Journal of Archaeological Science 38 (2011) 2731e2739Contents lists available at ScienceDirectJournal of Archaeological Sciencejournal homepage: http://www.elsevier.com/locate/jasToolstone constraints on knapping skill: Levallois reduction with two differentraw materialsMetin I. Eren a, b, *, Stephen J. Lycett b, Christopher I. Roos a, C. Garth Sampson caDepartment of Anthropology, Southern Methodist University, Heroy Building (ISEM), Dallas, Texas 75275 0336, USADepartment of Anthropology, University of Kent, Canterbury, UKcDepartment of Anthropology, Texas State University, USAba r t i c l e i n f oa b s t r a c tArticle history:Received 6 February 2011Received in revised form4 June 2011Accepted 14 June 2011Lithic raw material constraints are widely assumed to be a determining factor of ﬂaked stone toolmorphology, but this assumption remains largely untested. We conducted a controlled experiment todetermine whether a knapper’s growing replication skills would be hindered if the toolstone used wasswitched from large ﬂakes of an easily worked chert to nodules of less tractable one. Two batches ofPreferential Levallois cores were knapped, an earlier series made from standardised large ﬂakes ofsediments dominated by chalcedonic quartz followed by a more challenging one using variably-shaped,cortical nodules of microcrystaline quartz that varies in the completeness of quartz replacement ofcalcite and dolomite. Skill level markers were designed to measure the knapper’s ability to achievea series of set goals. These were quantiﬁed and subjected to statistical testing. In all but one test,signiﬁcant increases in skill could be detected from the earlier to the later batch of reductions, despite thedrop in toolstone quality. Signiﬁcant improvements in the consistency of the knapper’s output could alsobe detected. However, the switch to a more challenging, nodular chert did require extra shaping, whichresulted in more waste. This masked visible progress towards producing a less costly core. Overall, ourresults do not support the assumed primacy of toolstone constraints over other factors in inﬂuencing themorphology of ﬂaked stone tools.Ó 2011 Elsevier Ltd. All rights reserved.Keywords:Lithic raw material constraintsExperimental archaeologyMiddle PaleolithicFlintknapping skillPreferential Levallois1. IntroductionIt has long been recognized that ﬂaking stone tools is a reductiveprocess in which slivers of stone are knapped from a larger rock (e.g.Evans, 1872: 13; Frison, 1968; Shott, 2010: 276e277). It is equallywell known that isotropic rocks, i.e. those free of cleavage planes orother inclusions that inhibit the free passage of energy, are the onesmost suitable for successful initiation and control of conchoidalfracture. The best exhibit properties that combine elasticity, brittleness, hardness, and homogeneity (Goodman, 1944; Inizan et al.,1999; Whittaker, 1994). The exact combination of these propertiesvaries from one potentially suitable raw material (toolstone) toanother, and it is reasonable to suspect that the exact mix of suchproperties in any particular toolstone could affect the form ofPalaeolithic artefacts made from it (Archer and Braun 2010; Dibble,* Corresponding author. Department of Anthropology, Southern MethodistUniversity, Heroy Building (ISEM), Dallas, Texas 75275 0336, USA. Tel.:þ1440 933 7155.E-mail addresses: [email protected] (M.I. Eren), [email protected] (S.J.Lycett), [email protected] (C.I. Roos), [email protected] (C.G. Sampson).0305-4403/ e see front matter Ó 2011 Elsevier Ltd. All rights reserved.doi:10.1016/j.jas.2011.06.0111985; Goodman, 1944; Isaac, 1972; Jones, 1979). Untested assumptions about toolstone constraints reached their apogee at the socalled Movius Line e a putative subcontinental pattern of artefactvariation where bifacial and prepared core reduction is present inIndia and to the west, but is all but absent farther east. The purportedlack of better-quality toolstone east of the Line, although oncea popular explanation, has now fallen out of favour (Bar-Yosef et al.,In press; Brantingham et al., 2000; Lycett and Bae, 2010; Nortonet al., 2006; Schick, 1994). While there have been repeated calls toconsider carefully the effects of raw material parameters and theirinﬂuences on artefact form and assemblage composition (Clegg,1977; Isaac, 1972; Kuhn, 1992), an understanding of how thesefactors play out at the assemblage level remains limited.Two different sets of toolstone properties are thought toinﬂuence lithic artefact form (e.g. Goodman, 1944; Jones, 1979;Clark, 2001: 1). The ﬁrst set is internal, i.e. the mechanicalﬂaking properties already listed. The second set is external,namely the form (size, shape, surface regularity, and presence ofcortex) of the initial nodule, block, or blank from which the ﬂakesare struck. External parameters have been claimed to be a primarycause of differences between whole assemblages made of chert
