Precise dating of the Middle-to-Upper Paleolithic transition in Murcia (Spain) supports late Neandertal persistence in Iberia

previous item | next item

Precise dating of the Middle-to-Upper Paleolithic transition in Murcia (Spain) supports late Neandertal persistence in Iberia

Material Information

Precise dating of the Middle-to-Upper Paleolithic transition in Murcia (Spain) supports late Neandertal persistence in Iberia
Series Title:
Zilhão, João
Anesin, Daniela
Aubry, Thierry
Badal, Ernestina
Cabanes, Dan
Kehl, Martin
Klasen, Nicole
Lucena, Armando
Martín-Lerma, Ignacio
Martínez, Susana
Matias, Henrique
Susini, Davide
Steier, Peter
Wild, Eva Maria
Angelucci, Diego E.
Villaverde, Valentín
Zapatal, Josefina
Publication Date:
Physical Description:
2 online resources


Subjects / Keywords:
Neanderthals ( lcsh )
Radiocarbon dating ( lcsh )
Luminescence dating ( lcsh )
Caves ( lcsh )
serial ( sobekcm )
Europe -- Spain -- Región de Murcia


The late persistence in Southern Iberia of a Neandertal-associated Middle Paleolithic is supported by the archeological stratigraphy and the radiocarbon and luminescence dating of three newly excavated localities in the Mula basin of Murcia (Spain). At Cueva Antón, Mousterian layer I-k can be no more than 37,100 years-old. At La Boja, the basal Aurignacian can be no less than 36,500 years-old. The regional Middle-to-Upper Paleolithic transition process is thereby bounded to the first half of the 37th millennium Before Present, in agreement with evidence from Andalusia, Gibraltar and Portugal. This chronology represents a lag of minimally 3000 years with the rest of Europe, where that transition and the associated process of Neandertal/modern human admixture took place between 40,000 and 42,000 years ago. The lag implies the presence of an effective barrier to migration and diffusion across the Ebro river depression, which, based on available paleoenvironmental indicators, would at that time have represented a major biogeographical divide. In addition, (a) the Phlegraean Fields caldera explosion, which occurred 39,850 years ago, would have stalled the Neandertal/modern human admixture front because of the population sink it generated in Central and Eastern Europe, and (b) the long period of ameliorated climate that came soon after (Greenland Interstadial 8, during which forests underwent a marked expansion in Iberian regions south of 40°N) would have enhanced the “Ebro Frontier” effect. These findings have two broader paleoanthropological implications: firstly, that, below the Ebro, the archeological record made prior to 37,000 years ago must be attributed, in all its aspects and components, to the Neandertals (or their ancestors); secondly, that modern human emergence is best seen as an uneven, punctuated process during which long-lasting barriers to gene flow and cultural diffusion could have existed across rather short distances, with attendant consequences for ancient genetics and models of human population history.
Original Version:
Volume 3, Issue 11
Original Version:
51 p.
Original Version:
136 p. (supplement)

Record Information

Source Institution:
University of South Florida
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
K26-05471 ( USFLDC DOI )
k26-5471 ( USFLDC Handle )
K26-05472 ( USFLDC DOI )
k26-05472 ( USFLDC Handle )

USFLDC Membership

University of South Florida
Karst Information Portal

Postcard Information



This item is only available as the following downloads:

Full Text


PrecisedatingoftheMiddleto-UpperPaleolithictransition inMurcia(Spain)supports lateNeandertalpersistencein IberiaJoãoZilhãoa , b , c ,* ,DanielaAnesind,ThierryAubrye,ErnestinaBadalf, DanCabanesg,MartinKehlh,NicoleKlasenh,ArmandoLucenac, IgnacioMartín-Lermai,SusanaMartínezc,HenriqueMatiasc,DavideSusinid , j, PeterSteierk,EvaMariaWildk,DiegoE.Angeluccid,ValentínVillaverdef, JosefinaZapatalaInstitucióCatalanadeRecercaiEstudisAvançats(ICREA),PasseigLluísCompanys23,08010Barcelona,SpainbUniversitatdeBarcelona,Departamentd ’ HistòriaiArqueologia,FacultatdeGeografiaiHistòria,c/Montalegre6, 08001Barcelona,SpaincUNIARQ – CentrodeArqueologiadaUniversidadedeLisboa,FaculdadedeLetrasdeLisboa,Universidadede Lisboa,AlamedadaUniversidade,1600-214Lisboa,PortugaldUniversitàdegliStudidiTrento,DipartimentodiLettereeFilosofia,viaTommasoGar14,38122Trento,ItalyeParqueArqueológicodoValedoCôa,FundaçãoCôaParque,RuadoMuseu,5150-610VilaNovadeFozCôa,PortugalfUniversitatdeValència,DepartamentdePrehistòria,ArqueologiaiHistòriaAntiga,Av.BlascoIbañez28,46010 València,Spain,Av.BlascoIbañez28,46010València,SpaingDepartmentofAnthropology,RutgersUniversity,BiologicalSciencesBuilding,32BishopStreet,NewBrunswick, NJ,08901,USAhUniversityofCologne,InstituteofGeography,Albertus-Magnus-Platz,50923Cologne,GermanyiUniversidaddeMurcia,ÁreadePrehistoria,FacultaddeLetras,CampusdeLaMerced,30071Murcia,SpainjUniversitàdiSiena,DipartimentodiScienzefisiche,dellaTerraedell'Ambiente,StradaLaterina8,53100Siena,ItalykVERA(ViennaEnvironmentalResearchAccelerator)Laboratory,FacultyofPhysics – IsotopeResearchandNuclear Physics,UniversityofVienna,Währingerstra ß e17,1090Wien,AustrialUniversidaddeMurcia,ÁreadeAntropologíaFísica,FacultaddeBiología,CampusUniversitariodeEspinardo, 30100Murcia,Spain *Correspondingauthor. E-mailaddress: (J.Zilhão). Received: 12April2017 Revised: 25August2017 Accepted: 19October2017 Citeas:JoãoZilhão, DanielaAnesin, ThierryAubry, ErnestinaBadal,DanCabanes, MartinKehl,NicoleKlasen, ArmandoLucena, IgnacioMartín-Lerma, SusanaMartínez, HenriqueMatias, DavideSusini,PeterSteier, EvaMariaWild, DiegoE.Angelucci, ValentínVillaverde, JosefinaZapata.Precisedating oftheMiddle-to-Upper PaleolithictransitioninMurcia (Spain)supportslate Neandertalpersistencein Iberia. Heliyon3(2017)e00435. doi: 10.1016/j.heliyon.2017. e00435 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


AbstractThelatepersistenceinSouthernIberiaofaNeandertal-associatedMiddle Paleolithicissupportedbythearcheologicalstratigraphyandtheradiocarbon andluminescencedatingofthreenewlyexcavatedlocalitiesintheMulabasinof Murcia(Spain).AtCuevaAntón,MousterianlayerI-kcanbenomorethan37,100 years-old.AtLaBoja,thebasalAurignaciancanbenolessthan36,500years-old. TheregionalMiddle-to-UpperPaleolithictransitionprocessistherebyboundedto thefirsthalfofthe37thmillenniumBeforePresent,inagreementwithevidence fromAndalusia,GibraltarandPortugal.Thischronologyrepresentsalagof minimally3000yearswiththerestofEurope,wherethattransitionandthe associatedprocessofNeandertal/modernhumanadmixturetookplacebetween 40,000and42,000yearsago.Thelagimpliesthepresenceofaneffectivebarrierto migrationanddiffusionacrosstheEbroriverdepression,which,basedonavailable paleoenvironmentalindicators,wouldatthattimehaverepresentedamajor biogeographicaldivide.Inaddition,(a)thePhlegraeanFieldscalderaexplosion, whichoccurred39,850yearsago,wouldhavestalledtheNeandertal/modern humanadmixturefrontbecauseofthepopulationsinkitgeneratedinCentraland EasternEurope,and(b)thelongperiodofamelioratedclimatethatcamesoonafter (GreenlandInterstadial8,duringwhichforestsunderwentamarkedexpansionin Iberianregionssouthof40°N)wouldhaveenhancedthe “ EbroFrontier ” effect. Thesefindingshavetwobroaderpaleoanthropologicalimplications:firstly,that, belowtheEbro,thearcheologicalrecordmadepriorto37,000yearsagomustbe attributed,inallitsaspectsandcomponents,totheNeandertals(ortheirancestors); secondly,thatmodernhumanemergenceisbestseenasanuneven,punctuated processduringwhichlong-lastingbarrierstogeneflowandculturaldiffusioncould haveexistedacrossrathershortdistances,withattendantconsequencesforancient geneticsandmodelsofhumanpopulationhistory. Keyword:Archaeology1.IntroductionIntheAquitainebasinandthePyrenees,theMiddlePaleolithic(MP)Mousterian cultureisfollowed,insuccession,bytheChâtelperronian,theProtoaurignacian andtheAurignacianI(a.k.a.EarlyAurignacian).InIberia,theseinitialphasesof theUpperPaleolithic(UP)arerepresentedintheCantabrianstripandinCatalonia butremainunknowntotheSouthoftheEbrobasin.Basedontheseobservations, the “ EbroFrontier ” modelhypothesizesthat(a)inValencia,Murcia,Andalusia, Gibraltar,theMesetanhinterland,andPortugal,thecorrespondingchronostratigraphicslotisoccupiedbyalate-persistingMousterianand(b)thepatternis explainedbythemajorbiogeographicaldividethattheEbrobasinwouldhavebeen atthattime( Zilhão,1993 ; Zilhão,2000 ; Zilhão,2006a ; Zilhão,2009 ). ArticleNo~e00435 2 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


ThepaleontologicalandancientDNA(aDNA)evidenceindicatesthat,inEurope, extensiveadmixtureoccurredatthetimeofcontactbetweenaboriginal Neandertalsandin-dispersinggroupsofmodernhumans,resultinginthe former ’ seventualassimilation( Smithetal.,2005 ; Trinkaus,2007 ; Pääbo,2015 ). TheauthorshipoftheChâtelperronian,theProtoaurignacian,andtheothersocalled “ transitional ” industriesfromthistimeremainsdebated( Highametal., 2010 ; Caronetal.,2011 ; Hublinetal.,2012 ; TrinkausandZilhão,2013 ; Zilhão, 2013 ; Zilhãoetal.,2015 ; Welkeretal.,2016 ).InWesternEurasia,however,the MousterianisexclusivelyassociatedwiththeNeandertals,whiletheAurignacian IandthesucceedingAurignacianII(a.k.a.EvolvedAurignacian),whichextend fromAsturiasintheWesttonorthernIsraelintheEast,areassociatedwith modernhumansonly( Vernaetal.,2012 ).Inthiscontext,thebroader paleoanthropologicalsignificanceofthe “ EbroFrontier ” modelresidesinthe implicationthatNeandertalspersistedinSouthernandWesternIberialongerthan everywhereelse. Withinthemodel,thechronologicalboundariesoftheMiddlePaleolithic/ Neandertalpersistencepatternaregivenbythedifferenceinagebetweenthe earliestarcheologicalcultures(ortheirphases)that,oneachsideoftheEbro divide,areunambiguouslyassociatedwithmodernhumans:totheNorth,the AurignacianI;totheSouth,theAurignacianII.Giventhecurrentlyaccepted datingoftheseassemblagetypes( Highametal.,2011 ; Banksetal.,2013a ; Banks etal.,2013b ),thelagimplicated(i.e.,thedurationofthe “ EbroFrontier ” pattern) is,attheleast,ofthreemillennia,between40,000and37,000yearsago. ThenumberofoccurrencessubstantiatingthatIberianregionstotheSouthofthe Ebrodividewereoccupiedbyalate-persistingMousterianwhilethosetotheNorth wereoccupiedbytheAurignacianIis,however,limited.Thispaucityof occurrenceshasledtoalternativereadingsoftheevidencewherebythelate persistenceisapparent.Insuchreadings,the “ EbroFrontier ” patternwouldstem frominsufficientinformationontheearlyUpperPaleolithic,aggravatedby(a) MiddlePaleolithic-associatedradiocarbondatingresultsthatwouldbeinaccurately young,and(b)ambiguityinthedefinitionofthestonetoolassemblagesimplicated ( Woodetal.,2013 ). Conversely,ithasbeenarguedthatnoAurignacianexistsinSouthernandWestern Iberia,theirUpperPaleolithicbeginningwiththeGravettian( delaPeña,2013 ). Suchviewsimplythat(a)theMousterianpersistedevenlonger( Finlaysonetal., 2006 ; Finlaysonetal.,2008 ),or(b)afteraNeandertalextinctionevent,Southern andWesternIberiaremaineduninhabiteduntilmodernhumanreoccupation ( Bradtmölleretal.,2012 ; Galvánetal.,2014 ).Inthesescenarios,theroleof biogeographicaldivideplayedbytheEbrobasinundercertainclimaticand ArticleNo~e00435 3 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


environmentalconditionswouldnothavecontributedtoobservedpatternsinany significantmanner. Re-datingandcriticalexaminationofoldsitesandcollections( Kehletal.,2013 ; Woodetal.,2013 )haveadvancedthesedebates.Thescopeofthemanyempirical issuesinvolved,however,requirestheexcavationofnewsiteswiththepotentialto settlethekeypointsofcontention.Here,wereportontheprogressmadeinthat directionresultingfromadecadeoffieldworkinMurcia,SoutheastSpain. Whenspecificallycited,individualradiocarbonresultsaregivenasprovidedbythe datinglaboratory,i.e.,expressedinuncalibratedradiocarbonyearsBeforePresent (BP).Throughout,however,thediscussionisframedincalendarterms,i.e.,in yearsorthousandsofyears(ka)beforethetimeofmeasurementforU-seriesand luminescencedates,andincalibratedyearsorthousandsofyearsBPfor radiocarbondates.2.Results 2.1.SiteformationanddatingWeexcavatedthreelocalities < 2kmapartwithintheMulabasin( Angeluccietal., 2017 ).TheSupplementaryInformation(SI)Appendixprovidesasuccinct geographicaldescriptionofthearea,aswellasextensivemonographic presentationsofthesites ’ stratigraphicsequences,dating,humanoccupation features,andstonetoolassemblage s.Thesitesare:CuevaAntón(CA; 38°03 51.84 N,01°29 47.20 W),FincaDoñaMartina(FDM;38°04 43.21 N, 01°29 25.13 W),andAbrigodeLaBoja(ADB;38°04 43.37 N,1°29 23.17 W) ( Fig.1 ;Figs.S1.1 … S1.2). CuevaAntón(SIappendix,chapter2; Fig.2 )isacavelocatedinthevalleyof RiverMula( Zilhãoetal.,2010a ; Angeluccietal.,2013 ; Zilhãoetal.,2016 ). Sandwichedbetweenbasalpalustrinedeposits(complexFP)andwell-bedded inundationsiltsandsandsaccumulatedinrecenttimesduringperiodsof submersionbythereservoiroftheLaCiervadam(complexDD),thesitecontains athickUpperPleistocenesuccession(complexAS).Thebaseofthissuccession (sub-complexesAS2-AS5)isanalluvialfillofMIS(MarineIsotopeStage)5age thatfeaturesdiscreteanthropogeniclensesrecordingshort-livedoccupation episodes „ thelastofwhichislayerII-l.Afteranerosionalhiatus,broadly coincidentwithMIS4,theaccumulationofalluviuminsidethecave „ representedbythebasallayers(I-i,I-j,II-a,II-candII-b; Fig.2 )oftheAS1 sub-complex „ resumedbrieflyinMIS3.LayerI-k,anarcheologicallyfertile brecciamade-upofwalldegradationdebris,capstheAS1deposit,whosesurfaceis erosional.PreviousworkhasplacedthebasalMIS5alluviuminthe72 … 85kaage range( Burowetal.,2015 ; Zilhãoetal.,2016 )andtheMIS3alluviumandbreccia ArticleNo~e00435 4 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


inthe35.1 … 37.7kaagerange( Table1 ; Zilhãoetal.,2016 ).Here,thefocuslieson layerI-k ’ ssiteformationprocessandstonetoolassemblagecomposition,upon whichlieitsassignmenttotheMiddlePaleolithic. FincaDoñaMartina(SIappendix,chapter3; Fig.3 )andLaBoja(SIappendix, chapter4; Figs.4and5 )arerock-shelterslocatedintheRamblaPerea[ ( F i g . _ 1 ) T D $ F I G ] Fig.1. TheMulabasinsites.a.LocationofthelateMiddlePaleolithicsitesofSouthernandWestern IberiarelativetotheEbrobasin(1.CuevaAntón;2.SimadelasPalomas;3.Gorham ’ sCave;4.Gruta daOliveira;5.FozdoEnxarrique).b.LocationoftheMulabasinsitesina2013orthophoto. Source: ;CA,CuevaAntón;FDM,FincaDoñaMartina;ADB, AbrigodeLaBoja);asthecrowflies,thedistancebetwe enCuevaAntónandtheRamblaPerearock-sheltersis 1670m.c.TheRamblaPerearock-sheltersfromupstream(2009).d.ThetailoftheLaCiervareservoir,with CuevaAntónseenfromNortheast(2007),after( Zilhãoetal.,2016 ),withpermissionfromElsevier.e.LaBoja attheendofthe2016fieldseason;theredlinesinthee xcavationgriddenotethere ferencecross-sectionsin Fig.4.f.FincaDoñaMartina ’ sexcavationtrenchattheendofthe2016fieldseason. ArticleNo~e00435 5 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


