A FURTHER ASSESSMENT OF PALEOINDIAN SITE-USE AT BONFIRE SHELTER Ryan M. Byerly, JudithR. Cooper,David J. Meltzer,MatthewE. Hill, and JasonM. LaBelle
In Byerly et al. (2005) we explored the hypothesis that the Paleoindian componentat Bonfire Shelter was the result of a jump kill. Our efforts involvedextensive mappingand GIS analysis, a re-examinationof the Paleoindian-age bison assemblage, and consideration of the geomorphic history of the canyon in which the site is located. Weconcluded that the preponderance of evidence indicated the Paleoindian-age bison remainsat BonfireSheltermarkeda processing site as Binford (1978) suggested, ratherthan a primarykill locality as originally interpreted(Dibble 1968). Bement(this issue) raises several concerns about our analysis and discussion, including that we omit pertinent informationrelevant to the interpretation of the site. His comments,however,resultfrom a misreading of our discussion and a misconstrual of the data set, as we explain in this response. En Byerly et al. (2005) exploramosla hipotesis que el componentepaleoindio de BonfireShelterfuese el resultadode una matanzade salto. Nuestros esfuerzos incluyeronla preparacionde extensivosmapas y un andlisis SGI (GIS), una reexaminacion de la asambleabisontedefecha paloeoindia y la evaluacionde la historiageomorfologicadel canon en que se encuentra el sitio. Concluimosque la mayorparte de la evidenciaindicaque los restosbisontesdefecha paleoindia en BonfireShelter demuestranquefue un sitio de procesamientotal como sugirio Binford(1978), en vez de ser una localidad de una matanza primariatal como originalmentesepropuso (Dibble 1968). Bement(this issue)plantea variosproblemassobre nuestroandlisis, entreotras cosas que omitimosla informacionpertinenteque tiene que ver con la interpretaciondel sitio. Sin embargo, su comentarioes el resultadode unamala interpretacionde nuestradiscusiony los datos, tal comoexplicamosen esta respuesta.
et al. (2005) explored whetherthe Paleoindian-agebisonbone assemblageat BonfireShelter(41VV218) was the result of ajumpkill at thatspot,as originallyinterpreted (Dibbleand Lorrain1968), or alternativelywas a localitywherecarcasspartsof animalskilledelsewhere were transportedand processed (Binford 1978:475).Ourstudyinvolvedthreeelements:(1) fieldmappingandGeographicInformationSystem (GIS) analysis of the terrain,in search of drive lanes or topographicfeaturesthat might indicate how animalsweredriventowardthecliff edge, and the feasibilityof doing so; (2) a thoroughreanalysis of thebisonremainsrecoveredin the 1963-1964 excavations,to see if the skeletalelementpatterns indicateda primarykill or secondaryprocessing; and (3) a study of the geomorphichistoryof the canyonin whichtheshelteris located,to determine
how muchof the LateGlacialvalley fill was since removed, a matter relevant to the question of whetherbison carcasssegmentscould have been hauledinto this shelter,now -18 m abovethe valley floor. Ourresults,in brief,indicatedthis would have been an ideal spot for jumping bison relativeto otherareasin the vicinityof the shelter.However, it maynothavebeenso used.Reanalysisof thebison remainsindicatedthatskeletalpartspresentat the site were moresuggestiveof a processinglocality. we expresslystatedthis did not speak Importantly, to howthebisonwerekilled,onlythattheseremains didnotindicatethiswastheprobablekilllocus.And althoughwe were unable to securely determine where the valley floor was in Late Glacialtimes, evidencesuggestedit was perhapsjust a few vertical metersbelow the shelterentrance,and hence
Ryan M. Byerly, Judith R. Cooper, and David J. Meltzer Departmentof Anthropology,SouthernMethodistUniversity, Dallas, TX 75275 (