2732M.I. Eren et al. / Journal of Archaeological Science 38 (2011) 2731e2739that may exhibit a similar mix of internal properties (Ashton andMcNabb, 1994).Despite its seemingly logical basis, the case for toolstoneconstraints on artefact form is hardly supported by experiment, andthere are “few studies that actually document the impact of theproperties of raw materials on the way in which particular tooltypes were produced” (Holdoway and Stern, 2004: 55). Indeed,several recent case studies suggest that raw material may not bethe dominant factor in regulating the ﬁnal form of knapped artefacts. Archer and Braun (2010), for instance, undertook threedimensional geometric morphometric analyses of Acheulean handaxes from Elandsfontein (South Africa) and compared them toexperimentally-produced replicas. Their comparisons reveal thatmuch of the shape variation between handaxes could not beaccounted for solely by raw material. A more emphatic refutation ofthe argument may be found at Castel di Guido (Italy) where handaxes were knapped in both stone and bone. Here, Costa’s (2010)morphometric analysis of outline form failed to detect any statistical difference in plan shape between stone and bone handaxesdespite the fundamental differences in ﬂaking properties of the twomaterials used. In a similar manner, Clarkson’s (2010) multivariatemorphometric comparison of cores from the Howiesons Poorttechnocomplex (South Africa) revealed that classiﬁcation scoresbased on geographic region were more accurate than those basedon raw material (72.8 versus 46.0% accuracy, respectively). Whilethese results do not deny a role for raw material in causing differences in cores between regions, they do indicate that “raw materialdifferences would appear to be subservient to other causes ofvariation” Clarkson (2010: 53).It would appear, then, that the jury is still out on whether or nottoolstone quality exercises any signiﬁcant constraint on resultingtool form. We doubt that the question will be resolved conclusivelyby ever-more sophisticated analyses of prehistoric materials,because there always remain too many unknowns. To ﬁll these gapsrequires so large a body of reﬁtted material as to be beyond ourpresent means and abilities. To circumvent the problem wedesigned a replication experiment which documented how theimproving knapping skills of an individual knapper (Eren) might beinterrupted when the toolstone on which he was honing his technique was switched to a material that was less tractable in both itsinternal and external properties.The experiment focused on Eren’s ability to execute a particularly challenging knapping routine e that of Levallois coreproduction. Levallois reduction was common during the MiddlePalaeolithic and has been recorded at sites in Africa, Europe, andwestern Asia (Dibble and Bar-Yosef, 1995 and chapters therein).Being part of a group of ‘prepared core’ reduction strategies, there isdisagreement on how to deﬁne Levallois (Chazan, 1997; Dibble andBar-Yosef, 1995; Sandgathe, 2005; Schlanger, 1996; Van Peer, 1992).Traditional deﬁnitions are rooted in the concepts of preparation orpredetermination of the ﬂake products, but more recent deﬁnitionssee only a reduction sequence that results in a core with speciﬁc‘volumetric’ properties (Boëda, 1995; Chazan, 1997; Brantinghamand Kuhn, 2001). Here, Levallois reduction results in a bifacialcore on which both faces are hierarchically related while displayingdistinctive ﬂaking patterns. One (lower) face sets up a series ofdihedral striking platforms (Bradley, 1977), while the other (upper)surface is for the removal of ﬂakes initiated from those dihedralplatforms to render that surface suitably convex. The ultimate‘Levallois ﬂake(s)’ run parallel to a plane of intersection that dividesthe two faces. The upper surface exhibits distal and lateralconvexities, which facilitate control over the size and shape of theLevallois ﬂake(s) removed just prior to re preparation or discard(Boëda, 1995; Chazan, 1997; Eren and Bradley, 2009; Van Peer,1992). Archaeological examples of Levallois may vary widely insize, but tend to conform to a restricted set of shape properties inline with this ‘volumetric’ concept of Levallois (Lycett et al., 2010).Having learnt the rudiments of Levallois reduction, Erenproduced a series of 100 Levallois cores from a pre-selected set