( Zilhãoetal.,2010b ; Lucenaetal.,2012 ).Intheregionallandscape,thistributary ofRiverMulalikewisecommunicatesthelowlandsoftheMurcialittoralwiththe plateausandmountainrangesextendingnorthwardtotheMesetanhinterland.Both sitesfeaturestratigraphicsuccessionswhereabasalMiddlePaleolithicisoverlain bylongUpperPaleolithicsequences.Thepreservationisgoodforshellbutpoorto-nilforbone,andcharcoalisabundant „ eventhough,atFincaDoñaMartina,[ ( F i g . _ 2 ) T D $ F I G ] Fig.2. CuevaAntón.a.Siteplanandexcavationgrid.b.Cross-sectionillustratingthepositionoflayer I-k „ sandwichedbetweentheDDreservoir-inundationsiltsandthebasalalluviumofsub-complex AS1(hererepresentedbylayersI-i,I-jandII-a).c.ViewfromtheWestattheendofthe2011field season;thelayerlabelsdesignatetheunitswhosesurfaceisexposedineachsector.d.Viewfromthe Eastattheendofthe2012fieldseason.Elevationsareinmasl.Figs.2a,2cand2dafter( Zilhãoetal., 2016 ),withpermissionfromElsevier. ArticleNo~e00435 6 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


chemicallyweathered(leadingtoradiocarbonresultsthatareminimumagesonly; TablesS3.1-S3.2). Layer8ofFincaDoñaMartinayieldedalithicassemblagewhoseAurignacian affinities(Figs.S3.31-S3.32)areconsistentwiththelayer ’ sstratigraphicposition betweenMousterianlayer9andGravettianlayers7band6/7.AtLaBoja,the excellentpreservationofcharcoalandthesub-centimeterprecisionwithwhich mostarcheo-stratigraphicunits „ designatedOH(OccupationHorizons) „ could beseparatedprovidedforalargeseriesofradiocarbonresultsthat,aburrow sampleexcepted,areinfullstratigraphicorder( Table2 ;TableS4.1;Fig.S4.8). ThebasalMousteriandatesbeyond44kaandisburiedunderathick,multi-ton, roof-collapsedslab.Thesitewasre-occupied,intheAurignacian,oncethisslab wascoveredbytheaccumulationofthesedimentformingtheIL(Intermediate Level)4unit.Otherwisearcheologicallysterile,IL4includessomepostdepositionallyintrudedmaterialandyieldedadateofca.41ka.Thisdateprovides a terminuspostquem fortheca.75cm-thickAurignaciansequence,whichissealed byanotherlarge,roof-collapsedslab.Radiocarbondatingplacesthethreebasal Aurignacianhorizons(OH18-OH20)withinthe34.9 … 38.2kaintervalandthethree upperones(OH15-OH17)withinthe33.9 … 35.6kainterval. SedimentsamplesfromtheMousterian(OH21-OH23)andtheAurignacian (OH17-OH18)ofLaBojawerealsodatedbyOpticallyStimulatedLuminescence (OSL)( Table3 ;Figs.6 … 8;Fig.S4.9).Themultiple-graindatingofthequartzand feldsparmineralsplacesthesequencebetween32.6±1.9ka(C-L3906),forOH17, and59.9±6.8ka(C-L3901),forthebaseofthedeposit,belowOH23.These luminescenceagesareincompleteagreementwiththeradiocarbonresultsforthe correspondingAurignacianandMousterianhorizons. TheagesoftheLateMousterianinlayerI-kofCuevaAntónandoftheEvolved AurignacianinOH18-OH20ofLaBojaoverlap( Fig.9 ).Astheoccupationevents recordedatthesesitesareofshortduration,apossibleinterpretationofthispattern isthatthetwoassemblagetypescoexistedintheregionforanextendedperiod, Table1. CuevaAntón.ABOx-SCradiocarbondatingresultsforsub-complexAS1(after Zilhãoetal., 2016 ).TheageshavebeencalibratedagainstIntCal13( Reimeretal.,2013 )inCalib7.0.4( Stuiverand Reimer,1993 );thecalibratedagesaregivenas95.4%probabilityintervals.SampleTaxonFieldunitLayerOxA 13C[ ‹ ]Yield(mg)%Yld%CAgeBPAgecalBP I20-3ConiferI-kI-ktop26346±27035067 … 36245 G21-4 Juniperus sp.dec4I-kbase22625 21.08.6a8.7a77.932330±25035627 … 36826 E21-11 Juniperus sp.dec5aII-a22019 22.76.436.075.632390±28035594 … 37055 J19-7 Pinus sp.I-k/II-dII-b21244 22.311.7a12.1a88.432890±20036314 … 37714aThesevaluesareestimatedasonlyapproximatelyhalfofthesampleremainingafterthewetchemistrywaspre-combusted. ArticleNo~e00435 7 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


duringwhichtheirmakerswouldhavemadeinfrequent,alternatingincursionsinto theRiverMulaandRamblaPereavalleys.Ifso,MiddlePaleolithicmaterialought toexistwithinthebasalAurignacianofLaBojaas(a)discrete,interstratified lenses,or(b)isolatedelementsmixedintheOH18-OH20assemblages.Asneither isthecase,theregionalcontemporaneitybetweenthebearersofthetwokindsof stonetooltechnologiesmusthavebeenshort-lived.Therefore,thedatingoverlap mustprimarilyreflectthestatisticaluncertaintyinherenttoradiometricdating.[ ( F i g . _ 3 ) T D $ F I G ] Fig.3. FincaDoñaMartina.a.3Dmodeloftheaccumulation(foranextendeddiscussion,seetheSI appendix);thelabelsdenotethedifferentstratigraphicunitsrecognized.b.Thestratigraphicsuccession inthetrench ’ swesternwall.Elevationsareinmasl. ArticleNo~e00435 8 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Underthesepriors,CA/I-kandADB/OH18-OH20canbetreatedastwo consecutivephasesoftheregionalchrono-stratigraphicsequence. WhetherthecharcoalfoundinlayerI-kofCuevaAntónisanthropogenic, environmentallyaccumulated,orboth,cannotbeascertained.However,thebasal AS1alluviumconsistsoflensesoffine,sandy-siltyalluviumdepositedinquick successionduringlow-energyinundationevents;suchkindsofeventsarealso largelyresponsibleforthematrixoftheI-kbreccia( Angeluccietal.,2013 ).This record ’ sresolutionimpliesthatanytemporaldifferencethatmayhaveexisted betweenhumanoccupationandcharc oaldepositionmustbenegligible. Nevertheless,tobeconservative,theageoftheLateMousterianinlayerI-kis bestconstrainedusingthe terminuspostquem representedbytheunderlyingunits, layersII-aandII-b. ThatlayersII-aandII-bprovideindeedarobustmaximumageforthehuman occupationoflayerI-kisintimatedbythearcheologicalsterilityofthebasalAS1 alluvium,towhichthosetwolayersbelong.Suchsterilityprecludesinterpreting theartefactassemblageinoverlyinglayerI-kasinheritedviasomesortoflocal[ ( F i g . _ 4 ) T D $ F I G ] Fig.4. LaBoja.Thearcheo-stratigraphicsequence.Trenchcross-sectionsasrecordedattheendofthe 2013fieldseason(foranextendeddiscussion,seetheSIappendix).Elevationsareincmbelowdatum. ArticleNo~e00435 9 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


post-depositionalprocess.Inaddition,(a)thestratigraphicintegrityoftheAS1 packageisaccreditedbytheabsenceofdisturbancefeaturesacrossitstotal thicknessandentireexcavatedextent,and(b)themodeofaccumulationoflayer I-kimpliesthatitsartefactcontentcannothavebeeninheritedviafluvialtransport fromanearlierMiddlePaleolithicsitelocatedelsewhereinthelandscape.The stonetoolrefits( Fig.10 ;Fig.S2.18),whichdocumenton-siteproduction,[ ( F i g . _ 5 ) T D $ F I G ] Fig.5. Thebasal,MousterianandAurignaciansectionsoftheLaBojasequence.Elevationsareincm belowdatum.a.TheOH19doublehearthingridunitT3atexposureofthefeature ’ stop(above, orthorectifiedplanview)andbase(below,obliqueviewfromtheoppositeangle).b.Orthorectifiedplan viewoftheOH19hearthingridunitU4;theprovenienceofthesamplethatestablishedthishorizon ’ s radiocarbonageisindicatedbythereddiamond.c.Stratigraphiccross-sectionsrepresentingthebasal partsofthesequenceextantattheendofthe2014fieldseason;thepreservationofintacthearthsand/or extensivelensesofanthropizedsedimentallowssub-centimeterdiscriminationofoccupationfloors (OH)separatedbyintermediatelevels(IL);thelatteraresterileoronlycontainpost-depositionally intrudeditems(OH21-23areMousterian,OH15-20areAurignacian,OH13-OH14areEarly Gravettian). ArticleNo~e00435 10 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Table2. LaBoja.Radiocarbondatingresults.CalibrationusedCalib7.0.4againstIntCal13( StuiverandReimer,1993;Reimeretal.,2013 ).Unless otherwisestated,sampleswereABA-pretreated.TheVERAlab 13CvaluesweredeterminedforthegraphitizedsampleswiththeAMSsystem.See TableS4.1foradditionaldetail.HorizonSampleLab#AgeBPAgecalBP(2 ) 13C[ ‹ ]Observations burrow2008-775OxA-201166959±337694 … 7918 23.72 Oleaeuropaea OH12010-27VERA-536312605±45 … 21.2±1.1 Juniperus sp. VERA-5363_212585±40 … 20.5±1.1repeat VERA-5363_av12594±3014745 … 15136 … average OH1/OH22008-774VERA-5212a12965±4015295 … 15706 21.4±0.7 Pinusnigra OH32013-868VERA-593713290±4015793 … 16156 24.9±1.5 Pinusnigra/sylvestris OH42014-846VERA-608015390±50 … 20.3±1.5 Juniperus sp. VERA-6080ABOx15320±45 … 19.3±1.2ABOx,nosteppedcombustion VERA-6080_av15351±3318522 … 18740 … average OH52012-385VERA-578816580±7019755 … 20228 20.5±0.9 Juniperus sp. OH62010-183VERA-5364a16990±7020255 … 20704 19.5±0.5 Juniperus sp. VERA-5364b17430±7020801 … 21310 15.1±0.7 Juniperus sp. OH72010-225VERA-536519390±100 … 20.9±0.6 Juniperus sp. VERA-5365_219240±90 … 19.0±0.9repeat VERA-5365_av19307±6722996 … 23509 … average OH92014-1270VERA-608120440±90 … 19.2±1.6 Juniperus sp. VERA-6081ABOx20350±90 … 21.8±1.0ABOx,nosteppedcombustion VERA-6081_av20395±6424252 … 24840 …average 2012-1522VERA-585020580±10024434 … 25155 22.0±0.9 Juniperus sp. OH102010-316VERA-536620980±12025031 … 25617 21.5±0.6 Juniperus sp. VERA-5366_220830±110 … 22.0±0.5repeat VERA-5366_av20898±81 …… average( Continued ) ArticleNo~e00435 11 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Table2 .( Continued )HorizonSampleLab#AgeBPAgecalBP(2 ) 13C[ ‹ ]Observations VERA-5366HS20640±110 … 20.9±0.6humicacids OH112008-760VERA-521320980±11024976 … 25511 25.4±0.9 Juniperus sp. VERA-5213HS21060±110 … 22.7±0.5humicacids 2014-2578VERA-615220754±10524577 … 25343 20.9±0.9 Juniperus sp. VERA-6152HS20457±105 … 21.3±1.1humicacids burrow2012-178VERA-585120610±110 … 23.7±1.0 Juniperus sp. VERA-5851_220720±100 … 19.5±3.7repeat VERA-5851_av20670±7424551 … 25215 … average OH122012-175VERA-585223530±15027434 … 27899 23.7±1.0 Juniperus sp. VERA-5852HS21870±130 … 19.6±1.2humicacids OH132012-622VERA-578927260±23030895 … 31483 21.9±0.8 Juniperus sp. VERA-5789HS26760±230 … 21.8±0.7humicacids OH152014-2903VERA-615330548/+363/ 34733891 … 35137 20.3±1.8 Juniperus sp. OH162014-3046VERA-615430686/+355/ 34033989 … 35289 22.9±1.4 Juniperus sp. OH172012-1518VERA-5853HS29300/+300/ 290 … 21.0±1.4humicacids 2014-3129VERA-6155HS29230/+298/ 287 … 17.7±1.7humicacids 2014-3184VERA-615630918/+359/ 34334165 … 35561 26.8±1.6 Juniperus sp. OH182012-1352VERA-585432080/+420/ 40034948 … 37011 20.9±1.0 Juniperus sp. VERA-5854HS30090/+320/ 310 … 23.2±1.2humicacids OH192014-3348VERA-615733290/+494/ 466 … 22.4±1.6 Juniperus sp. VERA-6157ABOxSC33179/+482/ 455 … 23.2±1.4ABOx,steppedcombustion VERA-6157_av33233±33536491 … 38396 … average 2014-3421VERA-6158HS32331/+439/ 417 … 26.1±1.9 Juniperus sp. OH202012-1382VERA-585532890/+430/ 410 … 22.6±1.4 Juniperus sp.( Continued ) ArticleNo~e00435 12 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Table2 .( Continued )HorizonSampleLab#AgeBPAgecalBP(2 ) 13C[ ‹ ]Observations VERA-5855ABOxSC33170/+470/ 450 … 24.4±2.2ABOx,steppedcombustion VERA-5855_av33017±31036321 … 38191 … average VERA-5855HS31490/+370/ 350 … 23.5±1.2humicacids IL42012-1481VERA-585637160/+680/ 620 … 25.9±1.4 Juniperus sp. VERA-5856ABOxSC37154/+710/ 660 … 19.6±1.5ABOx,steppedcombustion VERA-5856_av37157±47240794 … 42356 … average VERA-5856HS31960/+670/ 620 … 22.2±1.2humicacids OH222013-384VERA-589946500/+2400/ 1800beyondcurve 24.1±4.8 Pinusnigra/sylvestris VERA-5899HS40820/+1090/ 960 … 24.5±1.3humicacids 2013-330VERA-590046900/+2400/ 1800beyondcurve 21.1±2.9 Pinusnigra/sylvestris VERA-5900HS45700/+2100/ 1700 … 26.9±1.8humicacids OH232013-258VERA-590143300/+1600/ 130044181 … 49611 23.3±1.5 Juniperus sp. VERA-5901HS46200/+2200/ 1700 … 19.7±1.2humicacids 2013-361VERA-5902HS42800/+1400/ 1200 … 21.4±3.1 Pinusnigra/sylvestris; humicacids ArticleNo~e00435 13 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Table3. LaBoja.Doseratedata,equivalentdosevaluesandluminescenceages.Thecosmicdosewascalculatedafter PrescottandHutton(1994) ;the conversionfactorsof Guérinetal.(2011 andanassumedwatercontentof5±2%wereused.Theinternalbetadoseratecontributionofthefeldsparsamples wascalculatedbyassumingapotassiumcontentof12.5±0.5%,after HuntleyandBaril(1997) ,andana-valueof0.12±0.02.LabcodeMineralGrainsize( m)Accepted/measuredaliquots(N)U(ppm)Th(ppm)K(%)Doserate(Gy/ka)RSD(%)AgemodelDe(Gy)Age(ka) SampleLBJ6(2.3mbelowsurfaceofcross-section);OH17 C-L3906Quartz100 … 15055/563.14±0.161.71±0.150.37±0.011.35±0.0416AM43.9±2.332.6±1.9 SampleLBJ5(2.5mbelowsurfaceofcross-section);OH18 C-L3905Quartz100 … 15039/403.09±0.161.53±0.140.33±0.011.28±0.0430AM45.9±3.235.8±2.8 K-FIR50100 … 20025/253.09±0.161.53±0.140.33±0.012.02±0.2132AM51.1±3.933.7±4.0 K-FpIRIR290100 … 20025/253.09±0.161.53±0.140.33±0.012.02±0.2132AM91.8±7.545.4±5.6 MAM75.5±7.537.4±5.3 SampleLBJ4(3.7mbelowsurfaceofcross-section);OH21 C-L3904Quartz100 … 15040/453.54±0.181.44±0.120.30±0.011.33±0.0424AM68.4±5.651.5±4.5 K-FIR50100 … 20012/123.54±0.181.44±0.120.30±0.012.11±0.2013AM65.9±4.140.9±5.7 K-FpIRIR290100 … 20021/213.54±0.181.44±0.120.30±0.012.16±0.2126AM131.2±10.060.9±7.4 SampleLBJ3(3.9mbelowsurfaceofcross-section);OH22 C-L3903Quartz100 … 15031/323.39±0.181.51±0.140.30±0.011.30±0.0451AM46.7±4.936.0±3.9 SampleLBJ2(4.1mbelowsurfaceofcross-section);OH23 C-L3902Quartz100 … 150103/1313.36±0.171.61±0.130.32±0.011.31±0.0447AM64.6±4.449.3±3.7 K-FIR50100 … 20013/133.36±0.171.61±0.130.32±0.012.09±0.2020AM59.9±4.541.4±6.1 K-FpIRIR290100 … 20015/153.36±0.171.61±0.130.32±0.012.03±0.2014AM128.7±7.460.3±6.7 SampleLBJ1(4.1mbelowsurfaceofcross-section);basal C-L3901Quartz100 … 15019/203.55±0.191.58±0.140.30±0.011.33±0.0420AM80.6±6.657.7±3.2 K-FIR50100 … 20013/133.55±0.191.58±0.140.30±0.012.11±0.206AM75.5±4.053.7±6.6 K-FpIRIR290100 … 20015/153.55±0.191.58±0.140.30±0.012.16±0.2115AM129.6±8.159.9±6.8 F=feldspar;K=Potassium;Th=Thorium;U=Uranium;AM=ArithmeticMean;De=equivalentdose;IR50=infraredstimulatedluminescencesignalat50°C;MAM=MinimumAge Model;pIRIR290=post-infraredinfraredstimulatedluminescencesignalat290°C;RSD=relativestandarddeviation. ArticleNo~e00435 14 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


corroboratethehomogeneity,integrity,andinsitunatureofboththeartefact assemblageanditsstratigraphiccontext.Therecanbenodoubt,therefore,that,at CuevaAntón,thehumanactivityrecordedinlayerI-kpost-datesthetimeof depositionoflayersII-aandII-b. AtLaBoja,theageofthesuccessful,hearth-collectedsamplefromOH19(20143348;33,233±335BP,VERA-6157_av)isstatisticallyindistinguishablefrom thatobtainedforimmediatelyunderlyingOH20andrepresentsadirectrecordof humanactivity.OH19andOH20bothcontaindiagnosticallyUpperPaleolithic, specificallyAurignacian,tool-kits.Thus,theirdatingsetsanunambiguous terminusantequem fortheendoftheregion ’ slatestMiddlePaleolithic. Underthisreasoning,theearliestpossibleageofCuevaAntón ’ slatestMousterian is37.1ka,andtheyoungestpossibleageofLaBoja ’ sAurignacianis36.5ka,in calendaryears.Theyellowbandin Fig.9 representstheintervalboundedbythese dates.Itwaswithinthisintervalthat,afteracoexistenceandinteractionperiodof unknownduration,theregion ’ sNeandertal-associatedLateMousterianwas replacedbythemodernhuman-associatedEvolvedAurignacian.[ ( F i g . _ 6 ) T D $ F I G ] Fig.6. LaBojaOSLdating.Representativeequivalentdosedistributionsofthedatedquartzand feldsparsamples.Thedistributions,displayedasabanicoplots( Dietzeetal.,2016 ),whichcombinea scatterplotwithakerneldensityestimate,areforsampleC-L3901,takenatthebaseofthesequence, immediatelybelowOH23.Thedashedlineisthearithmeticmeanequivalentdose.Theplotswere generatedusingRLuminescencepackageversion0.7.3( DietzeandKreutzer,2017 ).a.quartz.b. feldspar(IR50).c.feldspar(pIRIR290). ArticleNo~e00435 15 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


