development of each experimental debitage assemblage, under the influence of different transport patterns, to be observed in precise detail. These results are immediately important on two levels. First, they clearly demonstrate that

Table 7 Results of the experimental simulation testing the ability of the cortex ratio (CR), volume ratio (VR) and flake to core ratio (F:C) to respond to artefact transport in a situation where cores are preferentially removed from an assemblage produced from cortical nodules. Cores remaining 29 27 25 23 21 19 17 15 13 11 9 7 5 3 1 0

Est. No. Nodules

CR

Initial Modified CR

Modified CR

Actual CR

VR

Actual VR

F:C

34 33 31 31 32 30 29 29 28 26 25 22 21 19 13

1.08 1.10 1.12 1.16 1.23 1.26 1.30 1.39 1.46 1.55 1.65 1.78 2.03 2.53 3.66 e

1.03 1.03 1.04 1.05 1.07 1.08 1.09 1.12 1.14 1.16 1.18 1.21 1.27 1.36 1.54 e

0.97 0.97 0.96 0.95 0.93 0.93 0.92 0.90 0.88 0.86 0.85 0.83 0.79 0.73 0.65

1.06 1.05 1.02 0.97 0.95 0.93 0.89 0.87 0.83 0.82 0.79 0.74 0.72 0.70 0.65 e

0.93 0.91 0.89 0.86 0.81 0.80 0.77 0.72 0.68 0.65 0.60 0.56 0.49 0.39 0.27 e

0.98 0.96 0.93 0.85 0.79 0.78 0.74 0.70 0.63 0.63 0.59 0.54 0.50 0.45 0.40 e

15.22 16.35 17.66 19.20 21.02 23.24 25.97 29.43 33.96 40.14 49.06 63.07 88.30 147.17 441.50 e

52

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

Table 8 A core volume diminution test for all cores at each removal increment in a situation where cores are removed from an assemblage produced from cortical nodules. The count and average volume of cores is given for three cortex categories; 0%, 1e50% and 51e100%. All volumes are in centimetres cubed (cm3). Cores remaining

29 27 25 23 21 19 17 15 13 11 9 7 5 3 1 0

0%

1e50%

51e100%

Count

Average

Count

Average

Count

Average

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 e

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 e

17 16 15 14 12 11 10 8 7 5 4 4 3 1 1 e

197.97 120.06 126.93 124.53 108.70 116.42 123.99 110.18 121.01 158.26 188.54 188.54 158.76 75.10 75.10 e

12 11 10 9 9 8 7 7 6 6 5 3 2 2 0 e

150.99 211.65 213.81 188.56 188.56 203.58 197.43 197.43 156.95 156.95 147.55 122.13 170.16 170.16 0.00 e

artefact transport is capable of producing a profound effect on stone artefact assemblage formation. Second, that the cortex ratios, volume ratios, flake to core ratios, non-cortical to cortical flake ratios, and flake/core diminution tests changed in close correspondence with artefact transport induced assemblage change in each simulation, even when raw material form varied (simulation 2). This suggests the set of indices is capable of determining the influence of stone artefact transport on the formation of debitage assemblages to the extent where some characteristic patterning can be assigned to certain transport patterns. When the flake and core transport simulation results are compared, the combined indices reflect some major assemblage structural differences: 1. In situations where flakes are removed, the cortex and volume ratio will drop below one whereas, when cores are removed, the cortex ratio will initially rise above one while the volume ratio will drop below one. It is only following application of the modifications proposed to rectify an unexpectedly high cortex ratio in simulation 3 that cortex ratio drops below one in congruence with core transport. 2. When flakes are transported, the (modified) cortex ratio will tend to be lower than the volume ratio. When cores are transported, the opposite is true. This is because more surface area is proportionally removed when flakes are transported while more volume is proportionally removed when cores are transported. 3. As shown in the simulations 1 and 2, when flakes are transported, the flake to core ratio will decrease while the noncortical to cortical flake ratio will increase. Respective low and high ratios for each can indicate (large) flake transport. As shown in simulation 3, when cores are transported, the flake to core ratio will rise while the non-cortical to cortical flake ratio will remain constant. A respective high and low ratio for each can indicate core transport. 4. If flaking occurred in situ, the diminution tests should show a progressive decrease in flake and core size with decreasing proportions of cortex (i.e. 51e100%, 1e50% and 0% cortex categories) but, for example, if large flakes are transported, then the flakes in the 51e100% cortical class may be under-represented and smaller or similar in size compared to those in the 1e50% class. This is precisely what is observed when flakes are transported in simulations 1 and 2 where flakes from the 51e100%

