J Archaeol Method Theory (2016) 23:379–398 DOI 10.1007/s10816-015-9245-z

A Statistical Examination of Flake Edge Angles Produced During Experimental Lineal Levallois Reductions and Consideration of Their Functional Implications Metin I. Eren 1,2 & Stephen J. Lycett 3

Published online: 24 March 2015 # Springer Science+Business Media New York 2015

Abstract Recent studies have indicated that Levallois-style core reduction offered potential practical benefits to hominin populations. However, none of these studies have yet considered one of the most important functional attributes of flake tools, which is edge angle. To address this shortcoming, we statistically examined flakes produced experimentally during “classic” or “lineal” Levallois core production and reduction. The primary aim of our analyses was to statistically test the null hypothesis of “no difference” between the edge angles of “Levallois” products and the flakes involved in their production. We employ existing edge angle analytical techniques and develop new comparative methodologies to assess flake blank standardization through the case of Levallois core reduction. Having determined the statistical properties of our experimental Levallois reductions, we thereafter evaluated to what extent edge angles produced may, or may not, have been useful to prehistoric hominins. Our analyses demonstrated that the experimentally produced Levallois edge angles were indeed statistically different from the flakes involved in their production. These differences were evident both in terms of relatively higher (i.e., more obtuse) edge angles than debitage flakes and in being significantly less variable around their higher mean edge angles compared to debitage flakes. However, based on current knowledge of how flake edge angle properties relate to issues of functionality, such differences would not have been detrimental to their functionality. Indeed, the edge angle properties they possess would have provided distinct benefits to hominins engaged in their

* Metin I. Eren [email protected] * Stephen J. Lycett [email protected] 1

Department of Anthropology, University of Missouri, Columbia, MO 65211, USA

2

Department of Archaeology, Cleveland Museum of Natural History, Cleveland, OH 44106, USA

3

Department of Anthropology, University at Buffalo, The State University of New York, Amherst, NY 14261, USA

380

Eren and Lycett

manufacture. Most notably, Levallois-style core organization and reduction would have provided hominins with a reliable means of consistently producing flakes (i.e., “Levallois flakes”) possessing average flake angles that are beneficial in terms of providing a viable cutting edge yet not being so acute as to be friable upon application. Hence, edge angle properties join an array of other features that provide logical motive for why hominins may have organized core production and reduction around Levalloisstyle patterns at various times and places during the Mid-Late Pleistocene. Keywords Middle Palaeolithic . Levallois . Edge angles

Introduction Middle Palaeolithic (or in Africa “Middle Stone Age”) lithic artifacts classified as “Levallois” play a prominent role in archaeological discussions of the Pleistocene behavioral record. In recent years, for example, Levallois flakes and cores have been embroiled in debates concerning hominin behavioral evolution (e.g., Moncel et al. 2011; Hovers and Belfer-Cohen 2013; Kuhn and Hovers 2013; Tryon and Faith 2013; Wurz 2013), the evolution of particular hominin species (Hublin 2009), skill learning (Eren et al. 2011a, b), technological convergence (Lycett 2009; Adler et al. 2014), and cognition (Wynn and Coolidge 2004, 2010). Such recent debates are part of long tradition, with academic discussions of Levallois now extending back to well over a century ago (Spurrell 1884; Commont 1909; Smith 1911). Bordes (1950: 21) famously defined Levallois reduction as a means of producing flakes “predetermined by special preparation prior to detachment from the core” (translation from Schlanger 1996: 231). Indeed, the “standard” view of Levallois is that it was a means of core reduction undertaken purposefully and strategically on the part of hominins (see e.g., Van Peer 1992; Boëda 1995; Schlanger 1996; Pelegrin 2005; Wynn and Coolidge 2010), even if its (presumable) benefits have frequently been left unspecified. However, some workers have challenged the veracity of such statements, claiming that Levallois flakes were not “intended” products, fashioned by prehistoric knappers who set forth with a clear “plan” to manufacture these (merely) “archaeologically conspicuous” artifacts (e.g., Dibble 1989; Noble and Davidson 1996; Sandgathe 2004). In an attempt to consider this issue further, Eren and Lycett (2012) used a combined experimental and morphometric approach to determine whether the production of flakes removed from classic (“lineal”) Levallois cores might logically have been motivated by particular practical/functional considerations. Experimentally replicated Levallois cores and flakes, based on Boëda’s (1995) volumetric conception of lineal (i.e., “classic” or “preferential”) Levallois, were used to meet these aims. Boëda’s (1995) “volumetric concept” invokes a series of specific criteria, whereby the core is organized bifacially in terms of two distinct surfaces that intersect at the core’s margin forming a “plane of intersection.” Under this scheme, the Levallois flaking surface exhibits distal and lateral convexities, which are alleged to facilitate some control over the size and shape of the final flake(s) removed from that surface (Fig. 1). Boëda’s (1995) “volumetric” concept of Levallois can be thought of as a model of core organization and reduction, in which the core’s surfaces and the flakes removed from those surfaces play particular roles. Flakes that are involved in the initial structuring of the core and/or physically build convexity on the Levallois flaking surface,

A Statistical Examination of Flake Edge Angles

381

Fig. 1 A preferential Levallois core experimentally knapped by M.I.E. Column 1 on the far left depicts the side-view of the Levallois core before preferential Levallois flake (PLF) removal. Column 2 shows the Levallois core’s ventral face (top) and proximal end (bottom). The proximal end is struck with a percussion blow to remove the PLF. Column three again shows the ventral face of the Levallois core, but this time with the PLF removed. Column 4 on the far right shows the removed PLF held in a person’s hand

therefore, can be classified (for analytical purposes) as one general category of flake, while those that directly remove convexity from the Levallois flaking surface can be classified as a further category of flake (i.e., “Levallois flakes”). As Chazan (1997: 724) has noted, flakes classified by analysts as Levallois through application of this volumetric concept are, therefore, not defined a priori by their physical characteristics but by their position and role in the overall reduction sequence. The use of experimental assemblages allows retention of all flakes, and the position of each flake in the overall reduction sequence is known with certainty. Hence, the use of experimental assemblages in Eren and Lycett’s (2012) analyses, whereby the various analytical elements could be defined based on Boëda’s (1995) model, was undertaken deliberately in order to avoid problems associated with arbitrarily categorizing archaeological flakes as Levallois based solely a priori on visible properties (see e.g., Pepère 1986). An experimental approach also facilitated the generation of flake samples large enough (n=642 flakes) that several inferential statistical analyses could be used. Eren and Lycett (2012) used a morphometric scheme involving 15 variables that enabled multivariate statistical analysis to be applied, while size-adjustment of the data enabled issues of flake size and shape to be disentangled analytically. Eren and Lycett’s (2012) analyses aimed to determine specifically whether the flakes struck from classically shaped Levallois cores (produced and defined in the manner described) were not only statistically distinguishable from other flakes but also distinguishable specifically in such a way that might—on functional grounds—have made them logically “preferable” to the makers of Levallois cores, as is so often suggested. Hence, Eren and Lycett (2012) used size-adjusted data to determine whether the putative “preferential Levallois flakes” or “PLFs” were distinguishable from other flakes from the same core, even when size-adjusted (i.e., their large size alone would not influence the results). Via multivariate discriminant function analysis (DFA), they aimed to identify which particular attributes might unite Levallois flakes as a coherent entity and, moreover, which particular attributes appear to be relatively more standardized in Levallois as opposed to the flakes produced during their manufacture. Hence, a specific aim of Eren and Lycett’s (2012) analyses was to determine whether the properties that might statistically distinguish putative Levallois flakes from other flakes