2.2.CompositionoftheartefactassemblagesJaramaVI,acavesiteintheIberianhinterlandoncethoughttospantheMP-UP transition,illustrateswellhowissuesofdefinitionareasmuchimplicatedinthe Neandertalpersistencedebateasthoseofdatingaccuracyandsampleassociation:[ ( F i g . _ 7 ) T D $ F I G ] Fig.7. LaBojaOSLdating.Analyticaldata.a.Representativequartzdoseresponseanddecaycurve forsampleC-L3905.b.Preheatplateautestsindicatingthattheequivalentdoseisindependentfrom temperaturetreatmentbetween:180and240°C(C-L3901,square);220and280°C(C-L3904,circle); 180and280°C(C-L3905,triangle);240and280°C(C-L3906,invertedtriangle).c.Doserecovery testsshowingthatalaboratorygivendosewasbestrecoveredusingatemperatureof180°Cforsamples C-L3901andC-L3905andof260°CforsamplesC-L3904andC-L3906.d.PriorIRstimulation temperaturetestscarriedoutforfeldsparsampleC-L3905indicatingaplateaubetween80and180°C; 80°Cwaschosenasprior-IRstimulationtemperature.e.RepresentativefeldsparpIRIR290dose responseanddecaycurvesofsampleC-L3905.f.DosedistributionoffeldsparsampleC-L3905 displayedasabanicoplot;thedashedlineistheMAMequivalentdose. ArticleNo~e00435 16 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


uponcloserexamination,the “ UpperPaleolithic ” stonetoolsretrievedinthelevels cappingthesite ’ sPleistocenesuccessionturnedouttobeofMousterianaffinities instead( Kehletal.,2013 ).Clearly,therobustnessoftheMulabasin ’ schronology alsodependsonwhethertheartefactassemblagesassociatedwiththedated samplesdorepresentthetwosidesoftheregionaltransition. Figs.10and11 illustratethekeyaspectsoflithictechnologysupportingour assignments:methodofcorereduction,andtypeofblankthatproductionis designedfor. InlayerI-kofCuevaAntón,thefollowingmethods,whichareexclusivetothe MiddlePaleolithic,arefound(Figs.S2.17-S2.19):Centripetal,Levalloisor Discoid,corereduction,representedbyacore,refittedflakes,anddebris;Discoid, representedbyimportedcore-trimming,ordeliberatelyovershot,naturallybacked flakesbearingnotchedordenticulatededges;Kombewa,representedbyacore discardedinaninitialstageofthereduction;andLevallois,representedbyan importedlaminarflake. InLaBojaOH18-OH20,onlytwomethods,bothunknownintheregionalMiddle Paleolithic,arefound(Figs.S4.39-S4.43):prismaticfortheextractionofblades andbladelets,representedbycores,débitage,andrefittedsets;andcarinated/nosed “ scraper ” reduction,alsoincludingrefittedsetsandrepresentedbyallstepsofthe[ ( F i g . _ 8 ) T D $ F I G ] Fig.8. LaBojaOSLdating.Age(±1 )vsdepthplotofluminescencedates.Filledsymbols:quartz OSLresults.Opensymbols:feldsparIR50results.Half-opensymbols:feldsparpIRIR290results. ArticleNo~e00435 17 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


sequence(initiallargecoreforlong-and-thickbladesusedasblanksforthe extractionoftheintendedbladelets,theabandonedbladeletcores,thebladelets themselves,andthewasteproducedasthe “ scraper ” frontwasreduced,trimmed andreconfigured).TheDufourbladeletin Fig.10 isatypicalexampleoftheRocde-Combesubtype,anindexfossiloftheEvolvedAurignacian.Itcomesfrom OH17,butthisandothersubtypesofDufourbladeletsoccurthroughtheOH15OH20sequence(Figs.S4.41-S4.43).Theyarealsopresent,alongsidethe[ ( F i g . _ 9 ) T D $ F I G ] Fig.9. ChronologyoftheMiddle-to-UpperPaleolithictransitionintheMulabasinsites.Plotof calibratedradiocarbondates(95.4%probabilityintervals)fortheAurignacianofLaBojaandforthe Mousterian(layerI-k)andimmediatelyunderlyingalluvium(layersII-aandII-b)ofCuevaAntón.The verticalyellowbanddenotestheintervalduringwhichthetransitiontookplace:between36.5ka,the youngestpossibleageofLaBoja ’ sAurignacianinOH19-20,and37.1ka,theoldestpossibleageofthe CuevaAntónMousterianasprovidedbythelayerII-a terminuspostquem .Thecomparisonwiththe globalproxies( Rasmussenetal.,2014;Sánchez-Goñietal.,2008,2013 )showsthat,intheMulabasin, thetransitioncoincideswiththeendofalongandmildtemperatephase,GreenlandInterstadial8. ArticleNo~e00435 18 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


characteristiccarinated/nosed “ scrapers ” /cores,inlayer8ofFincaDoñaMartina (Figs.S3.31-S3.32).InOH15andOH16ofLaBoja,backedmicroliths(Fig.S4.43, nos.4 … 5)appearforthefirsttimealongsidethesecharacteristicAurignacianitems, suggestingthattheemergenceofthesucceedingGravettianlikelycorrespondstoa technologicaltransitionwithnomajordiscontinuityinpopulation,demography,or settlement. Well-stratifiedPortugueseexamplesshowthatspecializedsiteoccupancymay generatelithicassemblagesthat,despitetheirUpperPaleolithicage,lackthe period ’ sdiagnostics.ThisevidencequestionsautomaticassignmenttotheMiddle Paleolithicofsimilarassemblages,themoresoiftheyaresmall( Woodetal., 2013 ).However,unlikelayerI-kofCuevaAntón,thosePortugueseassemblages alsolackMiddlePaleolithicdiagnostics:theycontainnoitems(eithercoresor[ ( F i g . _ 1 0 ) T D $ F I G ] Fig.10. BlankproductionanddiagnosticstonetoolsacrosstheMiddle-to-UpperPaleolithictransition inthetheMulabasinsites.a.Centripetalcoreforsmallflakes,withrefits(CuevaAntón,layerI-k, Mousterian).b.Multi-stepreductionsequencefortheproductionofbladelets(LaBoja,OH20, Aurignacian):preparation(1)orre-preparation(1 )ofaprismaticcorefortheextractionoflong,thick blades(2),followedbypreparationofsuchlaminarblanksascarinatedornosed “ scrapers ” (3), extractionofbladeletsfromthe “ scraperfront ” (4),andeventualdiscardoftheexhausted “ scraper ” /core (5);thebluecirclesdenotestepsrepresentedintherefit,thewhitecirclesdenotestepsrepresentedby removalscarsoramongtheblock ’ sunrefittedmaterial.c.longbladewithminor,proximalbreak(La Boja,OH20,Aurignacian).d.LaminarLevalloisflake,representingalateralremovalafterthe extractionofapreferentialflakeinaLevalloisrecurrentreductionsequence(CuevaAntón,layerI-k, Mousterian).e.CharacteristicallytwistedDufourbladeletoftheRoc-de-Combesubtypeextractedfrom acarinatedornosed “ scraper ” /core(LaBoja,OH17,Aurignacian). ArticleNo~e00435 19 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


[ ( F i g . _ 1 1 ) T D $ F I G ] Fig.11. CorereductionmethodsacrosstheMiddle-to-UpperPaleolithictransitionintheMulabasin sites.a.Simplified,schematicrenditionoftheapproachtocorereductionrepresentedbytherefitted materialfromMousterianlayerI-kofCuevaAntón( Fig.10 a);therefittingunitdocumentsthe endpoint,priortodiscard,ofthecentripetalproductionofsmallflakesfromacorepreviouslyexploited ArticleNo~e00435 20 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


blanks)indicatingthattheDiscoid,LevalloisandKombewareductionmethods wereinuseatthetimeofproduction.Acaseinpointistheassemblagefromthe EE15occupationsurfaceoftheLagarVelhorock-shelter(N=593) ( Almeidaetal.,2009 ).Here,theidiosyncrasyrelatestothesituationalcontext (reductionofimmediatelyavailablequartzitecobblesfortheexpedientproduction ofcuttingedgesusedincarcass-processingtasks),andisofnowiderchronostratigraphicconsequence. Themutuallyexclusivepresence/absenceofdiagnostictechnologiesintheMula basinsitesstandsdespitedifferencesinassemblagesizeofuptotwoordersof magnitude,andisconsistentlyseenacrosstime( Table4 ).Inthisregard,theLate MousterianinlayerI-kofCuevaAntónisnodifferentfromtheMiddlePaleolithic assemblageofMIS5agefoundinthesite ’ slayerII-l(TablesS2.2 … S2.5). Likewise,theequivalentlysmallsizeoftheEarlyGravettianassemblagesin OH13-OH14ofLaBoja(TablesS4.22 … S4.25)isnoimpedimentfortheirfully UpperPaleolithicnaturetomanifestitselfthroughsuchdiagnosticsasbladelets extractedfrombothprismaticand “ burin ” core-types,the “ burins ” themselves,and eventhetechnocomplex ’ sindexfossil(amicrogravettepoint).Muchthesame appliestoLaBoja ’ sAurignacianassemblages(TablesS4.10 … S4.21).AtFinca DoñaMartina,thelowerresolutionofthestratigraphicsequencemeansthateach unitsamples,andaveragesout,muchlongertimeintervals.Yet,itremainsthat(a) LevalloisandDiscoidcoresandblanks,sidescrapers,anddenticulatesarefound togetherinthissite ’ sbasallayer9(TablesS3.3 … S3.5,Figs.S3.29 … S3.30)butnot inoverlyinglayers8,7band6/7,while(b)thereverseistrueofprismatic, carinated/nosed “ scraper ” and “ burin ” core-types,endscrapers,orbladelettools (TablesS3.7 … S3.12;Figs.S3.31 … S3.33). Thevariationinthesizeandcompositionoftheseassemblagesisprimarilydueto localfactors.AtCuevaAntón,thepatchesofdrysedimentavailableforsettlement insidethecaveduringthetimeofformationoflayersII-landI-kwererestricted andsurroundedbyinundatedorboggyriversideterrain(Figs.S2.11,S2.16).As shownbythetaphonomyoftheabundantrabbitbone,thesitefunctionedasan eagle-owlroostthroughout,whichisinconsistentwithfrequentorintensivehuman presence( Sanchis,2012 ; Zilhãoetal.,2016 ).Likewise,thespatialrestrictionsto habitationcausedbyamassiveroofcollapseexplainthesmallsizeoftheartefact scatteraroundthehearthinLaBoja ’ sOH13horizon(Fig.S4.21). forsimilarblanksandinsimilarmanner(asindicatedbytheshapeandradialpatterningoftheflaking scars).b.Simplified,schematicrenditionofthecorereductionmethodsrepresentedintheEvolved Aurignacian(OH20)ofLaBoja( Fig.10 b-c);twotypesofbladesareextractedfromprismaticcores „ thin,tobeusedasatoolorasablankforaretouchedtool,andthick,tobeusedasablankforbladelet coresofthecarinatedornosedkind;thus,thelatter ’ sintendedend-productsarebladeletsobtained separately,notattheendofacontinuous,blade-then-bladeletcorereductionsequence. ArticleNo~e00435 21 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Thespectrumofactivitiesreflectedintheuse-weardataforlayerI-kofCueva Antónislimitedtowood-working(TableS2.5;Fig.S2.19),whichisinkeeping withthehighlytransientnatureoftheoccupation(s).IntheRamblaPereasites, raw-materialeconomypatternsindicatenosignificantchangeinsitefunction acrossthetransition.Intheresidentialversuslogisticalbalanceofhunter-gatherer settlement-subsistencesystems „ asgaugedbytherelativeimportanceof domestic-versushunting-relatedstonetools „ thescalesweresomewhattippedin favorofthelatterintheEarlyGravettianandtheAurignacianofFincaDoña Martina,butnotintheAurignacianofLaBoja(SIappendix,chapters3 … 4). FortheRamblaPerearock-shelters,lateralvariationbetweentwoadjacent archeologicalsitesthat,inthelivingpast,musthavefunctionedasasingle, spatiallyextensivelocusofhumanactivity,sufficestoexplainthecontrasts Table4. CuevaAntónandLaBojastonetools.Assemblagesizeversus representationofthediagnosticlithics.CAADBaCategoriesbDiagnosticsII-lI-kOH20OH19OH18OH17OH16OH15OH14OH13 Cores MPKombewa … 1 …………………… centripetal11 …………………… UPcarinated/nosed …… 2212 ………… burin ………… 12 … 1 … 1 prismatic …… 42361112 … Unretouchedblanks MPKombewa1 ……………………… Levallois11 …………………… UPblades …… 18 …… 278 …… bladelets …… 376942511863 Formaltools MPsidescrapers6 ……………………… denticulates … 1 …………………… UPendscrapers …… 1 … 1 … 1 ……… bladelettools …… 32111471 … Totalc26141795969285371772214 Totald342045314620292315432318235aOH15-OH20,Aurignacian,OH13-OH14,EarlyGravettian(IL4andIL3itemscountedunderOH20 andOH13,respectively).bMP=MiddlePaleolithicdiagnostics;UP=UpperPaleolithicdiagnostics.cDebris(chippageandchunks),manuportsandhammerstonesexcluded.dDebrisincluded. ArticleNo~e00435 22 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


betweencoevallithicassemblages.Throughtime,acrosstheregionalMP-UP transition,theuse-wearevidenceshowsthatthedifferencesareprimarilyofa techno-typologicalnature.Hide-working,wood-working,defleshingandtheuseof projectilesaredocumentedinboththeMousterianandtheAurignacian (TablesS3.6,S4.9; Figs.12 … 14 ;Figs.S3.30 … S3.32,S4.37 … S4.38,S4.41). However,(a)hideswereprocessedwithsidescrapersintheMousterianbutwith[ ( F i g . _ 1 2 ) T D $ F I G ] Fig.12. MiddlePaleolithicwood-workingtoolsintheMulabasinsites.a.DenticulatefromCueva Antón(layerI-k).b.UnretouchedblankfromLaBoja(OH23).c.DenticulatefromLaBoja(OH23). Theinsetsshowcharacteristicmicroscopicpolish.Notethesimilarityofthetwodenticulates,both madeonorange-segmentordiscoid-overshotblanks;denticulatesofthiskindareentirelyabsentfrom toptobottomofthelongandcompleteUpperPaleolithicsequencesofLaBojaandFincaDoñaMartina (foradditionaldetail,seetheSIappendix). ArticleNo~e00435 23 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


endscrapersintheAurignacian,and(b)projectileswerearmedwithsingle,axiallymountedpointsintheMousterianbutwithmultiple,laterally-mountedmicrolithic elementsintheAurignacian.Inshort,synchronicfunctionalvariabilitycannot explainthedifferencesinlithictechnologyuponwhichwehaveassignedthestone toolassemblagesoftheMulabasinsitestoeithertheMiddleortheUpper Paleolithic. Ochreisofteninvolvedintheprocessingofhides,asdocumentedbyresidueona MousteriansidescraperfromFincaDoñaMartina(Fig.S3.30).Nosuchresidues werefoundinthelithicsfromlayerI-kofCuevaAntón.Thus,thepigmentcoverof[ ( F i g . _ 1 3 ) T D $ F I G ] Fig.13. Hide-workingtoolsacrosstheMiddle-to-UpperPaleolithictransitionatFincaDoñaMartina.a. EndscraperfromAurignacianlayer8.b.SidescraperfromMousterianlayer9.Theinsetsshow characteristicmicroscopicpolish(foradditionaldetail,seetheSIappendix). ArticleNo~e00435 24 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


theassociatedscallopshell( Fig.15 ;Fig.S2.20)cannotrepresentaccidentalor post-depositionalstainingbyironoxidesbroughtinforhide-processingtasksor locallyproducedbydiageneticprocesses.Muchthesameappliestotheornamental shellassemblageofquitedistinctcompositionfoundintheAurignacianofLaBoja (TableS4.2; Fig.15 ;Figs.S4.32,S4.34).Thisassemblagefeaturesubiquitousred ochrestainingeventhoughnonewasfoundinthe78stonetoolsfromOH15-OH20[ ( F i g . _ 1 4 ) T D $ F I G ] Fig.14. ProjectiletechnologyacrosstheMiddle-to-UpperPaleolithictransitionintheMulabasinsites. AxialpointsintheMousterian,compositepointsarmedwithcutting,laterallymounted,microlithic elementsintheAurignacian.a.MousterianpointfromFincaDoñaMartina(layer9).b.marginally backedbladeletfromLaBoja(OH16).c.DufourbladeletfromFincaDoñaMartina(layer8).Theinsets showcharacteristicmicroscopicstriationsgeneratedbyimpact(foradditionaldetail,seetheSI appendix). ArticleNo~e00435 25 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


examinedforuse-wear(TableS4.9).Thesefindingsfurtherstrengthenthe symbolicinterpretationpreviouslyadvancedforCuevaAntón ’ sochredscallop ( Zilhãoetal.,2010a ).3.Discussion 3.1.DatingaccuracyAtLaBoja,thearcheologicalsequence ’ sradiocarbonchronologyisindependently supportedbytheOSLdatingofthebasalMousterianandoftheAurignacian.At CuevaAntón,layerI-kcouldnotbeOSL-datedfortwomainreasons:(a)priorto[ ( F i g . _ 1 5 ) T D $ F I G ] Fig.15. OrnamentalshellacrosstheMiddle-to-UpperPaleolithictransitionintheMulabasinsites.a. Pecten half-valvefromMiddlePalaeolithiclayerI-kofCuevaAntón(after Zilhãoetal.,2010a );the reddishcoloroftheinternalsideisnatural;remnantsofanorangecolorantmadeofgoethiteand hematitearevisibleinthesidethatwaspainted(theexternal,whitishone).b … g.perforatedand/or ochre-stainedbivalveandgastropodshell(allatthesamescale)fromtheAurignacianofLaBoja(for additionaldetailandtaxonomicidentifications,seetheSIappendix). ArticleNo~e00435 26 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


20th-centuryburialbysiltsaccumulatedduringintermittentperiodsofsubmersion undertheLaCiervareservoirthelayerwasexposedasasurfaceforan undeterminedamountoftime,implyingsignificantuncertaintywithregardsto environmentalradiationparameters;and,(b)coupledwithitslimitedthicknessin thecross-sectionsexposedatthetimeofsampling,itshighstonecontent(layerI-k isaclast-supportedbrecciawithfewfines)madethislayerinappropriatefor luminescencedating( Burowetal.,2015 ). Fromwithintheradiocarbonmethoditself,theCuevaAntónandLaBojacharcoal samplespassedallthereliabilitytestscurrentlyavailable.Thedatesallowingusto boundtheMulabasin ’ sMousterian-to-Aurignaciantransitionbelongtolongseries ofresultsthatarefullystratigraphicallyconsistent,bothinternally(withineach site)andexternally(acrosssitesandwiththebroader,regionalandsupra-regional framework). AtLaBoja,thehumicfractionwasalsomeasuredtoassessthepotentialimpactof contamination.Theaccuracyofthechronologyobtainedonthefractionprocessed withtheABA(Acid-Base-Acid)treatmentissupportedby(a)theidenticalresults obtainedwheneverthedatingofindividualsampleswasrepeated,and(b)thelack ofstatisticaldifferencebetweentheresultsobtainedforindividualsamples processedwithbothABAandABOx-SC(Acid-Base-OxidationwithStepped Combustion)(basedon Birdetal.,1999 ). AtCuevaAntón,theABAprotocolwasfoundtoslightlyunderestimatetheageof thesamples,andthesuccessrateofABOx(26%;fiveoutof19)waslowerthanat LaBoja( Zilhãoetal.,2016 ).However,theCuevaAntónsamplessurvivingthe ABOx-SCpretreatmenthadahigh%C,which,following Rebolloetal.(2011) ,isa goodindicatorthatthematerialthatsurvivedwaswellpreserved.Inaddition, giventheaggressivenessofthetreatment,thepercentageoffailedsamplesisnot unexpected;similarrateshavebeenreportedwhenusingABOx-SCforsamples derivedfromcontextsdatedtobroadlythesametimeinterval( BrockandHigham, 2009 ).3.2.ThelatestMiddlePaleolithicsouthoftheEbroThedatingworkcarriedoutatthesiteofSimadelasPalomas( Fig.1 ,no.2),onthe coastofMurcia,ca.60kmtotheSoutheastofCuevaAntón,providesfurther supportforthelatepersistenceoftheMiddlePaleolithicintheregion „ inthis case,withdiagnosticNeandertalremainsfoundstratigraphicallytogetherwiththe lithics( Walkeretal.,2008 ;TrinkausandWalker,2017).Correctunderstandingof thesignificanceofthedatesobtainedatthiskeysiteishinderedbythesamples ’ proveniencenotationsreferringtoarbitraryhorizontalspitsthatdonotreflectthe stratigraphiclayoutofthesequence „ somethingmisunderstoodby Woodetal. (2013) and SantamaríaanddelaRasilla(2013) ,althoughexplicitlystatedin ArticleNo~e00435 27 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