[email protected]) Matthew E. Hill Departmentof Anthropology,Universityof Arizona,Tucson,AZ 85721 Jason M. LaBelle Departmentof Anthropology,ColoradoState University,Fort Collins, CO 80523 AmericanAntiquity,72(2), 2007, pp. 373-381 Copyright©2007 by the Society for AmericanArchaeology 373
374
AMERICANANTIQUITY
hunters who had made a kill nearby were not impededby the currentlysteepslope. Bementcriticizedportionsof ourstudy,leading with the suggestionwe omitted"pertinentinformationthatis directlyrelevantto the interpretation of the site and to an evaluationof [our] conclusions."As we had explicitly called for more evidenceto resolvethe use of this site in Paleoindian times(Byerlyet al. 2005), Bement'scomments,as a participantin the work at BonfireShelterin the early 1980s, are welcome. That said, they do not significantlyadvancetheissue,sincetheyarebased of ourargumentsandavailon a misinterpretation abledata.Althoughwe arecontentto let ourpaper standon its merits,we takethisopportunityto clarify the natureof the argumentand evidence, and expandon ourresults. Bement(thisissue) assertswe committeda logical fallacyby reasoningthatthe Paleoindian-age bonebedat BonfireShelter(BB2) "didnot result fromajumpbecauseno otherPaleoindian jumpsite has everbeen found."Hadwe actuallyreasonedin thatmanner,he wouldhavebeenrightto makethat claim. However,we did not do so. To be sure,we illustratedthe distributionof bison jump kills in time and space (Byerlyet al. 2005:Figures1, 13). Yet,thismerelyreinforcesthepoint,madelongago by Dibble(BonfireShelter'soriginalinvestigator), that Paleoindianbison jumps are scarce on the SouthernPlains(Dibble 1970:252;see also Forbis 1969:91).We then went on to agree with Dibble thatjust becausea phenomenonis rare,thatis not evidence againstany particularlocality, say BB2 atBonfireShelter,being anexpressionof thatphenomenon (i.e., a bison jump; Byerly et al. 2005:597). Bement'scommentaboutour reasoning is thereforeincorrectandirrelevant. Moreover,if usingthelackof Paleoindianbison jumps as evidenceagainst BB2 being a jump is a fallacy,thenwhy is it logical for Bementto use the Late Archaicbison bonebedat Bonfire(BB3) as evidence/or a Paleoindianjump?(e.g., "Byerlyet al. fail to explorethe implicationsof BB3 whichis undeniablytheresultof LateArchaicbisonjumps" [Bementthis issue]). The logic cuts bothways, so too doesthefallacy.Forthatreason,we alsoexplicitly omittedBB3 fromour discussion. The scarcityof Paleoindianbison jumps does warranttwo questions:first,is therean alternative hypothesisthatmight accountfor the facts of the
[Vol. 72, No. 2, 2007]
case? Second,if the facts are as they appear(i.e., BB2 resultedfroma jumpkill), why does such an anomalouspatternexistinthearchaeological record, and how might it be explained (Byerly et al. 2005:626; see also Forbis 1969; Frison 1991)? Some obvious hypothesesexplain the latter.For example,repeatedjumpkillson theNorthernPlains (e.g., Frison1991)wereprobablytheresultof intensificationstrategiesgearedtowardsthereliefof seasonallyconstrainedresourcelimitations,such that long-termnutrientstoragein the form pemmican productionwasimplementedto subsidizeleanwinter months (e.g., Reeves 1990; Todd 1987). The absenceof such sites on the SouthernPlainsindiin theregionwerenever catesthathunter-gatherers faced with the same limitations,andthereforedid not needto use similaracquisitionstrategies. But, of course,one mustresolvethe firstquestionbeforeturninganalyticalattentionto thesecond. Thatis whywe analyzedtheBB2 faunalassemblage, the surfacetopography, andthegeomorphichistory of thecanyon,andmaintaineda firewallbetweenthe issue of the largerpatterningin the archaeological record,andthe empiricalevidencefor/againstBB2 beingthe