382

Eren and Lycett

(if any) could logically be tied to current knowledge regarding the functional utility of particular flake forms. In essence, these analyses did not merely ask whether the Levallois flakes produced in their experiments had a particular form, or even merely if these flakes were “standardized”; what they aimed to identify is whether these flakes had particular attributes which could logically have motivated hominins to deliberately engage in their production and so deliberately produce Levallois flakes and cores for practical reasons. Using this experimental, morphometric, and multivariate statistical approach, Eren and Lycett (2012) found that Levallois flakes defined in the manner described were indeed statistically distinguishable from other flakes and could be classified as a consistent and coherent group in statistically robust terms. The properties most responsible for this statistical pattern were the possession of a moderate thickness that is evenly distributed across a broad cross section of the flake, and also, on average, greater overall symmetry. As Eren and Lycett (2012) noted, all of these properties represent functionally desirable features in flake tools, such as greater capacity for retouch (Turq 1992; Kuhn 1994), robusticity, and a balance and evenness of weight distribution during use (Simão 2002). Hence, on logical grounds, Levallois reduction conforming to classic conceptions of such artifacts (e.g., Boëda 1995) would have provided hominins with a reliable means of producing flakes with particular properties that would have provided a range of practical and functional benefits. Note that these arguments concerning the “functional optimality” of Levallois flakes (or PLFs) are not based on generalized statements of “standardization.” Indeed, Levallois flakes may be quite variable in some aspects of their morphology, such as outline form (e.g., Picin et al. 2014). Rather, quite specific statistical properties of their physical variation—most notably thickness properties and symmetry—are constrained in ways that are different to the flakes involved in their production. This means that in a multivariate analysis (DFA), they can be identified as a discrete category of flake at levels statistically above chance. In addition to these factors, Levallois reduction may also have provided particular economic advantages in terms of raw material exploitation. Using mathematical modeling, Brantingham and Kuhn (2001) presented a novel view of Levallois, which invoked economic factors in specific terms. In their study, Brantingham and Kuhn (2001) modeled initial starting nodules as two-dimensional ellipses of varying dimensions. Then, by varying factors such as the initial platform position for a series of longitudinally removed flakes, and the acuteness of the angle of such cores, they were able to mathematically model how much “waste” would be generated by varying core shapes. Brantingham and Kuhn found that steep-angled cores, conforming broadly to Boëda’s (1995) volumetric definition, minimized both the amount of raw material wastage, while also maximizing the amount of cumulative cutting edge that would be produced for the longitudinally removed flakes. Hence, this modeling work provided evidence that Levallois reduction, as seen in the archaeological record, had been influenced by economic considerations. Generating such abstract mathematical models is a vital endeavor given that they facilitate the precise and formal consideration of archaeologically relevant factors in ways that are difficult to achieve solely through archaeological data (Clarke 1972; Lycett and Eren 2013a). However, as Brantingham and Kuhn (2001) themselves were keen to note, these models do not necessarily take full account of all the practical

A Statistical Examination of Flake Edge Angles

383

difficulties and material challenges involved in reproducing Levallois-style reductions in stone. In particular, Brantingham and Kuhn (2001: 749) noted that the hypothetical “nodules” in their model were two-dimensional in form. Moreover, a distinct element of Boëda’s (1995) volumetric definition is that between each phase of longitudinal flaking, the primary flaking surface of the core is reprepared in order to rebuild properties of distal and lateral convexity (Van Peer 1992; Boëda 1995). Brantingham and Kuhn’s (2001) models did not, however, take account of this requisite repreparation of core surfaces. This may be particularly important given that, as Brantingham and Kuhn (2001: 758) noted, their models are “realistic only where organized flaking surfaces can be achieved with a minimum of preparation.” Hence, Brantingham and Kuhn’s (2001) model would imply that the initial stages of Levallois core preparation are the most wasteful and that, thereafter, the relationship between cutting edge length and waste is, at the very least, stabilized across subsequent stages of Levallois removal. However, Brantingham and Kuhn’s (2001) models did not assess the economy of repreparation, which is potentially detrimental given that as the reduction sequence proceeds, the knapper is dealing, incrementally, with less material mass and may encounter new challenges as a result. Given these points, Lycett and Eren (2013b) examined these specific factors using a series of experimentally produced Levallois reduction sequences, under conditions that more closely approximate the material reality and challenges of knapping faced by Levallois-producing hominins. A total of 3957 flaking events were considered in their analyses, using chert nodules of varying shapes and dimensions, and six specific measures of economy to examine Levallois reduction across successive phases. Lycett and Eren’s (2013b) analyses found that key assumptions of Brantingham and Kuhn’s (2001) mathematical model could be upheld. That is, once the initial Levallois core morphology is established, the economic costs of mass loss can, at the very least, be stabilized across all subsequent stages. Moreover, in some cases, removing a subsequent Levallois flake was found to be statistically more economical in terms of mass lost through preparation than for the preceding flake. In other words, although repreparation of convexities takes place as Levallois reduction proceeds, the raw material costs need not necessarily ever be large enough that it results in significantly greater levels of material usage than in preceding stages. Hence, these experimental results supported the assumptions underlying Brantingham and Kuhn’s (2001) models and, in turn, lend support to the hypothesis that economic considerations may logically have played a part in motivating the prehistoric production of Levallois-style reduction sequences. In sum, various lines of research have recently supported the long-held view that Levallois reduction, and its products were motivated by practical factors, most specifically by delineating more clearly what those potential motivating factors were. Brantingham and Kuhn (2001: 758–759) noted that “coinciding optima,” in terms of multiple functional/adaptive advantages, might explain the widespread geographical and temporal spread of Levallois (see also, Adler et al. 2014). Combining all of the considerations described above, therefore, although factors relating to both raw material economy and the utility of flakes will vary (from high optimality to low optimality) depending on the particular method of core reduction utilized, Levallois reduction scores highly in both of these particular variables. Archaeological examples of Levallois assemblages will, however, inevitably display a continuous range of