[ ( F i g . _ 1 6 ) T D $ F I G ] Fig.16. SimadelasPalomasdeCabezoGordo,UpperCutting.a.Schematicdrawingofthe stratigraphy[after( Walkeretal.,2008 )( Walkeretal.,2012 ),modified].b.Compositemosaicview overthenorthandeastwallsoftheUpperCuttingexcavationtrenchduringthe2007fieldseason.c. Schematicpositionoftheradiocarbon-andU-series-datedsamplesrelativetostratigraphyandarbitrary horizontalspitsofprovenience(2a-to-2l). ArticleNo~e00435 28 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Walkeretal.(2008) .Whentheactualstratigraphyisconsidered,thedating results „ obtainedbyradiocarbononburntbonetreatedwiththeABAprotocol,UseriesonboneusingDiffusion/Adsorption(D/A)assumptions,andmulti-grain quartzOSLonsediments „ aremutuallyconsistent( Fig.16 ). TheU-seriesdatesforSimadelasPalomasshowthat(a)theaccumulationofthe lowercementeddepositcontainingarticu latedNeandertalskeletons(unitD)took placepriorto46.4ka,(b)providea terminuspostquem of53.5kaforthe accumulationoftheoverlyingdepositcontainingfragmentaryNeandertal remains(unitsA-BandE),and(c)suggestforthebaseofthelatteranage youngerthan45.3ka.TheOSLresultislesspreciseand,becauseofthe existenceofremnantsofanoldersedimentaryfillbrecciatedagainstthewalls androofofthecave,couldbeaffectedby incompletebleaching;evenso,when its95.4%probabilityinterval(45.3 … 64.1ka)isconsidered,itagreeswiththe U-seriesresults. Takentogether,theOSLandU-seriesdatesareinturnconsistentwiththetwo radiocarbondatesfromsamplesretrievedatthesamestratigraphicelevationor higherupinunitsA-BandE.Theuppermostradiocarbonresult(OxA-10666)is fromafaunalfragmentcementedtoadiagnosticNeandertalmandiblethatwas(a) foundhalf-waythroughtheunitAdepositand(b)overlainbyca.50cmof sedimentcontainingnothingbutdiagnosticMiddlePaleolithicstonetoolsand diagnosticNeandertalremains.AsOxA-10666translatesintoacalibratedage withinthe38.6 … 42.0kainterval,theSimadelasPalomasevidencestrongly indicates,inlinewiththeCuevaAntónpattern,thattheMiddlePaleolithic persistedintheregionwellbeyond40 … 42ka.Inaddition,itshowsthatsuchalatepersistingMousterianisindeedaNeandertal-associatedtechnocomplex.Thereis noreason,therefore,toquestionthattheassociationpertainsinthoseotherpartsof IberiawherestratigraphyanddatingsupportpersistenceoftheMiddlePaleolithic intothesametimerange:GibraltarandPortugal. AtGorham ’ sCave(Gibraltar; Fig.1 ,no.3),anuncalibrateddateof32,280±420 BP(OxA-7857)wasobtainedforacharcoalsamplerecoveredinstratigraphic associationwithdiagnosticMiddlePaleolithicstonetoolswithinContext24ofthe NaturalHistoryMuseum ’ s(NHM)1995 … 1998excavations( PettittandBailey, 2000 ).InMiddlePaleolithiclayerIVoftheGibraltarMuseum ’ s1999 … 2005 excavationsattherearofthecave,anuncalibrateddateof32,330±390BP(OxA10230)wasobtainedinthesamelaboratory,andaseparatesetofsamplesyielded uncalibrateddatesrangingbetween23,780±540BP(Beta-185345;2 )and 32,560±780BP(Beta-196771;2 )( Finlaysonetal.,2006 ; Finlaysonetal., 2008 ).Incalendaryears,theseresultsimplypersistenceoftheMiddlePaleolithic inGibraltaruntilatleast36.0 … 37.8ka(the95.4%probabilityintervalofthe calibrationofBeta-196771). ArticleNo~e00435 29 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


TheBetasamplesfromGorham ’ sallunderwentthestandardABAtreatment,but theyoungeronesprobablyreflectstratigraphicintrusionbecause,attherearofthe cave,aseveralmillennia-longhiatusmakesfordirectcontactbetweenMousterian layerIVandSolutreanlayerIII( ZilhãoandPettitt,2006 ).TheOxAresults,inturn, comefromsamplesprocessedwiththegentlerRRtreatment,whichdoesnot includeabasewash( Brocketal.,2010 ; Woodetal.,2013 ). EventhoughOxA-10230wasalargepineconescalethat,per BronkRamseyetal. (2002) ,madeforreliabledatingmaterial, Woodetal.(2013) assumethattheRR treatmentwasinsufficienttoremoveallcontaminationfromtheGorham ’ sOxA samples.Basedonthisassumption,theyarguethatnoconfidencecanbeplacedin thenotionthatthesite ’ sMiddlePaleolithicsignificantlypost-dates40 … 42ka. However,theydidnottesttheRRresultsviaprocessingofremainingmaterialin storage,orofnewsamples,withABAorABOx-SC(theyreportnoadditional charcoaldating,onlyfailedattemptsatextractingcollagenfromassociatedanimal bone).Inaddition,theRR-treatedcharcoalsamplesfromtheNHMexcavations collectedlowerdownintheGorham ’ ssequencereturnedresultsasoldas51,700± 3300BP(OxA-7790).Ifthelatterweretobetakenasabyproductofincomplete decontaminationproducingafiniteresultforasampleofinfiniteradiocarbonage, theunremovedcontaminant,ifmodern(i.e.,F14C=1),couldrepresentnomore than0.16%ofthemeasuredcarbon.ForOxA-10230,modelingsuchalevelof contaminationshiftstheuncalibratedradiocarbonresultfrom32,330to33,069BP, whichis,giventhestandarddeviation,statisticallythesamething. Againstthisbackground,arguingthathigherlevelsofcontaminationcharacterized thesamplescomingfromtheupperpartofGorham ’ sMousteriansequence(but onlythose...)wouldbespecialpleading.Themoresobecausethegeneral reliabilityoftheOxAresultsfortheGibraltarsites ’ RR-processedcharcoal samplesisotherwiseimplied,inthecaseofstratigraphicunits53 … 55ofVanguard Cave,bytheiragreementwiththeluminescenceagesobtainedforthesame deposit:radiocarbon ’ sRRresultswerebetween41,800±1400BP(OxA-6998) and54,000±3300BP(OxA-6891),OSL ’ swas46.32±3.30ka(OxL-1029) ( PettittandBailey,2000 ). InPortugal,layer8oftheGrutadaOliveiracavesite( Fig.1 ,no.4)yieldedan unquestionablyMiddlePaleolithicstonetoolassemblage( Marksetal.,2001 ).Its radiocarbondatingonburntbonetreatedwithABAatGroningenandwithRRat Oxfordyieldedstatisticallyindistinguishableresultsof,respectively,31,900±200 BP(GrA-10200)and32,740±420BP(OxA-8671)( AngelucciandZilhão,2009 ). Incalendarterms,thesetworadiocarbonresults,whichtranslateintoa95.4% probabilityintervalcomprisedbetween35.3and38.2ka,arestatisticallyidentical tothreeU-series(D/A)datesonbonefromthesamelayer( Hoffmannetal.,2013 ). ArticleNo~e00435 30 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Thetimespanindicatedbytherich,single-occupationMousterianopen-airsiteof FozdoEnxarrique,neartheSpanishborder( Fig.1 ,no.5),isthesame(Raposo, 1995).Here,theweightedaverageofthedatesobtainedbyU-seriesonthetooth enamelofonebovidandtwohorsesamplesis33.6±0.5ka.Theaccuracyofthis chronologyisdependentontheuncertainvalidityoftheEarlyUptakeassumption underpinningthecalculationoftheages,whilethenatureoftheassociation betweenthedatedfaunalremainsandthestonetoolsisanopenissue.Indeed,per BrugalandRaposo(1999),thesite ’ sfaunalassemblageisprimarilyanatural riversidethanatocenosis,withonlythecervidcomponentbearingmarksindicative ofahumanactivity-relatedaccumulation.Thetwomulti-grain,K-feldsparOSL resultssinceobtainedatthesiteforthebaseofthealluvialsandswithinwhichthe archeologicalleveliscontained(theT5unitofthelocalterracestaircaseofthe Tagus)are,therefore,abetter,iflesspreciseestimateofthetimeofdepositionof thestonetoolassemblage.At34.8±1.3and38.5±1.6ka(aftercorrectionfor anomalousfading)( Cunhaetal.,2008 ),theOSLresultssupportanagepost-dating 40kaforthesite ’ soccupation „ and,thus,thattheMiddlePaleolithicpersistedin interiorIberiabeyondthetimeofemergenceoftheEarlyAurignacianinthe CantabrianstripandnorthernCatalonia.3.3.TheearliestUpperPaleolithicsouthoftheEbroThepersistenceofaNeandertal-associatedMiddlePaleolithicfromIberia ’ s MediterraneanSoutheasttoitsAtlanticseaboardimpliesthatarcheological manifestationsofthemodernhuman-associatedAurignacianInotbefoundacross thesameterritory.Suchisindeedthecase.Neitherstratigraphicunitscontaining diagnosticassemblagesnorisolatedindexfossilsoftheEarlyAurignacianhavebeen identifiedinthelongcavesequencesspanningtheMP-UPtransitionknowninthose partsofthepeninsula:CovaBeneito(Valencia),CuevaBajondillo(Andalusia), Gorham ’ sCave(Gibraltar),andGrutadoCaldeirão(Portugal)( Zilhão,2006a ).At thesesites,andatothersthatareeitheropen-air,single-occupationlocalities,orlacka basalMiddlePaleolithic,theearliestUpperPaleolithicistheAurignacianII(Evolved Aurignacian)orIII … IV(a.k.a.LateAurignacian). Technologically,theAurignacianIIisdefinedbythedébitageofcarinatedand thick-nosed “ scrapers ” /coresproducingcharacteristicallytwistedblankstransformedintoDufourbladeletsviainverseoralternateretouch,whilethe AurignacianIII … IVischaracterizedbythepredominanceofcarinatedandother “ burin ” typesofbladeletcores.However,asdemonstratedatLaBoja,the microlithicdiagnosticsoftheAurignacianIIpersisttotheendoftheAurignacian sequence.Therefore,intheabsenceofreliabledating,orofatechnologically representativeassemblageofcoresanddébitageproducts,thepresenceofsuch microliths,eventhoughsufficienttoexcludeappurtenancetotheAurignacianI, doesnotexcludeassignmenttotheAurignacianIII … IV.Whenstratigraphic ArticleNo~e00435 31 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


sequencesarenotresolvedtothelevelofdetailseenatLaBoja,thepossibilitythat assemblagescontainingDufourbladeletscorrespondtopalimpseststhatsubsume bothphases(AurignacianIIandIII … IV)cannotbeexcludedeither. InMediterraneanSpain,theassemblagesfromBeneito,therock-shelterofRatlla delBubo( IturbeandCortell,1992 ),andthecavesiteofCovadeMallaetes( Fortea andJordá,1976 ),allinValencia,andfromBajondillo,areexamplesofclearly post-AurignacianIcollectionsthatcannotbepreciselyassignedtooneofthe succeedingphasesofthetechnocomplex.IntheBeneitoandRatlladelBubo assemblages,whichremainundated,backedelementsarefoundalongsidethe characteristicDufourbladelets.Thiscoexistencehasledsometoquestionthe validityoftheindustrialdiagnosis,ortheintegrityofthesedimentarycontexts( de laPeñaandVega,2013 ).However,basedontheevidencefromhorizonsOH15OH16ofLaBoja,thecoexistencesuggestsinsteadthattheBeneitoandRatlladel BuboassemblageseitherareLateAurignacianorincludeacomponentbelonging tothatphase.TheMallaetescontextlacksdiagnosticstonetoolsbutyielded lozengebonepointsinassociationwithaconventionalcharcoaldateof29,690± 560BP(KN-I/926).TheBajondillocontextcontainsdiagnosticssuggestiveofthe AurignacianIIandisdatedto33,690±1195BP(Ua-17150)and32,770±1065 BP(Ua-18050);however,giventheinadequatenatureofthesamples(of “ sediment andcharcoal ” )andtheimprecisionoftheresults,appurtenancetothesucceeding AurignacianIII … IVcannotbeexcluded.Arelatedproblemexistswiththetwo large,well-studiedstonetoolassemblagesfromtheopen-airAurignaciansitesof theRioMaiorbasin,inPortugal:GatoPreto ’ sisofAurignacianIIaffinitiesbutis datedbyThermoluminescence(TL)andthereforewithalarge95.4%probability interval,30.3 … 45.9ka;andValedePorcos ’ s,technologicallyofAurignacian III … IVaffinities,remainsundated( Zilhão,2006b ). Ithasbeenproposedthatthediagnosticmicrolithictool-typeoftheLate Aurignacianisanelongated,straightvariantoftheDufourbladeletpointedby alternateretouch( Zilhãoetal.,2010c ).Thisvariantisknownfromlayer2ofthe caveofPegodoDiabo,inPortugal,andfromthedisturbed,surficialdeposits cappingtheMousteriansequenceofCuevadeZafarraya,inAndalusia.Atthe Portuguesesite,thePleistocenefaunaassociatedwiththesmallassemblageofsuch DufoursyieldedfourAMSradiocarbondatesontoothsamplestreatedwithboth theLonginandtheultrafiltrationprotocols.Underthestringentcriterionof consideringreliableonlythosesamplesforwhichboththestandardgelatin productionandthe>30kDa(thousandsofDaltons)ultrafilteredproduction yieldedstatisticallyidenticalresults,thePegodoDiabodepositaccumulated between29,090±270BP(VERA-4047)and30,260+330/-320BP(VERA4050).TheearlierresultoverlapsthoseforOH15-OH16ofLaBoja,butthelater oneextendstherangeforanothermillennium,untilca.33ka.Becausethedated faunaisnon-anthropogenic,however,itcannotbeascertainedwhetherthe “ Pego ArticleNo~e00435 32 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


doDiabopoints ” (a)standfora “ Final ” phase,datingbeyond34.0ka,ofthe AurignaciantechnocomplexinWesternIberia,astheyoungerresultmightsuggest, or(b)areacomponentoftheca.34.0 … 35.5kaLateAurignacian,asindicatedby theearlierresult.Ifthesecondhypothesisisretained,theimplicationwouldbethat themicrolithictool-kitoftheLateAurignacianwasmorediversethansofar documentedinValenciaandMurcia. Beitasitmay,theMulabasinsitessufficetodemonstratethat,by36.5 … 37.1ka,the AurignacianIIwasalreadypresentinIberianregionstotheSouthoftheEbrobasin. Thisintervalisthesameduringwhich,basedonBayesianmodelingofavailable dates, Banksetal.(2013b) foundthatthetransitionfromtheEarlytotheEvolved AurignacianhadoccurredtotheNorth.Thistechnologicaltransitionwouldtherefore seemtohavebeenconcomitantwithaprocessofsettlementexpansion:inNorthern Europe,towardtheBritishIslesandequivalentlatitudesofGermanyandPolandthat, duringthepreviousphase,hadbecomedevoidofhumanoccupation;inIberia,toward thelandsbeyondtheEbrobasin,eventuallyleadingtoreplacementoftheirlatepersistingMousterianandtheassimilationofitsNeandertalmakers.The “ Ebro Frontier ” modelprovidesabiogeographicalandpaleoecologicalframeworkforthe interpretationofthesedevelopmentsintermsofpopulationhistory.3.4.The “ Ebrofrontier ”InIberia,theEbrobasinnowadaysliesattheinterfacebetweentwobiogeographic regionsdefinedafterthedistributionofplantcommunities:Eurosiberianand Mediterranean( Rivas-Martínez,1987 ).Theseparationrunsalongthesouthern foothillsoftheCantabro-Pyreneanmountainsbut,duringtheUpperPleistocene,its veryexistenceandlatitudinalplacementmusthavebeendependentontheperiod ’ s highlyvariableandfrequentlyoscillatingclimates. DuringMIS4,Eurosiberiansteppe-tundraenvironmentsspilledintoandbeyondthe EbrobasinwellintotheIberiancore.Thisisshownbythedistributionofwoolyrhino andmammothfinds:alongtheMediterraneancoast,downtotheLlobregatdelta,near Barcelona;incentralIberia,asfarWestastheManzanaresvalley(Madrid)andasfar SouthasthenorthernflanksoftheSierraNevada(Granada)( Dauraetal.,2013 ). DuringtheLastGlacialMaximum(LGM),Europe ’ sUpperPleistocenecoldfauna (mammoth,woolyrhino,bison,reindeer)wasagainpresentinCatalonia,the Cantabrianstrip,andpartsofthenorthernMesetabutabsentfromValencia,Murcia, Andalusia,andPortugal.Thesedifferencesinthecompositionofthelargeherbivore faunaimplysignificantenvironmentalgradientswithinthepeninsuladuringMIS4 andtheLGM,albeitonesthat(a)didnotfollowthepresentEurosiberian/ Mediterraneandivide,and(b)giventhesharedaspectsofstonetooltechnologyand thewidespreadhomogeneityinrockartstylesobservedthroughtheGravettianand mostoftheSolutreanallthewayfromPortugal,intheWest,totheRhonevalley,in ArticleNo~e00435 33 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