cortex category return an average size similar to the 1e50% cortex category as a result of flake transport. 5. Building on the last point, under flake transport conditions, a core diminution test should also show a progressive increase in size with cortex categories because cores are not transported. The same is true for the flake diminution test under conditions of core transport. These results suggest that the major structural differences (especially with regard to assemblage surface area and volume composition) created by flake and core transport in the experimental simulations, as reflected by the applied set of indices, can profitably be used to distinguish between the two transport patterns in archaeological assemblages. Although it is relatively straight forward to distinguish between flake and core transport patterns (also see Eren and Andrews, 2013), the results also indicate that it is not as simple to distinguish between different flake transport patterns. The results from simulations 1 (the removal of flakes from an assemblage produced from fully cortical nodules) and 2 (the removal of flakes from an assemblage produced from partially cortical nodules) are similar. However, the simulation results suggest two differences can be utilised to distinguish between these: 1. When flakes are transported from assemblages produced from partially cortical nodules, the non-cortical to cortical flake ratio tends to be higher. Excluding the 90% removal stage, the noncortical to cortical flake ratio is, on average, 21.69% higher in simulation 2 than in simulation 1. 2. The difference between the volume and cortex ratios is greater in simulation 2 in comparison to simulation 1. On average the volume ratio is 14.68% larger than the cortex ratio in simulation 1 while, in simulation 2, the volume ratio is on average, 25.78% larger. The difference exists because assemblages created from partially cortical nodules retain less cortex, but not necessarily less volume, than those produced from cortical nodules. These are particularly important differences because, although it cannot be conclusively demonstrated here, in archaeological assemblages formed by flake transport, these two distinguishing conditions will most likely be even more pronounced. This is because many archaeological assemblages will be more heavily reduced than the experimental assemblage used in this paper. As specified above, the assemblage used for simulation 2 is still highly cortical. A more heavily reduced archaeological assemblage will produce a much larger difference between the cortex and volume ratios (the cortex ratio will be much lower) and a higher noncortical to cortical flake ratio (as non-cortical flakes will be more numerous). In addition, it could also be expected that very few flakes will occur in the 51e100% cortex category producing a very low average size for that category in the diminution test. This can be borne out by future research. 6.2. The effect of reduction technique and intensity As previously demonstrated (Dibble et al., 2005:549e552; Douglass, 2010:143), the cortex and volume ratios will be unaffected by reduction intensity as, regardless of how heavily an assemblage is reduced, the total amount of volume and cortical surface area will remain the same (i.e. more assemblage surface area and volume will just be stored in flakes instead of cores if reduction intensity is high). As discussed above, current research (see Ditchfield et al., 2014; Douglass et al., 2008; Holdaway et al., 2004; Phillipps and Holdaway, in press) also suggests differences in reduction technique will also produce little influence on the

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

cortex and volume ratios. However, reduction technique and intensity will affect the flake to core ratio, non-cortical to cortical flake ratio and flake/core diminution tests. Firstly, with regard to differing reduction techniques (or strategies), it may not always be true that, for example, an assemblage will become less cortical as reduction progresses due to the way in which different techniques manipulate the reduction process (for a certain product) which, in turn, may create different proportions of cortex in an assemblage (see Holdaway and Stern (2004) for examples of different reduction techniques). This means, due to reduction technique, an assemblage may appear disproportionally non-cortical or cortical due to the manner in which it was produced. Equally, it may have a disproportionate number of flakes to cores or cores to flakes. As such, when investigating artefact transport patterns, a disproportionate amount of cortex, flakes or cores cannot immediately be attributed to the influence of artefact transport. It is important that this is borne in mind when interpreting the flake to core, non-cortical to cortical flake ratio and flake/core diminution tests. Second, with regard to reduction intensity, the flake to core ratio, non-cortical to cortical flake ratio and flake/core diminution tests all correlate with reduction intensity and have been used to measure it (e.g. Holdaway et al., 2004; Roth and Dibble, 1998). For example, as reduction progresses, more flakes become present in an assemblage relative to the number of cores while proportionally more noncortical flakes will become present relative to cortical flakes. This means the flake to core ratio, non-cortical to cortical flake ratio and flake/core diminution tests cannot immediately be taken as indicators for artefact transport. In the same vein, given the correlation of the flake to core ratio, non-cortical to cortical flake ratio and flake/ core diminution tests with artefact transport demonstrated in this paper, nor can they be immediately taken as measures for assemblage reduction intensity without also assessing artefact transport. To establish the extent to which the flake to core ratio, noncortical to cortical flake ratio and flake/core diminution tests represent artefact transport as opposed to assemblage reduction intensity (and vice-versa), their application together with the cortex and volume ratios is crucial. The cortex and volume ratios only measure artefact transport meaning the relative effect of artefact transport on the flake to core ratio, non-cortical to cortical flake ratio and flake/core diminution tests can be assessed before they too are used to distinguish artefact transport patterns or measure reduction intensity. 6.3. Further observations In Section 4.1.1 it was stated that estimates for assemblage volume were not proportionally increased by 10% (like simulations 1 and 2) for the core transport simulation (simulation 3) because, in comparison to surface area, proportionally more volume is removed from an assemblage under conditions of core transport cancelling out the effect of an over-approximation in surface area. The reasoning behind this is illustrated by Tables 3 and 7 where, when flakes are transported away from an assemblage (Table 3), comparatively little (non-cortical) volume is removed from an assemblage in comparison to core transport (Table 7). This creates a situation where the cortex ratio rises quickly under conditions of core transport (Table 7). Since, under these conditions, the cortex ratio is multiplied by observed assemblage volume to calculate expected assemblage volume (as above) there is no need to proportionally increase this amount even further by 10% (like simulations 1 and 2). Said another way, in simulation 3, we are dealing primarily with the movement of volume and not surface area. In simulations 1 and 2, we are dealing primarily with the movement of surface area and not volume which requires the need to

