Society for American Archaeology

The Analysis of Cutmarks on Archaeofauna: A Review and Critique of Quantification Procedures, and a New Image-Analysis GIS Approach Author(s): Yoshiko Abe, Curtis W. Marean, Peter J. Nilssen, Zelalem Assefa, Elizabeth C. Stone Source: American Antiquity, Vol. 67, No. 4 (Oct., 2002), pp. 643-663 Published by: Society for American Archaeology Stable URL: http://www.jstor.org/stable/1593796 Accessed: 02/01/2010 16:32
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=sam. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Society for American Archaeology is collaborating with JSTOR to digitize, preserve and extend access to American Antiquity.

http://www.jstor.org

REPORTS

THE ANALYSIS OF CUTMARKS ON ARCHAEOFAUNA: A REVIEW AND CRITIQUE OF QUANTIFICATIONPROCEDURES, AND A NEW IMAGE-ANALYSIS GIS APPROACH
YoshikoAbe, CurtisW. Marean,Peter J. Nilssen, Zelalem Assefa, and ElizabethC. Stone

Zooarchaeologists utilize a divere set of approachesfor quantifyingcutmarkfrequencies. The least quantitativemethodfor cutmarkanalysis relies on composite diagrams of cutmarksoverlain on drawings of skeletal elements (diagramaticmethods). Todate, interpretationsof these data have generally relied on qualitativeand subjectiveassessments of cutmark frequency and placement. Many analysts count the number fragments that have a cutmark,regardless of the number of cutmarkson the fragments (fragment-countdata). Others count the numberof cutmarks(cutmark-countdata). Both can be expressed as simple counts (NISP data), or as a count of some more-derivedmeasure of skeletal element abundance (MNE data). All of these approachesprovide differenttypes of data and are not intercomparable.Several researchershave shown thatfragmentationof specimens impacts the frequency of cuts, and we show here thatfragmentation impacts all these currentapproaches in ways that compromisecomparative analysis whenfragmentation differs between assemblages. Weargue that cutmark frequencies from assemblages with differing levels offragmentation are most effectively made comparable by correcting thefrequency of cutmarksby the observed surface area. Wepresent a new method that allows this surface area correction by using the image analysis abilities of GIS. Thisapproachovercomesthefragmentationproblem.Weillustratethepower of this techniqueby comparing a highlyfragmented archaeological assemblage to an unfragmentedexperimentalcollection. Los zooarqueologosutilizan diversos metodospara cuantificarla frecuencia de huellas de corte. El metodo menos cuantitativo para el andlisis de huellas de corte utiliza diagramascompuestosde este tipo de huellas que se sobreponena dibujosde elementos esqueleticos (metodo diagramdtico).Hasta el dia de hoy, la interpretaci6n de estas observaciones se ha basado en evaluaciones cualitativas y subjetivas de la frecuencia y posicion de huellas de corte. Muchos investigadorescuentan el numerode fragmentos6seos con huellas de corte, sin considerarel nlimerode huellas en los mismosfragmentos(metodode conteo defragmentos). Otrosinvestigadorescuentansimplementeel numerode huellas de corte (metodode conteo de huellas de corte).Ambos se pueden expresarya sea como cuantificacidnsimple (datos de NISP), 6 como una medidaderivadade abundanciade elementos 6seos (datos de MNE). Todosestos metodosofrecendistintostipos de observaciones,los cuales nos son comparablesentresi. Variosinvestigadoreshan mostradoque lafragmentaci6n osea afecta lafrecuencia de huellas de corte. Nosotros mostramosen este articulo que lafragmentacion6sea influyenotablementeen todos los m6todosusados hasta el momento,y que, debido a ello, se arriesgan los andlisis comparativoscuando el grado defragmentaci6nosea es distintoentre las colecciones a comparar. Proponemos que cuandoel grado defragmentaci6n6sea varia entrelas colecciones, lafrecuencia de huellas de cortepodria ser comparada en forma mds efectiva al corregir la referidafrecuencia con la medida del area de la superficie observada. Nosotros presentamosun metodonuevo que permite estandarizarlafrecuencia de huellas de cortepor drea de superficiea traves del uso de andlisis de imagen con GIS. Este metodosupera el problemade lafragmentacidnosea. Aqui mostramossu potencial al comcon otra colecci6n experimental parar una colecci6n osea altamentefragmentada nofragmentada.

he analysisof cutmarks on skeletalelements is a standardresearch endeavor in zooarchaeology and has been used to addressa studiesof cutmarks have varietyof topics.Generally, T

been used to reconstruct butcherystrategies,which then are used to addressmore wide-ranging topics of greater interest. Weusetheterm"butchery" to refer to the actionstakento rendera carcassinto usable

Yoshiko Abe and Zelalem Assefa * Interdepartmental DoctoralProgramin AnthropologicalSciences, SUNY at Stony Brook, Stony Brook, NY 11794-4364 Curtis W. Marean * Instituteof HumanOrigins, Departmentof Anthropology,PO Box 872402, Arizona State University, Tempe,AZ 85287-2402 Peter J. Nilssen * Departmentof Archaeology,Iziko - SouthAfrican Museum, P.O. Box 61, Cape Town, 8000, SouthAfrica Elizabeth C. Stone * Departmentof Anthropology,SUNY at Stony Brook, Stony Brook, NY 11794-4364 AmericanAntiquity,67(4), 2002, pp. 643-663 Copyright? 2002 by the Society for AmericanArchaeology
643

644

AMERICAN ANTIQUITY

[Vol.67, No. 4, 2002]

but portions(Lyman1987), often for consumption, fortool manufacture theproduction of rawmaterials can also be a single or relatedgoal. Butcheryby cutting for consumptiontypically involvesskinning,disarticulation, defleshing,andin some cases removalof periosteum.Hammerstone percussionof skeletal elements is also technically but thatprocess generallyhas the goal of butchery, ormake skeletalpartsto accessmarrow fragmenting Hamthemmoreeasily boiledfor greaserendering. merstonepercussionleaves a markthat,to a trained et (Blumenschine analyst,is distinctfroma cutmark al. 1996). Our focus here is on cutmarksand their analysis. in human Studiesof cutmarks figureprominently havestudwherezooarchaeologists originsresearch ied patterningin cutmarksto investigatewhether Plio-Pleistocene hominids were hunters or scavengers(Binford1981,1985,1988; Bunn1981,1991; BunnandKroll 1986, 1988;Potts 1983, 1988;Potts and Shipman1981; Shipman1986, 1988; Shipman and Rose 1983). Over time this dichotomous to helpidengavewayto usingcutmarking approach of carcass in the where consumption sequence tify hominidsregularlyfit (Capaldo1995, 1998b; Selhave also vaggio 1994, 1998). Studiesof cutmarks been used in modemhumanoriginsresearchto test nonBinford'ssuggestionsthateven late-occurring of modemhominidswereprimarily scavengers large ungulates(Chase1986, 1988;GraysonandDelpech 1994; Marean1998; Mareanet al. 2000a; Marean andKim 1998;Milo 1994, andAssefa1999;Marean Stiner 1998; 1994). However,cutmarkstudieshave been conducted in manyotherresearchendeavorswherethe reconto of butchery struction processesis seen as relevant other behavioraltraits.The optimisticview is that the butchery processvarieswith the intendeduse of will be expressedin a carcass,andthatthis variation the placement and frequency of cutmarkson the skeleton,allowingus to infer carcassuse from cutmarkstudies.For example,it has been arguedthat butcheryshould vary between contexts where the goals are immediate consumptionversus storage (Binford1978). Binford'swell-knownutilitymodel for skeletalelementchoice, haddiffering predictions and researchershave anticipatedthat cutmarking should also vary widely between, for example, an versusmoreintenor"gourmet unselective strategy" sive utilization(Binford 1984). Zooarchaeologists

havehopedto identifyfilletingversusskinning(Binford 1981;Shipman1981;ShipmanandRose 1983; Wilson 1982). Yellen (1991) has arguedthat some butcherypatternshave a "style"that could be culthus holding out the possibility turallydetermined, thatbutchery patterns mayprovideways to examine ethnicity. This high promisehas been frustrated by several factors.First is a lack of detailed observationsof andits resultant Therearemany butchery patterning. moder bones (Binford 1981, studiesof butchered 1983;Gifford-Gonzalez 1989;Gifford 1984;Crader observed andCrader 1977),butnoneof theseactually the act of butcherythat producedmarks,and then A linkedthose specificactionsto specificcutmarks. recent study thatfilmed butcheryactionsclose-up, thisunambiguous thusproviding linkage,foundthat illusordefleshing marks manyof thedisarticulation tratedin Binford (1981), and regularlyused as a guide to butcheryanalysis, are not unambiguous of specificactivities(Nilssen2000). Thus indicators the strictcausal linkagebetweenobservedspecific or for behaviors(such as cuttingfor disarticulation defleshing) and their traces (such as cutmarkson ends versus shafts),called for by Giffordarticular Gonzalez (1991), is not yet fully developedin the literature. Another critical problem is the diversity of approachesfor recording the cutmarksand then Itis safeto saythatthere theirfrequency. quantifying is no acceptedmethodfor either.Ourreviewof the literatureshows that recordingcutmarkscan take two paths.One is to recorda count and description of the cutmarksonto a database.This can be done eitherat a grosslevel (how manyareon a specimen) or a finerlevel (whereon the specimenthey occur A second along with a diagnosisof theircharacter). onto a is cutmarks to draw diagramof a approach will a or we call what bone, template. The two be can easily combined,and probably approaches often are. the analystmust arerecorded Once the cutmarks choose a way to quantify,analyze,and presentthe andherethereis also a wide varidatain publication, This step is problematicdue to of ety approaches. in convariation the potentialfor wide interanalyst is a severeproblemfor comvention.This variation becauseit makes studiesin zooarchaeology parative databetween to compare if notimpossible, it difficult, have defined Justas zooarchaeologists researchers.