resultof a jumpkill. Inregardto thatempiricalevidence,we deemed it usefulto reanalyzethebisonfromthe 1963-1964 excavations ratherthan rely on published data, which would do little more than replicateother studies(e.g., Binford1978:475).Werealizedthese dataarebiased,most notablythatthey come from a relativelylimited area of the site (Byerly et al. 2005:605).Thosebiases,of course,equallyhinder of the site. We also did not originalinterpretations from the 1980s excavations reanalyzethe bison notmateriallyimpact as that should but, explained, since that later ourconclusions, samplewas small (Numberof IndividualSpecimens [NISP] = 51; Bement 1986:26)andrepresentsa minorpercentage (< 2.5 percent)of the totalremainsfromBB2 (Byerlyetal. 2005:605). Based on thatanalysis,we calculateda Mimimum Numberof Individuals (MNI) of 24 to 27 Bison antiquus(Byerlyet al. 2005:625).Nowhere do we suggestthisis the"totalorultimate"(Bement this issue) numberof animalsin BB2. The originalinvestigatorsarrivedat a similarMNIestimate, but then arbitrarilymultipliedit by four, on the undemonstrated assumptionthatonly half the aniand malspreserved thatunexcavatedportionsof the
COMMENTS
thetaluscone,containedskeletons site,particularly in the sameor greaterdensityandproportion(e.g., Lorrain1968:84).Thisassumptionrestson thecontentionthat the bison enteredthe sheltervia the notch in the cliff face. As argued(Binford 1978; Byerlyet al. 2005), thatmay not be the case at all. BB2 bone andlithics(Byerlyet al. 2005:Figure1) arehighly concentratedin the southernend of the shelter, and certainly greater numbers are not expectedin the unexcavatednorthernportionsof the shelter. Indeed, Dibble notes the bonebed thinnedandelementfrequenciesdiminishedaway from the excavations(Dibble 1968:29), a pattern Bementlikewise observed(1986:19).Regardless, themattermustbe resolvedby testing,notthrough unfoundedassertionsabouta relationshipbetween skeletalcountsand site area. Our analysisof dentalpatternsindicatedBB2 probablyrepresenteda single summerkill (Byerly et al. 2005:612). Bement(this issue) pointsto the difficultyof makingsuch a seasonalassignment, andassertswe failto considerthecriteriathatwould indicatemore than one event took place, namely "thepresenceof stratifieddeposits,differentialfaunal taphonomies,or artifactassemblage differences." We agree the identification of a single componentvia dentalcohortclusteringis a difficult assessment,especiallywith as smalla sample as thatfromBB2. But it is hardlyunheardof (e.g., we do notpreclude Toddet al. 1992).Furthermore, the possibility that the dental data may indicate "multiple,closely timed deathevents"(Byerly et to denal. 2005:612), nor limit our interpretation tal data. We consideredthe possibility,raisedby Dibble (1968) and assertedby Bement (this issue), that BB2 was subdivisibleintoseveralstrata.However, Dibble (1968:29-30) observedthatBB2 is divisible into threestrataonly on the southernslope of the talus cone and not at all in the "centraland northernareasof its occurrence"(Dibble1968:29). Lorrain(1968:92),in fact, cites the lack of distinguishablestrataacrossBB2 as a reasonwhy bones werenot separatelyanalyzed,andwhy seasonality was likewisenot assessedper stratum.These facts unfortunatelynegate any investigationof "differentialfaunaltaphonomies"(Bementthis issue) in the extantassemblage. For that matter,althoughBement (this issue) suggeststhe "collapseof threestratainto one does
375