theEast,didnotrepresentsignificantbarrierstothemovementofpeople,the circulationofobjects,ortheexchangeofideas. Wealsoknowthat,duringperiodsofextremeariditysuchastheepisodeoficeberg dischargeknownasHeinrichStadial(HS)4,whichlastedforafewcenturies aroundca.40ka,thekindsofsemi-desertenvironmentsnowadaysconfinedto northernAlmeriaandsouthernMurciaexpandedtotheMesetanhinterlandandthe badlandsofthemiddleandupperEbrobasin( d ’ ErricoandSánchez-Goñi,2003 ; Sepulchreetal.,2007 ).Conversely,duringperiodsofmilder,wetterclimatic conditionssuchasGreenlandInterstadial(GI)8(ca.38.2 … 36.6ka),mountain forestsandwoodedlandscapesunderwentaverysignificantexpansionbelowthe latitudeof40°N( Fletcheretal.,2010 ).Judgingfromwhathappenedinthe Holocene,duringsuchmilderperiodshumansettlementmusthaveretractedtothe resource-richerlittoralareas,leadingtothebreaking-upofexchangeand communicationnetworks,andfavoringtheemergenceofcultural/biological isolates. Basedonthisevidence,the “ EbroFrontier ” modelhypothesizesthatsteppe-tundra environmentswouldhavebeencontinuouslypresentinNorthernIberiathroughthe entireMP-UPtransitionprocessandthat,duringthisperiod,theEbrobasinwould havefunctionedasamajorphysicalandbiogeographicaldividedueto:(a)the establishmentofsemi-desertconditionsinthebasinitself,thenorthernflanksof theIberianRange,andtheMesetanhinterland,inHS4,and(b)thedevelopmentin adjacentlandstotheSouthandWest,bothbeforeandafterthisextremearidity event,ofextensivemountainforestsandopenwoodlands.Atpresent,this hypothesisremainsdifficulttotest,becausethepaleoenvironmentaldataavailable areinsufficienttoreconstruct,withthespatialandtemporalresolutionrequired,the impactoftheseclimaticoscillationsontheecosystemsoftheterritoryacrosswhich theenvironmentalgradientdeveloped.However,thedivergentcultural-historical trajectoriesfollowedeithersideofthe “ EbroFrontier ” afterca.45ka „ namely, thefailureoftheChâtelperronian,theProtoaurignacianandtheAurignacianIto extendsouthward „ doimplythepresenceofamajor,long-lastingbarrierto migration,geneflowanddiffusion. ThespreadoftheAurignacianIIintoSouthernandWesternIberiasignalsthe disappearanceoftheconditionsunderpinningtheprecedingpatternofcultural divergence,whatevertheircause.Thatpaleoenvironmentalfactorsmusthave playedaroleisnonethelessintimatedbythetemporalcoincidenceofthe replacementofIberia ’ slate-persistingMousterian(ca.36.5 … 37.1ka)withthe globalclimatictransitionfromGI8(thelongestandmildestofallMIS3 insterstadials)toGreenlandStadial(GS)8(a “ normal ” coldphase)( Rasmussen etal.,2014 ).Duringthistransitionalperiod,theEurosiberiansteppe-tundra couldandlikelydidbegintospreadintotheIberiancore,whilethecharcoalfrom ArticleNo~e00435 34 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


sub-complexAS1ofCuevaAntón(12%cryophilouspines,85%steppictaxa,3% riversidetaxa; Zilhãoetal.,2016 : Fig.8 ,SITable2)indicatesadescentofthe montanepineforestbeltfromabove1100mtobelow400m,inagreementwiththe neardisappearanceofMediterraneanforesttaxaseenatthistimeinthedeep-sea pollenrecord( Fig.9 ). ThepresenceofamajorbiogeographicalgradientalongtheEbrobasinacquires broaderpaleoanthropologicalsignificancebecauseoftheperiodwhenithappened tobeseparatingmodernhumansandNeandertals.Inandofitself,however,the existenceatthattimeofsuchagradient,withattendantimplicationsfordiffusion andexchange,innowayshouldbemistakenforsomethingexceptionalorunique. AftertheLGM,forinstance,theEbrobasinwouldcometoseparatemoderns (BadegoulianandEarlyMagdalenian)fromothermoderns(UpperSolutreanand Solutreo-gravettian)foracomparableduration „ threetofourmillennia( Banks etal.,2009 ).Conversely,priorto42katheEbrobasinhadalreadybeenseparating Neandertals(Châtelperronian)fromotherNeandertals(Mousterian) „ andmay wellhavecontinuedtodosoforanothercouplethousandyearsifNeandertalswere alsoinvolvedinthemanufactureoftheProtoaurignacian. TheProtoaurignacianiswelldocumentedalongtheshoresoftheCantabrianSea, fromtheBasquesitesofIsturitzandLabekoKobaintheEasttotheAsturiansiteof LaViñaintheWest( Zilhão,2006a ).Eventhoughnoarcheologicallyassociated diagnostichumanremainshavesofarbeenfoundacrosstheProtoaurignacian ’ s entiregeographicalrange(BulgariatonorthernSpain)andtemporalspan(39 … 42 ka),thegenomeoftheOase1adultmaleshowsthathehadhada “ pure ” Neandertalancestoronlyfourtosixgenerationsback( Fuetal.,2015 ).Combined withtheageofthefossil(directlydatedbyradiocarbonto37.1 … 41.4ka)( Trinkaus etal.,2013 ),thisgenomicevidenceimpliesastrongprobabilityofoverlapbetween NeandertalsandatleastthebeginningsoftheProtoaurignacian.Thelatter ’ s industrially “ intrusive ” characteristicsandsimilaritywiththeNearEastern,modern human-associatedEarlyAhmariansuggestanintrinsicrelationtomodernhuman immigration.ThetechnologicalinnovationstheProtoaurignacianstandsfor, however,couldwellhavediffusedintoNeandertalterritorywellinadvanceofthe arrivaloftheadmixturefront.Sincenoevidenceexiststhatan “ archeological culture=humantype ” equationappliestotheProtoaurignacian,itremainsentirely plausible,therefore,thatitwasalsomadebyvariouslymixedNeandertal-modern human,oreven “ pure ” Neandertalpopulations „ andespeciallysointheWest ( TrinkausandZilhão,2013 ; Zilhãoetal.,2015 ). IfNeandertalswerealsoinvolvedinthemakingoftheProtoaurignacian,thenitis onlyinAurignacianItimes,after40ka,thattheEbrobasinrepresenteda Neandertal/modernhuman “ frontier. ” Ifso,theemergenceofsucha “ frontier ” wouldhavebeenbroadlycoincidentalwiththe39.9kaexplosionofthePhlegraean ArticleNo~e00435 35 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Fieldscaldera,whoseashfall-outblanketedvaststretchesofItalyandSoutheastern Europe,severelydisruptingfoodchainsforanextendedperiod „ thehighest trophiclevels,includinghumanhunters,beingmostimpacted.Forthepopulations ofWesternEurope,whichwasnotdirectlyaffected,themainconsequenceofthe explosionwouldhavebeentobringaboutareleasefromtheconstraintsof demographicpressureinducedacrossthecontinentallandmassbytheprevious millenniaofpopulationgrowthandNeandertalassimilation.Inthisscenario,the explosionwouldhaveconstitutedahistoricallycontingentbutsignificantfactor contributingtoexplainwhyMiddlePaleolithicNeandertalspersistedforsolongin theterritoriesofEurope ’ sFarWest( Zilhão2009 ; Fitzsimmonsetal.,2013 ; Marti etal.,2016 ; Giaccioetal.,2017 ). Whether,atthetimeofthiscatastrophicevent,theNeandertal/modernadmixture fronthadalreadyreachedthePyreneesandtheCantabrianstripforquitesome timeorhadjustarrivedthereremainsanopenissue.But,whicheverthecase,the explosion ’ simpactonthemodernhumanpopulationsofCentralandEastern Europewouldhavestalledthewestwardexpansionofthefrontafterca.40ka.Ifa biogeographicalgradientwasthenextantacrosstheEbrobasin,thedemographic crisiscausedbythePhlegraeanFieldsexplosionwouldhaveenhancedthat gradient ’ seffect.Andif,withthereturntonormalstadialconditions,following theendofGI8,thateffectceasedtooperate,itwouldhavedonesoatatime whenreplenishmentoftheCentral/EasternEuropeansinkcreatedbythe explosionwouldalsohaveresetdemographicpressureovertheperipheries. ForNorthernEurope,theconsequencewouldhavebeenresettlement.ForIberia, itwouldhavebeentheeventualassimilationofthelastofEurope ’ sNeandertals, aspostulatedbythe “ EbroFrontier ” model.Bothexpectationsaremetbythe empiricalrecord.4.ConclusionsThetechnologicalanduse-wearevidencerejectsinterpretinglayerI-kofCueva AntónandoccupationhorizonsOH20andOH19ofLaBojaasdistinctstructural posesofasingle,multifacetedsystem.Putanotherway,thesmalllithicassemblage inlayerI-kofCuevaAntóncannotbeinterpretedasafunctionallyspecialized,or activity-specificfaciesoftheregion ’ sEvolvedAurignacian.Instead,layerI-kof CuevaAntónandoccupationhorizonsOH20andOH19ofLaBojastandfor concretemanifestationsofmutuallyexclusive,long-lastingtechnologieswhose succession,ratherthanagradualtra nsition,trulyconsistedofanabrupt replacement.Astheefficiencyofstonetoolproductionintermsofcuttingedge perunitofmassisidenticalinbothtechnologies( MullerandClarkson,2016 ),the parsimoniousreadingofthisreplacementprocessisthatitrepresentsamajor break,withdemicunderpinnings,inregionalculturaltrajectories. ArticleNo~e00435 36 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Theevidencefromstonetooltechnologyandthestratigraphiclayoutofsitesisthat thepatternderivedfromthehigh-precisionMulabasindatacanbeextrapolatedto allIberianregionstotheSouthoftheEbrobasin.Intheseregions,artefact assemblagesattributabletotheearliestphasesofWesternEurope ’ sUpper PaleolithicaremissingfromstratifiedsitesthatcontaindepositsspanningtheMPUPtransition,andhaveneverbeenfoundassingle-component,open-aircontexts. Inaddition,noisolatedoccurrencesoftheirindexfossils(e.g.,Châtelperronian points/knives,orAurignaciansplit-basedbonepoints)haveeverbeenreported amongsurface,mixed,orpost-depositionallydisturbeddeposits.Fromthebasics ofPrehistoricArcheology,i.e.,fromtheculture-stratigraphicreasoningproviding theframeworkforallitschronologies,theonlyinferencethatonecanderivefrom thispatternisthat,southwardoftheEbrobasin,alate-persistingMousterian occupiesthetimeslotinwhichtheAurignacianIisfoundelsewhere.The radiocarbonevidenceisentirelyconsistentwiththisnotion,whichavailable luminescenceandU-seriesindependentlysupport,andwhichnootherkindsof radiometricdatingresultshavesofarcountered. AcorollaryofthesefindingsisthatNean dertalspersisteduntilca.37kaacross SouthernandWesternIberia „ whichcarriesimplicationsfortheauthorshipof allotheraspectsoftheseregions ’ archeologicalrecord.F orinstance,giventheir datingandarcheologicalassociations,therecanbenoquestionthatthepainted/ perforatedshellsfromCuevaAntónandCuevadelosAviones,aswellasthe abstractengravingand ornamentaluseofraptorfeathersdocumentedat Gorham ’ sCave,standformanifestatio nsofNeandertalsymbolism( Zilhão etal.,2010a ; Finlaysonetal.,2012 ; Rodríguez-Vidaletal.,2014 ).Knowing thatminimumagesof40.8kaforareddiskand37.3kaforahandstencilhave beenobtainedatElCastillocave(Cantabria)( Pikeetal.,2012 ),andthatsuch motifsexistinExtremaduranandAndalusiansites,itiseasytoseehowthe “ EbroFrontier ” patternmayalsobearimplicationsfortheauthorshipofcave paintings. RecentadvancesinthefieldofGeneticsincreasinglymakeitclearthat,intheLate PleistoceneofEurasia,thecontinentalextensionofratherhomogeneous archeologicalculturesissuperimposedoncomplexancestrypatchworks( Mallick etal.,2016 ; Paganietal.,2016 ).Thiscanbeexplainedbyapatternoflongdistancediffusionandculturalresilience,whichmaintainednetworksoverthe long-term,combinedwithextendedperiodsofgeographicalisolation,which conservedregionalgeneticvariants.The “ EbroFrontier ” effectmakesthis mechanismapparentevenintherefugiaofSouthernEuropeandespeciallysoat thetimeoftheMP-UPtransition.Thisvisibilityisduetowhenthefrontierformed andforhowlongitlasted,bothallowingtheeffecttobepicked-upwiththecurrent resolutionofdatingtechniques.Likely,however,similar,broadlycoevalbut chronometricallylessvisibleLatePleistocenefrontiersmusthaveexistedinother ArticleNo~e00435 37 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


partsofAsiaandEurope,aswellasduringtheearlierphasesoftheprocessof modernhumandispersalintothesecontinents. Theresultswereportherehighlighttheneedforproperintegrationofthe biologicalandthearcheologicalevidencewhenreconstructingLatePleistocene populationhistories.Alllinesofevidencearenowconvergingtosupport replacement-through-admixture,orAssimilation,asthebestexplanationforthe disappearanceoftheNeandertalandotherarchaicphenotypes.TheIberian evidencesuggeststhiswasatime-transgressiveevolutionaryoutcomestemming fromdynamic,complexandgeographicallyunevenprocesses „ apunctuated historyinwhichthelong-termmaintenanceofpan-continentalnetworksofgene flowandculturalexchangedidnotexcludetheoccurrenceofextendedperiodsof significantgeographicalisolation.5.Materialsandmethods 5.1.ArcheologicalexcavationandanalysisExcavationproceededthroughdécapagealongobservedboundaries,whether natural(e.g.,theinterfacewiththeunderlyinggeologicalstratigraphy)or anthropogenic(e.g.,thebaseofdistinctoccupationfloorsstackedupwithina singlenaturalstratigraphicunit),withsubdivisionswhennecessary.Findswere piece-plottedwiththehelpofalaserlevel,tothenearestcentimeter,againstgrid andsitedatum.Use-wearanalysisofstonetoolswasbasedondifferential interferencecontrastmicroscopy,carriedoutwithaBHMJOlympusmodel(at× 200or×400magnification),andfollowedstandardrecommendationsforthe cleaningandpreparationofthematerial.Largesamplesofthesedimentwere floatedfortherecoveryofpaleobotanicaldata;theremainderwasentirelydrysievedusingtwo-sievestacks(2and1mmmesh-sizes).Theanalysisofpollen, charcoal,molluskshellandanimalbonefollowedstandardprotocols.Stratigraphic cross-sectionsweregeologicallydescribed,drawnanddigitallyrecorded,aswere thesurfacesexposedateachstepofthedécapageprocess.AtFincaDoñaMartina, theDStrectchplug-inforImageJwasusedtohighlightcolorcontrastsandproduce printsusedinthefieldtohelpwiththedécapageofstratigraphicinterfaces.Photo mosaicswereassembledusingPTGUIorMicrosoftICEandorthorectifiedwith theUniversityofVenice ’ sRDFsoftware.Elevationmapsand3Dmodelswere producedwithSurfer.Undisturbedsoilandbulksedimentsampleswerecollected formicromorphological,phytolithandbiomolecularanalysis.5.2.RadiocarbondatingOnlysecurelyprovenanced,taxonomicallyclassifiedcharcoalsampleswere submittedfordating.AllsamplesweretreatedwiththeABAprotocol,andthe humicfractionsofseveralsampleswerealsomeasured( Wildetal.,2008 ).For ArticleNo~e00435 38 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


FincaDoñaMartina ’ s,amildertreatmentwasusedforsome,duetopoor preservation;inmostcases,onlythehumicfractioncouldbedated.Theresultsfor thissitearethereforeallminimumages.AtCuevaAntón,theABAtreatment provedinsufficienttoremoveallcontamination,butthechronologyoflayerI-k reportedhereisentirelybasedonresultsobtainedforsamplesthatwereprocessed withtheABOx-SCprotocol( Zilhãoetal.,2016 ).Tocheckifasimilarproblem existedatLaBoja,someofitssampleswerealsoprocessedwithABOx-SC,in paralleltothestandardABAtreatment( Wildetal.,2008 )andusingamodified versionoftheproceduregivenin( Brocketal.,2010 ),i.e.acidandbasetreatment at60°C.Inaddition,tocontrolfortheaccuracyofindividualmeasurements,some ABA-treatedsamplesweredatedtwice.TheABOx-SCresultsandtherepeatswere inallcasesstatisticallyindistinguishablefromtheoriginalABAdate.Whenmore thanoneresultforasinglecharcoalfragmentwasobtained,thecorresponding averagewasused.CalibrationwascarriedoutwiththeINTCAL13curveinCalib 7.0( StuiverandReimer,1993 ; Reimeretal.,2013 ).The Fig.9 plotwasprepared inOxCal4.2.4( BronkRamsey,2009 ).5.3.LuminescencedatingTheADBsampleswereextractedfrommacroscopicallyhomogeneoussilt-rich deposits(Fig.S4.9).Duetotheunconsolidatednatureoftrenchwalls,itwas decidednottodrivemetalcylindersintothesediment;instead,thesampleswere extractedwithaknife,incompletedarkness.Coarsegrainquartz(100 … 150 m) andpotassiumfeldspar(100 … 200 m)wereextractedusingconventionalsample preparationtechniques( Kehletal.,2016 ).Allmeasurementswerecarriedouton anautomatedRisøTL/OSLDA20readerequippedwithacalibrated90Srbeta sourceandanEMI9235photomultiplier.Multiple-grainquartzsampleswere measuredusingthesingle-aliquotregenerative-doseprotocol(SAR)( Murrayand Wintle,2000 ; MurrayandWintle,2003 ),includingsignalstimulationbyblue diodes(470nm,FWHM=20)andsignaldetectionthroughaHoyaU340filter. Theinitial0.8softhesignalminusabackgroundofthelast5swasusedforquartz dating.Preheatplateauanddoserecoverytestswerecarriedouttocheckthe suitabilityofthemeasurementprotocol.Single-grainquartzdatingwasnotfeasible becauseoflowsignalintensities. Multiple-grainpotassiumfeldsparsamplesweremeasuredusingthepost-infrared infraredstimulatedluminescencesignalmeasuredat290°C(pIRIR290)( Thiel etal.,2011 ).Stimulationwascarriedoutwithinfrareddiodes(870nm,FWHM= 40),andthesignalsweredetectedthroughaninterferencefilter(410nm).The initial4softhesignalminusabackgroundofthelast20swasusedinthepIRIR dating.PriorIRstimulationtemperaturetestsanddoserecoverytests(24hHönle Sol2bleaching)werecarriedouttochecktheperformanceofthemeasurement protocol.Equivalentdoseswerecalculatedusingthearithmeticmean(AM),except ArticleNo~e00435 39 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