REPORTS

645

thatstanlandmarks andmeasurements anatomical dardize their approachto osteometrics (Driesch must also strivefor stan1976), zooarchaeologists in otherforms of data.Importantly, dardization not we needto knowthat only shouldtherebe standards, areeffective. these standards In this paperwe begin with a briefreviewof the main approaches used by analyststo quantifycutWe then review the impactof bone fragmarking. mentation on theeffectiveness of thesequantification We that these currentanalytical procedures. argue do not overcome approaches adequately problems Recentdevelcausedby differential fragmentation. orimage-analysis we software, opmentsin graphical the for of hold believe, promise solving many problems inherentin quantifyingcutmarks. Finally,we with GIS softwarethatwe presenta new approach thinkovercomesthe main problemsin quantifying We illustratethis method and analyzingcutmarks. with an applicationto two radicallydifferentcollections, a fragmentedMiddle Stone Age (MSA) faunalcollection and an unfragmented experimental collection.Ourintentis not to reviewthe issues of how to define a cutmark,because thathas been discussed in detail elsewhere(Blumenschineet al. 1996;Fisher1995;Shipman1981).Also, we do not examinetheissue of how to diagnosehow manycutmarksarepresenton a fragment, suchas whetherto recordmultiplestriaeas one or moreindividual cutmarks(Lyman1987). Rather, ourgoal is to address the issue of how to recordandthenanalyzethe numbers afterone has diagnoseda cutmark, decidedon a count,andis readyto recordandultimatelyquantify the samplefor meaningfulbehavioralanalysis, such as comparison to modem controlassemblages or otherarchaeological assemblages. Review of Methods for Quantifying Cutmark Frequency The literature on cutmarkanalysis is vast, and our goal here is not to review it all but ratherto distill from thatliterature the basic methodsfor quantifying cutmark frequencies. The least quantitative methodforcutmark analysisrelieson compositediaoverlainon drawingsof skeletal gramsof cutmarks elements (Binford 1988; Grayson and Delpech 1994:Figure10; Landon1996;Marshall1990). We call these"diagramatic methods," and,to date,interof thesedatahavegenerally reliedon qualpretations itative and subjective assessments of cutmark

frequencyandplacement.It is our impressionfrom discussionswith many zooarchaeologists that they often begin theirrecordingof cutmarks by drawing on bonediagrams. cutmarks However,in an attempt to be more quantitative and test for statisticalsignificance, zooarchaeologiststypically use various methodsforcountingandsummarizing cutmark frenever the data quencies, actually using diagrammatic for anythingother than presentationand impressionableanalysis. When analystspresentcutmarkdata,they typi(see cally choose to count one of two observations Table 1). Many analystscount the numberof fragmentsthathavea cutmark, of the number regardless of cutmarks on the fragments, andwe referto these as "fragment-count" data.Thesecanbe expressedas a simplecountof fragments thatarecutmarked, what = we call "NISPdata"(NISP number of identifiable specimens),oras a countof some morederivedmeasure of skeletal element abundance,what we call "MNEdata." The MNE (minimumnumberof elements)underliesmost otherderivedmeasuressuch as the MNI (minimumnumberof individuals)or MAU (minimum number of animalunits),so we use thistermgenerallyto referto all MNE-derived measures.Analystsoften choose to presentthe dataas a andtypicallytheseoccuras theNISPcutproportion, markeddividedby the totalNISP,or the MNE cutmarkeddividedby the totalMNE. can countthe frequency of Alternatively, analysts individual on specimenswithina skeletal cutmarks element,and/orwithina definedregion(suchas the proximalend or the middleshaft).We referto these as"cutmark-count" data. Cutmark-count dataareoften or expressedas both a raw value,or as a proportion index.As withfragment-count data,the analystmay in the employthe NISP or MNE as the denominator in the permutations calculation, proportion resulting listedin Table1. Thereis no standard in zooarchaeto use, so as is indicated in ology as to whatapproach our discussionbelow, much of the publishedliteraturepresents datathatis only directlycomparable to a narrow sampleof otherstudies. One of the simplest approachesto quantifying and otherforms of surfacemodification, cutmarks, has developed within the field of early hominid research. This approach was firstused by Blumenschine (1988) in his analysis of hammerstone percussion and carnivoretooth marks,and was later extendedto cutmarks (Capaldo1995, 1997, 1998a,

646

AMERICAN ANTIQUITY Table 1. Approachesto CutmarkQuantification. Expressed as NISP NISP cutmark-count data NISP fragment-countdata

[Vol.67, No. 4, 2002]

Counts of Cutmarks Counts of Cut Fragments

Expressed as MNE MNE cutmark-count data MNE fragment-countdata

et al. 2000a;Selvaggio 1994, 1998). 1998b;Marean The goals in these studiesareprimarily two: identiof hominids within the carcass the fying position access versus late consumption sequence (early the and and amount access), measuring intensityof actionon thefaunalassemblagefollowing carnivore hominid discard.These approachesfocus only on long bone fragmentsthat are classified into three types: epiphyseal (technicallyarticular),near-epiphyseal,andshaft(see Blumenschine1988fora definition).The fragments may or may not be grouped skeletal element for areonly by analysis.Fragments not to classes (Brain1981), taxon, assigned body-size and only ungulatelong bones are used. Fragments as having a markof a certaintype or are tabulated not (numbersof marksper specimen are not utilized), so this approachproducesNISP fragmentcountdata. An example of NISP cutmark-count data is Stiner's (1994) presentationof Paleolithic faunas fromItaly.Stiner(1994) providesskeletaldiagrams of anatomically completetaxa with the frequencies of cutmarks indicatedon the skeleton(by numbers propointingto a skeletalelement).Such diagrams of the intensityof cutvide a usefulvisual summary ting at variousanatomicallocations. Stiner(1994) alsoprovidestablesof cutmarked fragments pertotal NISP assignedto a specific taxon, so her presentation includes both cutmark-countand fragmentcountdata,all in NISP. Giffordet al. (1980) andBunnand Kroll(1986) simthataresuperficially provideskeletaldiagrams ilar to those in Stiner(1994), but the presentedvalfrom a ues are frequenciesof cutmarkedfragments particularskeletal element. These are NISP fragfromStiner's ment-count data,andthusverydifferent in presentaskeletaldiagrams, despitethe similarity tion.Giffordet al. (1980) also presentstablesof cutbuttheyaresegregated marked by skeletal fragments, element and body-size category,not species, and thus are not directlycomparableto Stiner'stables data are common in (1994). NISP fragment-count since the 1960s (BunnandKroll 1986; the literature Frison 1970; Gilbert 1969; Guilday et al. 1962;

Marean1992;Parmalee1965;Wheat 1979). While thedataarecarefully described andpresented in each of thesestudies,thevariation in analytical procedures makescomparative analysissomewhatdifficult. It has become increasingly commonfor analysts to presentandanalyzecutmark datacorrected by the MNE, or a measurederivedfrom it (Binford1984; Graysonand Delpech 1994; Milo 1998). Binford (1984) was one of the firstanalyststo presentcutandhis tableson theKlasies mark datain thismanner, River fauna provide an MNE on cutmarked fragmentsby bovid size class and skeletalelement,and then a total MNE. Milo's analysis of the Klasies River fauna provides similar data (1998:Table 2:104): the MNI calculatedfrom all fragmentsper skeletalelementandportion, andtheMNIcalculated Milo fromjust those fragments thatare cutmarked. of an index cutmarks also (1998) employs per anatomicalzone (in this case joints), and these are data.By proclearlyindexedMNE cutmark-count the he broadens vidingboth, potentialusefulnessof his presenteddata. His reasonsfor calculatingthe numbers indexareclear(Milo 1998:109):"Absolute becauseof disparities of markscannotbe compared in skeletalpartrepresentation." Note thatin all cases is a derivedmeasureof skeletalelehis denominator meaandhe uses it becausederived mentabundance, sures "partly circumvent the problem posed by differentialfragmentation" (Milo 1998:102). Milo does not elaborateon this statement,but Bartram (1993) provided a thoughtfuldiscussion, and we buildon thatbelow. The Fragmentation Problem to The use of derivedmeasuresis often undertaken overcomeor at least minimizethe primaryanalytical problemfacing the analysis of cutmarks:cutfrom aresensitiveto fragmentation markfrequencies both human and nonhuman processes (Bartram 1993). Bartramdiscussed the problems with cutthatbelow,but mark-count data,andwe summarize not for fragment-count data, so we extend his disas well. cussion to those approaches of the fragmentaFigure1 beginsthe illustration

REPORTS

647

A)

B)

Percussion n ,t \ i Hammerstone
I

/ ;;'5

.S

Expressed as: MNE NISP Raw Proportion Raw Proportion CutmarkCount FragmentCount 3 1 3.0 1.0 3 1 3.0 1.0