384

Eren and Lycett

variability around these parameters, for example, through “recurrent” strategies and/or due to knapping errors (e.g., Delagnes 1995; Schlanger 1996; Meignen et al. 2009). In this sense, classic or lineal Levallois would be unlikely to represent the “average,” but a theoretical optimum (Lycett and Eren 2013a). The key point is that according to the various results reviewed above, reduction schemes that more closely approximate the theoretical “optimum” that classic Levallois reduction can provide would―on average―have presented advantages to hominins in terms of economic variables and in terms of producing highly serviceable flakes, at least on a comparative basis with non-Levallois reduction strategies. There is one key factor, however, that is currently missing from such discussions. Inevitably, the basic functionality of stone “cutting tools” ultimately depends on some part of the tool having a suitably acute “edge angle” for this purpose. Indeed, for this reason, edge angle is a property of stone tools that has received consideration in ethnoarchaological work (e.g., Gould et al. 1971; White and Thomas 1972; White et al. 1977; Gould 1980) and, in turn, has frequently been given consideration in examination of archaeological material (e.g., Wilmsen 1968; Ferguson 1980; Jensen 1986; Hayden et al. 1996; Borel et al. 2013). Despite the fundamental role that edge angle plays in the basic functionality of a stone cutting tool, none of the aforementioned studies contending that Levallois core organization and reduction results in beneficial flake properties (in statistical terms) have yet considered the issue of edge angle. When comparing the edge angles of different analytical categories of flakes, the null hypothesis will inevitably be one of “no difference” between them. Based on previous work assessing experimentally produced Levallois flakes, however, we might expect putative PLFs to possess edge angles that differ from flakes involved in their production. Moreover, given previous arguments purporting a functional “optimality” for Levallois, we might expect that any statistical properties they possess can logically be linked to beneficial functional considerations. At the very least, the statistical properties of the edge angles of Levallois flakes should not be detrimental to functional concerns if contentions of optimality concerning other aspects of their physical tendencies are to be upheld. Here, therefore, we use experimentally produced lineal Levallois reduction sequences to address this issue empirically. Our primary aim is to statistically test the null hypothesis of no difference between the edge angles of PLFs and the flakes involved in their production. Having determined the statistical properties of our experimental Levallois reductions, we thereafter evaluate to what extent edge angles produced may, or may not, have been useful to prehistoric hominins engaged in this style of lithic reduction.

Materials and Methods Experimental Assemblage In order to procure the experimental flake sample examined here (n=525 flakes), one of us (MIE) undertook a series of multiple Levallois reductions (total n=75) on a series of 25 nodules of Texas (Fredericksburg) chert. Hence, from each nodule, one to five PLFs were removed with a mean of three PLFs per nodule. The cores produced during these

A Statistical Examination of Flake Edge Angles

385

reductions conformed fully to Boeda’s (1995) volumetric criteria for Levallois (Fig. 1), and the PLFs removed on average 50 % of a core’s upper surface area, with a standard deviation of 13 %. A series of 3D geometric morphometric analyses have previously (Eren and Lycett 2012) determined that these experimental Levallois cores accurately replicate the form of genuine archaeological specimens classified as “preferential Levallois cores” according to Boëda’s (1995) widely used criteria. Direct, hard hammer percussion was used exclusively throughout each reduction. Following Bradley (1977; see also Eren and Bradley 2009; Eren and Lycett 2012), we defined “ventral” flakes as those removed from the face from which the PLFs are removed and refer to flakes removed from the non-PLF surface as “dorsal” flakes. From each separate Levallois reduction stage, all debitage flakes from the dorsal and ventral surfaces, as well as each PLF, were bagged separately and labeled. The experimental assemblage consisted of 3957 total flakes. The analyzed experimental assemblage is the same used in Eren and Lycett (2012) and Lycett and Eren (2013b), and further details about the assemblage can be found in those references. Flake Sampling Protocol A total sample of 525 experimentally produced flakes were examined in this study, which was comprised of 75 PLFs, 222 dorsal flakes, and 228 ventral flakes. Only flakes >2 cm in maximum length were measured. A maximum of six flakes per PLF were measured—up to three removed from the ventral surface and up to three removed from the dorsal surface during each PLF’s production. Wherever the total number of potentially measurable flakes from a surface exceeded three specimens, three flakes were sampled randomly using a random number generator (http://www.randomizer. org/). Edge Angle Measurements On every flake, we measured edge angle at five locations (Fig. 2). To determine these locations, we first drew a line that bisected the flake’s axis of percussion. The intersection of this line with the flake’s distal edge determined the first edge angle location (distal edge angle). Orthogonally to this axis, we drew another line, which determined two more edge angle measurement locations (right edge angle and left edge angle). Two more lines bisecting the vertical and horizontal axes were drawn which is where the final two edge angle measurements were recorded (right distal edge angle and left distal edge angle). From the 75 PLFs and 450 debitage flakes, we aimed to have 2625 edge angle measurements, but we encountered a small number (3.6 %) of locations that had been broken during the process of flake removal. Our final data set was thus comprised of 2534 total edge angle measurements across the 525 flakes. The breakdown of recorded edge angles per flake category and per location can be found in Table 1. To measure each edge angle (θ), we used the “caliper method” of Dibble and Bernard (1980), which can be implemented using standard needle-point calipers

386

Eren and Lycett

Fig. 2 a Edge angles were recorded from five locations on a PLF. To determine these locations, a line was drawn that bisected the flake’s axis of percussion. The intersection of this line with the flake’s distal edge determined the first edge angle location (distal edge angle). Orthogonally to this axis, another line was drawn, which determined two more edge angle measurement locations (right lateral edge angle and left lateral edge angle). Two more lines bisecting the vertical and horizontal axes were drawn which is where the final two edge angle measurements were recorded (right distal edge angle and left distal edge angle). b For the “smoothness” analysis (see main text for description), the absolute value of the difference of each set of adjacent edge angles was calculated

(Fig. 3). By taking flake thickness at a predetermined distance from the edge of the flake, the method proceeds on the basis of a straightforward trigonomic function:    −1 0:5T ; θ ¼ 2 tan D where D is a known distance perpendicular to the edge (vertex) of the flake (here set at 3 mm), T is the thickness of the flake at the predetermined point from the flake edge, and θ is the computed edge angle measured in degrees (°) (Dibble and Bernard 1980: 360). Analyses and Statistical Tests We statistically examined the edge angles of PLFs against those of dorsal and ventral flakes in three different sets of analyses. Table 1 Edge angle measurements per flake Flake category

Edge angle location

PLF

Dorsal

Ventral

Total

Distal

73

209

205

487

Right lateral

75

217

220

512

Left lateral

75

213

217

505

Right distal

75

218

224

517

Left distal

75

217

221

513

Total edge angle measurements

373

1074

1087

Total sample=2534

A Statistical Examination of Flake Edge Angles

387

Fig. 3 Computation of edge angles using the Bcaliper method.^ Where, D is a known distance perpendicular to the edge (vertex) of the flake, T is the thickness of the flake at the predetermined point from the flake edge, and θ is the computed edge angle, where θ=2[tan−1(0.5T/D)] (redrawn and modified after Dibble and Bernard (1980) and Key and Lycett (2015))