forsampleC-L3905,forwhichwealsousedtheminimumagemodel(MAM) ( Galbraithetal.,1999 ).Additionally,infraredstimulatedluminescencemeasured at50°C(IR50)wasapplied( Wallingaetal.,2000 ; Preusser,2003 ),andthesignal wascorrectedforanomalousfadingusingtheapproachesof Auclairetal.(2003) and HuntleyandLamothe(2001) . DataanalysiswascarriedoutusingtheRluminescencepackage( Burow,2017; Kreutzer,2017;Kreutzeretal.,2017 ).Theradionuclideconcentrationsofthe surroundingsedimentsweremeasuredusinghighresolutiongamma-rayspectrometry.ThedoseratewascalculatedusingDoseRateandAgeCalculator(DRAC) ( Durcanetal.,2015 ),andincludedconversionfactors( Guérinetal.,2011 )andan assumedwatercontentof5±2%.Theinternalbetadoseratecontributionofthe feldsparsampleswascalculatedbyassumingapotassiumcontentof12.5±0.5% ( HuntleyandBaril,1997 ).Thecosmicdoseratewascalculatedafter Prescottand Hutton(1994) .Dosedistributionsaredisplayedasabanicoplots( Dietzeetal., 2016 )( Figs.6and7 ).Equivalentdosescalculatedwiththearithmeticmeanandthe CentralAgeModel(CAM)arestatisticallyindistinguishableat1 andfinallythe arithmeticmeanwasused. AtypicaldoseresponsecurveandadecaycurveareshownforquartzsampleCL3905( Fig.7 a).Preheatplateautests( Fig.7 b)indicatedthattheequivalentdoseof thequartzisindependentfromtemperaturetreatmentintheranges180 … 240°C(CL3901),220 … 280°C(C-L3904),180 … 280°C(C-L3905),and240 … 280°C(CL3906).Doserecoverytestsshowedthatalaboratorygivendosewasbest recoveredusingatemperatureof180°CforsamplesC-L3901andC-L3905andof 260°CforsamplesC-L3904andC-L3906( Fig.7 c).PriorIRstimulation temperaturetestscarriedoutforfeldsparsampleC-L3905indicatedaplateau between80°Cand180°C( Fig.7 d).Laboratorydoseswererecoveredwitharatio ofthemeasuredtothegivendoseof1.07±0.06(aresidualdoseof5Gyafter24h ofbleachingintheHönleSol2solarsimulatorwassubtracted).Arepresentative doseresponsecurveforthisfeldsparsampleisshownin Fig.7 eandthedose distributionin Fig.7 f. Thelaboratoryexperimentsconfirmedthesuitabilityofthemeasurementprotocols forbothquartzandfeldsparminerals.ExceptforsampleC-L3903,thequartzOSL ageestimatesareinstratigraphicorder,scatterbetween57.7±3.2kaand32.6± 1.9ka,andareconsistentwiththeradiocarbonagesobtainedforthesameunits. pIRIR290andIR50datingwascarriedouttoinvestigateifthequartzOSLsignal waslikelytobefullybleachedatthetimeofdeposition.Aninternalcrosscheckof thetwomineralsisadvisable( Murrayetal.,2012 )becausethepIRIR290andIR50signalsbleachslowerthanthequartzOSLsignal( Buylaertetal.,2012 ). Comparisonofthemeanageestimatesofallthreeluminescencesignalsshows goodagreementbetweenthequartzOSLandfeldsparIR50andpIRIR290agesof ArticleNo~e00435 40 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


sampleC-L3901.ForsampleC-L3905,thequartz(35.8±2.8ka)andIR50age estimatesareyoungerthanthepIRIR290age(45.4±5.6ka),whichindicates incompletebleachingofthefeldsparpIRIR290signalatdeposition.Thisis supportedbythegoodagreementofthequartzOSLandfeldsparIR50resultswith thecalibratedradiocarbonage(34.9 … 37.1ka;VERA-5854)obtainedforthesame stratigraphicunit,confirmingcompletebleachingoftheOSLandIR50signals. ApplyingaMAMtothefeldsparpIRIR290datasetresultsinanageof37.4±5.3 ka,whichdemonstratesthattheMAMsuccessfullyextractsindividualequivalent dosevaluesfromthedistributionthatarelikelytobefullybleachedatdeposition. ForsamplesC-L3902andC-L3904,thepIRIR290ageestimatestendto overestimatethequartzandIR50results.Itwasnotpossibletoextractthose individualequivalentdosesfromthedistributionthatarelikelytohavebeen completelybleachedpriortodepositionusingtheMAM.QuartzsampleC-L3903 appearstobeunderestimatedcomparedtotheunderlyingsamplesandwevalue thisresultasanoutlier.Declarations AuthorcontributionstatementJoãoZilhão:Conceivedanddesignedtheexperiments;Performedtheexperiments; Analyzedandinterpretedthedata;Wrotethepaper. DiegoAngelucci,ValentinVillaverde,JosefinaZapata:Conceivedanddesigned theexperiments;Performedtheexperiments;Analyzedandinterpretedthedata. DanielaAnesin,ThierryAubry,ErnestinaBadal,DanCabanes,MartinKehl, NicoleKlasen,ArmandoLucena,IgnacioMartín-Lerma,SusanaMartínez, HenriqueMatias,DavideSusini,PeterSteier,EvaMariaWild:Performedthe experiments;Analyzedandinterpretedthedata.CompetingintereststatementTheauthorsdeclarenoconflictofinterest.FundingstatementArchaeologicalfieldworkandresearchatCuevaAntónandtheRamblaPerearockshelterswasfundedbytheDirecciónGeneraldelMedioNaturaldelaRegiónde Murcia,theMunicipalityofMula,theUniversityofMurcia,theFundaciónSéneca (Murcia),theMinisteriodeCienciaeInnovación(grantsHAR2011-24878, HAR2014-52671-PandCGL2012-34717),theGeneralitatValenciana(grant PROMETEOII/2013/016),theExcellenceResearchProjectsProgramofthe AndalusianGovernment(grantP11-RNM-7033),andtheLeakeyFoundation.The GermanResearchFoundation ’ s(DFG)projectCRC806( “ OurWaytoEurope. ArticleNo~e00435 41 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Culture-EnvironmentInteractionandHumanMobilityintheLateQuaternary ” ) fundedtheluminescencedating.AdditionalinformationSupplementarycontentrelatedtothisarticlehasbeenpublishedonlineat http://dx. UniversityofValencia,J.ZilhãooftheSERPresearchgroup(SGR2014-00108)of theUniversityofBarcelona,andN.KlasenofCRC806'sprojectF2(Application ofluminescencedatingtechniquesingeoarchaeologicalstudies).Weareindebted totheMuseodeArteIbéricoElCigarralejoinMulaforlogisticalsupport.Will Banks,CatherinePerlès,PaulPettitt,AlistairPike,andErikTrinkausread preliminaryversionsofthemanuscriptand/orprovidedusefulcomments.We thankSahraTalamo,theothertwo,anonymousreviewers,andtheeditor,fortheir carefulreadingoftheoriginalsubmissionandthemanyusefulcommentsand suggestions.Asusual,anyerrorsoromissionsareourown.ReferencesAlmeida,F.,Moreno-García,M.,Angelucci,D.E.,2009.Fromunderthe bulldozer ’ sclaws:theEE15LateGravettianoccupationsurfaceoftheLagar Velhorock-shelter.WorldArchaeol.41,242 – 261.doi: 00438240902843790 . Angelucci,D.,Zilhão,J.,2009.Stratigraphyandformationprocessesofthelate pleistocenedepositatGrutadaOliveira,AlmondaKarsticSystem,TorresNovas Portugal.Geoarchaeology24,277 – 310.doi: . Angelucci,D.,etal.,2013.FormationprocessesatahighresolutionMiddle Paleolithicsite:CuevaAntón(Murcia,Spain).Q.Int.315,24 – 41.doi:http://dx.doi. org/10.1016/j.quaint.2013.03.014 . Angelucci,D.,etal.,2017.Ataleoftwogorges:LateQuaternarysiteformation andsurfacedynamicsintheMulabasin(MurciaSpain).QuatIntdoi:http://dx.doi. org/10.1016/j.quaint.2017.04.006 . Auclair,M.,Lamothe,M.,Huot,S.,2003.Measurementofanomalousfadingfor feldsparIRSLusingSAR.Radiat.Meas.37,487 – 492.doi: 10.1016/S1350-4487(03)00018-0 . Banks,W.E.,Zilhão,J.,d ’ Errico,F.,Kageyama,M.,Sima,A.,Ronchitelli,A., 2009.InvestigatinglinksbetweenecologyandbifacialtooltypesinWestern ArticleNo~e00435 42 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Europeduringthelastglacialmaximum.J.Archaeol.Sci.36,2853 – 2867.doi: . Banks,W.E.,d ’ Errico,F.,Zilhão,J.,2013a.Human-climateinteractionduringthe earlyUpperPaleolithic:testingthehypothesisofanadaptiveshiftbetweenthe Proto-AurignacianandtheearlyAurignacian.J.Hum.Evol.64,39 – 55.doi:http:// . Banks,W.E.,d ’ Errico,F.,Zilhão,J.,2013b.Revisitingthechronologyofthe Proto-AurignacianandtheearlyAurignacianinEurope:AreplytoHighametal. ’ s commentsonBanksetal.(2013).J.Hum.Evol.65,810 – 817.doi: 10.1016/j.jhevol.2013.08.004 . Bird,M.I.,etal.,1999.Radiocarbondatingof ‘ old ’ charcoalusingawetoxidation, stepped-combustionprocedure.Radiocarbon41,127 – 140.doi: 10.1017/S0033822200019482 . Bradtmöller,M.,Pastoors,A.,Weninger,B.,Weniger,G.C.,2012.Therepeated replacementmodel – rapidclimatechangeandpopulationdynamicsinLate PleistoceneEurope.Q.Int.247,38 – 49.doi: quaint.2010.10.015 . Brock,F.,Higham,T.,2009.AmsradiocarbondatingofPaleolithic-agedcharcoal fromEuropeandtheMediterraneanrimusingABOX-SC.Radiocarbon51, 839 – 846.doi: . Brock,F.,Higham,T.,Ditchfield,P.,BronkRamsey,C.,2010.Current pretreatmentmethodsforamsradiocarbondatingattheOxfordRadiocarbon AcceleratorUnit(ORAU).Radiocarbon52,103 – 112.doi: 10.1017/S0033822200045069 . BronkRamsey,C.,2009.Bayesiananalysisofradiocarbondates.Radiocarbon51, 337 – 360.doi: . BronkRamsey,C.,Higham,T.F.G.,Owen,D.C.,Pike,A.W.G.,Hedges,R.E.M., 2002.RadiocarbondatesfromtheOxfordAMSsystem:datelist31.Archaeometry 44(Suppl.1),1 – 149.doi: . Burow,C.,etal.,2015.Luminescencedatingoffluvialdepositsintherockshelter ofCuevaAntón,Spain.Geochronometria42,107 – 125.doi: 10.1515/geochr-2015-0010 . Burow,C.,2017.calc_MinDose:applythe(un-)loggedminimumagemodel (MAM)afterGalbraithetal.(1999)toagivenDedistribution.Functionversion 0.4.4.In:Kreutzer,S.,Dietze,M.,Burow,C.,Fuchs,M.C.,Schmidt,C.,Fischer, M.,Friedrich,J.(Eds.),Luminescence:ComprehensiveLuminescenceDatingData ArticleNo~e00435 43 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Analysis.RPackageVersion0.7.3,. calc_MinDose.html . Buylaert,J.P.,etal.,2012.Arobustfeldsparluminescencedatingmethodfor MiddleandLatePleistocenesediments.Boreas41,435 – 451.doi: 10.1111/j.1502-3885.2012.00248.x . Caron,F.,d ’ Errico,F.,DelMoral,P.,Santos,F.,Zilhão,J.,2011.Therealityof NeandertalsymbolicbehaviorattheGrotteduRenne,Arcy-sur-Cure.PLoSOne6, e21545.doi: . Cunha,P.P.,Martins,A.A.,Huot,S.,Murray,A.S.,Raposo,L.,2008.Datingthe TejoriverlowerterracesintheRódãoarea(Portugal)toassesstheroleoftectonics anduplift.Geomorphology102,43 – 54.doi: . d ’ Errico,F.,Sánchez-Goñi,M.F.,2003.Neandertalextinctionandthemillennial scaleclimaticvariabilityofOIS3.Q.Sci.Rev.22,769 – 788.doi: 10.1016/S0277-3791(03)00009-X . Daura,J.,etal.,2013.TerrassesdelaRieradelsCanyars(Gavà,Barcelona):the landscapeofHeinrichStadial4northoftheEbrofrontierandimplicationsfor modernhumandispersalintoIberia.Q.Sci.Rev.60,26 – 48.doi: 10.1016/j.quascirev.2012.10.042 . delaPeña,P.,2013.ThebeginningoftheUPintheBaeticMountainarea(Spain). Q.Int.318,69 – 89.doi: . delaPeña,P.,Vega,G.,2013.TheEarlyUpperPalaeolithicpuzzlein MediterraneanIberia.Quartär60,85 – 106.doi: . Dietze,M.,Kreutzer,S.,2017.plot_AbanicoPlot:functiontocreateanAbanico PlotFunctionversion0.1.10.In:Kreutzer,S.,Dietze,M.,Burow,C.,Fuchs,M.C., Schmidt,C.,Fischer,M.,Friedrich,J.(Eds.),Luminescence:Comprehensive LuminescenceDatingDataAnalysis.RPackageVersion0.7.3,. cran/Luminescence/man/plot_AbanicoPlot.html . Dietze,M.,etal.,2016.Theabanicoplot:visualisingchronometricdarawith individualstandarderrors.Q.Geochronol.31,12 – 18.doi: 10.1016/j.quageo.2015.09.003 . Durcan,J.A.,King,G.E.,Duller,G.A.T.,2015.DRAC:doserateandage calculatorfortrappedchargedating.Q.Geochronol.28,54 – 61.doi:http://dx.doi. org/10.1016/j.quageo.2015.03.012 . ArticleNo~e00435 44 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Finlayson,C.,etal.,2006.LatesurvivalofNeandertalsatthesouthernmost extremeofEurope.Nature443,850 – 853.doi: . Finlayson,C.,etal.,2008.Gorham ’ sCave,Gibraltar — thepersistenceofa Neanderthalpopulation.QuatInt181,64 – 71.doi: quaint.2007.11.016 . Finlayson,C.,etal.,2012.Birdsofafeather:Neandertalexploitationofraptors andcorvids.PLoSOne7,e45927.doi: pone.0045927 . Fitzsimmons,K.E.,Hambach,U.,Veres,D.,Iovita,R.,2013.TheCampanian Ignimbriteeruption:newdataonvolcanicashdispersalanditspotentialimpacton humanevolution.PLoSONE8,e65839.doi: pone.0065839 . Fletcher,W.J.,etal.,2010.Millennial-scalevariabilityduringthelastglacialin vegetationrecordsfromEurope.Q.Sci.Rev.29,2839 – 2864.doi: 10.1016/j.quascirev.2009.11.015 . Fortea,J.,Jordá,F.,1976.LaCuevadeLesMallaetesylosproblemasdel PaleolíticoSuperiordelMediterráneoEspañol.Zephyrus26(28),129 – 166 . Fu,Q.,etal.,2015.AnearlymodernhumanfromRomaniawitharecent Neanderthalancestor.Nature524,216 – 219.doi: . Galbraith,R.F.,Roberts,R.G.,Laslett,G.,Yoshida,H.,Olley,J.M.,1999.Optical datingofsingleandmultiplegrainsofquartzfromJinmiumrockshelter,Northern Australia:partI,experimentaldesignandstatisticalmodels.Archaeometry41, 339 – 364.doi: . Galván,B.,etal.,2014.NewevidenceofearlyNeandertaldisappearanceinthe IberianPeninsula.J.Hum.Evol.75,16 – 27.doi: jhevol.2014.06.002 . Guérin,G.,Mercier,N.,Adamiec,G.,2011.Dose-rateconversionfactors:update. AncientTL29,5 – 8 . Giaccio,B.,Hajdas,I.,Isaia,R.,Deino,A.,Nomade,S.,2017.High-precision14C and40Ar/39ArdatingoftheCampanianIgnimbrite(Y-5)reconcilesthetime-scales ofclimatic-culturalprocessesat40ka.Sci.Rep.7,45940.doi: 10.1038/srep45940 . ArticleNo~e00435 45 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Higham,T.F.G.,etal.,2011.PrecisiondatingofthePalaeolithic:anew radiocarbonchronologyfortheAbriPataud(France),akeyAurignaciansequence. J.Hum.Evol.61,549 – 563.doi: . Higham,T.F.G.,etal.,2010.ChronologyoftheGrotteduRenne(France)and implicationsforthecontextofornamentsandhumanremainswithinthe Châtelperronian.Proc.Natl.Acad.Sci.U.S.A.107,20234 – 20239.doi:http:// . Hoffmann,D.L.,Pike,A.W.G.,Wainer,K.,Zilhão,J.,2013.NewU-seriesresults forthespeleogenesisandthePalaeolithicarchaeologyoftheAlmondakarstic system(TorresNovas,Portugal).Q.Int.294,168 – 182.doi: 10.1016/j.quaint.2012.05.027 . Hublin,J.J.,etal.,2012.RadiocarbondatesfromtheGrotteduRenneandSaintCésairesupportaNeandertaloriginfortheChâtelperronian.Proc.Natl.Acad.Sci. U.S.A.109,18743 – 18748.doi: . Huntley,D.,Baril,M.,1997.TheKcontentoftheK-feldsparsbeingmeasuredin opticaldatingorinthermoluminescencedating.AncientTL15,11 – 13.doi:http:// . Huntley,D.,Lamothe,M.,2001.UbiquityofanomalousfadinginK-feldsparsand themeasurementandcorrectionforitinopticaldating.Can.J.EarthSci.38, 1093 – 1106.doi: . Iturbe,G.,Cortell,E.,1992.ElAuriñacienseevolucionadoenelPaísValenciano: CovaBeneitoyRatlladelBubo.In:Utrilla,P.(Ed.),Aragón-litoralMediterráneo: intercambiosculturalesdurantelaprehistoria:enhomenajeaJuanMaluquerde Motes.InstituciónFernandoelCatólico,Zaragoza,pp.129 – 138 . Kehl,M.,etal.,2013.LateNeandertalsatJaramaVI(CentralIberia)?Q.Res.80, 218 – 234.doi: . Kehl,M.,etal.,2016.SiteformationandchronologyofthenewPalaeolithicsite SimadeLasPalomasdeTeba,SouthernSpain.Q.Res.85,313 – 331.doi:http://dx. . Kreutzer,S.,2017.calc_FadingCorr:Applyafadingcorrectionaccordingto Huntley&Lamothe(2001)foragiveng-valueandagiventc.Functionversion 0.4.2.In:Kreutzer,S.,Dietze,M.,Burow,C.,Fuchs,M.C.,Schmidt,C.,Fischer, M.,Friedrich,J.(2017).Luminescence:ComprehensiveLuminescenceDating DataAnalysis.Rpackageversion0.7.3. . Kreutzer,S.,Dietze,M.,Burow,C.,Fuchs,M.C.,Schmidt,C.,Fischer,M., Friedrich,J.,2017.Luminescence:ComprehensiveLuminescenceDatingData ArticleNo~e00435 46 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Analysis.Rpackageversion0.7.3. . Lucena,A.,etal.,2012.LaocupaciónsolutrensedelabrigodeLaBoja(Mula, Murcia,España).ETF(Arqueol)5,447 – 454.doi: etfi.5.2012.8290 . Mallick,S.,etal.,2016.TheSimonsGenomeDiversityProject:300genomesfrom 142diversepopulations.Nature538,201 – 206.doi: nature18964 . Marks,A.,Monigal,K.,Zilhão,J.,2001.ThelithicassemblagesoftheLate MousterianatGrutadaOliveira,Almonda,Portugal.In:Zilhão,J.,Aubry,Th., Carvalho,A.F.(Eds.),LespremiershommesmodernesdelaPéninsuleIbérique. InstitutoPortuguêsdeArqueologia,Lisboa,pp.145 – 154 . Marti,A.,Folch,A.,Costa,A.,Engwell,S.,2016.Reconstructingtheplinianand coignimbritesourcesoflargevolcaniceruptions:anovelapproachforthe CampanianIgnimbrite.Sci.Rep.6,21220.doi: srep21220 . Muller,A.,Clarkson,C.,2016.IdentifyingMajorTransitionsintheEvolutionof LithicCuttingEdgeProductionRates.PLoSOne11,e0167244.doi:http://dx.doi. org/10.1371/journal.pone.0167244 . Murray,A.S.,Wintle,A.G.,2000.Luminescencedatingofquartzusingan improvedsingle-aliquotregenerative-doseprotocol.Radiat.Meas.32,57 – 73.doi: . Murray,A.S.,Thomsen,K.J.,Masuda,N.,Buylaert,J.P.,Jain,M.,2012. Identifyingwell-bleachedquartzusingthedifferentbleachingratesofquartzand feldsparluminescencesignals.Radiat.Meas.47,688 – 695.doi: 10.1016/j.radmeas.2012.05.006 . Murray,A.S.,Wintle,A.G.,2003.Thesinglealiquotregenerativedoseprotocol: potentialforimprovementsinreliability.Radiat.Meas.37,377 – 381.doi:http://dx. . Pääbo,S.,2015.Thediverseoriginsofthehumangenepool.Nat.Rev.Genet.16, 313 – 314.doi: . Pagani,L.,etal.,2016.Genomicanalysesinformonmigrationeventsduringthe peoplingofEurasia.Nature538,238 – 242.doi: . Pettitt,P.B.,Bailey,R.M.,2000.AMSradiocarbonandluminescencedatingof Gorham ’ sandVanguardcaves,Gibraltar,andimplicationsfortheMiddletoUpper ArticleNo~e00435 47 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