Expressed as: MNE NISP Raw Proportion Raw Proportion 3 2 .375 .25 3 1 3.0 1.0

Figure 1. a) Three cutmarks on a complete bovid femur, and b) the same femur fragmented by hammerstone percussion, both showing the resulting counts and proportions using the various approaches discussed in the text. Small arrows on a) indicate the position of cut marks.

tionproblem.Figurela showsa femurthathas been of its flesh, andtherearethreecuts on it. butchered At this pointin the taphonomic historyof this specimen, the frequency of cutmarksand cutmarked fragmentsis 100 percent with either an NISP or MNE approach. Thefrequencies wouldvaryby portion (proximalend, proximalshaft,and so on) and would be accurately reflectedby both an NISP and MNE approach. Zooarchaeologists often create experimental assemblages of skeletal elements undercontrolledcircumstances butchered for comto collections (Milo 1998), parison archaeological and this example illustratesthe conditionof those assemblages. FigureIb showsthe samefemurbrokenby hammerstonepercussionfor marrowaccess. This simple, and almost universal,stage in the taphonomic andprohistoryof a long bone lowersthefrequency portionof cutmarksand cutmarked fragmentsdramaticallyif the NISP is used. It is safe to say that evenminimalpostbutchery will lower fragmentation the frequencyof cuts with any NISP approach and makesNISP approaches nearlyuseless for any type of comparative colanalysisbetweenarchaeological lectionsandunfragmented collections. experimental accounts showus thatpeoImportantly, ethnographic

ple fragmentbones at differentlevels of intensity dependingon their ultimate goal (Bartram1993; Binford 1978, 1981;Yellen 1991). For example,if bone boiling for grease renderingis a goal, then bones areoften moreintenselyfragmented to expeditetheboilingprocess(Binford1978, 1981).Bones of largeranimalsmay be more heavily fragmented to fitintopotsforboiling(Marshall 1990).Ithas also been shownthatbonesfromdifferent-sized animals are differentially hammerstone fragmentedduring percussion (Bartram1993; Bartramand Marean 1999): generally, the larger the animal the more intensely the bones are fragmented for nutrient ThismeansthatNISP-based extraction. methods will falsely illustratecutmarkfrequenciesacross body sizes. MNEapproaches workbetterin thisexampleand overcomethefragmentation as indicated problem by the equality between the proportioncalculations before and afterhammerstone percussion(see also Bartram the cutmark 1993). If one were calculating eitheras fragment-count orcutmark-count frequency, data, or by whole bone or portions, the MNE assessment of thecutapproach providesanaccurate markfrequenciesin this situation. We can takethis andin twodirections thatmore exampleevenfurther,

648

AMERICAN ANTIQUITY

[Vol. 67, No. 4, 2002]

A)

Expressed as: MNE NISP Raw Proportion Raw Proportion Count 5 Cutmark FragmentCount 3
B)

1.66 1.0

5 3

1.66 1.0

_sl-

.I

,3

^'

!;^U:;>

Expressed as: MNE NISP Raw Proportion Raw Proportion Cutmark Count 2 FragmentCount 2 .666 .666 2 2 .666 .666

Figure 2. a) Cutmarks on 3 proximal bovid femora broken by hammerstone percussion prior to attrition by sedimentary processes, and b) the same three proximal bovid femora after attrition by sedimentary processes. The surviving fragments are shown in dark outline while the original bone is shown in gray outline as a ghost image. Small arrows on a) indicate the position of cut marks.

realisticallyportraythe realitiesof the taphonomic bone assemblages. historyof archaeological Figure 2a shows threeproximalfemorabroken two withtwo cutmarks percussion, by hammerstone This represents the preand one with one cutmark. have state,beforesedimentary processes depositional

Figure2b shows an examimpactedthe fragments. ple of how the femorain 2a wouldtypicallysurvive fragmentation and bone loss from sedimentary processes. It is widely believed that fragmentation by sedimentary processes is density mediated (Grayson 1989; Klein 1989; Lyman 1984, 1985,

REPORTS

649

in archae1992)andthefemoralheadsurvivesbetter probably ological sites than the greatertrochanter, denserthan becausethefemoralheadis significantly thegreater trochanter (Lamet al. 1999).Inthisexamandthe MNE approach ple, boththe NISP approach databetweenthe predefail to providecomparable positional and postdepositionalstates with either or cutmark-count data.In all cases, fragment-count of cutmarks or lowers the fragmentation proportion cutmarked fragments. to This resultis soberingfor anyoneattempting of cutmark conductcomparative analyses frequency. Wheninterpreting cutmark fromarchaefrequencies to have some type of ological sites it is imperative This is controlassemblageto aid in interpretation. often eitheran unfragmented experimental sample, or a humanlyfragmented samethnoarchaeological not that was fragmented by sedimentary ple processes. Neitherof these two types of comparative sampleswill providecomparable cutmarkfrequenciesto archaeological samples,becauseof the Thisproblem canbe extrapfragmentation problem. olated to attemptsto comparetwo archaeological sites of differing levels of fragmentation, and betweenanimalsof differingbody size: the cutmark values simply will not be comparable. Thebasicproblem canbe summarized As simply. of bone such as destruction, processes postdiscard carnivorescavenging and diagenesis, increasingly a cutmarked bone,fragmentation fragment generally decreasesthe numberof cutmarked fragmentsand cutmark countsrelativeto totalfragments, andthus reducestherelativefrequencies in NISPapproaches, ineffectivefor comparmakingany NISP approach ative studiesof cutmark The fragmentafrequency. tion process also moves more fragmentsinto the unidentifiable categoryanddestroysless-densebone Both of theseprocesseseffectivelyreduce altogether. the amountof bone surfaceareastudiedby the anato surviveas a funclyst. Sinceboneportionsappear tion of density,MNE countswill resisttheimpactof andremainhigh since the denseporfragmentation tion is preserved,and countedby the analyst,again andagain,while the less denseareasof thebone are lost through attrition. Unfortunately for MNE approaches,the larger the sample the greaterthe in Figure3. biasingeffect. This is illustrated for Imaginefive flat squaresurfacessubstituting the surfaceof five bone fragments(Figure3a). We dividethatsurfaceintofoursampleareas, arbitrarily

A)

B)

Destruction DensityMediated

Figure 3. a) Five flat surfaces representing the bone surface area of five bones of the same skeletal element, divided into four zones, each with one cutmark, but with zone 1 having higher density than zones 2-4, and b) the survival of these zones following density-mediated destruction.

each a squarewith one cutmark in it. Cutmarks are distributed across that surface such systematically that each sample area has one cutmark.Like our bones, this flat surfacehas varyingdensity,and in this example sample area one (the darkestarea)is the densestarea(similarto the femoralhead in the bone surfacewas exampleabove).If this imaginary to subjected taphonomic processesof bone destruction(Figure3b), sampleareaone thenpreserves better than all others.OurMNE for the five flat bone surfaceswill alwaysbe high relativeto the amount of bone surfaceareapreserved becausesamplearea one continuesto preserveand overlapwith sample area one in other surfaces.The MNE may in fact remainfive while mostof the othersampleareasare degraded.Statedanotherway, as density-mediated destructionattacksthe bone, the MNE counts on sample area one will decreaseat a lesser rate than theothersampleareas.If we wereto countcutmarks and divide by MNE, our resultingestimatewould the frequencyof cutmarks relvastlyunderestimate

650

AMERICAN ANTIQUITY

[Vol. 67, No. 4, 2002]

ativeto the originalsurface.This is why MNE data arenotimmuneto thefragmentation problem.However, if we dividedthe numberof cutmarks by the preserved surface area, our cutmarkfrequencies would closely matchthe originalfrequencies.This situationshouldhold if cutmarking does not preferentiallyoccur,ornotoccur,in denseregionsof bone. We haveno reasonto believethatcutmarking intenvarieswithbonedensity, buteven sity systematically if it does, the GIS methodwe describebelow has a meansto overcomethis problem. of a cutmark Theresultis thatthelikelihood being andcountedby an analystis a functionof preserved of bonesurface andrecorded. theamount areastudied As more surfaceareais studiedandrecorded by an cutmarks will be and vice versa. more found, analyst, Thusif we cancorrect thenumber of cutmarks by the amountof examinedsurfacearea,much as demographersstandardize populationsize by estimating then we can standardize cutmark population density, numbersbetween sites, between body sizes, even betweenanalyststhatmay differin theirchoice and abilities to identify and record fragmentedbone. as samplesize (andthussurfacearea)is Importantly, the increased, frequencies by this method generated areless susceptible to chanceoccurrences thatcould skew the results.The key is, of course,thatwe have some way of recording the amountof examinedsurface areaandhow manycutmarks (orotherformsof surfacemodification) were found. Rapson (1990) recognized the fragmentation problem and recommended correcting cutmark countsby surfaceareawiththefollowingprocedure: calculatethe frequency of cutmarks perunitareafor thetotalcutmarks each specimenby multiplying per specimenby 1,000 and then dividingby specimen area,resultingin a figurethatestimatedthe number of cutmarks per 1,000 mm2of bone surfaceareafor each specimen,thencalculatethe meansurfacearea and the mean numberof cutmarks per unit areaby taxonomicgroup.Rapson'smethodfor estimating surfaceareaper specimeninvolvedmultiplyingthe maximum lengthof a faunalspecimenby anapproximatemeasureof specimenwidth.Specimenwidth is estimatedby measuringthe distancebetweenthe surface andits oppositeaspect maximum weathering (the "weatheringprofile height"). Rapson argued freof cutmark thatthisapproach allowscomparison samples, fragmented quenciesbetweendifferentially in his case bighornsheep andbison.