53

proportionally increase volume by 10% in an attempt to rectify the over-estimate of surface area identified by Lin et al. (2010) when using the metric and ordinal measures for observed assemblage cortical surface area (see Section 4.1.1). The simulation results also provide some level of confidence for both the volume ratio and the modifications made to the cortex ratio. As stated above, should these approaches (and their assumptions) be reasonable, the simulation results should substantiate this. Indeed, the simulation results provide good support for both the volume ratio and the cortex ratio modifications. For the volume ratio, the assumption was that the cortex ratio is a reasonable indicator of artefact transport (Section 4.2), essentially allowing an estimate for expected assemblage volume. This assumption is supported by the lack of statistical difference between calculated and actual volume ratios throughout simulations one e three. Importantly, the same can be said for the cortex ratio modification which supports the assumption that the estimate for expected assemblage volume (produced for the volume ratio) is a more reasonable indicator of original assemblage volume than observed assemblage volume (Section 4.2). As such, its use in place of observed assemblage volume to calculate original nodule size for the cortex ratio is substantiated. Significantly, in comparison, when the unmodified cortex ratios are compared against actual values, there is a significant difference (Tables 3, 5 and 7). Nevertheless, it is still significant that the cortex ratio, without any modification, changes consistently with artefact transport in each simulation. This clearly demonstrates its ability to reflect artefact transport. Some further modifications were made in some parts of Section 5 where it is important to recognise that there may be some danger in circularity (mostly for simulation 3) where modifications may not make ideal use of independent data or independently derived equations. For simulation 3, it is still possible to distinguish between flake and core transport without making these modifications (due to the respective rise and fall observed for the cortex and volume ratios) but, then again, the modifications do produce statistically accurate results when compared to actual values. Future analyses should focus on addressing whether these issues have an impact on the methods and their product (including the adjustments proposed in Section 4.2). Overall, the different and characteristic response of each ratio to different transport patterns, along with the lack of significant statistical difference between the experimental cortex and volume ratios when compared against actual values (even when raw material form varies (simulation 2)), illustrates the flexibility and robusticity of the indices in combination. Clearly, these indices are analytically valuable when used together while, if used alone, their ability to distinguish between transport patterns may be lost. Finally, in a manner similar to Ditchfield et al. (2014) as well as Phillipps and Holdaway (in press), the volume ratio could also serve as a basis for calculating the number of cores transported under a situation where core transport has been positively identified from debitage assemblages. Application of the volume ratio allows the original number or nodules to be estimated (Table 7) which, minus the number of cores remaining in the assemblage, can provide an estimate for the number transported. In theory, a similar figure could be produced for the number of flakes transported where the size of transported flakes is known. For example, if large flakes, especially those which show signs of use, remain in the assemblage (e.g. Holdaway, 2004; Dibble, 1997), these could serve as proxies for those which were transported to produce the minimum number of transported flakes. 7. Implications and conclusions The aim of this paper, set out in the introduction, was to provide