Comparison of All Edge Angles Across Flake Categories We first statistically compared the edge angle values of all PLFs against those computed for both dorsal and ventral flakes. In other words, we compared the 373 PLF edge angles to the 1074 dorsal edge angles and the 1087 ventral edge angles. The data in each of our three flake populations were continuous, but Sharpio-Wilk tests indicated that they were not normally distributed (PLF, Shapiro-Wilk=0.978, p<0.001; dorsal, Shapiro-Wilk=0.968, p<0.001; ventral, Sharpiro-Wilk=0.957, p<0.001). Hence, to test for statistically significant (α=0.05) differences in edge angles of PLFs compared to the dorsal and ventral flakes, we used nonparametric Mann-Whitney U tests. This is a conservative statistical procedure that requires only minimal assumptions of the data (Dytham 2011: 119). Here, we report p values based on Monte Carlo simulation with 10,000 random assignments. Two tests were undertaken: one comparing PLFs to dorsal flakes and a second comparing PLFs to ventral flakes. To test for statistically significant differences in the variation of edge angles of different flake categories, we undertook D′ AD tests, which test for significant differences in coefficient of variation (CV) values. CVs are a simple descriptive statistic that can be used to express variation in a continuously distributed variable. The CV is simply the standard deviation of a range of values divided by the mean of those values (Sokal and Rohlf 1995: 57–59). It is convenient to multiply the computed ratio by 100 in order to express relative degrees of variation as a percentage. The D′ AD is a nonparametric statistical procedure that was originally designed for use in clinical trials as a means of testing for significant differences in CV values of patient responses (Feltz and Miller 1996). The applicability of the method for analyzing archaeological data was later highlighted by Eerkens and Bettinger (2001), and it has subsequently been used in archaeological contexts by Lycett and Gowlett (2008) and Okumura and Araujo (2014).

388

Eren and Lycett

Comparison of Averaged Edge Angles Across Flake Categories We next compared the three flake categories in terms of each flake’s average edge angle. In other words, we averaged the edge angle measurements for each PLF (n=75), dorsal flake (n=222), and ventral flake (n=228) and then compared the flake categories. Although the PLF dataset in this instance conformed to a normal distribution (Shapiro-Wilk=0.981, p=0.339), the dorsal and ventral populations did not (dorsal, Shapiro-Wilk = 0.974, p < 0.001; ventral, Shapiro-Wilk = 0.947, p<0.001). We thus again used nonparametric Mann-Whitney U tests to test for statistically significant (α=0.05) differences in edge angles of PLFs compared to the dorsal and ventral flakes. Congruent with the previous set of analyses, reported p values were based on Monte Carlo simulation with 10,000 random assignments. Again, two tests were undertaken: one comparing PLFs to dorsal flakes and a second comparing PLFs to ventral flakes. Edge Angle Organization (“Smoothness”) Along an Edge Our final analysis involved consideration of the organization (or smoothness) of edge angles along an edge. In this analysis, we were concerned not with the variation of edge angles across flake categories per se, but instead whether that variation was structured differently along the edges of PLFs compared to dorsal or ventral flake edges. Figure 4 illustrates what we mean by this. There are two hypothetical stone tool edges depicted, A and B. Each edge has exactly the same variability of edge angle (i.e., CV=56.3 % in both cases). However, edge B exhibits a systematic pattern that edge A does not. This organized edge B patterning results in a smaller average difference between adjacent edge angles than does the disorganized edge A pattern. To understand edge angle organization along an edge, we calculated the absolute value of the difference between each set of adjacent edge angles on every flake (Figure 2). Once again, none of our three flake populations conformed to a normal distribution in terms of these data (PLF, Sharpiro-Wilk=0.886, p<0.001; dorsal, Shapiro-Wilk=0.886, p<0.001; ventral, Sharpio-Wilk=0.864, p<0.001). In order to test the null hypothesis of no difference (α=0.05) between any of the three different flakes groups, we thus used a nonparametric Kruskal-Wallis test (Dytham 2011: 142).

Results Basic Comparisons and Descriptive Statistics for Different Flake Categories Table 2 shows the mean and standard deviations for the three different categories of flakes. Figure 5 shows box plots that also indicate how edge angles compare across the three different flake categories. It is notable even at this stage that median edge angles recorded for PLFs are higher than those of the other two flake categories (Fig. 5), with mean edge angles for PLFs also subsequently higher than for the other two flake categories (Table 2).

A Statistical Examination of Flake Edge Angles

389

Fig. 4 Comparison of edge angle organization (or Bsmoothness^) along the edge of a flake. Two hypothetical stone tool edges are depicted: flake edges A and B. Each edge has exactly the same overall variability of edge angles (i.e., CV=56.3 % in both cases). However, flake edge B exhibits a systematic pattern that flake edge A does not. This more organized (or Bsmoother^) transition between angles along the edge of flake B results in a smaller average difference between adjacent edge angles compared to the relatively Bdisorganized^ pattern along the edge of flake A

Comparison of All Edge Angles Across Flake Categories Our first set of analyses assessed whether statistically significant differences existed between edge angles of PLFs compared to the two debitage flake categories. MannWhitney U tests indicated highly significant differences between PLFs and dorsal flakes (U=9.637, Monte Carlo p<0.0001) and also between PLFs and ventral flakes (U=7.539, Monte Carlo p<0.0001). Indeed, although PLFs and ventral flakes are Table 2 Mean and standard deviations for computed flake edge angles in the three different flake categories

PLF

Dorsal

Ventral

45.35

32.21

29.39

Standard deviation 13.63

13.63

12.75

Mean edge angles CV

0.30 (30 %) 0.423 (42.3 %) 0.434 (43.4 %)

390

Eren and Lycett

Fig. 5 Box plot showing basic patterns of edge angle variation across all three flake categories. Median values in each case are indicated by the horizontal lines across each 25–75 percentile box. Whiskers mark largest data point ≤1.5 times box range. Outliers (circles) indicated

removed from the same surface of a Levallois core, statistical examination of their edge angles reveals that they are significantly different. The CV values for edge angles of the PLFs, dorsal, and ventral flake categories were 30.0, 42.3, and 43.4 %, respectively. Two D′ AD tests comparing PLFs to dorsal flakes (D′ AD=41.61, p<0.001) and PLFs to ventral flakes (D′ AD=46.90, p<0.001) indicated that these CV values were significantly lower in the case of the PLFs compared to the other two flake categories. Hence, these statistical results indicate that PLFs are significantly different from the debitage flake categories in terms of their edge angles. In sum, PLFs are statistically associated with higher edge angle values, and PLF edge angles are significantly less variable around their higher mean values than either ventral or dorsal flakes. A Comparison of Averaged Edge Angles Across Flake Categories Our second set of analyses assessed whether statistically significant differences existed between edge angles of PLFs compared to the two debitage flake categories when edge angle measurements were averaged per flake. Figure 6 shows box plots that indicate, in general terms, how the averaged edge angles compare across the three different flake categories. It is again notable that median edge angles for PLFs are higher than for those of the other two flake categories. Again, Mann-Whitney U tests indicated highly significant differences between PLFs and dorsal flakes (U=3029, Monte Carlo p <0.0001) and also between PLFs and ventral flakes (U = 1979, Monte Carlo p<0.0001). The CV values of averaged edge angles for PLFs, dorsal, and ventral flake categories were 22.7, 32.3, and 31.4 %, respectively. Two D′ AD tests comparing PLFs to dorsal flakes (D′ AD=9.72, p<0.002) and PLFs to ventral flakes (D′ AD=8.43, p<0.004) indicated that these CV values were significantly lower in the case of the PLFs compared to the other two flake categories.