ofblanks (ﬂakes) of high quality (i.e. with good mechanical ﬂakingproperties) and of uniformly unchallenging size and shape (Erenet al., 2011). A series of attributes designed to capture exhibitedskill level and regularity of product were recorded on thesereductions. Having mastered the skills after 20 months of practise,a second set of Levallois reductions was then recorded, usinga chert of lower quality ﬂaking properties and of more unwieldy,nodular morphology. Following the orthodoxy of the toolstoneconstraints argument, we expected that the switch from the higherto lower quality chert would stiﬂe any measurable improvement inEren’s skill levels and/or lead to greater irregularity of product.Accordingly, we tested these predictions via a series of statisticalanalyses.2. Materials and methods2.1. Knapping the Levallois coresEren was taught to prepare the sub-circular Levallois tortoisecore (Fig. 1a) and to detach from its upper surface the ﬁrst ‘preferential Levallois ﬂake’ (Bradley, 1977), hereafter called the ‘Levallois ﬂake’ (Fig. 1b). We chose this reduction because it iswidespread in the Middle Palaeolithic (MP) record, and requiresskill to execute. MP knappers are known to have sometimes reshaped the upper surface of the core after the ﬁrst Levallois ﬂakeremoval and struck off subsequent ﬂakes (Schlanger, 1996). ThisFig. 1. (a) Levallois tortoise core after the ﬁrst Preferential Levallois ﬂake removal; (b)ﬁrst Levallois ﬂake, arrow indicating axis of percussion; (c) Levallois ﬂake reﬁtted to itsparent core. The platform facets have been highlighted; the arrow points to the pointof percussion. This specimen was knapped by M. Eren.
M.I. Eren et al. / Journal of Archaeological Science 38 (2011) 2731e2739could be repeated 2e3 more times before the much-reduced corewas abandoned (Van Peer, 1992). We elected to halt the reductionafter the ﬁrst detachment to avoid having to grapple with theunknown effects of shrinking core size on skill level e a topic forfuture study.Having learnt the rudiments of Levallois tortoise core reductionin December 2007 from his tutor (Bruce Bradley), Eren thenrepeated the exercise 100 times, starting with blanks of the sameraw material and form (Eren et al., 2011). This was done in batchesover three months in 2008, with pauses of days or weeks betweenwork-sessions. All ﬂakes were removed using solely hard stonehammer percussion techniques. Each set of ﬂaking debris wasbagged and labelled along with the parent tortoise core and theLevallois ﬂake. After a further 20 months of almost daily knappingpractise, a second batch of 27 reductions was undertaken in 2010for comparison with the ﬁrst set.2.2. Raw materialsAt present, the relationship between the chemical, petrological,and microstructural properties of toolstones and their suitabilityfor knapping is only coarsely known. While it may be possible toquantify differences between various types of chert in terms oftexture, lustre, translucency, and colour (e.g. Luedtke, 1992:62e66), lithic analysts are currently able to do little more thanqualitatively assign raw materials to ‘high’ versus ‘low’ knappingpotential (but see Callahan, 1979 for a semi-quantitative effort).Additionally, the combination of several interacting variables (e.g.elasticity, brittleness, hardness, and homogeneity) may be moreimportant than the isolated expression any single variable(Whittaker, 1994), and unique combinations of such variables maydiffer across different rock types or even within a single type.Stone toolmakers rely (and relied) on macroscopic propertiesand strike-testing to identify differences between raw materials.We primarily employed this same simple strategy to distinguish‘highly workable’ from ‘less workable’ toolstone. This strategy isjustiﬁable for two reasons. First, visual ‘in hand’ tests of toolstoneproperties we have applied are precisely those that would havebeen accessible to hominins when making decisions regarding rawmaterial exploitation. Second, our study is designed to determine ifhypotheses of raw material constraints e based, as they frequentlyare, on the macroscopic determinations of raw material quality (e.g.Morgan and Renne, 2008; Noll and Petraglia, 2003) e are valid. Ourstudy is designed, in part, to determine if such assessments meritreconsideration, at least as incontrovertible rules that may beapplied in all archaeological situations.The external properties of the raw material (morphology andcortex) appear to present fewer difﬁculties. There is generalagreement among lithic specialists that regular nodules presentfewer difﬁculties for the knapper than irregular ones, and noncortical ﬂa