54

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

a set of indices capable of distinguishing between different stone artefact transport patterns from debitage assemblages. The cortex ratio, volume ratio, flake to core ratio, non-cortical to cortical flake ratio and the flake/core diminution test were put forward as possible candidates to, in combination, meet this aim. These were applied, at set increments, to three different artefact transport simulations constructed from an experimental assemblage. The results suggest that, for debitage assemblages, the suite of indices can distinguish between different stone artefact transport patterns even when raw material form varies, meaning the results are in correlation with the major aim of this paper. This was possible due to the significant influence artefact transport behaviour (as simulated in this paper) exerted on assemblage formation. Characteristic assemblage structural changes were produced by the transport simulations making it possible to distinguish one pattern from the others. As Holdaway and Douglass (2012:124) note, intention (in transport and anticipated use) is apparent not by what is extant in the assemblage, but by what is missing. The analyses presented here have also built on, and provide support for, previous applications of the cortex and volume ratios (e.g. Ditchfield et al., 2014; Douglass et al., 2008; Douglass, 2010; Holdaway et al., 2008). By examining artefact transport in a variety of experimental, geographic and technological contexts, the previous applications provided a robust basis from which to expand the methods. Previously, application of the cortex ratio mostly required an assemblage be produced from cortical nodules. The results of simulation 2, however, show that it is possible to expand this ratio to assemblages produced from partially cortical nodules. The supplementation of the cortex and volume ratios with the flake to core ratio, non-cortical to cortical flake ratio and flake/core diminution tests to distinguish artefact transport patterns also provides a new and more robust basis for investigating artefact transport patterns. These indices have been expanded beyond their usual application to measure reduction intensity (see Holdaway et al., 2004) which makes some progress towards supporting their validity. It should be reaffirmed that there are other approaches for calculating the same indices as presented here (e.g. Ditchfield et al., 2014; Douglass, 2010; Douglass and Holdaway, 2011; Lin et al., 2015; Phillipps, 2012; Phillipps and Holdaway, in press) and, like this approach, these should be used under appropriate circumstances. Importantly, however, regardless of calculation method, their responses to artefact transport patterns should remain similar to those observed in this study. The overall success in experimental application suggests the indices, combined, may have archaeological utility for distinguishing artefact transport patterns from debitage assemblages. Given the commonality of debitage assemblages in the archaeological record, it may be possible to provide detailed reconstructions of past hunteregatherer artefact transport patterns as part of technological organisation and its change overtime in varying environmental, socio-economic and functional contexts. Since mobility also placed certain restraints on technology (e.g. Binford, 1979, 1980; Nelson, 1991; Wallace and Shea, 2006), knowing what was commonly transported from debitage assemblages may also help improve inferences about past human mobility. This can be particularly useful when complemented with approaches which focus on exotic raw materials and retouched tools (e.g. Clarkson, 2002a; Close, 2000; Nash et al., 2013). Application to archaeological assemblages will, of course, bring additional challenges. The applications in this paper occur under controlled conditions whereas archaeological assemblages can be a complex and composite product of various processes (e.g. Holdaway and Wandsninder, 2008; Vaquero, 2008). If possible, these should be intensively investigated. Taphonomic processes, for example, can alter archaeological assemblage compositions and so,

their relative effect on assemblage formation should be accounted for prior to the application of most methods, including those presented in this paper. The temporality of surface assemblages, and their formation, could also be approached, for example, through Bayesian analysis (e.g. de Pablo and Barton, 2015) or by dating the surfaces upon which assemblages occur (e.g. Fanning et al., 2008; Holdaway and Fanning, 2010). Lin et al. (2015) have also addressed some challenges concerning the cortex ratio, specifically focussing on a means for establishing statistical confidence in archaeological application. This includes the proportion of an archaeological assemblage required to produce statistical confidence in cortex ratios. Further, the transport behaviours simulated in this paper only represent a sample of those which may be present archaeologically. For example, the transport of partially cortical cores represents another real possibility. Ditchfield et al. (2014) have examined an archaeological assemblage formed by partially cortical core transport and found that, prior to their correction, both cortex and volume ratios returned values above one. This could be a distinguishing feature of partially cortical core transport as corresponding cortex and volume ratios above one do not feature in any of the well-controlled simulation results presented here (Tables 3, 5 and 7). As such, future archaeological applications will undoubtedly result in further refinements. Archaeological assemblages that are likely to see application in the near future include those located on Barrow Island (Veth et al., 2014) and south-western Tasmania (Ditchfield, 2011), both located at opposite ends of Australia. There are numerous avenues for future research related to the application and development of the methods presented here. Future research should consider specifically investigating assemblages where a large proportion of highly cortical <25 mm (maximum dimension) flakes are present, such as resharpening assemblages. In such a case the inclusion of flakes below the 25 mm maximum dimension cutoff may be important. Archaeologically, the use of core frequency as an indicator for the number of nodules to calculate the cortex ratio should also see greater assessment. This is because nodules may fracture into multiple cores (or core fragments) which would cause core frequency to over-inflate the number of nodules. Further, Lin et al. 's (2015) approach for deducing statistical confidence in the cortex ratio could be expanded to the volume ratio. Finally, further experimental simulations, aimed at testing the applicability of the indices to assemblages formed by varying raw material (size, shape, quality etc.), different knappers and transport behaviour combinations would be beneficial. Acknowledgements Matthew Douglass and Sam Lin are both thanked for generously allowing access to this experimental data-set, in terms of earlier iterations of these analyses and, particularly Sam Lin, for the present paper. This paper has also benefited from insightful comments provided by Peter Veth, Jane Balme, Thomas Whitley and Natasha Busher who read earlier drafts. Sam Lin and four anonymous reviewers provided excellent comments on this paper which lead to vast improvements. For that, they are gratefully thanked. Tiina Manne is thanked for reading and providing comments on segments of the paper. Simon Holdaway is also thanked for stimulating this research subject. References Andrefsky, W.J., 1994. Raw-material availability and the organisation of technology. Am. Antiq. 59, 21e34. Andrefsky, W.J. (Ed.), 2008. Lithic Technology: Measures of Production, Use and Curation. Cambridge University Press, Cambridge. Andrefsky, W.J., 2009. The analysis of stone tool procurement, production, and