A Statistical Examination of Flake Edge Angles

391

Fig. 6 Box plot showing patterns of mean edge angle variation across all three flake categories. Median values in each case are indicated by the horizontal lines across each 25–75 percentile box. Whiskers mark largest data point ≤1.5 times box range. Outliers (circles) indicated

Hence, these results again indicate that PLFs are significantly different from the debitage flake categories both in terms of their tendency toward higher edge angle values, as well as in terms of their lower variation around their higher mean and median values. Edge Angle Organization Along an Edge Our final analysis assessed whether there was more organization of PLF edge angles along an edge than there was for dorsal or ventral edge angles. Average differences between adjacent edge angles in the case of PLFs were 10.55°, while in the case of dorsal and ventral flakes, these were 9.40° and 9.16°, respectively. It is notable in this regard that in all three cases, average differences between adjacent flake angles were relatively low (i.e., on the order of 10°). A Kruskal-Wallis test indicated that there were no significant differences between PLFs and either dorsal or ventral flake populations in this regard (H=5.578, asymptotic [two-tailed] p=0.061).

Discussion Studies have indicated that Levallois-style core reduction may have offered practical benefits to hominin populations, both in terms of reduction economy defined in specific terms (Brantingham and Kuhn 2001; Lycett and Eren 2013b) and with regard to the production of flakes with functionally desirable properties (Eren and Lycett 2012; Lycett and Eren 2013a). However, none of the aforementioned studies have considered one of the most important functional attributes of flake tools, which is edge angle. Here, to address this shortcoming, we statistically examined a sample of n=525 flakes produced from experimental Levallois reduction sequences based around classic or lineal Levallois core production and reduction. The sample comprised “debitage” flakes

392

Eren and Lycett

produced from the ventral surfaces of cores (n=228) and the dorsal surfaces of cores (n=222), as well as n=75 putative “preferential Levallois flakes” or simply “PLFs.” Our basic dataset for analysis was comprised of 2534 total edge angle measurements across the 525 flakes. Our statistical analyses demonstrated that median edge angles were significantly higher (i.e., more obtuse) in the PLFs compared to either dorsally or ventrally removed debitage flakes. This statistical difference was found regardless of whether we considered the flake angle measurements separately or averaged the edge angles for each individual flake. What is particularly notable here is that even though ventral flakes and PLFs are removed from the same surface of the core, they are significantly different in terms of edge angles, emphasizing that with regard to this particular property, the two sets of flakes are fundamentally different due to their differing role and position in the overall reduction sequence. We also compared edge angles across flake categories in terms of their extent of overall variability, using the coefficient of variation (CV) statistic. Statistical testing demonstrated that PLFs were significantly less variable around their higher mean edge angle variables compared to either dorsally or ventrally removed debitage flakes. In other words, PLFs are more constrained and consistent in their higher values for edge angle properties than are the flakes involved in their production. Again, this difference was apparent regardless of whether we considered the flake angle measurements separately or averaged the edge angles for each flake. Our final statistical analysis considered whether adjacent edge angles (along the edge of each flake) exhibited statistically greater differences in any of the three different categories of flake. We considered this latter measurement an indication of the smoothness of edge angle properties along the edges of flakes. This analysis indicated no statistical differences across flake categories in these terms. In sum, our statistical analyses demonstrated that the experimentally produced PLF edge angles were significantly different from those of the flakes involved in their production. These differences were evident both in terms of relatively higher (i.e., more obtuse) edge angles than debitage flakes and in being significantly less variable around their higher mean edge angles compared to debitage flakes. Given previous arguments concerning the idea that Levallois-style reduction and its products may have helped “optimize” specifically desirable characteristics in flake tools, it is important to consider how the issue of edge angles relates to functional issues. As discussed by Key and Lycett (2015), the engineering literature suggests that the edge angle of cutting tools has implications in terms of cutting efficiency. The initiation of a “cut” involving a splitting of material is affected by the variable of “cutting stress” (force per unit area), a factor which is influenced by edge angle (Atkins 2009). More obtuse angles restrict the depth of penetration during cutting, increasing the surface area of the tool that is in contact with the material being cut, which will result in increased friction (Atkins 2009). With regard to stone cutting tools, increased edge angles will thus translate into a requirement on the part of the tool user to either increase the load applied or extend the time engaged in the process of cutting, either of which will require increased energy expenditure (Key and Lycett 2015). Ethnographic reports for flake stone tool edge angles have tended to provide mean estimates ranging from ~35° to 58° (e.g., Gould et al. 1971; White and Thomas 1972; White et al. 1977; Gould 1980). There is a recognition that the “performance characteristics” of a given archaeological artifact are an important factor for consideration (e.g., Schiffer and Skibo 1987;