cutmark countsby Rapson'ssuggestionto correct surfacearea,andhis use of a gross estimateof surface area,was anexcellentfirststep.Recentadvances in image-analysissoftware, and particularly GIS, allow us to go far beyondthat approach. GIS software makespossible the following procedures that thebestof themethodsdiscussedabove,and capture excel in severalcriticalways: 1) captureandcalculate the preservedsurface area of fragmentsin a muchmorerealisticmanner andthussolve the fragmentationdilemma, 2) precisely quantifythe freon a skeletalelementin anyway quencyof cutmarks desiredso thatpositional canbe askedflexquestions to a graphically database ibly,3) attacha descriptive recorded cutmark so thatcutmarks canbe further anaor other lyzed by type, length, angulation, any recordedvariable,4) allow diagrammatic represenof tationof cutmark placementin any permutation variables desired(taxon,provenience, cutmark type, and so on), and 5) if it is eventuallyfoundthatcutmarkingintensity systematicallyvaries with bone density,thismethodwill allowthe analystto restrict analysis to surface-area samples that are densityequivalent. We have developeda series of methodsthatutilize ArcView GIS software, the ArcView Spatial Analyst extension, and several modificationsand extensions written in Avenue (the ArcView programminglanguage) and MicrosoftVisual Basic. Thesematerials includeanArcViewprojectwiththe addedAvenuefeatures,an exampleprojectwith the andaVisualBasicprogram digitalimages,a manual, overcome several file management developed to problemsassociatedwith linkingArcViewto exterTherearealso instructions on how to nal databases. develop interfacesfor your own animalsof preferand splice them ence (hare,fish, bird,or whatever), code. All are availablegratis into ourprogramming from Mareanby request(curtis.marean@asu.edu), but you must alreadyhave a copy of ArcView,for which manyuniversities have site licenses. The Image-Analysis GIS Method our methodof cutmark To understand analysis,we mustfirstexplainhow we calculatethe MNE using the image-analysis GIS approach(Mareanet al. the 2001). Zooarchaeological systemsforestimating MNE, and other measuresderivedfrom it, can be classified into two types: fraction summation and overlapapproaches(Mareanet al. approaches

REPORTS

651

Overlap

Figure 4: a) Example of a femur fragment drawn onto a template, b) a second femur fragment where the MNE still equals 1, and c) a third femur fragment raises the MNE to 2. The darker shading indicates the area of overlap between the two specimens.

2001). Overlapapproaches(see Figure4) seek an actualcountof thenumber of overlapping fragments and thus a directestimateof the numberof skeletal elementsrepresented Howby a seriesof fragments. arecumever,these methods,when done manually, bersome,time consuming,and could be inaccurate when workingwith largecollections. We describeda new image-analysis GIS method et al. that utilizes GIS softwareto (Marean 2001) makethe overlapapproach more approachable and accurate. The simplestandmost completerecordof a fragmentwould be to drawits positionon a template.WiththeGISmethodwe usethemouseto draw the outlineof a fragmentonto an outlineof a combothof whicharevecplete bone,calleda template, tor images (Figure5). Our procedureis to do this butif neededan analystcould directlyto computer, forms of the templatesandenter generatehard-copy themlater.ArcViewtreatseach fragmentas a separate theme, and each theme is a relatedset of files thatincludea shapefile (thevectorimage)linkedto a tablefile (whereotherinformation, such as specimen number,can be stored).ArcView names the

files for you in a ratheruninformative way, so we havedevelopedaVisualBasicprogram thatrenames all files by the specimennumber, allowingeasy file The fragment vectoroutlinesarelater management. into bitmapsfor the variouscalculations translated describedbelow. Sittingon top of the templateduring theentryprocessis a high-resolution digitalphotographof the bone thathelps the analystprecisely andsurfacemodification on to positionthefragment thetemplate. shows an 5 of the Figure example fragmententryinterfacewe have designedfor femur. are (andothermodifications) Likewise,cutmarks entered mouse onto the and directlyby template digitalphotograph. Wedesignedthe systemso thatcutmarksare enteredtogetheronto a single theme for cutmarks. Eachcutmark is namedon the underlying databasetable by specimennumberand described cutmark by a varietyof variables(angulation, type, andso on).An analyst canaddto orlessentheamount of detailrecorded. Using varioussoftwareroutines,the positionof fragmentsin relationto each othercan be assessed andthe overlapscan be estimatedto the accurately,

652

AMERICAN ANTIQUITY

[Vol.67, No. 4, 2002]

*--

---

'r ri ---<
i V

!.. ... .... .

pl.

Tl,,

t-

-----

h64m, r eg,

'

il,

"i

T
'.

i:Rs
."....... .

... '

. ._. .

tons that

activate

Avenue

code we accomp....ish

have

written

to

variety

of

tasks.

For

example,

the

lightening

button

gener-

standard view of the screen, with a single fragment drawn in the 3 views it covers. Across the top tool bar are a series of but-

place the fragment. To the left of the image, with the checked boxes, are the lists of items in the view. FelfOOlis the current fragment polygon (vector image), and this has a data table underlying it where one can enter other information such as specimen number. Each separate fragment adds another polygon (FelfOO2,FelfP03, and so on). Felphoto.img is the digital photograph of the femur. All can be turned on and off to aid viewing and entry.

REPORTS

653

MNE 1 2 3' 4 6

7
8

Anterior

Lateral

Posterior Medial

MNEon a bovidfemur, fromMarean et al. 2001. for estimating Figure6. Resultsof the GISprocedure

pixel level. Each pixel has an MNE value, which is the numberof fragmentsthat overlappedthe position of thatpixel. The MNE map generatedby this image-analysisapproach(Figure6) is thus a compositedigitalrecordof the physicalposition,shape, and size of all identifiablefragments.It is also a

MNE map of the bone surface.Drawingeach fragmentprovidesthe mostcompleterecordof the identified archaeological fragmentsas is possible, aside theactualfragments fromextensivelyphotographing themselves,andis infinitelyeasierto analyze. The digitalimages, once entered,can be manip-

a)

b)

c)

2
Surface area for whole femur: 4,000 pixels Surface area represented by fragments: Surface area for whole bone:

MNE= 1 750 pixels

1 MNE = 0 | MNE=1

MNE = 2

Surface area represented by fragments: 1,250 pixels (1 x 750) + (2 x 250) = 1,250 pixels 4,000 pixels

% Surfacearea representedby fragments:1,250/ 4,000 x 100 = 31.3% of a femurtemplate. Thepixelcountfor thistemplate of percent surfaceareacalculation. a) Example Figure7. An example
is 4,000 pixels. b) Fragments are drawn onto the template. c) Example with fragments drawn onto the template. Darker shading indicates the area of overlap between two specimens. The pixel count of the area with fragments is calculated as 1,250 pixels.

654

AMERICAN ANTIQUITY

[Vol. 67, No. 4, 2002]

% Surface area represeritedIby fragments: 31.3% 10 cuts Total number of cutmar:ks: cuts / % surface area:

10 / 31.3% x 100 - 31.9

MNE = 0 MNE = 1 MNE = 2

Figure 8. Example of cuts / % surface area calculation (= CNC) using the femur example from Figure 7. The resulting value is the number of cuts that would occur on a single complete bone.

ulatedand measuredfor manypurposes.To obtain theMNEvaluefora skeletal element,onewouldlook for the pixel with the highest value, or numberof overlaps.One can also count the numberof pixels within a fragmentdrawing,which is a two-dimensional measureof the surfaceareaof the fragment The surfaceareaof multiple relativeto the template. fragmentsin a skeletalelement(such as those seen in theMNEmapin Figure6) canbe obtained by multiplying the surfacearea with the numberof overlaps. By dividingthe surfaceareaof the fragments by the total surface area of each skeletal element template,one can calculatethe percentsurfacearea preservedfor each skeletal element. A simplified exampleis given in Figure7, wherethe surfacearea is shownas a percentvaluerelative of the fragments to the whole skeletalelement.This correctedvalue can be used as a denominatorof other measures, such as the count of surfacemodification marks,to acrossdifferent skeletalelements makecomparisons anddifferentcollectionspossible. marksassociated with If the surface-modification identifiablefragmentsare recordedusing the same between skeletal-element template,the information Inother thesetwo setsof datacanbe linkedspatially. the of it is to words, possible identify position a surmarkon a skeletalelement,as well face-modification of the areawherethe suras the relativeabundance markis found.Anothersimplified face-modification exampleis given in Figure8, wherethe percentsurthenumin Figure7) corrects face area(ascalculated

ber of surface-modification marks,in this case cutmarks.The corrected numberof cutmarks (i.e., cutsurfacearea)is the estimated number marks/percent of cutmarks thatwouldbe foundon one wholebone, in this case the femur,as extrapolated fromthe preserved fragments.This procedureallows comparisons of differentiallyfragmentedassemblagesby correctingdirectlyfor preservedand recordedsurface area.Forconvenience,we will referto the corwith the acronymCNC. rectednumberof cutmarks to note thatthereis a slight overIt is important lap between differentviews of the same bone and, of course, this is a 2-D renderingof a 3-D object. The overlapoccursfromourdesireto maintain contraditional forms of orientation with (antesistency riorview, posterior view, etc.) as well as the need to illustrate clearly fragmentsand marks.This means that CNC comparisons,for examplebetween sites and analysts,must be done only between template entrysystemsthatarerelativelysimilarto eachother in theuse of views. Inotherwords,it wouldnotwork to compareCNC between an analystwho records andestimatesfemurCNCin five views fortheentire if comfemurandone who does so in four.However, were done within similar or regions zones parisons there would be no problem. as (such proximalend), Application of the Image-Analysis GIS Method We will illustratethe use of CNC with an archaeosample.We use logical sampleandan experimental these two samples to illustrate how this method

REPORTS

655

a
1 MNE = 0...... MNE 1...... 800 pixels 0 pixels 0 pixels

L
-......... J1 41 450 pixels 300 pixels 50 pixels ' A1 800 pixels 0 pixels 0 pixels Sum of pixel counts 2,250 pixels 800 pixels 150 pixels

200 pixels 500 pixels 100 pixels

MNE= 2 ......