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56 maintenance. J. Archaeol. Res. 17, 65e103. Andrews, B.N., Knell, E.J., Eren, M.I., 2015. The three lives of a uniface. J. Archaeol. Sci. 54, 228e236. Bamforth, D.B., 1986. Technological efficiency and tool curation. Am. Antiq. 51, 38e50. Bamforth, D.B., 2003. Rethinking the role of bifacial technology in Paleoindian adaptations on the Great Plains. In: Soressi, M., Dibble, H.L. (Eds.), Multiple Approaches to the Study of Bifacial Technologies. University of Pennsylvania, Museum of Archaeology and Anthropology, Pennsylvania, pp. 209e228. Beck, C., Taylor, A.K., Jones, G.T., Fadem, C.M., Cook, C.R., Millward, S.A., 2002. Rocks are heavy: transport costs and Paleoarchaic quarry behavior in the Great Basin. J. Anthropol. Archaeol. 21, 481e507. Berman, H., 1939. A torsion microbalance for the determination of specific gravities of minerals. Am. Mineral. 24, 434e440. Binford, L., 1979. Organization and formation processes: looking at curated technologies. J. Anthropol. Res. 35, 255e273. Binford, L.R., 1980. Willow smoke and dogs' tails: HuntereGatherer settlement systems and archaeological site formation. Am. Antiq. 45, 4e20. Bleed, P., 1986. The optimal design of hunting weapons: maintainability or reliability. Am. Antiq. 51, 737e747. Boulanger, M.T., Buchanan, B., O'Brien, M.J., Redmond, B.G., Glascock, M.D., Eren, M.I., 2015. Neutron activation analysis of 12,900-year-old stone artifacts confirms 450e510þ km Clovis tool-stone acquisition at Paleo Crossing (33ME274), northeast Ohio, U.S.A. J. Archaeol. Sci. 53, 550e538. Brantingham, J.P., Olsen, J.W., Rech, J.A., Krivoshapkin, A.I., 2000. Raw material quality and prepared core technology in Northeast Asia. J. Archaeol. Sci. 27, 255e271. Braun, D.R., 2006. The Ecology of Oldowan Technology: Perspectives from Koobi Fora and Kanjera South (unpublished Ph.D. thesis). The State University of New Jersey, New Jersey. Braun, D.R., Rogers, M.J., Harris, J.W.K., Walker, S.J., 2008a. Landscape-scale variation in hominin tool use: evidence from the developed Oldowan. J. Hum. Evol. 55, 1053e1063. Braun, D.R., Plummer, T., Ditchfield, P., Ferraro, J.V., Maina, D., Bishop, L.C., Potts, R., 2008b. Oldowan behavior and raw material transport: perspectives from the Kanjera Formation. J. Archaeol. Sci. 35, 2329e2345. Bretzke, K., Conard, N.J., 2012. Evaluating morphological variability in lithic assemblages using 3D models of stone artefacts. J. Archaeol. Sci. 39, 3741e3749. Cardillo, M., 2010. Some applications of geometric morphometrics to archaeology. In: Elewa, A.M., Elewa, T. (Eds.), Morphometrics for Nonmorphometricans. Springer, New York, pp. 325e341. Clarkson, C., 2002a. Holocene scraper reduction, technological organization and landuse at Ingaladdi Rockshelter, Northern Australia. Archaeol. Ocean. 37, 79e86. Clarkson, C., 2002b. An index of invasiveness for the measurement of unifacial and bifacial retouch: a theoretical, experimental and archaeological verification. J. Archaeol. Sci. 29, 65e75. Close, A.E., 1996. Carry that weight: the use and transportation of stone tools. Curr. Anthropol. 37, 545e553. Close, A.E., 2000. Reconstructing movement in prehistory. J. Archaeol. Method Theory 7, 49e77. De Pablo, J.F.-L., Barton, C.M., 2015. Bayesian estimation dating of lithic surface collections. J. Archaeol. Method Theory 22, 559e583. Delagnes, D., Roche, H., 2005. Late Pliocene hominid knapping skills: the case of Lokalalei 2C, West Turkana, Kenya. J. Hum. Evol. 48, 435e472. Dibble, H.L., 1997. Platform variability and flake morphology: a comparison of experimental and archaeological data and implications for interpreting prehistoric lithic technological strategies. Lithic Technol. 22, 150e170. Dibble, H.L., Schurmans, U.A., Iovita, R.P., McLaughlin, M.V., 2005. The measurement and interpretation of cortex in lithic assemblages. Am. Antiq. 70, 545e560. Ditchfield, K., 2011. Stone Artefact Transport and Human Movement in Pleistocene South-western Tasmania (unpublished M.A. thesis). The University of Auckland, Auckland. Ditchfield, K., Holdaway, S., Allen, M.S., McAlister, A., 2014. Measuring stone artefact transport: the experimental demonstration and pilot application of a new method to a prehistoric adze workshop, southern Cook Islands. J. Archaeol. Sci. 50, 512e523. Douglass, M.J., 2010. The Archaeological Potential of Informal Lithic Technologies: a Case Study of Assemblage Variability in Western New South Wales, Australia (unpublished Ph.D. thesis). The University of Auckland, Auckland. Douglass, M.J., Holdaway, S.J., 2011. Quantifying stone raw material size distributions: investigating cortex proportions in lithic assemblages from western New South Wales. Tech. Rep. Aust. Mus. 23, 45e57. Douglass, M.J., Holdaway, S.J., Fanning, P.C., Shiner, J.I., 2008. An assessment and archaeological application of cortex measurement in lithic assemblages. Am. Antiq. 73, 513e526. Elston, R.G., 1990. A cost-benefit model of lithic assemblage variability. In: Elston, R.G., Budy, E.E. (Eds.), The Archaeology of James Creek Shelter. University of Utah Press, Salt Lake City, pp. 153e163. Eren, M.I., 2013. The technology of Stone Age colonisation: an empirical, regionalscale examination of Clovis unifacial stone tool reduction, allometry, and edge angle from the North American Lower Great Lakes region. J. Archaeol. Sci. 40, 2101e2112. Eren, M.I., Andrews, B.N., 2013. Were bifaces used as mobile cores by Clovis foragers in the North American Lower Great Lakes region? An archaeological test of