notfurtherourunderstanding of BonfireSheltersite history or Folsom and Plainviewhunting strategies,"it is the only way the availablearchaeological data can be evaluated.But then perhapsthe issue is not so muchthe separationof these strata, but theiroverlap.Bement (this issue) assertsthat the presenceof "bothFolsom and Plainviewprojectile points suggests at least two temporaland societalgroups."That'shardlya compellingclaim on eithertheoreticalor empiricalgrounds.Projectile pointstyles arenot nicely layer-cakedin time; they can and do overlap(Sellet 2001). Moreover, they actuallydo at BonfireShelter.Grantingthat threedepositionalepisodescomposeBB2, Dibble reportsthat the lowest "stratum,"ComponentA, yieldedbothFolsomandPlainviewpoints(Dibble 1968:Table 2; also Cooper and Byerly 2005). Bement's(this issue) concernsaboutFolsom and Plainvieware thus misguided,even acceptingthe division. tripartitestratigraphic Moreover,acceptingthisdivisionrendersmeaninglessBement'sestimateof thetimespanningand separatingthe supposed multiple kills. Leaving aside the degree of confidenceone can have in a seriesof radiocarbonageswithstandarddeviations as largeas theonesfromBB2 (Holliday2000:251), and from a period characterizedby significant radiocarbonplateaus(i.e., the latterportionof the YoungerDryas; see Blackwell et al. 2006; Kitagawa and van der Plicht 1998), such an estimate necessarilyassumessedimentdepositionis linear over time. This reasoningis inconsistentwith the assertionthatthreeseparatedepositionalepisodes occurred. Although three BB2 deposits (A, B, and C; Bement1986:Figure5) wereidentifiedin the 1980s excavations,bone is only sparselypresentin A, withthe majorconcentrationlying at the B/C contact;no differentiationby burning(as seen in the 1960s)is noted(Bement1986).This,coupledwith the reworkingof Zone 2a underBB2 in the central andnortherndeposits(Dibble 1968:21),indicates thata directcorrelationbetweenBement's(1986) andDibble's(1968) stratigraphic sequencesis perhapsunfounded,andthattheBB2 separationcould representan isolatedpostdepositionalburnfeature buriedby one or more series of talus slope wash events(Byerlyet al. 2005:612). Bement (this issue), following Lorrain (1968:100), furtherargues that crania and pha-
376
AMERICANANTIQUITY
[Vol. 72, No. 2, 2007]
Table 1. SHEI for "High-SurvivalElements"in BB2. Element Femur Tibia Metatarsal Humerus Radius Mandible Skull Metacarpal
TPb
MNE SFUI MAU p. PiPnCpj)] SHEI 100.0 33 16~5 !l80 ^309 -.277 62.8 26 13.0 .142 14 7.0 37.0 .077 -.197 36.8 30 15.0 .164 -.296 .142 -.277 25.8 26 13.0 26 .142 -.277 11.5 13.0 9.1 5 5.0 -.159 .055 5.2 18 9.0 .098 -.228 Total 178 91.5 1.000 -2.021 .972a aTheevenness value fits only within the range of the bulk transportstrategymodeled by Faith and Gordon(2006) at sample size (MNE total) = 150. It also does not fit within the range of values for an unconstrainedor non-transportedbone assemblage, implying selective transport(Faithand Gordon2006). Element data from Byerly et al. (2005:Table2). bPj= ElementMAU/X MAU (Faithand Gordon2006); SHEI = -Z {p^ln^)] }/ln(X Element),I Element= 8 in this instance (Faithand Gordon2006); TP = TotalProducts(Byerly et al. 2005:Table2); SFUI = StandardizedFood Utility Index. 100.0 58.1 15.9 28.4 19.7 10.4 10.4 6.0
langesoccurin equallyhigherfrequencythanthey ought to were bison carcass parts differentially transportedto this spot for processing.Thereare two problemswith this argument:(1) the cranial MNI of 18 maybe inaccurate,and(2) utilitymodeling incorporatesrelative,not absolute,element abundance. In regardto the former,Lorrain'sMNI derivations are problematic. In her Table 9 (Lorrain 1968:80) cranialdata are presentedaccordingto specifiedor unspecifiedportion(maxillaandmiscellaneous) and anatomicallandmark(auditory bulla),perunit.Thesedataarefurthersummarized per elementand elementportionin her Tables18 and19 (Lorrain1968:100).ThemaximumperportioncranialMNIis thatforleft maxilla(MNI=11). The cranialMNI of 18 