3,200 pixels

Surface arearepresented by fragments: (1 x 800)+ (2 x 150)= 1,100pixels Surface areaforentire middle shaftzone: 3,200pixels
% Surface area represented by fragments: 1,100 / 3,200 = 34.4%
Figure 9. Example of percent surface area calculation using middle shaft zone of femur.

allows directcomparison of two samplesthatdiffer the archaeological widely in fragmentation: sample is highly fragmented and dominatedby shaft fragments,while the experimental samplederivesfrom whole bones. The archaeologicalsample is taken froma series of MiddleStoneAge (MSA) layersat Die KeldersCave 1 (DK1) in SouthAfrica (Avery et al. 1997; Mareanet al. 2000b). The DK1 MSA large mammal(body size 3 and 4) bovid remains were selected for this analysisbecause analysesof toothmarks andpercussionmarksdemonstrate that were the accumulators of these people primary remains et al. 2000b).Mostof theseremains (Marean derivefromeland (Taurotragus oryx)andare taken fromlayers10-13. Theexperimental sampleis from Nilssen's experimentalbutcheryof South African bovids (Nilssen 2000). The fragmentsand cutmarksfrom both studies were recorded usingArcViewGIS softwareversion 3.1 withtheArcViewSpatialAnalystextensionversion 3.0 on aWindows95/98 platform. Thefragment outlines were recordedas polygon vector images and converted to grids (pixel images) using the methodoutlinedin Mareanet al. (2001). The shape andpositionof each identifiedfragmentwas drawn

on a skeletal-element template,which was layered overa high-resolution of thebone, digitalphotograph to ensureaccurate The positioningof the fragment. cutmarks were recordedas line vectorimages with attacheddatatables thatdescribedthe character of each mark and the specimen numberof the fragment. The cutmarkswere enteredusing the same as thefragment, in orderto recordtheexact template spatialpositioning. CNC can be calculated for any zone within a skeletalelementthatthe analystwishes, and this is a realandnovel strength of thisapproach. Forexamone could calculate CNC for the whole femur ple, in an defined zone (as Figure8), arbitrarily (such as middle shaft, Figure 9), or zones defined by the specificsof anatomy(such as femoralhead).In our study, the long bones of both archaeologicaland into five arbicomparative sampleswere separated zones: trary, roughlyequal proximalend, proximal shaft, middle shaft, distal shaft, and distal end andSpencer1991).Thesezone boundaries (Marean do not follow epiphysealfusion lines or any other anatomical feature. Forbothsamples,thecutmarks wererecorded on that showed four views of each eleskeletal templates

656

AMERICAN ANTIQUITY

[Vol. 67, No. 4, 2002]

MNE 4

Totalfemur(fourviews).................... 16,000pixels Totalmiddleshaftportion (fourviews).. 3,200 pixels

Surface arearepresented by fourfemurs: 16,000x 4 = 64,000pixels Surface areaforwholebone: 16,000 pixels

% Surface area represented: 64,000 /16,000 = 400.0%

Surface arearepresented shaftzones: 3,200x 4 = 12,800pixels by fourmiddle Surface areaforonewholemiddle shaftzone: 3,200pixels
% Surface area represented: 12,800 / 3,200 = 400.0%

surface-area calculation for wholebones(fourwholefemora). of percent Figure10.Example

as wererecorded ment,andwithDK1 the fragments counts well (as in Mareanet al. 2001). The cutmark foreachzone were andthepixel countsof fragments summedacross the four views. Figure9 illustrates the summationof pixel countsfor the middle shaft zone. In actualpractice,the pixel countsareusually much largernumbersand quite unwieldy.The calculationmust also be done for all five zones. Using the ArcView summationfeature,one can produce, in one step,a summation tableof pixelcountsby zone and by MNE value for the whole skeletalelement, for further to spreadsheets which can be transferred analysis.The same calculationneeds to be done for butsincethecutmarks theexperimental assemblage, occuron whole bones we can calculatepercentsurface area with a much less-computer intensive the calculation of the approach. Figure10 illustrates percent surface area for a sample of four whole femurs.As this value in effect calculatesthe percentage of one whole bone or portionpresent,the percentageis, as expected,400 percent. foundwithina zone was Thenumber of cutmarks

calculatedusing the SpatialAnalyst extension of ArcView,for both assemblages.All cutmarksthat intersected a zone were counted.We used the intersect principlefor this analysis,butotherapproaches if a cutmark fallswithin areavailable. Usingintersect, is countedin both two adjacentzones, the cutmark if we are zones. This approach is most appropriate the of within to evaluate frequency cutting attempting in an these five zones zones. We used particular between attemptto capturediffering frequencies cutmarks would articular areas(wheredisarticulation areas(whereboth near-articulation be concentrated), disarticulation and defleshing marks would be found), and middle shaft areas (where defleshing cutmarks wouldbe concentrated). Nilssen conducted his experimentalbutchery study using both small and large African bovids activandfilmedthebutchery (Table 2). He observed ities by subsistencebutchersof the Karoo (South leftby modifications Africa),andstudiedthe surface these activities.Since all activitiesandactionswere in most directlyobservedandfilmed,each cutmark

REPORTS Table 2. Species and Sample Sizes of Nilssen's Experimental Assemblage. Species Used in Experiment Small Bovids (sizes 1 and 2): Steenbok(Raphiceruscampestris) Springbok(Antidorcasmarsupialis) Sample Size Humerus Radius Femur Tibia Humerus Radius Femur Tibia n Bones 12 12 16 14 18 18 18 18 n Cutmarks

657

688 Large Bovids (sizes 3 and 4): Blesbok (Damaliscus dorcas phillipsi) 619 Black Wildebeest(Connochaetesgnou) 225 Eland (Taurotragusoryx) 176 Note: The size 1 and 2 sample is not includedin this analysis, so we have not calculatedthe numberof cutmarks.

cases could be directlylinked to its exact activity. Nilssen controlledthe sequenceof butchery to simulatedifferingbutchery activitiessuch thatthe sample has some animals filleted without any andalso filletedafterdisarticulation. disarticulation, Whenall thesamplesanddiagnosesofcutmark function were combined,we were able to develop two differentsets of cutmarkfrequenciesrepresenting two differentbutchering activities.These represent the two differingbutchery models of quantitative strategies: 1.TheFilleting-Onlydataset removal represents of flesh without disarticulation and implies early access to a carcass when significantquantitiesof flesh are available.The cutmarksin this category thusrelateto the removalof meatin long stripssuitable for dryingand storage. 2. TheDisarticulation-and-Filleting dataset representsdisarticulation followedby meatremovaland implies early access to a carcass when significant disarquantitiesof flesh are available.Technically, ticulation couldoccurafterfilleting,buttheseexperiments were conducted such that disarticulation sigprecededfilleting.It is possiblethatthe cutmark natures coulddifferdepending on the sequence.The in thiscategorythusincludecutmarks cutmarks that could be fromdisarticulation and meatremoval. For both the experimentaland archaeological sampleswe includedbovid size 3 and4 femur,tibia, As separate andradius. humerus, templatesexist for the left and right elements,we have addedthe left andrightcutmark countsandpixel countsbeforethe CNC calculation. Forthe experimental sample,cutmarks were counted separatelyfor each of these zones. activity categories,across the five arbitrary Thesenumbers wereenteredintoa spreadsheet programanddividedby the numberof whole bones of each category(e.g., fourbones = 400 percent= 4) to

obtainthe CNC. The CNC correctionallows us to theDK sampleto theexperimental directly compare samples. Cutmark faunaaretypanalysesof archaeological interested in the of ically intensity cuttingin particularzones on skeletalelements.Forexample,in the debateover scavengingversushunting,it has regularly been arguedthat intensive cutting along the shaftof long bones suggestsremovalof flesh. More importantly, regularremovalof flesh is believed to indicateearly access to meatycarcasses,and therefore is consistentwith hunting(e.g., Binford 1981, 1988;Bunn and Kroll 1986, 1988). Similarly,with our experimental datawe would expect a Filletingto have highercutmark Only approach frequencies in the shaftareasversusthe ends relativeto a DisarTheDisarticulationticulation-and-Filleting strategy. and-Filletingstrategyshould have higherfrequencies of cutmarkson the ends near the joint areas, thegreater reflecting intensityof cuttingatthejoints. A key assumptionthat all zooarchaeologists make in this type of analysisis thatmoreintensivecutting (morecuttingactions)resultsin higherfrequencies of cutmarks on the bone surface,andwe follow that here. However,we know of no experiassumption mentaldocumentation of this assumption.Though Nilssen's experimental data could be analyzedfor such a test, it has not yet been done. The CNCfrequencies fromthe experimental Filand leting-Only Disarticulation-and-Filleting datasets show patternsthat on visual inspection distinctfromeach other(Figures11 and 12). appear In the figureswe have scaled the CNC to 100 percent for eachbone to facilitatevisualcomparison of therelative of Cutmarks on the ends intensity cutting. arerelativelymoreabundant in the Disarticulationwhile cutmarks in the shaft dataset, and-Filleting zones arerelativelymore abundant in the Filleting-