55

experimentally derived quantitative predictions. Am. Antiq. 78, 166e180. Fanning, P.C., Holdaway, S.J., Rhodes, E.J., 2008. A new geoarchaeology of Aboriginal artefact deposits in western NSW, Australia: establishing spatial and temporal geomorphic controls on the surface archaeological record. Geomorphology 101, 524e532. Flenniken, J.J., White, J.P., 1985. Australian flaked stone tools: a technological perspective. Rec. Aust. Mus. 36, 131e151. Gnaden, D., Holdaway, S., 2000. Understanding observer variation when recording stone artifacts. Am. Antiq. 65, 739e747. Goodyear, A.C., 1989. A hypothesis for the use of cryptocrystalline raw materials among Paleo-Indian groups of North America. In: Ellis, C.J., Lothrop, J.C. (Eds.), Eastern Paleoindian Lithic Resource Use. Westview Press, Boulder, pp. 1e9. Grosman, L., Smikt, O., Smilannsky, U., 2008. On the application of 3-D scanning technology for the documentation and typology of lithic artifacts. J. Archaeol. Sci. 35, 3101e3110. Henry, D.O., 1989. Correlations between reduction strategies and settlement patterns. Archaeol. Pap. Am. Anthropol. Assoc. 1, 139e155. Hiscock, P., 2002. Quantifying the size of artefact assemblages. J. Archaeol. Sci. 29, 251e258. Hiscock, P., Clarkson, C., Mackay, A., 2011. Big debates over little tools: ongoing disputes over microliths on three continents. World Archaeol. 43, 653e664. Holdaway, S.J., 2004. Continuity and change: an investigation of the flaked stone artefacts from the Pleistocene deposits at Bone Cave, Southwest Tasmania, Australia. In: Report of the Southern Forests Archaeological Project. Archaeology Program, vol. 2. School of Historical and European Studies, La Trobe University, Melbourne. Holdaway, S.J., Fanning, P.C., 2010. Geoarchaeology in Australia: understanding human-environment interactions. In: Bishop, P., Pillans, B. (Eds.), Australian Landscapes, Special Publication 346. The Geological Society, London, pp. 71e85. Holdaway, S.J., Stern, N., 2004. A Record in Stone: the Study of Australia's Flaked Stone Artefacts. Aboriginal Studies Press, Canberra. Holdaway, S., Wandsnider, L. (Eds.), 2008. Time in Archaeology: Time Perspectivism Revisited. University of Utah Press, Salt Lake City. Holdaway, S., Shiner, J., Fanning, P., 2004. Hunter-Gatherers and the archaeology of discard behaviour: an analysis of surface stone artefacts from Sturt National Park, western New South Wales, Australia. Asian Perspect. 43, 34e72. Holdaway, S., Shiner, J., Fanning, P., Douglass, M., 2008. Assemblage formation as a result of raw material acquisition in western New South Wales, Australia. Lithic Technol. 23, 1e16. Holdaway, S., Wendrich, W., Phillipps, R., 2010. Identifying low level food producers: detecting mobility from lithics. Antiquity 84, 185e194. Holdaway, S., Douglass, M., 2012. A Twenty-First Century Archaeology of Stone Artifacts. Journal of Archaeological Method and Theory 19, 101e131. Kelly, R.L., 1988. The three sides of a biface. Am. Antiq. 