was apparentlyderivedby addingthe individuallyderivedMNIs for left and rightmaxillaas well as thatforauditorybulla(Lorrain 1968:80). Since auditorybulla are temporal features,not maxillary,andbecauseit is unknown if maxillaweremaximallydistinguishedby age or overlappinganatomicalfeatures(see Hill 2001) in Lorrain'sanalysis, we consideredthe maximum maxillaryestimateof 11 (Byerlyet al. 2005:Table 2) to be the best minimumestimatefor BB2 crania, andnot the originallyreported18. Nonetheless,thiswouldstillimplythat11 of 24 46 (~ percent)expectedcraniaarepresent.This is indeed a large numbergiven thatcraniaought to be rareat a processinglocality.It is not, however, absolutecranialabundancethatis importantto our transport modeling,butrathertherelativefrequency (as percent MinimumAnimal Units [MAU]) of
craniarelativeto thatof otherelements,in thiscase, high-utilitylimb bones. As noted, comparisonof ourderivedrelativeelementfrequenciesto standard bison utilityindices best fit a bulk transportstrategy (rs= .54,/? = .05). However,traditionalmethodsof utilitymodeling relying on statisticalcorrelationand pattern "eye-balling"(Faith and Gordon2006) are subjective means of evaluatingtransportstrategies (Byerly et al. 2005:616-618; also Beaver 2004; FaithandGordon2006; Rogers2000; Rogersand Broughton2001). Faithand Gordon(2006) show that rank-order(rs) correlationsfor bulk utility strategies,in particular,do not reliably indicate relationshipsbetween economic utility and elementfrequencyatlow samplesizes (i.e., Minimum Numberof Elementstotals[MNE]),suchthatType II errorsdramaticallyincreasewith low element abundance(FaithandGordon2006). Accordingly,and following Faithand Gordon (2006), we reevaluateouroriginaltransportinferences here (Tables1 and 2) in termsof the Shannon EvennessIndex (SHEI),using our published MNE data(Byerlyet al. 2005:Table2). Thesedata indicatea cranialMNE of five, basedon left incisive counts (NISP = 5, MNE = 5), left maxilla counts(NISP= 7, MNE= 3), andleftpetrouscounts (NISP= 5, MNE = 5). Unconstrainedassemblages should be characterizedby an evenness of one, indicatingno differentiationin relative element abundance,with evenness declining across bulk andunbiasedassemblages,andgourmetstrategies of elecharacterized by highlyunevendistributions ments (Faith and Gordon 2006; also see Todd
COMMENTS
Table2. Spearman's(rs) Correlationsfor "High-Survival Elements"in BB2. Correlations InferredStrategy p rs %MAUvs. SFUI ^59 ~A3 WeakBulk %MAUvs. TPMRa .74* .04 Bulk %MAUvs. TP .83* .02 Bulk aUsesthe cranialTP value for mandibles(Emerson 1993). Utility values presentedin Table 1. *p < .05
1983:Table24). At a samplesize of 178 (totalMNE) for "highsurvivalelements"(e.g.,femora,tibiae,metatarsals, humeri,radii,metacarpals,mandibles,andcrania; see FaithandGordon2006), the SHEIfor BB2 is onlyconsistentwitha bulktransportstrategywhere high-utility upper limbs were selectively transported over low-utility crania and metapodials (FaithandGordon2006:Table4). These datasupportthehypothesisthatBB2 representstheremains of transported bison carcassportions(Byerlyet al. also Binford 2005; 1978).Of course,suchevidence is only one partof the equationof site-use;hence ourexplorationof skeletalelementarticulationand distribution,and our acknowledgmentof fluvial activityas a possibletaphonomicagentgoverning those variables within BB2 (Byerly et al. 2005:613). In regardto the latter,Bement(thisissue) cites elementco-association,most notablythe scapula concentrationin N60/W50, andthe skullandaxial in N60/W50andN50/W50elementconcentration N50/W60 at the base of the talus cone (Lorrain 1968:Table17, Figure40), as testimonythese animals died in place.Yet,these concentrationswere not interpretedas such by Lorrain.Rather,they weretakenas indicatingdeliberatestackingrelated to "intensive"butchery(Lorrain1968:96).Lorrain commentsthat: in the lowerbed (BoneBed Thearticulations 2), which were very few in number,usually consistedof onlytwobones,whereastheupper bonebed [BB3]oftencontainedarticulations of anentirelimboranentirevertebralcolumn and,in at least one case, of an entireyoung bison [1968:93]. Indeed, only seven unspecifiedarticulationsare cited (Lorrain 1968:Table 16), including two unfused femur/innominatearticulations(Lorrain