PaleolithictransitioninIberia.In:Stringer,C.,Barton,R.N.E.,Finlayson,C. (Eds.),NeandertalsontheEdge:150thAnniversaryConferenceoftheForbes ’ QuarryDiscovery,Gibraltar,OxbowBooks,Oxford,pp.155 – 162 . Pike,A.W.G.,etal.,2012.U-seriesdatingofPaleolithicartin11cavesinSpain. Science336,1409 – 1413.doi: . Prescott,J.R.,Hutton,J.T.,1994.Cosmicraycontributionstodoseratesfor luminescenceandESRdating:Largedepthsandlong-termtimevariations.Radiat. Meas.23,497 – 500.doi: . Preusser,F.,2003.IRSLdatingofK-richfeldsparsusingtheSARprotocol: comparisonwithindependentagecontrol.AncientTL21,17 – 23.doi:http://dx.doi. org/10.7892/boris.86653 . Rasmussen,etal.,2014.Astratigraphicframeworkforabruptclimaticchanges duringtheLastGlacialperiodbasedonthreesynchronizedGreenlandice-core records:refiningandextendingtheINTIMATEeventstratigraphy.Q.Sci.Rev. 106,14 – 28.doi: . Rebollo,N.R.,etal.,2011.Newradiocarbondatingofthetransitionfromthe MiddletotheUpperPaleolithicinKebaraCave,Israel.J.Archaeol.Sci.38, 2424 – 2433.doi: . Reimer,P.J.,etal.,2013.IntCal13andMARINE13radiocarbonagecalibration curves0-50000yearscalBP.Radiocarbon55,1869 – 1887.doi: 10.2458/azu_js_rc.55.16947 . Rivas-Martínez,S.,1987.MemoriadelmapadeseriesdevegetacióndeEspaña1: 400.000.MinisteriodeAgricultura,PescayAlimentación,Madrid . Rodríguez-Vidal,J.,etal.,2014.ArockengravingmadebyNeandertalsin Gibraltar.Proc.Natl.Acad.Sci.U.S.A.111,13301 – 13306.doi: 10.1073/pnas.1411529111 . Sánchez-Goñi,M.F.,etal.,2008.ContrastingimpactsofDansgaard-Oeschger eventsoverawesternEuropeanlatitudinaltransectmodulatedbyorbital parameters.Q.Sci.Rev.27,1136 – 1151.doi: . Sánchez-Goñi,M.F.,etal.,2013.Air-seatemperaturedecouplinginwestern Europeduringthelastinterglacial-glacialtransition.Nat.Geosci.6,837 – 841.doi: . Sanchis,A.,2012.LoslagomorfosdelPaleolíticomedioenlavertiente mediterráneaibérica.Humanosyotrospredadorescomoagentesdeaportey ArticleNo~e00435 48 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


alteracióndelosrestosóseosenyacimientosarqueológicos.Diputaciónde Valencia,Valencia . Santamaría,D.,delaRasilla,M.,2013.DatandoelfinaldelPaleolíticomedioenla PenínsulaIbérica:Problemasmetodológicosylímitesdelainterpretación. TrabajosPrehist70,241 – 263.doi: . Sepulchre,etal.,2007.H4abrupteventandlateNeanderthalpresenceinIberia. EarthPlanet.Sci.Lett.258,283 – 292.doi: epsl.2007.03.041 . Smith,F.H.,Jankovi ,I.,Karavani ,I.,2005.Theassimilationmodel,modern humanoriginsinEurope,andtheextinctionofNeandertals.Q.Int.137,7 – 19.doi: . Stuiver,M.,Reimer,P.J.,1993.Extended14CdatabaseandrevisedCALIB3.014Cagecalibrationprogram.Radiocarbon35,215 – 230.doi: 10.1017/S0033822200013904 . Thiel,C.,etal.,2011.Luminescencedatingofthestratzingloessprofile(Austria) – testingthepotentialofanelevatedtemperaturepost-IRIRSLprotocol.Q.Int. 234,23 – 31.doi: . Trinkaus,E.,2007.EuropeanearlymodernhumansandthefateoftheNeandertals. Proc.Natl.Acad.Sci.U.S.A.104,7367 – 7372.doi: pnas.0702214104 . Trinkaus,E.,Constantin,S.,Zilhão,J.,2013.LifeandDeathatthePes teracu Oase:ASettingforModernHumanEmergenceinEurope.OxfordUniversity Press,Oxford . Trinkaus,E.,Zilhão,J.,2013.PaleoanthropologicalImplicationsofthePes teracu OaseandItsContents.In:Trinkaus,E.,Constantin,S.,Zilhão,J.(Eds.),Lifeand DeathatthePes teracuOase:ASettingforModernHumanEmergenceinEurope. OxfordUniversityPress,Oxford,pp.389 – 400 . Verna,Ch.,Dujardin,V.,Trinkaus,E.,2012.TheEarlyAurignacianhuman remainsfromLaQuina-Aval(France).J.Hum.Evol.62,605 – 617.doi:http://dx. . Walker,M.J.,etal.,2008.LateNeandertalsinSoutheasternIberia:Simadelas PalomasdelCabezoGordo,Murcia,Spain.Proc.Natl.Acad.Sci.U.S.A.105, 20631 – 20636.doi: . Walker,M.J.,etal.,2012.TheexcavationofburiedarticulatedNeanderthal skeletonsatSimadelasPalomas(Murcia,SESpain).Q.Int.259,7 – 21.doi:http:// . ArticleNo~e00435 49 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Wallinga,J.,Murray,A.S.,Wintle,A.G.,2000.Thesingle-aliquotregenerativedose(SAR)protocolappliedtocoarse-grainfeldspar.Radiat.Meas.32,529 – 533. doi: . Welker,F.,etal.,2016.Palaeoproteomicevidenceidentifiesarchaichominins associatedwiththeChâtelperronianattheGrotteduRenne.Proc.Natl.Acad.Sci. U.S.A.113,11162 – 11167.doi: . Wild,E.M.,etal.,2008.14CDatingoftheUpperPaleolithicSiteatKremsHundssteiginLowerAustria.Radiocarbon50,1 – 10.doi: S0033822200043332 . Wood,R.,etal.,2013.Radiocarbondatingcastsdoubtonthelatechronologyof theMiddletoUpperPalaeolithictransitioninsouthernIberia.Proc.Natl.Acad. Sci.U.S.A.110,2781 – 2786.doi: . Zilhão,J.,1993.LepassageduPaléolithiquemoyenauPaléolithiquesupérieur danslePortugal.In:Cabrera,V.(Ed.),ElOrigendelHombreModernoenel SuroestedeEuropa.UNED,Madrid,pp.127 – 145 . Zilhão,J.,2000.TheEbrofrontier:amodelforthelateextinctionofIberian Neanderthals.In:Stringer,C.,Barton,R.N.E.,Finlayson,C.(Eds.),Neandertalson theedge:150thAnniversaryConferenceoftheForbes ’ Quarrydiscovery, Gibraltar,OxbowBooks,Oxford,pp.111 – 121 . Zilhão,J.,2006a.ChronostratigraphyoftheMiddle-to-UpperPaleolithic TransitionintheIberianPeninsula.Pyrenae37,7 – 84 . Zilhão,J.,2006b.TheAurignacianofPortugal:areappraisal.In:Maíllo,J.M., Baquedano,E.(Eds.),MisceláneaenhomenajeaVictoriaCabrera,Vol.I,Museo ArqueológicoRegional,AlcaládeHenares,pp.372 – 395 . Zilhão,J.,2009.TheEbrofrontierrevisited.In:Camps,M.,Szmidt,C.(Eds.),The Mediterraneanfrom50,000to25,000BP:TurningPointsandNewDirections. OxbowBooks,Oxford,pp.293 – 311 . Zilhão,J.,2013.Neandertal-ModernhumancontactinwesternEurasia:issuesof dating,taxonomy,andculturalassociations.In:Akazawa,T.,Nishiaki,Y.,Aoki, K.(Eds.),DynamicsofLearninginNeandertalsandModernHumansVolume1: CulturalPerspectives.Springer,Tokyo,pp.21 – 57.doi: 978-4-431-54511-8_3 . Zilhão,J.,Pettitt,P.,2006.OnthenewdatesforGorham'scaveandthelate survivalofIberianNeanderthals.BeforeFarm.2006(3),95 – 122.doi:http://dx.doi. org/10.3828/bfarm.2006.3.3 . ArticleNo~e00435 50 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Zilhão,J.,etal.,2010a.Symbolicuseofmarineshellsandmineralpigmentsby IberianNeandertals.Proc.Natl.Acad.Sci.U.S.A.107,1023 – 1028.doi:http://dx. . Zilhão,J.,etal.,2010b.DosabrigosdelPaleolíticosuperiorenRamblaPerea (Mula,Murcia).In:Mangado,X.(Ed.),ElPaleolíticosuperiorpeninsular. NovedadesdelsigloXXI.UniversitatdeBarcelona,Barcelona,pp.97 – 108 . Zilhão,J.,etal.,2010c.PegodoDiabo(Loures,Portugal):datingtheemergenceof anatomicalmodernityinwesternmostEurasia.PLoSOne5,e8880.doi:http://dx. . Zilhão,J.,Banks,W.,d ’ Errico,F.,2015.IstheModernvs.Neandertaldichotomy appropriateanylongerforthetechnocomplexesoftheMiddle-to-UpperPaleolithic transition?Abstracts,EuropeanSocietyfortheStudyofHumanEvolution, London,September2015p215 . Zilhão,J.,etal.,2016.CuevaAntón:amulti-proxyMIS3toMIS5a paleoenvironmentalrecordforSEIberia.Q.Sci.Rev.146,251 – 273.doi:http:// . ArticleNo~e00435 51 2405-8440/©2017TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBYlicense (


Precise dating of the Middle to Upper Paleolithic transition in Murcia (Spain) supports late Neandertal persistence in Iberia SUPPLEMENTARY INFORMATION Chapter 1. Geographical setting and archeological context Chapter 2. The cave/rock shelter of Cueva Antn Chapter 3. The rock shelter of Finca Doa Martina Ch ap ter 4. The rock shelter of La Boja


Chapter 1. Geographical setting and archeological context 1 .1 . THE MULA BASIN Cueva Antn and the Rambla Perea rock shelters are situated in the Mula basin , which occupies the geographical cent er of the Autonomous Region of Murcia, Spain (Fig. S1.1). The basin is a ca.700 km tectonic depression — a graben filled with >10 00 m thick Miocene marls — delimited t o the North, South , and West , by the Sierras of Ricote, Espua, and Cambrn, respectively . These mountain ranges ris e above 1000 m , while t he elevation of the basin itself decreases gradually from ca.400 m along the western border to ca.200 m along the e ast ern border . T he basin drains to the valley of River Segura, which communicates it with the littoral plains of Murcia and Campos de Cartagena extending to East and South ( Fig. S1. 1A ). The mean annual temperature is 19.7 C . The mean annual rainfall ranges are 350500 mm, in the Northwest , and 250 350 mm, in the lower elevation badlands extending downstream of the town of Mula. The dominant soil temperature regimes are thermic, and soil moisture regimes are aridic to xeric (Garca-Corts et al., 1999). The current vegetation is a xerophytic brushwood with Artemisia herbaalba, Rosmarinus officinalis and Stipa tenaccisima. Aleppo pine ( Pinus halepensis ) and juniper ( Juniperus phoenicea, J. oxicedrus ) thrive in areas with deeper soils, while Tamarix , Nerium oleander and Phragmites occur along river margins. T he basin’s extant surface morphology is structured by (a) ridges of diverse carbonate lithology (limestone, conglomerate, breccia, calcarenite) resulting from Pliocene and Pleistocene tectonics and (b) the incision of the streams that traverse the basin on their way to the Segura, namely River Mula and its tributaries. The Pleistocene deposits formed in as sociation with these landforms — cave and rock shelter fills, slope deposits, alluvial terraces — are well known for their potential to preserve rich record s of human occupation . Indeed, b y the last decade of the 20th century, Upper Pleistocene archeo logical sites had already been documented to th e Southeast (e.g., Sima de las Palomas de Cabezo Gordo; Walker et al., 2012), West (e.g., the Cueva Negra del Estrecho del Ro Qupar; Walker et al., 2016), and North (e.g . , the cave art sites around Cieza; Salmern et al., 1999). It was not until the 1990 s, however , that the basin ’s Paleolithic archeo logy was first revealed — at Cueva Antn ( Martnez Snchez, 1997). 1. 2. THE RESEARCH PROJECT In 1929, River Mula was dammed just upstream of the town of the same name , resulting in the formation of the La Cie rva reservoir. In the 1990s, the raising of the dam’s wall triggered an assessment of environmental impact . A mong the mitigation measures recommended, that assessment mentioned the archeological testing of a large cave located at the tail of the artificial lake, locally known as Cueva Antn ( Fig. S1.1 B). Eventually carried out in 1991, this salvage operation uncovered Middle Paleo lithic occupation horizons buried in a thick, wellstratified fluviatile sequence . Based on correlation with a 5 7 m terrace dated elsewhere, the excavators estimated that the site’s fill had accumulated between 38 and 40 ka (Martnez Snchez, 1997) .


Interest in Cueva Antn was revived in 200 5 , in the context of preliminary reconnaissance work for a project targeting the Middle to Upper Paleo lithic transition in Iberia (Zilho and Villaverde, 2008; Zilho et al., 2010a). The reasons were twofold: (a) the nature of the sedimentary envelope warranted an expectation of high stratigraphic integrity for both artefacts and ecofacts, while (b) the age suggested by the 1991 study implied that the site’s archeolog ical content might well date to the time range of relevance to test the hypothesis of late Neanderthal persistence in Iberian regions located to the S outh of the Ebro drainage (the Ebro Frontier m odel; Zilho, 1993). Cueva Antn was therefore selected for additional excavation and analysis related to the Middle Paleo lithic component of the project. The Rambla Perea gorge, a deeply incised, meandering valley located ca. 1.5 km n orth of Cueva Antn , was surveyed a t the same time. The watercourse flowing along the bottom of this gorge is the main left tributary of River Mula. P rior to diversion for flash flood control, this stream was active yearround because the permanent spring it originates in, Fuente Caputa, is found in correspondence with a tectonically active escarpment and drains a large underground aquifer . In the initially surveyed section ( Fig. S1. 1 C) , the valley runs almost straight, following a tectonic alignmen t. This disposition determines an asymmetric transverse profile and, along its northern slope, the presence of outcrops of Upper Miocene biocalcarenite that form near vertical, continuous, up to 30 mhigh rock walls . Such morphology favo r ed the formation o f rock shelters , while the ye arround availability of water, control led by local geolog y more than annual rainfall, was unlikely to have been significantly affected by past climatic oscillations . Combined , these factors suggested that, for humans, the Rambla Perea would have been a focal point of the landscape . Hence, it could be expected to contain archeological sites upon which to base the project’s Upper Paleo lithic component. In deed, flints of unambiguous Upper Paleo lithic affinities were seen at the base of one of the gorge’s rock faces during a first visit to the valley carried out on December 11, 2005 ; th is locality was thence designated Finca Doa Martina (FDM), after the owner of the farmstead below ( Fig. S1. 2). Based on the observation that a large and deep but emptied rock shelter had formed downstream in the same bedrock strata, it was inferred that one could also exist, even though completely buried by slope deposits , at the foot of a rock face situated in intermediate position ( Fig. S1. 1 C). S ubsequent testing eventually confirmed the inference, and t his new locality was thence designated as Abrigo de La Boja (ADB; Fig. S1.2 ) , after the name locally given to the abundant wormwood shrubs covering the surface of the site and adjacent slopes . Th e La Cierva reservoir inundated a similar gorge, El Corcovado, akin to the Rambla Perea even though shorter in length . The original depth of the El Corcovado incision can be pictured from the ca.100 m difference in elevation between riverbed and adjacent terrain observed at the damming site . Between this point and Cueva Antn, at the tail of the reservoir, River Mula bridged, over ca.1 km, a d ifference in elevation of some 50 m. In the regional landscape, th ese two nearby gorges play a similar role — communicating the Mula basin with the mid elevation limestone plateau extending to the North of the Sierra de Ricote ridge ( Fig. S1. 1 A) . D uring the Pleistocene, for game and humans moving to, or coming back from this higher terrain, El Corcovado and Rambla Perea would have provided the shortest route s to do so .


Coupled with their proximity, the near identical geographical setting of the three sites targeted by the project meant that access to raw-material and food resources could be held constant in terms o f assessing whatever variation might be observed when comparing their archeo logical records. Thus, any aspects of change through time in technology, spatial patterning of the occupations , or functionality not accounted for by parallel change in the sites’ morphology (e.g., due to erosion, collapse , or filling up) could be assumed to relate to c hange in culture, social organization and settlementsubsistence system — i.e., the realms the Middle to Upper Paleo lithic transition primarily manifests itself in . G iven the favo rable conditions, funding was sought, secured, and a decade of field work ensued, b eginning in 2006 at Cueva Antn with the cleaningup of the debris accumulated at the bottom of the 1991 trench and the re analysis of its exposed cross section s. The time range covered by the correlation and combination of the stratigraphic sequences exposed extends from the Last Interglacial to the very end of the Pleistocene, i.e., from ca.90 to ca. 10 ka . D etails on the results obtained so far can be found in Zilho et al. (2010 b , 2016), Lucena et al. (2012), Angelucci et al. ( 201 3 , 2017 ), Romn et al. (2013), Burow et al. (2015), and references therein. The chapters that follow provide overviews of each of the three sequences and contain supporting, detailed information on the issues of site formation, dating and assemblage definition underpinning current debates on the Middle to Upper Paleo lithic transition in Iberia and beyond (Zilho, 2006; Zilho and Pettitt, 2006; Finlayson et al., 2008; Zilho et al., 2 010c; Kehl et al., 2013; Wood et al., 2013). 1.3. REFERENCES ANGELUCCI, D.; ANESIN, D.; SUSINI, D.; VILLAVERDE, V.; ZAPATA, J.; ZILHO, J. (2013) — Formation processes at a high resolution Middle Paleolithic site: Cueva Antn (Murcia, Spain) . Quaternary I nternational, 315, p. 24 41. ANGELUCCI, D.; ANESIN, D.; SUSINI, D.; VILLAVERDE, V.; ZAPATA, J.; ZILHO, J. (201 7 ) — A tale of two gorges: Late Quaternary site formation and surface dynamics in the Mula basin (Murcia, Spain). Quaternary International ( ). BUROW, C.; KEHL, M.; HILGERS, A.; WENIGER, G.C.; ANGELUCCI, D. E.; VILLAVERDE, V.; ZAPATA, J.; ZILHO, J. ( 2015 ) — Luminescence Dating of Fluvial Deposits in the Rock Shelter of Cueva Antn, Spain. Geochronometria, 42, p. 107 125. FINLAYSON, C.; FA, D. A.; JIMENEZ ESPEJO, F.; CARRI”N, J. S.; FINLAYSON, G.; GILES PACHECO, F.; RODRGUEZ VIDAL, J.; STRINGER, C. B.; MARTINEZ RUIZ, F. (2008) — Gorham’s Cave, Gibraltar — The persistence of a Neanderthal population. Quaternary International, 181, p. 64 71. GARCA CORTS, .; GALLEGOVALCARCE, E.; BARETTINO FRAILE, D. (eds.) (1999) — Atlas del medio natural de la Regin de Murcia. Murcia, ITGE an d Regin de Murcia. KEHL, M.; BUROW, C.; HILGERS, A.; NAVAZO, M.; PASTOORS, A.; WENIGER, G.C.; WOOD, R.; JORD PARDO, J. F. (2013) — Late Neanderthals at Jarama VI (Central Iberia)? Quaternary Research, 80, p. 218 234.