A Statistical Examination of Flake Edge Angles

393

O’Brien et al. 1994; Beck 1995; Pool and Britt 2000). Performance characteristics can be defined as activity-specific interactions that will influence how an artifact will perform in a given task in relation to a set of functional properties (Skibo and Schiffer 2001:143), or more precisely, they are “the behavioral capabilities that an artifact must possess in order to fulfill its functions in a specific activity” (Schiffer and Skibo 1987: 599). Inevitably, therefore, “performance characteristics are strongly influenced by an artifact’s formal properties” (Schiffer and Skibo 1987: 31). As others (e.g., Schiffer and Skibo 1987: 30) have noted, the concept is not unlike that of “functional morphology” in the study of animal anatomy, which links directly the form of an organism—or a particular trait of that organism—to its role in various functions important to that animal (e.g., Kardong 2006). In the case of human-manufactured products, performance characteristics have obvious implications for concepts of artifact “design” or simply the idea that people in the past may have organized behavioral and manufacturing activities in ways that “engineered” the morphology of artifacts to deliberately influence their “performance” in a given activity (Schiffer and Skibo 1987; O’Brien et al. 1994; Bleed 2001; Skibo and Schiffer 2001). A reasonable prediction of “intent” underlying particular archaeological patterns is that they vary in ways that might logically have been beneficial in practical terms. While this does not offer “proof” of intent, it does at least address the issue of logical consistency with a hypothesis invoking “intent” (Eren and Lycett 2012; Eren et al. 2013). Given these matters, what do the statistical patterns identified here imply for functional considerations relating to Levallois reduction and its flake products? Perhaps the first point of note here is the range of flake edge angles produced during the experimental Levallois reductions—everything from ~5° in the most acute flakes to ~75° in the more obtuse examples, even excluding outliers. Hence, in principle, Levallois-style reduction is capable of producing flakes with a wide range of angles from which a prospective tool user might make a selection. However, the statistical analyses undertaken here identified two definite statistical patterns in the various flakes. That is, PLFs exhibited statistically higher (i.e., more obtuse) edge angles than debitage flakes and were significantly less variable around their higher mean edge angles than the debitage flakes. It has sometimes been suggested (largely based on enthographic examples of the type of noted above) that flakes exceeding edge angles of ~50° degrees might be unsuitable as cutting tools (e.g., Wilmsen 1968; Ferguson 1980; Jensen 1986). However, recent experiments have suggested that such statements are incorrect and that cutting efficiency in flakes ranging to at least 70° are as efficient as those with more acute angles, especially in larger flakes that facilitate the application of greater loading (Key and Lycett 2015). Notably, the mean edge angle for the PLFs (i.e., 45°) analyzed here falls squarely within the range (i.e., ~35°–58°) reported in several ethnographic studies examining mean edge angles in flake tools (Gould et al. 1971; White and Thomas 1972; White et al. 1977; Gould 1980). Indeed, at least 50 % of PLFs examined here would fall directly within the range of these reported averages due to their relatively low levels of variation around the mean compared to debitage flakes. At the very least, therefore, there appears little evidence that the edge angles observed here in the experimental PLFs would be of detriment to their use as functional tools. However, there is a second factor relating to functional issues that must be considered in regard to the edge angles of stone flake tools: the robustness of the edge and its subsequent resistance to breakage and blunting through microfracture (Tringham et al.

394

Eren and Lycett

1974). This factor is important not only because of blunting (i.e., functionality), but it also dictates how frequently a tool edge must be resharpened and thus relates to the potential use-life (or attrition rate) of an individual flake (Collins 2008). Durability of flake tools is an issue that is suggested to have affected hominin decisions and behavior (e.g., in regard to raw material selection) from the Oldowan onward (Braun et al. 2009). Experiments have indicated that flake tools with more acute edge angles are more susceptible to edge microfracture (i.e., attrition through the removal of small flakes) than are flakes with more obtuse angles (Tringham et al. 1974: 180). Likewise, recent experiments involving the trampling of flakes have established a firm relationship between edge angles and attrition potential, with acute angles showing higher levels of damage than more obtuse edge angles (McPherron et al. 2014). It is plausible, therefore, that ethnographically reported preferences for flake tools with mean angles of ~35°–58° represent a broadly “optimal” trade-off between increased surface area contact (i.e., decreased cutting stress) initiated by not choosing more acute flake angles, and the decrease in edge damage and/or attrition rate that will accompany choosing flakes with relatively more obtuse angles. Regardless of this, given the experimental data pertaining to edge damage, the PLFs examined here would certainly be less at risk of microfracture than debitage flakes due to their statistically greater edge angle values. In turn, the use-life of PLFs compared to other flakes would be relatively more extended. Our results also relate to a further factor that is worthy of consideration. Recent finds have suggested that hominins inhabiting the site of Nor Geghi 1 (NG1) in Armenia around 300 thousand years ago initiated Levallois-style reduction sequences in apparent instance of independent technological convergence with African hominins (Adler et al. 2014). It is important to consider, however, why an apparent instance of technological innovation may have occurred given the specifics of the circumstance. Notably, the NG1 assemblage is made entirely on obsidian, the majority (93.7 %) of which is from sources only 2–8 km from the site (Adler et al. 2014). However, 3.2 % of the obsidian is from 70 km away, while 0.3 % is from 120 km away. Part of the reason, therefore, may be in relation to issues of mobility (Kuhn 1994), which are particularly pertinent with regard to Levallois given its practical advantages in terms of durable, serviceable flake products (Eren and Lycett 2012). However, the potentiality for Levallois reduction strategies to be initiated as a strategic response to particular raw material properties is a point that has been noted elsewhere (Brantingham et al. 2000). Trampling experiments by McBrearty et al. (1998) using both flint and obsidian flakes led them to note that obsidian flakes were more susceptible to edge damage, which they contend is due to the increased brittleness of obsidian. Given the results reported here, therefore, the brittleness of obsidian and susceptibility to edge damage may have been an encouraging factor in hominins at NG1 experimenting with Levallois-style reduction schemas given that increased edge angles would have helped reduce damage and sustain artifact use-lives in these particular circumstances. One set of our results that failed to identify any statistical difference between edge angle properties of PLFs and debitage flakes was, however, our analysis of edge smoothness. We have already noted, however, that across all three flake categories mean values for our measure of differences between adjacent angles were broadly comparable and relatively low (i.e., on the order of 10° in all three cases). Minimally, therefore, if this property has any influence on the functional or performance

A Statistical Examination of Flake Edge Angles

395

characteristics of flakes, PLFs would not have been disadvantaged compared to debitage flakes. However, given that all three categories of flake appeared to show comparable levels of smoothness, it may be fruitful if future research considers reduction strategies that result in lower values of smoothness than we have recorded for Levallois reduction. One intriguing possibility is that the Upper Paleolithic transition to prismatic punch blades, and then pressure blades, was not necessarily because of blank morphology per se but because that blank morphology permits an increased smoothness of edge angles. Indeed, it may be worth considering if earlier instances of “blade” production (e.g., through Levallois strategies) led to different levels of edge smoothness than we have found. However, such speculations require empirical assessment via further work, an endeavor that will also need to determine what the performance benefits of smoothness are, if any.

Conclusions Here, we have demonstrated that Levallois flakes removed during classic lineal Levallois-styles reduction are indeed statistically different from the flakes involved in their production in terms of edge angles. However, based on current knowledge of how flake edge angle properties relate to issues of functionality, such differences would not be detrimental to their functionality. Indeed, the edge angle properties they possess would have provided distinct benefits to hominins engaged in their manufacture. Most particularly, Levallois-style core organization and reduction would have provided hominins with a reliable means of consistently producing flakes (i.e., Levallois flakes) possessing average flake angles that are optimal in terms of providing a viable cutting edge, yet not being so acute as to be weak and friable upon application. Hence, edge angle properties join an array of other features relating to reduction economy and beneficial flake properties that provide logical motive for why hominins may have organized core production and reduction around Levallois-style patterns at various times and places during the Pleistocene. Acknowledgments We are grateful to Noreen von Cramon-Taubadel for assistance and comments on this paper. This research was conducted whilst MIE was on a Leverhulme Trust Early Career Fellowship. MIE is currently supported financially by a University of Missouri College of Arts and Sciences Post-Doctoral Fellowship.