53, 717e734. Kelly, R.L., Todd, L.C., 1988. Coming into the country: early Paleoindian hunting and mobility. Am. Antiq. 53, 231e244. Key, A.J.M., Lycett, S.J., 2014. Are bigger flakes always better? An experimental assessment of flake size variation on cutting efficiency and loading. J. Archaeol. Sci. 41, 140e146. Kuhn, S.L., 1992. On planning and curated technologies in the Middle Paleolithic. J. Anthropol. Res. 48, 185e214. Kuhn, S.L., 1994. A formal approach to the design and assembly of mobile toolkits. Am. Antiq. 59, 426e442. Kuhn, S.L., 1995. Mousterian Lithic Technology: an Ecological Perspective. Princeton University Press, Princeton. Lin, S.C.H., Douglass, M.J., Holdaway, S.J., Floyd, B., 2010. The application of 3D laser scanning technology to the assessment of ordinal and mechanical cortex quantification in lithic analysis. J. Archaeol. Sci. 37, 694e702. Lin, S.C., Rezek, Z., Braun, D., Dibble, H.L., 2013. On the utility and economization of unretouched flakes: the effects of exterior platform angle and platform depth. Am. Antiq. 78, 724e745. Lin, S.C., McPherron, S.P., Dibble, H.L., 2015. Establishing statistical confidence in Cortex Ratios within and among lithic assemblages: a case study from the Middle Paleolithic of southwestern France. J. Archaeol. Sci. 59, 89e109.
pez-Ortega, E., Rodríguez, X.P., Vaquero, M., 2011. Lithic refitting and movement Lo connections: the NW area of level TD10-1 at the Gran Dolina site (Sierra de Atapuerca, Burgos, Spain). J. Archaeol. Sci. 38, 3112e3121. McPherron, S.P., Dibble, H.L., 1999. Stone tool analysis using digitised images: examples from the Lower and Middle Palaeolithic. Lithic Technol. 24, 38e52. Meltzer, D., 1989. Was stone exchanged among eastern North American Paleoindians? In: Ellis, C.J., Lothrop, J.C. (Eds.), Eastern Paleoindian Lithic Resource Use. Westview Press, Boulder, pp. 11e39. s, J.M., 2015. Measuring retouch intensity in lithic Morales, J.I., Lorenzo, C., Verge tools: a new proposal using 3D scan data. J. Archaeol. Method Theory 22, 543e558. Morrow, T.A., 1996. Bigger is better: comments on Kuhn's formal approach to mobile toolkits. Am. Antiq. 61, 581e590. Nash, D.J., Coulson, S., Staurset, S., Ullyott, J.S., Babutsi, M., Hopkinson, L., Smith, M.P., 2013. Provenancing of silcrete raw materials indicates longdistance transport to Tsodilo Hills, Botswana, during the Middle Stone Age. J. Hum. Evol. 64, 280e288. Nelson, M.C., 1991. The study of technological organisation. Archaeol. Method Theory 3, 57e100. Odess, D., Rasic, J.T., 2007. Toolkit composition and assemblage variability: the implications of Nogahabara I, Northern Alaska. Am. Antiq. 72, 691e717.