658

AMERICAN ANTIQUITY

[Vol. 67, No. 4, 2002]

Table 3. Mann-WhitneyU Test and Spearman'sRank CorrelationStatisticsBetween the ExperimentalDatasets and DK1. ExperimentalDatasets Filleting Both Limbs Disarticulationand Filleting Both Limbs Filleting Forelimb,Disarticulation and Filleting Hindlimb Disarticulationand Filleting Forelimb, Filleting Hindlimb U 342 243 268 315 p < .0001 < .0001 < .0001 < .0001 Rs .34 .21 -.08 .47 p > .05 > .05 > .05 < .05

We closely followingourexpectations. Onlydataset, would expect that the intensity of cutting within zones would co-vary weakly between the experimentaldatasetssince 40 percentof the zones (the articularends) are being treateddifferentlyin the butchery procedure, while 60 percent should be treatedsimilarly.In otherwords,there shouldbe a weak tendencyfor cutmark frequenciesto increase anddecreaseconsistentlybetweenthe two datasets. The Spearman's Rank correlation coefficient between pooled forelimband hindlimbdatasetsis consistentwith this expectation(rs= .46, p < .05). Even thoughwe would expect some discontinuities in the way cutmarkfrequenciesincreaseand decrease between the experimental datasets, we would expect the overallcutmark frequenciesto be similarbetweenthetwo datasets. Thisis becausethe butcheris using the same tools (metalknives) and the carcassesunderthe same conditions butchering of biltong)in (for consumptionand the production both experimental datasets.This null hypothesisis between difference supported by a lackof significant themediansof thetwopooledforelimbandhindlimb datasets(Mann-Whitney U = 119,p = .367). of cutmarking associatedwith Giventhepatterns the two differentactivities,we can now examinethe relationbetween the DK1 patternand the experidataset mentaldatasetsandask which experimental is most similarto the DK1 dataset.To examinethis question, we producedfour experimentalmodels from the two experimentaldatasets.Two of these experimentalmodels precisely reflect the experia Filletingstratin thattheyrepresent mentaldatasets to both limbs, and a egy applied consistently to both Disarticulation-and-Filleting strategy applied areposlimbs.However,atleasttwo otherstrategies sible,produced by applyingFilletingto the forelimb to the combinedwith Disarticulation-and-Filleting hindlimb,and vice versa.

the overallCNC frequenciesin the Interestingly, DK1 datasetare significantlyhigher than all four models (Table3). Nilssen's butchers experimental used metalkniveswhile the overwhelmingly dominantlithicrawmaterial atDK1 is quartzite, andmost of these are large unretouchedflakes and blades to con2000). This resultwould appear (Thackeray firmuntestedexpectations thatmetalbutcherymay resultin overallfewercutmarks thanstonetoolbutchery (Fisher1995; Lyman1987). In our experience, stone tools become rapidlyless effective as a result of dullingedges andfat adhering to the surface,neither of which is as significantwhen using metal of thecutmark knives.Thecorrection frequencies by surfaceareaallows us to measurethis differencein intensity of cutting quite precisely. Such abilities may in the futurehold out the possibilityof directly and reductionintensity relatingtool resharpening andmeasuring (Dibble 1995)to cutmark frequency, the impact of changes in raw materialquality on butchery efficiency. on the DK1 faunalassemblagehad Ourresearch as one if its goals the testing of Binford's (1984) hypothesisthat MSA people in SouthAfrica were not behaviorallymoder and tended to scavenge here in our large antelope of the size represented the River Binford that Klasies assemsample. argued a showed of that failed to blage pattern butchery resemblethatproduced in that moder butchers, by cutmarks were relativelyrarein areasof bones that have large amountsof flesh. The DK1 large bovid forelimbpattern visually shows a compellingsimito the Disarticulation-and-Filleting larity experimentaldataset(Figure 11), in that cuttingis more abundant at the tight humeral-radialjoint. The hindlimbvisuallyresemblesthe Filletingdatasetin its distribution (Figure12). Thus, we would expect thatthe DK1 CNC frequencieswould increaseand decrease across the long-bone zones most consis-

REPORTS

659

9.4 6.7

1 4.3 j 1.7

HU proximal epiphysi,

7 26.3

HU proximalshaft ] 15.2

6.3
8

1
I 5

I 10.7 HU distal shafti 14.2

HU mid shaft

5.3

15.7
4 3 *1 1.5

HU distal epiphysis 7]
RA proximal epiphysiL RA proximal shaft RA mid shaft RA distal shaft RA distal epiphysis

25.5

3 1.7
3.8 2.9 3.8 0.1
op
'^o 0 0 0
0?

i1 45.1 ] 18.6 ] 16 i 11.5

s ]5
,' ,
0 Io 0 0

0p

tn

0s v

0o

0 0

Filleting only Disarticulation followed by Filleting

DKi bovids (Sizes 3 and 4)

Figure 11. Cutmark frequencies across the length of forelimb long bones, expressed as percentages (each skeletal element adds up to 100 percent). Cutmark frequencies (data labels) are the CNC values.

4.9 9.4
7.2
E 9.l

,
I3.3

18 8 12.3

FE proximalepiphysis, 5.6
FE proximal shaft :

FE mid shaft
FE distal shaft

12.3 24.5

3 6
9
^

4.5

FE distal epiphysis

20.6 14.1

TI proximalepiphysis1 6.7 TI proximalshaft t1

1_

14.8 18.5 15.7 6

0
O

2.1

I 1.3 3.8 oI
0 0 0 0

TI mid shaft TI distal shaft

aI
o
0.

TI distal epiphysis s]
c <

0 VI
O

-0 0 O~~~~~r.

Filleting only Disarticulation followed by Filleting

DK1 bovids (Sizes 3 and 4)

Figure 12. Cutmark frequencies across the length of hindlimb long bones, expressed as percentages (each skeletal element adds up to 100 percent). Cutmark frequencies (data labels) are the CNC values.

660

AMERICAN ANTIQUITY

[Vol. 67, No. 4, 2002]

forelimb tentlywiththeDisarticulation-and-Filleting andFilletinghindlimb model.The correlation analyin Table3 supports sis presented thishypothesis. In summary, throughthe use of the CNC analysis we were able to directlyandquantitatively comin long-bonezones parethe frequencyof cutmarks betweena highly fragmented fossil bone collection (DK1) and an actualisticassemblageof cutmarked whole bones. This allowed us to discovertwo patterns with the following likely conclusions: 1) the DK1 MSA people butcheredsize 3 and 4 animals such thatthe cuttingintensityco-variedacrosslong bones in a way that significantlycorrelateswith a model where moder people disarticulated and filleted the forelimband filletedthe hindlimb,and 2) that the use of stone tools at DK1 likely caused a per unitbone higheroverallfrequencyof cutmarks surface.Clearly,the calculation of CNC (madepossible by the image-analysisGIS approach),when combined with highly controlledactualisticdata, allows us to significantlyimproveupon traditional to cutmark analysisthatrely on qualitaapproaches tive assessmentsof cutmarkpatterningguided by anecdotalknowledgeof butcherypatterns. Discussion and Conclusions havedevelopeda wide varietyof Zooarchaeologists for approaches recordingand presentingcutmark data, and they vary accordingto what is counted or cutmarked andthe way the (cutmarks fragments) valuesareexpressed(NISPor MNE).Ourreviewof data: the literature findsfive basictypesof presented MNE NISP cutmark-count data, data, diagrammatic cutmark-count data,NISP fragment-count data,and MNE fragment-count data.Thereis no widespread the as to themosteffectiveway to present agreement of as is an concern undercurrent and there also data, To to the effectivenessof thesedifferent approaches. ourknowledge,only Bartram (1993) has provideda detaileddiscussionof the problemssufferedby cutmarkanalysisas a resultof bone fragmentation. dataprovidesa usefuloverallview Diagrammatic butit is nearlyuseless of the placementof cutmarks, to go beyond for comparative that attempts analysis and Furthermore, impressions. qualitative subjective can be very misleaddiagrammatic representations ing in regardto cutmarkintensityif there is substantialintra-bonedifferentialsurvival.Cutmarks can be absent from locations on the bone simply in becausethose locationsare not well represented