377
1968:Table14). However,the degreeorlackof articulationmay have little to do with butcheryactivity,and may speak to other taphonomic variables. Lorrain (1968:93) refers to the "mixingof the bones by waterthatpouredthroughthenotch"to explainthe low abundanceof mandiblesin N50/W70.Weconsideredthepossibleroleof waterin influencingthe distributionandrepresentation of skeletalelements (Byerly 2005:613).Althoughdetaileddataon the force and volume of flow from the notch during periodsof high-intensityoutputdo not exist, there areindicationsthatwater-flowoccurredwithinthe shelteranddistributedportionsof the BB2 deposit (Byerlyet al. 2005:613-614). Bement (this issue) assertsotherwise,but the archaeologicalpatternshe cites are equivocalon this issue, includingthe "spoke-like"arrangement of bonesfroma smallportion(just~ 9 m2)of northern portion of the rockshelter excavated in 1983-1984. We talliedthe orientationof the bone elementsin thisapparentfeature(dataderivedfrom Bement1986:Figure13)in 10°incrementsbetween 180° and 360° (to remove mirrorimaging; see Lyman1994:178).Theirdistributionwas thenstatisticallycomparedto a randommodel (following Kreutzer1988; Meltzer 2006). Expected values were derivedby dividingthe total samplesize (in this case 24 elements)by the numberof possible classes (n = 18), modelingan assumedeven distribution.The resultingstatistic(%2= 13.50,/?= .70) showsthatthedifferencebetweentheobservedand modeledexpectedvalues is insignificant.In nonstatisticalterms, bone orientationsshow no preferredalignmentor pattern. In anycase, the criticaltaphonomicquestionof what pre- and postdepositionalprocesses influenced bone distributionpatternswill need to be resolvedbasedon a muchlargerareathanthissmall portionof the shelter.Here,too, it mustbe recognizedthatfluvialactionwas probablynot uniform acrossthe site since some areas,like those nearest the rearwall of the shelter,wherethe only hearth featureis preserved,or far from the pour-off,are betterprotectedthanothers.Accordingly,we suggest the "bonebedas mosaic"concept (Toddand Rapson1999), incorporatingan explorationof all taphonomicvariables,is a far more illuminating in perspectivethana simplesearchfor "patterning culturalremains"(Bementthis issue). It is via the
378
AMERICANANTIQUITY
[Vol. 72, No. 2, 2007]
Table 3. SignificantSpearman's(rs) Correlationsfor Utility among SampledBison Bonebeds. TF TF MAR SFUI SKF P Assemblage21 MNEb P P P P rs rs rs rs rs CSP 643 !54 ^73* !()3 T64 X)9 ^43 40 ^43 40 3l -.71* .05 308 -.83* .04 -.90** .00 -.94** .00 -.94** .00 GN(U) -.77 .07 -.67 .07 840 -.83* .01 -.89* .02 -.89* .02 GN(L) .08 -.28 .51 367 -.50 .21 -.90* .02 .02 -.75 -.90* AB(AB) BB2 .13 178 .74* .04 .84* .04 .84* .04 .70 .13 .59 aAB(AB)= Agate Basin, Agate Basin; BB2 = Bonfire Shelter,Bonebed 2; CSP = Casper;GN(L) = Glenrock,lower; GN(U) = Glenrock,upper.Data from Byerly et al. (2005:Table2), Hill (2001:Table3.21), and Todd (1987:Table2). bTotalMNE for "high-survivalelements"listed in Table 1. cUtilitydata summarizedin Byerly et al. (2005:Table2); TP = Total Products,TF = TotalFat, SKF = Skeletal Fat, MAR = MarrowFat, SFUI = StandardizedFood Utility Index. */?