56

K. Ditchfield / Journal of Archaeological Science 65 (2016) 44e56

Parker, D.J., 2011. The Complexity of Lithic Simplicity: Computer Simulation of Lithic Assemblage Formation in Western New South Wales, Australia (unpublished M.A. thesis). The University of Auckland, Auckland. Phillipps, R.S., 2012. Documenting Socio-economic Variability in the Egyptian Neolithic through Stone Artefact Analysis (unpublished Ph.D. thesis). The University of Auckland, Auckland. Phillipps, R.S., Holdaway, S.J., 2015. Estimating core number in assemblages: core movement and mobility during the Holocene of the Fayum, Egypt. J. Archaeol. Method Theory. http://dx.doi.org/10.1007/s10816-015-9250-2 (in press). Picin, A., Vaquero, M., Weniger, G.-C., Carbonell, E., 2014. Flake morphologies and patterns of core configuration at the Abric Romaní rock-shelter: a geometric morphometric approach. Quat. Int. 350, 84e93. Potts, R., 1991. Why the Oldowan? Plio-Pleistocene toolmaking and the transport of resources. J. Anthropol. Res. 47, 153e176. Roth, B.J., Dibble, H.L., 1998. Production and transport of blanks and tools at the French Middle Paleolithic site of Combe-Capelle Bas. Am. Antiq. 63, 47e62. Shackley, M.S., 2011. X-Ray Fluorescence Spectrometry (XRF) in Geoarchaeology. Springer, New York. Shiner, J.I., 2006. Artefact discard and accumulated patterns in stone artefact assemblage composition in surface archaeological deposits from Pine Point and Langwell Stations, western New South Wales. Rangel. J. 28, 183e195. Shiner, J.I., 2008. The intensity of raw material utilisation as an indication of occupational history in surface stone artefact assemblages from the Strathbogie Ranges, central Victoria. Aust. Aborig. Stud. 2, 80e92. Shiner, J., Holdaway, S., Allen, H., Fanning, P., 2007. Burkes Cave and flaked stone assemblage variability in western New South Wales, Australia. Aust. Archaeol. 64, 35e45. Shott, M.J., 1986. Technological organisation and settlement mobility: an ethnographic examination. J. Anthropol. Res. 42, 15e51. Shott, M.J., 1996. An exegesis of the curation concept. J. Anthropol. Res. 52,

259e280. Shott, M.J., 2014. Digitising archaeology: a subtle revolution in analysis. World Archaeol. 46, 1e9. ten Bruggencate, R.E., Milne, S.B., Fayek, M., Park, R.W., Stenton, D.R., 2015. Characterization of chert artifacts and two newly identified chert quarries on southern Baffin Island. Lithic Technol. 40, 189e190. Terradillos-Bernal, M., Rodríguez, X.-P., 2012. The Lower Palaeolithic on the northern plateau of the Iberian Peninsula (Sierra de Atapuerca, Ambrona and La Maya I): a technological analysis of the cutting edge and weight of artefacts. Developing an hypothetical model. J. Archaeol. Sci. 39, 1467e1479. Thomsen, K., 2004. Numericana [Web site]. http://www.numericana.com/answer/ ellipsoid.htm (accessed 24.07.10.). Torrence, R., 2001. Hunter-Gatherer technology: macro- and microscale approaches. In: Panter-Brock, C., Layton, R.H., Rowlyey-Conwy, P. (Eds.), Hunter-Gatherers: an Interdisciplinary Perspective. Cambridge University Press, Cambridge, pp. 73e98. Toth, N., 1985. The Oldowan reassessed: a close look at early stone artefacts. J. Archaeol. Sci. 12, 101e120. Toth, N., 1987. Behavioural inferences from early stone artefact assemblages: an experimental model. J. Hum. Evol. 12, 763e787. VanPool, T.L., Leonard, R.D., 2011. Quantitative Analysis in Archaeology. John Wiley & Sons, Ltd, West Sussex. Vaquero, M., 2008. The history of stones: behavioural inferences and temporal resolution of an archaeological assemblage from the Middle Palaeolithic. J. Archaeol. Sci. 35, 3178e3185. Veth, P., Ditchfield, K., Hook, F., 2014. Maritime deserts of the Australian northwest. Aust. Archaeol. 79, 156e166. Wallace, I.J., Shea, J.J., 2006. Mobility patterns and core technologies in the Middle Paleolithic of the Levant. J. Archaeol. Sci. 33, 1293e1309.