the faunalassemblageor analysis.The GIS method we havepresented hereovercomes thatproblem with its abilityto presentan MNE densitymap across a bone surfacealongwitha compositediagram of cutmarkplacement. Moreimportantly, the GISmethod can quantitatively correctthe intensityof cutmarksurfaceareain anyway the anaing by thepreserved lyst desires. data and NISP fragmentNISP cutmark-count countdataareextremelysensitiveto fragmentation becausewithincreasing thetotalnumfragmentation ber of fragmentsrises more rapidlythan the total of eithercutmarks or cutmarked number fragments. of archaeological NISP cutmark data Comparisons to unfragmented collectionssimplyare experimental not effectiveunless the archaeological collectionis of Likewise,comparison completelyunfragmented. databetweenarchaeological NISP cutmark sites of evenlimitedvariable will be spurious. fragmentation MNE cutmark-count and MNE fragment-count data have been arguedto be resistantto the fragmentation effect.Whilein some cases theMNEmay effect, it does not overmollify the fragmentation come it. This is becauseas processesof attrition act on a bone, they tend to destroythe less-dense portions first, leaving the denser portions preserved. MNEs based on these denserportionswill remain andcutmarked high,whilemanycutmarks fragments are lost to the analysiseitherbecausethey are renor are deredunidentifiable throughfragmentation, completelydestroyed. cutmark Accuratelyquantifying frequenciesis a to surface-area similar problem samplingproblems Forexamareain otherdisciplines. involvingsurface to when habitats estimate popple, ecologistssample ulation size, the proper correction is population havetriedto use dividedby area.Zooarchaeologists the MNE as a proxy for area,but it does not work do notallhaveanequalchance becauseboneportions of surviving attritionalprocesses. Rapson (1990) suggesteda way to attemptan aerialestimate,but the methodis fairlycoarseandprovidesdataof limvalue.TheGISmethoddoes anexcelitedanalytical for preserved surfaceareaand lentjob of correcting makes possible the meaningfulcomparisonof cutmarkfrequenciesbetweenheavily,lightly,and unsamples. fragmented The GIS image-analysis approachhas several It providesa multitudeof possible other strengths. that go beyondany otherextantapproach. analyses

REPORTS

661

The GIS methodallows the analystunlimitedflexibility in the definitionof zones to calculatesurface For example,if areaand count cutmark frequency. the analystwantsto examinethe intensityof cutting on just the femoralhead,thenthe femoralheadcan If theanaas thezoneforquantification. be identified of wishes to the lyst compare intensity slicing and marks withinjust the femoralhead, thatis hacking easy as well becausethe GIS attachesa datatableto each individual andone can restrict cutmark, analyses withqueryfunctionson thedatatables.Thisability to stratify analysesby datacharacteristics applies to anypossibledata-table information suchas taxon, or anything else. size, age, provenience, The GIS methodprovidesdataarchivalabilities thatgo farbeyondanyotherapproach. Any approach forthatrecordscutmark placementwith a database matimmediately loses the exactplacementof a cutmark.The GIS methodforeverrecordsthe location of a cutmark in a precisemanner. Other diagrammatic record cutmark in thisway as placement approaches data well, buteachmarkmustby codedto a separate tableif one wishes to recordspecificinformation on the nature of that cutmark, and this creates an extremelyclumsy recordingsystem. With the GIS is drawndirectlyon thecomputer method,a cutmark and one mouse click opens a data table template, attachedforeverto thatcutmark. We have focused our attentionhere on cutmark recordingand analysis,but clearlythis methodcan be expanded to almostanytypeof bone-surface modification. For example, one could record burning in damageon bone surfacesto discoverpatterning that could reflect burning differing approachesto of bonesurfaceareacouldreveal cooking.Estimates in theintensity differences of fragmentation andbone processing.Locationalanalysisof toothmark placement could reveal taxon-specificfeeding patterns among moder control assemblages, and then be to fossilcollections. Thereareprobably applied many otherzooarchaeological thatgo beyond applications the productionof MNEs (Mareanet al. 2001) and cutmarks.Bones fragmentinto specimensthat are in shapeandnongeometric in outline. unpredictable Foryearswe havebeen attempting to describebone fragmentsand their surfacemodificationsas numbers on a database,when in realitythe problemhas will always been one of image. Zooarchaeologists undoubtedly benefit as image analysis software becomes increasinglypowerfulanduser-friendly.

We thankGraham Acknowledgments: Averyand the rest of the staff in the Department of Archaeology, South African Museum, for their help duringthe analysis of the Die Kelders Cave 1 fauna. Thanks to Antonieta Jerardinofor the Spanish abstract.The Die Kelders analysis was funded by NSF grant SBR-9727491 and a Wenner-Gren grant to Marean, and the development of the GIS software was funded by NSF grant SBR-9727668 to Marean.

References Cited
P. Goldberg,F. E. Grine,R. G. Klein, Avery,G., K. Cruz-Uribe, M. J. Lenardi, C. W.Marean, W.J. Rink,H. P. Schwarcz, A. I. Thackeray, andM. L. Wilson 1997 The 1992-1993excavationsat the Die KeldersMiddle andLater StoneAge CaveSite,SouthAfrica.Journal of Field Archaeology24:1-29. L. E. Bartram, 1993 An Ethnoarchaeological Analysis of Kua San Ph.D. disser(Botswana)Bone Food Refuse. Unpublished of Anthropology, tation,Department Universityof Wisconsin at Madison. L. E., andC. W. Marean Bartram, 1999 Explainingthe "KlasiesPattern": KuaEthnoarchaeology, theDie KeldersMiddleStoneAge Archaeofauna, Long Bone Fragmentation and CarnivoreRavaging.Journal of ArchaeologicalScience 2:69-29. Binford,L. R. 1978 NunamiutEthnoarchaeology. Academic Press, New York. 1981 Bones: Ancient Men and Modern Myths. Academic Press,New York. 1984 TheFaunalRemains from KlasiesRiverMouth.Academic Press,New York. 1985 HumanAncestors:ChangingViews of theirBehavior. JournalofAnthropological Archaeology4:292-327. 1988 Fact and Fiction aboutthe Zinjanthropus Floor:Data, Arguments, and Interpretations.CurrentAnthropology 29:123-135. R. J. Blumenschine, 1988 An Experimental Modelof theTimingof Hominidand influenceon Archaeological Carnivore Bone Assemblages. JournalofArchaeologicalScience 15:483-502. R. J., C. W. Marean,and S. D. Capaldo Blumenschine, 1996 BlindTestson Inter-Analyst andAccuCorrespondence of Cut Marks,PercussionMarks, racy in the Identification and Carnivore ToothMarkson Bone Surfaces.Journalof ArchaeologicalScience 23:493-507. Brain,C. K. 1981 The Hunters or the Hunted? University of Chicago Press,Chicago. Bunn,H. T. 1981 Archaeological EvidenceforMeat-Eating by Plio-Pleistocene Hominids from Koobi Fora and Olduvai Gorge. Nature291:574-576. 1991 A TaphonomicPerspective on the Archaeology of Human Origins. Annual Review of Anthropology 20: 433-467. Bunn,H. T., andE. M. Kroll 1986 SystematicButcheryby Plio-PleistoceneHominidsat Olduvai Gorge, Tanzania. Current Anthropology 27:431-452. 1988 Fact and Fiction aboutthe Zinjanthropus Floor:Data, Current 29: Arguments,and Interpretations. Anthropology 135-149.

662

AMERICAN ANTIQUITY

[Vol. 67, No. 4, 2002]

Capaldo,S. D. HominidandCarnivore BehaviorfromDual1995 Inferring Patterned Archaeofaunal Ph.D. Assemblages.Unpublished of Anthropology, dissertation, Department RutgersUniversity. Determinations of Carcass 1997 Experimental Processingby at FLK22 (ZinPlio-PleistoceneHominidsandCarnivores Journalof Human janthropus),OlduvaiGorge, Tanzania. Evolution33:555-597. Marks 1998a Methods, andModelsforInferring Hominid and Carnivore Behavior. Journal of Human Evolution 35:323-326. of Dual-Patterned 1998b Simulatingthe Formation ArchaeofaunalAssemblageswith Experimental ControlSamples. Journalof ArchaeologicalScience 25:311-330. Chase,P. G. 1986 TheHuntersof CombeGrenal:Approachesto Middle PaleolithicSubsistencein EuropeBAR International Series 286, Oxford. 1988 Scavengingand Huntingin the Middle Paleolithic.In Eurasia, editedby UpperPleistocenePrehistory of Western H. L. Dibble andA. Montet-White, pp. 225-232. The UniMuseum,Philadelphia. versityof Pennsylvania D. C. Crader, 1983 RecentSingle-Carcass Bone Scattersandthe Problem of "Butchery" Sites in the ArchaeologicalRecord.In Animals andArchaeology: Huntersand theirPrey,editedby J. Clutton-Brock andC. Grigson,pp. 107-141. British Archaeological Reports163,Oxford. Dibble, H. L. 1995 Middle Paleolithic ScraperReduction:Background, andReview of the Evidenceto Date. Journal Clarification, of ArchaeologicalMethodand Theory2:299-368. Driesch,A. von den 1976 A Guide to the Measurementof Animal Bones from Sites.Harvard Archaeological PeabodyMuseum University, of ArchaeologyandEthnologyBulletin 1, Cambridge. Fisher,J. W., Jr. in Zooarchaeology. 1995 Bone SurfaceModifications Journal of ArchaeologicalMethodand Theory2:7-68. Frison,G. BuffaloJump, 1970 TheGlenrock 48C0304: LatePrehistoric and Butcheringon the NorthPeriodbuffaloProcurement westernPlains.PlainsAnthropologist, MemoirNo. 7:1-55. D. Gifford-Gonzalez, 1989 EthnographicAnalogues for InterpretingModifed Bones:Some CasesfromEastAfrica.InBoneModification, editedby R. BonnichsenandM. H. Sorg,pp. 179-246. Center for the Studyof the FirstAmericans,Orono,Maine. 1991 Bones Are Not Enough:Analogues, Knowledge,and in Zooarchaeology. JournalofAnthroInterpretive Strategies pological Archaeology10:215-254. Gifford,D. P., andD. C. Crader Faunal 1977 A Computer CodingSystemforArchaeological Remains.American 42:225-238. Antiquity Gifford,D. P.,G. Isaac,andC. M. Nelson at Prolonged 1980 Evidence for Predationand Pastoralism NeolithicSitein Kenya. Drift:A Pastoral Azania15:57-108. Gilbert,M. 1969 Some Aspects of Diet and Butchering Techniques Indiansin SouthDakota.PlainsAnthroamongPrehistoric pologist 14:277-294. Grayson,D. K. Bone Destruction, andReverseUtility 1989 Bone Transport, Curves.JournalofArchaeologicalScience 16:643-652. Grayson,D. K., andF. Delpech 1994 The Evidencefor MiddlePaleolithicScavengingfrom