References Adler, D. S., Wilkinson, K. N., Blockley, S., Mark, D. F., Pinhasi, R., Schmidt-Magee, B. A., Nahapetyan, S., Mallol, C., Berna, F., Glauberman, P. J., Rasczynski-Henk, Y., Wales, N., Frahm, E., Jöris, O., MacLeod, A., Smith, V. C., Cullen, V. L., & Gasparian, B. (2014). Early Levallois technology and the Lower to Middle Paleolithic transition in the Southern Caucasus. Science, 345, 1609–1613. Atkins, T. (2009). The science and engineering of cutting: the mechanics and processes of separating, scratching and puncturing biomaterials, metals and non-metals. Oxford: Butterworth-Heinemann. Beck, C. (1995). Functional attributes and differential persistence of Great Basin dart forms. Journal of California and Great Basin Anthropology, 17, 222–243.

396

Eren and Lycett

Bleed, P. (2001). Artifice constrained: what determines technological choice? In M. B. Schiffer (Ed.), Perspectives on Technology (pp. 151–162). Albuquerque: University of New Mexico Press. Boëda, E. (1995). Levallois: a volumetric construction, methods, a technique. In H. L. Dibble & O. Bar-Yosef (Eds.), The Definition and Interpretation of Levallois Technology (pp. 41–68). Madison: Prehistory Press. Bordes, F. (1950). Principes d’une méthode d’étude des techniques de debitage et de la typologie du Paléolithique ancien et moyen. L'Anthropologie, 54, 19–34. Borel, A., Gaillard, C., Moncel, M.-H., Sala, R., Pouydebat, E., Simanjuntak, T., & Sémah, F. (2013). How to interpret informal flake assemblages? Integrating morphological description, usewear and morphometric analysis gave better understanding of the behaviors of Anatomically Modern Humans from Song Terus (Indonesia). Journal of Anthropological Archaeology, 32, 630–646. Bradley, B., (1977). Experimental Lithic Technology with Special Reference to the Middle Palaeolithic. Unpublished Ph.D. dissertation, Department of Archaeology, Cambridge University. Brantingham, P. J., & Kuhn, S. L. (2001). Constraints on Levallois core technology: a mathematical model. Journal of Archaeological Science, 28, 747–761. Brantingham, P. J., Olsen, J. W., Rech, J. A., & Krivoshapkin, A. I. (2000). Raw material quality and prepared core technologies in Northeast Asia. Journal of Archaeological Science, 27, 255–271. Braun, D. R., Plummer, T., Ferraro, J. V., Ditchfield, P., & Bishop, L. C. (2009). Raw material quality and Oldowan hominin toolstone preferences: evidence from Kanjera South, Kenya. Journal of Archaeological Science, 36, 1605–1614. Chazan, M. (1997). Redefining Levallois. Journal of Human Evolution, 33, 719–735. Clarke, D. L. (1972). Models and paradigms in contemporary archaeology. In D. L. Clarke (Ed.), Models in Archaeology (pp. 1–60). London: Methuen. Collins, S. (2008). Experimental investigations into edge performance and its implications for stone artefact reduction modelling. Journal of Archaeological Science, 35, 2164–2170. Commont, V. (1909). L’industrie moustérienne dans la region de nord de la France. Congrés Préhistorique de France, 5ième session. Bureaux de la Société Préhistoire de France, Paris, pp 115–157. Delagnes, A. (1995). Variability within uniformity: three levels of variability within the Levallois system. In H. L. Dibble & O. Bar-Yosef (Eds.), The Definition and Interpretation of Levallois Technology (pp. 201– 213). Madison: Prehistory Press. Dibble, H. L. (1989). The implications of stone tool types for the presence of language during the Lower and Middle Palaeolithic. In P. Mellars & C. Stringer (Eds.), The human revolution: behavioural and biological perspectives on the origins of modern humans (pp. 415–432). Edinburgh: Edinburgh University Press. Dibble, H. L., & Bernard, M. C. (1980). A comparative study of basic edge angle measurement techniques. American Antiquity, 45, 857–865. Dytham, C. (2011). Choosing and using statistics: a biologist’s guide (3rd ed.). Oxford: Blackwell Science. Eerkens, J. W., & Bettinger, R. L. (2001). Techniques for assessing standardization in artifact assemblages: can we scale variability? American Antiquity, 66, 493–504. Eren, M., & Bradley, B. (2009). Experimental evaluation of the Levallois “core shape maintenance” hypothesis. Lithic Technology, 34, 119–125. Eren, M. I., & Lycett, S. J. (2012). Why Levallois? A morphometric comparison of experimental ‘preferential’ Levallois flakes versus debitage flakes. PLoS ONE, 7, e29273. doi:10.1371/journal.pone.0029273. Eren, M. I., Bradley, B., & Sampson, C. G. (2011a). Middle Paleolithic skill-level and the Individual knapper: an experiment. American Antiquity, 76, 229–251. Eren, M. I., Lycett, S. J., Roos, C., & Sampson, C. G. (2011b). Toolstone constraints on knapping skill: Levallois reduction with two different raw materials. Journal of Archaeological Science, 38, 2731–2739. Eren, M. I., Patten, R. J., O’Brien, M. J., & Meltzer, D. J. (2013). Refuting the technological cornerstone of the Ice-Age Atlantic crossing hypothesis. Journal of Archaeological Science, 40, 2934–2941. Feltz, C. J., & Miller, G. E. (1996). An asymptotic test for the equality of coefficients of variation from k populations. Statistics in Medicine, 15, 647–658. Ferguson, W. C. (1980). Edge-angle classification of the Quininup Brook implements: testing the ethnographic analogy. Archaeology and Physical Anthropology in Oceania, 15, 56–72. Gould, R. A. (1980). Living Archaeology. Cambridge: Cambridge University Press. Gould, R. A., Koster, D. A., & Sontz, A. H. L. (1971). The lithic assemblage of the Western Desert Aborigines of Australia. American Antiquity, 36, 149–169. Hayden, B., Franco, N., & Spafford, J. (1996). Evaluating lithic strategies and design criteria. In G. H. Odell (Ed.), Stone tools: theoretical insights into human prehistory (pp. 9–50). New York: Plenum Press. Hovers, E., & Belfer-Cohen, A. (2013). On variability and complexity. Current Anthropology, 54, S337–S357. Hublin, J.-J. (2009). The origin of Neandertals. Proceedings of the National Academy of Sciences USA, 106, 16022–16027.