CoucheVIII, GrotteVaufrey(Dordogne,France).Journal of ArchaeologicalScience 21:359-375. andD. P.Tanner Guilday,J. E., P.W. Parmalee, 1962 AboriginalButcheringTechniquesat the Eschelman Site (36 La 12), Lancaster PennsylCounty,Pennsylvania. vaniaArchaeologist32:59-83. Klein, R. G. 1989 WhyDoes SkeletalPartRepresentation Differbetween SmallerandLarger BovidsatKlasiesRiverMouthandother ArchaeologicalSites? Journal of ArchaeologicalScience 6:363-381. Lam,Y. M., X. Chen,and0. M. Pearson in Patterns 1999 Intertaxonomic of Bone Density Variability and the DifferentialRepresentation of Bovid, Cervid,and Equid Elements in the ArchaeologicalRecord.American 64:343-362. Antiquity Landon,D. B. 1996 FeedingColonialBoston:A Zooarchaeological Study. HistoricalArchaeology30: 1-153. Lyman,R. L. 1984 Bone Density and DifferentialSurvivorship of Fossil Classes.Journal 3:259-299. ofAnthropologicalArchaeology 1985 Bone Frequencies: Differential Transport, in Situ andthe MGUI.JournalofArchaeologicalSciDestruction, ence 12:221-236. 1987 Archaeofaunas and ButcheryStudies:A Taphonomic Perspective.In Advances in ArchaeologicalMethod and Vol. 10, editedby M. J. Schiffer,pp. 249-337. AcaTheory, demic Press,New York. of UtilityCurvesin Zooar1992 Anatomical Considerations chaeology.Journalof ArchaeologicalScience 19:7-22. C. W. Marean, 1992 Hunterto Herder:LargeMammalRemainsfrom the Hunter-Gatherer Ya Muto RockAt Enkapune Occupation shelter Review Rift,Kenya). (Central African Archaeological 10:65-127. 1998 A Critiqueof the Evidence for Scavengingby Neanderthals andEarlyModernHumans: New DatafromKobeh Cave (ZagrosMousterian) and Die Kelders(SouthAfrica MiddleStone Evolution 35:111-136. Age). Journal ofHuman Marean,C. W.,Y. Abe, C. J. Frey,andR. C. Randall and Taphonomic 2000a Zooarchaeological Analysis of the Die KeldersCave 1 Layers 10 and 11 Middle Stone Age Larger Mammal Fauna. Journal of Human Evolution 38:197-233. Marean,C. W.,Y. Abe, P. J. Nilssen, andE. C. Stone theMinimum Numberof SkeletalElements 2001 Estimating A Review andA New Image(MNE) in Zooarchaeology: American 66:333-348. Analysis GIS Approach. Antiquity Marean,C. W., andZ. Assefa 1999 Zooarchaeological Evidence for the FaunalExploitation Behaviorof Neandertals and EarlyModernHumans. 8:22-37. Evolutionary Anthropology C. W.,P.Goldberg, G.Avery,E E. Grine,andR. G. Klein Marean, andExcavations 2000b MiddleStoneAge Stratigraphy at Die KeldersCave 1 (Western CapeProvince,SouthAfrica):the 1992,1993, and 1995FieldSeasons.JournalofHumanEvolution38:7-42. Marean,C. W., and S. Y. Kim FaunalRemainsfromKobehCave(Zagros 1998 Mousterian Behavioral forNeanderthals Mountains, Iran): Implications and Early Modern Humans. Current Anthropology 39:S79-S114. Marean,C. W., andL. M. Spencer 1991 Impactof CarnivoreRavagingon Zooarchaeological Measures of Element Abundance. American Antiquity 56:645-658.

REPORTS Marshall,F 1990 CattleHerdsandCaprineFlocks. In EarlyPastoralists of Southwestern Kenya, edited by P. Robertshaw, pp. of EastAfrica,Nairobi. 205-260. BritishInstitute Milo, R. G. in Southern AfricanPreInteractions 1994 Human-Animal history:A MicroscopicStudyof Bone DamageSignatures. Ph.D. dissertation, of AnthropolDepartment Unpublished ogy, Universityof Chicago. 1998 Evidence for Hominid PredationAt Klasies River for the Behavior Mouth,SouthAfrica,andIts Implications of EarlyModem Humans.Journalof ArchaeologicalScience 25:99-133. Nilssen, P. J. 2000 An ActualisticButcheryStudyin SouthAfricaand Its of Carfor Reconstructing HominidStrategies Implications cass AcquisitionandButcheryin the UpperPleistoceneand Plio-Pleistocene.UnpublishedPh.D. dissertation,Departmentof Archaeology,Universityof CapeTown. P.W. Parmalee, 1965 The Food Economyof ArchaicandWoodland Peoples at the TickCreekCave Site, Missouri.MissouriArchaeologist 27:1-34. Potts,R. 1983 Foragingfor FaunalResourcesby EarlyHominidsat OlduvaiGorge,Tanzania. In Animalsand Archaeology:1. and C. Huntersand TheirPrey, editedby J. Clutton-Brock Grigson,pp. 51-62. BritishArchaeologicalReportsInternationalSeries 163, Oxford. at Olduvai. AldineDe Gruyter, 1988 EarlyHominidActivities New York. Potts,R., and P. Shipman 1981 Cutmarks Madeby StoneTools on Bones FromOlduvai Gorge,Tanzania. Nature291:577-580. Rapson,D. J. andProcessin Intra-Site 1990 Pattern SpatialAnalysis:Site Structural and FaunalResearchat the Bugas-HoldingSite. Ph.D. dissertation, of AnthropolUnpublished Department ogy, Universityof New Mexico, Albuquerque. Selvaggio, M. M. 1994 Carnivore ToothMarksandStoneToolButchery Marks On ScavengedBones:Archaeological Journal Implications. of HumanEvolution27:215-228.

663

1998 Evidencefor a Three-Stage Sequenceof Hominidand CarnivoreInvolvementwith Long Bones at FLK ZinjanJournalofArchaeologithropus,OlduvaiGorge,Tanzania. cal Science 25:191-202. Shipman,P. 1981 Applications of Scanning Electron Microscopy To In TheResearch Problems. Potential ofAnthroTaphonomic pological MuseumCollections,editedby A. M. Cantwell,J. B. Griffin, andN. A. Rothschild, pp. 357-385. Annalsof the New YorkAcademyof Sciences,Volume376, New York. 1986 Studies of Hominid-FaunalInteractionsat Olduvai Gorge.Journalof HumanEvolution15:691-706. 1988 ActualisticStudiesof AnimalResourcesand Hominid Activities.In ScanningElectronMicroscopy andArchaeology, edited by S. L. Olsen, pp. 261-285. BritishArchaeoSeries452, Oxford. logical ReportsInternational Shipman,P.,andJ. Rose andCarcass-Pro1983 EarlyHominidHunting,Butchering, to the Fossil Record.Jourcessing Behaviors: Approaches nal ofAnthropological Archaeology2:57-98. Stiner,M. C. A Zooarchaeological 1994 HonorAmongThieves: Studyof Neandertal Ecology.Princeton UniversityPress,Princeton, New Jersey. A. I. Thackeray, 2000 Middle Stone Age ArtifactsFrom the 1993 and 1995 Excavationsof Die KeldersCave 1, SouthAfrica.Journal of HumanEvolution38:147-168. Wheat,J. B. 1979 The JurgensSite. Plains Anthropologist, MemoirNo. 15. Wilson,M. C. 1982 CutMarksandEarlyHominids: EvidenceforSkinning. Nature298:303. Yellen,J. E. 1991 SmallMammals: SanUtilization andProduction !Kung of Faunal Assemblages.JournalofAnthropologicalArchaeology 10:1-26.

ReceivedSeptember10, 2001; RevisedMarch28, 2002; AcceptedApril2, 2002.