A Statistical Examination of Flake Edge Angles

397

Jensen, H. J. (1986). Unretouched blades in the Late Mesolithic of South Scandinavia: a functional study. Oxford Journal of Archaeology, 5, 19–33. Kardong, K. V. (2006). Vertebrates: comparative anatomy, function, evolution. Boston: MA, McGraw-Hill. Key, A. J. M., & Lycett, S. J. (2015). Edge angle as a variably influential factor in flake cutting efficiency: an experimental investigation of its relationship with tool size and loading. Archaeometry. doi:10.1111/arcm. 12140. Kuhn, S. L. (1994). A formal approach to the design and assembly of mobile toolkits. American Antiquity, 59, 426–442. Kuhn, S. L., & Hovers, E. (2013). Alternative pathways to complexity: evolutionary trajectories in the Middle Paleolithic and Middle Stone Age. Current Anthropology, 54, S176–S182. Lycett, S. J. (2009). Are Victoria West cores ‘proto-Levallois’? A phylogenetic assessment. Journal of Human Evolution, 56(2), 175–191. Lycett, S. J., & Eren, M. I. (2013a). Levallois lessons: the challenge of integrating mathematical models, quantitative experiments and the archaeological record. World Archaeology, 45, 519–538. Lycett, S. J., & Eren, M. I. (2013b). Levallois economics: an examination of ‘waste’ production in experimentally produced Levallois reduction sequences. Journal of Archaeological Science, 40, 2384–2392. Lycett, S. J., & Gowlett, J. A. J. (2008). On questions surrounding the Acheulean ‘tradition’. World Archaeology, 40, 295–315. McBrearty, S., Bishop, L., Plummer, T., Dewar, R., & Conard, N. (1998). Tools underfoot: human trampling as an agent of lithic artifact edge modification. American Antiquity, 63, 108–129. McPherron, S. P., Braun, D. R., Dogandžić, T., Archer, W., Desta, D., & Lin, S. C. (2014). An experimental assessment of the influences on edge damage to lithic artifacts: a consideration of edge angle, substrate grain size, raw material properties, and exposed face. Journal of Archaeological Science, 49, 70–82. Meignen, L., Delagnes, A., & Bourguignon, L. (2009). Patterns of lithic material procurements and transformation during the Middle Palaeolithic in western Europe. In B. Adams & B. S. Blades (Eds.), Lithic Materials and Palaeolithic Societies (pp. 15–24). Chichester: Blackwell. Moncel, M.-H., Moigne, A.-M., Sam, Y., & Combier, J. (2011). The emergence of Neanderthal technical behavior: new evidence from Orgnac 3 (Level 1, MIS 8), southeastern France. Current Anthropology, 52, 37–75. Noble, W., & Davidson, I. (1996). Human evolution, language and mind: a psychological and archaeological enquiry. Cambridge: Cambridge University Press. O’Brien, M. J., Holland, T. D., Hoard, R. J., & Fox, G. L. (1994). Evolutionary implications of design and performance characteristics of prehistoric pottery. Journal of Archaeological Method and Theory, 1, 259– 304. Okumura, M., & Araujo, A. G. (2014). Long-term cultural stability in hunter–gatherers: a case study using traditional and geometric morphometric analysis of lithic stemmed bifacial points from Southern Brazil. Journal of Archaeological Science, 45, 59–71. Pelegrin, J. (2005). Remarks about archaeological techniques and methods of knapping. In V. Roux & B. Bril (Eds.), Stone Knapping: the Necessary Conditions for a Uniquely HomininBehaviour (pp. 23–33). Cambridge: McDonald Institute Monographs. Pepère, M. (1986). Apport de la typométrie à la definition des éclats Levallois: l'exemple d'Ault. Bulletin de la Société préhistorique française, 83, 115–118. Picin, A., Vaquero, M., Weniger, G. C., & Carbonell, E. (2014). Flake morphologies and patterns of core configuration at the AbricRomaní rock-shelter: a geometric morphometric approach. Quaternary International, 350, 84–93. Pool, C. A., & Britt, G. M. (2000). Ceramic perspective on the Formative to Classic transition in Southern Veracruz, Mexico. Latin American Antiquity, 11, 139–161. Sandgathe, D. M. (2004). Alternative interpretation of the Levallois reduction technique. Lithic Technology, 29, 147–159. Schiffer, M. B., & Skibo, J. M. (1987). Theory and experiment in the study of technological change. Current Anthropology, 595–622. Schlanger, N. (1996). Understanding Levallois: lithic technology and cognitive archaeology. Cambridge Archaeological Journal, 6, 231–254. Skibo, J. M., & Schiffer, M. B. (2001). Understanding artifact variability and change: a behavioral framework. In M. B. Schiffer (Ed.), Perspectives on Technology (pp. 139–149). Albuquerque: University of New Mexico Press. Simão, J. (2002). Tools evolve: the artificial selection and evolution of Paleolithic stone tools. Behavioral and Brain Sciences, 25, 419. Smith, R. A. (1911). A Palaeolithic industry at Northfleet, Kent. Archaeologica, 62, 512–532.

398

Eren and Lycett

Sokal, R. R., & Rohlf, F. J. (1995). Biometry. Third Edition ed. W.H. Freeman & Co, New York. Spurrell, F. C. J. (1884). On some Palaeolithic knapping tools and modes of using them. Journal of the Anthropological Institute, 13, 109–118. Tringham, R., Cooper, G., Odell, G., Voytek, B., & Whitman, A. (1974). Experimentation in the formation of edge damage. Journal of Field Archaeology, 1, 171–196. Tryon, C. A., & Faith, J. T. (2013). Variability in the Middle Stone Age of eastern Africa. Current Anthropology, 54, S234–S254. Turq, A. (1992). Raw material and technological studies of the Quina Mousterian in Perigord. In: Dibble H, Mellars P (eds) The Middle Paleolithic: adaptation, behavior, and variability. The University Museum, University of Pennsylvania, Philadelphia, pp 75–85. Van Peer, P. (1992). The Levallois Reduction Strategy. Madison: Prehistory Press. White, J. P., & Thomas, D. H. (1972). What means these stones? Ethno-taxonomic models and archaeological interpretations in the New Guinea Highlands. In D. L. Clarke (Ed.), Models in archaeology (pp. 275– 308). London: Duckworth Press. White, J. P., Modjeska, N., & Hipuya, I. (1977). Group definitions and mental templates: an ethnographic experiment. In R. V. S. Wright (Ed.), Stone tools as cultural markers. Change, evolution and complexity (pp. 380–390). Camberra: Australian Institute of Aboriginal Studies. Wilmsen, E. N. (1968). Functional analysis of flaked stone artefacts. American Antiquity, 33, 156–161. Wurz, S. (2013). Technological trends in the Middle Stone Age of South Africa between MIS 7 and MIS 3. Current Anthropology, 54, S305–S319. Wynn, T., & Coolidge, F. L. (2004). The expert Neandertal mind. Journal of Human Evolution, 46, 467–487. Wynn, T., & Coolidge, F. L. (2010). How Levallois reduction is similar to, and not similar to, playing chess. In A. Nowell & I. Davidson (Eds.), Stone tools and the evolution of human cognition (pp. 83–103). Boulder: University Press of Colorado.