LUCENA, A.; MARTNEZ, S.; ANGELUCCI, D. E.; BADAL, E.; VILLAVERDE, V.; ZAPATA. J.; ZILHO, J. (2012) — La ocupacin solutrense del abrigo de La Boja (Mula, Murcia, Espaa) . Espacio, Tiempo y Forma. Serie I, Nueva poca. Prehistoria y Arqueologa, 5, p. 447 454. MARTNEZ SNCHEZ, C. (1997 ) — El yacimiento musteriense de Cueva Antn (Mula, Murcia) . Memorias de Arqueologa de la Regin de Murcia, 6, p. 18 47. ROMN, D.; ZILHO, J.; MARTN LERMA, I.; VILLAVERDE, V. (2013) — La ocupacin epimagdaleniense del abrigo de la Finca de Doa Martina (Mula, Murcia) , in DE LA RASILLA, M. (ed.) — F. Javier Fortea Prez. Universitatis Ovetensis Magister . Estudios en homenaje, Universidad de Oviedo/Mnsula Ediciones, p. 167 178. SALMER”N, J.; LOMBA, J.; CANO, M. (1999) — El arte rupestre paleoltico de Cieza: primeros hallazgos en la Regin de Murcia. Resultados de la I campaa de prospecciones Losares Almadenes 93 . Memorias de Arqueologa de la Regin de Murcia, 8, p. 93 111. WALKER, M. J.; L”PEZ MARTNEZ, M. V.; ORTEGA RODRIGEZ, J.; HABER URIARTE, M.; L”PEZ JIMNEZ, A.; AVILSFERNNDEZ, A.; POLOCAMACHO, J. L.; CAMPILLO BOJ, M.; GARCATORRES, J.; CARRI”N GARCA, J. S.; SAN NICOLSDEL TORO, M.; RODRGUEZ ESTRELLA, T. (2012 ) — The excavation of buried articulated Neanderthal skeletons at Sima de las Palomas (Murcia, SE Spain) . Quaternary International, 259, p. 7 21. WALKER, M. J.; ANESIN, D.; ANGELUCCI, D. E.; AVILS FERNNDEZ, A.; BERNA, F.; BUITRAGOL”PEZ, A. T.; FERNNDEZ JALVO, Y.; HABER URIARTE, M.; L”PEZ JIMNEZ, A.; L”PEZ MARTNEZ, M.; MARTN LERMA, I.; ORTEGA RODRIGEZ, J.; POLOCAMACHO, J. L.; RHODES, S. E.; RICHTER, D.; RODRGUEZ ESTRELLA, T.; SCHWENNINGER, J. L.; SKINNER, A. R. (2016) — Combustion at the late Early Pleistocene site of Cueva Negra del Estrecho del Ro Qupar (Murcia, Spain) . Antiquity, 90, p. 571 589. WOOD, R.; BARROSO-RUZ, C.; CAPARR”S , M.; JORD PARDO, J. F.; GALVN SANTOS, B.; HIGHAM, T. F. G. (2013) — Radiocarbon dating casts doubt on the late chronology of the Middle to Upper Paleo lithic transition in southern Iberia. Proceedings of the National Academy of Sciences USA, 110 (8), p . 27812786. ZILHO, J. (1993) — Le passage du Palolithique moyen au Palolithique suprieur dans le Portugal , in CABRERA, V. (ed.) — El Origen del Hombre Moderno en el Suroeste de Europa, Madrid, Universidad Nacional de Educacin a Distancia, p. 127 -1 45. ZILHO, J. (2006) — Chronostratigraphy of the Middle to Upper Paleolithic Transition in the Iberian Peninsula. Pyrenae, 37, p. 7 84. ZILHO J.; PETTITT, P. (2006) — On the New Dates for Gorham's Cave and the Late Survival of Iberian Neanderthals . Be fore Farming: [online version], 2006/3, article 3; [print version], 2006/3, p. 95122. ZILHO, J.; VILLAVERDE, V. (2008) — The Middle Paleolithic of Murcia . Treballs d’Arqueologia, 14, p. 229248.


ZILHO, J.; ANGELUCCI, D.; BADAL, E.; LUCENA, A.; MARTN, I.; MARTNEZ, S.; VILLAVERDE, V.; ZAPATA, J. (2010a) — Dos abrigos del Paleoltico superior en Rambla Perea (Mula, Murcia) , in MANGADO, X. (ed.) — El Paleoltico superior peninsular. Novedades del siglo XXI, Barcelona, Universidad de Barcelona, p. 97 108. ZILHO, J.; ANGELUCCI, D.; BADAL GARCA, E.; d’ERRICO, F.; DANIEL, F.; DAYET, L.; DOUKA, K.; HIGHAM, T. F. G.; MARTNEZ SNCHEZ, M. J.; MONTES BERNRDEZ, R.; MURCIA MASCAR”S, S.; PREZ SIRVENT, C.; ROLDN GARCA, C.; VANHAEREN, M.; VILLAVERDE, V.; WOOD, R.; ZAPATA, J. (2010b) — Symbolic Use of Marine Shells and Mineral Pigments by Iberian Neandertals . Proceedings of the National Academy of Sciences USA, 107 (3), p. 1023 1028. ZILHO, J.; DAVIS, S. J. M.; DUARTE, C.; SOARES, A. M. M.; STEIER, P.; WILD , E. (2010c) — Pego do Diabo (Loures, Portugal): Dating the Emergence of Anatomical Modernity in Westernmost Eurasia. PLoS ONE, 5 (1), e8880 (doi:10.1371/journal.pone.0008880). ZILHO, J; AJAS, A.; BADAL, E.; BUROW, Ch.; KEHL, M.; L”PEZ SEZ, J. A.; PIME NTA, C.; PREECE, R. C.; SANCHIS, A.; SANZ, M.; WENIGER, G.C.; WHITE, D.; WOOD, R.; ANGELUCCI, D. A.; VILLAVERDE, V.; ZAPATA, J. (2016) — Cueva Antn: A multi proxy MIS 3 to MIS 5a paleoenvironmental record for SE Iberia. Quaternary Science Reviews, 146, p. 251 273.


Chapter 2. The cave/rock -shelter of Cueva Antn 2.1. INTRODUCTION Cueva Antn (3803’52” N, 0129’47” W) is located on the right bank of the River Mula, in the external side of the tight meander through which , prior to construction of the La Cierva dam, the stream entered the now submerged El Corcovado gorge ( Fig. S1. 1 ; Fig. S2. 1). First investigated in 1991 in the context of a salvage operation (Martnez Snchez, 1997), the site was the target of extensive excavation between 2006 and 2012 . This work established the stratigraphic, chronometric and paleoenvironmental background to human occupation (Zilho et al., 2010, 2016; Angelucci et al., 2013; Burow et al., 2015) ( Figs. S2. 2-S 2.10) . Detailed information on the methodology followed in the excavation and analysis of the site can be found in the published literature . Here, the focus lies on its archeo logical content , specifically the presentation and discussion of the late Middle Paleolithic context found in uppermost layer I k. The evidence from the next archeologically fertile unit found as one moves down in the sequence, layer II -l , is also considered. Even though this unit significantly predat es layer I k, the comparison highlights the extent to which , regardless of change through time in settlement, subsistence, technology , or social organization, spatial constrain ts conditioned the site’s usefulness as a shelter for people, with attendant implications for the formation of an archeo logical record . Following Zilho et al. ( 2016) , fluviatile sediments began to accumulate in Cueva Antn during the early or middle part of Marine Isotope Stage (MIS) 5 . They hypothesize that the process would have been triggered by loss to erosion of the external wall of a large, preexisting karst fissure. Its interior having thus became exposed to penetration by the river, a fluviatile succession could accumulate inside and be subsequently preserved as a terrace under shelter deposit hanging a few me ters above the present streambed . Th is riveraccumulated deposit forms complex AS , which is sandwiched between basal, endokarst accumulated palustrine sediments ( complex FP ) and the silts of variable thickness that, in recent times, during periods of long term submersion by the La Cierva reservoir, blanketed the Pleistocene succession (complex DD) . Due to its location inside a karst feature , the sediments of the AS complex also include variable amounts, increasing with proximity to the back wall, of material derived from the degradation of the surrounding bedrock. A summary stratigraphic outline of the succession is provided in Table S2. 1, and the reference stratigraphic cross section is illustrated in Fig. S2. 4. The lateral and vertical variation observed across the excavated area of the site can be appreciated from the additional cross section and dcapage information provided in Figs. S2. 5S2. 9. A schematic 3D model of the accumulation process as reconstructed for the central part of the site is provided in Fig. S2. 10. Based on the erosional discontinuities identified, complex AS has been sub divided into five sub complexes that can be succinctly described as follows:


AS5 Basal s andy beach deposit containing the remain s of two stratigraphically wellseparated episodes of human occupation appearing as dense, spatially extensive, lenticular scatters of stone tools, animal bone and wood charcoal structured around wellpreserved hearth features — layers IIIi/j and II Ib/d. AS4 Sterile remnant of an eroded sandy beach deposit accumulated after a sedimentation hiatus denoted by the paleosoil formation seen in layer II -u , which caps AS5 . AS3 Sterile sands, silts and clays for the most part accumulated in a boggy riverside context. AS2 Gravel, sands and silts featuring , in layer IIl, a single episode of human occupation , represented by a low density scatter of stone tools contained in river-accumulated sands whose trampling eventually erased the original cross bedded structure . AS1 Sterile, riveraccumulated sequence of lens es of sand and silt , capped by a dense breccia — layer I k (laterally continuing, against the back wall, as the 1991 defined units I g and I h) — made of small, angular, walldegradation clasts packed in a sandy matrix and containing a low density scatter of stone tools, animal bones and charcoal. Coupled with the paleoe nvironmental information derived from terrestrial mollusk s, wood charcoal, and pollen, the OSL ages for the AS2 to AS5 sequence suggest that layer II l dates to the very end of MIS 5a , between 70 and 75 ka ( Fig. S2. 3). The chronology of sub complex AS1 is based on the radiocarbon dating of wood charcoal samples processed wit h the rigorous ABOx SC protocol. Consistency with stratigraphic order and the quality of the samples’ chemistry warrant the reliability of the se radiocarbon results , which place AS1 in the 35.1 37.7 ka interval ( Table 1; Zilho et al., 2016 : Table 2). Combined, the OSL and radiocarbon dat es impl y a hiatus of some 35 ,000 years during which no sedimentation occurred at the site . In agreement with the geometry of the units making up this part of the succession , the occurrence of such a hiatus is consistent with valley incision during MIS 4 and early MIS -3. A long the EW axis of the site, layer II l was found across the area taken down to its depth ( Fig. S2. 11 ) ; s agit t ally, however, it wedged out along the slope of the embankment formed as the channel migrated out ( to the inner side of the me ander, at the end of MIS 5a ) and downward ( forming the Cueva Antn terrace, during MIS -4). Inside the cave, it was not until Greenland Interstadial ( GI ) 8 that the fluviatile dynamics resumed, as shown by the texture of the basal AS1 units ( layers Ii, Ij, IIa, IIc and II -b) . Unlike AS2, however, these units contain no cobble, gravel or coarse sand . They are entirely made up of fine sediments , which is consistent with valley incision and a decrease in the energy of the accumulation relative to MIS 5 t imes : e ven though occasionally inundated by overflowing water s, the cave would now be too high above the streambed for the deposition of channel bottom material to be possible.


The basal AS1 alluvium (layers II a and II b) fills up a rill that, outward, ab uts the ridge of the talus created by post AS2 valley incision ( Figs. S2. 7, S2. 9S2. 10). This rill shows that erosional processes were active during the previous hiatus . M ostly , they would have consisted of runoff and limited overland flows, possibly originating in the large, upwardly oriented fissures and joints observed along the back wall, which may hav e functioned as temporary outl ets draining the plateau above . All of this is especially apparent in the western part of the site , to which l ayers I -i , I j and I k did not extend . The primarily non fluvial nature of layer I k and its chronological proximity to underlying layer II a mean that its delimitation can be taken as a proxy for the outline of the river margin through the time of accumulation ( Fi g. S2. 10) . After GI 7, a new round of valley incision left the fill of Cueva Antn exposed as ground surface ever since . Coupled with the impact of repeated inundation since the construction of the La Cierva dam, this longterm exposure must have implied the loss to erosion of a significant thickness of the uppermost Pleistocene deposit, well apparent in the outline of the DD/I k contact as seen in the J>I20 22 cross section ( Fig. S2. 5 ). That such losses were limited can be put down principally to the coar se, densely packed, and carbonate indurated nature of the unit . T he rim of large blocks and boulders accumulated in rows 18 19 of the grid that can be appreciated in Fig. S2. 7 must also have played a role in protecting the AS1 deposit against erosion . The components of this rim all lie at the interface between layer I k and the immediately underlying alluvium, suggesting that they stand for a single event of roof collapse , probably of seismic origin. Indeed, we know that multiple ton boulders fell at the ea stern end of the cavity before the beginning of the 2006 2012 work but after 1991 ( they are not visible in photos taken during that year’s excavation ). Like the walls and roof above the inundation line, the old collapse is uniformly soot stain ed because, in the 1930s , the cave sheltered a bread baking oven . The recently collapsed material, however, exposes fresh , non -st ained breaks. The event responsible for this latest episode of roof fall is likely to have been the seism of magnitude 4.8 (Richter scale) that hit the region on February 2, 1999 ( ). This earthquake had a rather close to the surface hypocenter (1.1 km deep), a nd an epicenter some 5 km north of the town of Mula, i.e., <2 km from Cueva Antn. The boulder rim skirting layer I k could have been produced in similar manner. No archeo logically relevant burrow features were identified during the excavation of Cueva Antn. T he large chamber well apparent in the e ast wall of the 1991 trench and whose inward continuation was clearly delimited in the 2011 excavation of the I20 “telephone booth” and adjacent squares ( Fig. S2. 5 ) is no exception. The mouth of that chamber was in the SE quadrant of grid unit H19 and cut through the upper part of AS2 down to layer II i, without reaching layers IIk and II l. Having affected neither layer I k nor layer IIl, this feature is therefore of no consequenc e for the assessment of the stratigraphic integrity of the archeo log y buried in the units of concern here .


2.2. THE ARCHEO LOGY OF LAYER II -l Excluding the ubiquitous rabbits, which, through the entire Cueva Antn sequence , are accumulated by the eagle owl and , hence, part of the natural background, the faunal remains retrieved in layer II -l total less than 30 specimens . Of these, only a handful could be taxonomically assigned — to Equus sp. and Cervus sp. (Zilho et al., 2016: Table 5). Surface condition s range from the slight polish and smoothing apparent in the equid tooth illustrated in Fig. S2. 12 to some significant rounding, in a few cases entailing loss of surface morphology . These features suggest an a lluvial origin. However, the fact that the deposit is affected by trampling confirms exposure as ground surface for some length of time and is consistent with humans having occupied the place once the accumulation of sediment s had ceased. Much the same is implied by t he condition of the associated stone tools ( Figs. S2. 13 S2. 15) . Even though patinated to varying degree s, they display fresh edges , indicating that they were discarded in situ, not transported — as otherwise implicated by the refit ting evidence . Together, this information is suggestive of the sporadic human occupation of a narrow stretch of frequently inundated riverside terrain. O f the 35 m area wh ere the excavation went down to or beyond the elevation of layer IIl, the latter only existed over some 29 m. I ts total extent probably is not much more than that because (a) layer II m can be observed outcropping directly under the DD deposit west of column Q of the grid, while (b) the inclination of the bedrock ( Fig. S2. 2) implies that layer IIl is unlikely to be present east of column G. We can therefore reconstruct the area available for settlement as a ca.50 m crescent shaped beach tucked under the cave’s roof and isolated from the surrounding slopes — except when lowered river levels left a dryland path between the wa ter and the vertical escarpment extending upriver of the cave opening ( Fig. S2. 1) . No archeology is reported from the 1991 excavation of layer II l. However, t he grid units excavated in 2006 2012 that are adjacent to that year’s trench yielded stone tool finds of substantial size (e.g., the K19 core and the J19 sidescraper of Fig. S2. 14 , nos. 1 2) . Therefore, the 1991 pattern is more likely to reflect expedient excavation in a salvage context than genuine absence and , w hen computing the density of the artefacts’ distribution ( Tables S2. 2S2. 3) , the 9 m then excavated ought to be excluded from consideration. Even so, the value is very low: 44 items over some 20 m, i.e., 2.2/m . Except for the immediately available limestone component (a tested cobble, refitted flake blanks , and one retouched tool; Fig. S2. 15), the lithic assemblage of layer IIl is mostly imported. Only 18% are knapping debris and, b earing in mind the recurrent inundation of the place , the six items of flint chippage in II -l may well be, like the large herbivore bones and teeth, a transported, natural component of the deposit: all are very small, ranging in mass between 0.02 and 0.36 g , and two are indeed rolled. On the other hand, the K19 14 ( Fig. S2. 14 , no. 1) centripetal/discoid core probably represents a lost object because t he flint variet y it is made of is not represented among the blanks . T his volume therefore appears not to have been reduced at the site — unless the activity took place in side the area of the 1991 trench , and the c orresponding products , byproducts and debris went unnoticed at the time.


T he only secure instance of on site flint knapping in layer II l is that documented by item P20 -7 , a sidescraper re sharpening flake ( Fig. S2. 13 ). A nother small flake of identical size and made on the same raw material (N20 5) , retrieved nearby, may have been produced during the same knapping event. The resharpened tool, however, was not found . Likely , it was taken away by its owner upon leaving the site. T he large, distolaterally broken J19 18 sidescraper is the only formal tool of the assemblage that features a break ( Fig. S2. 14, no. 2) . T he missing part was not found, so it cannot be ascertained whether the object fractured on or off site . Subsequ ent on site use, or re use, is nevertheless indicated by macros copic wear showing that the active part of the tool ( represented by the sidescraper retouched edge ) extended to and along the break . The remaining finds are either unretouched flakes and flake fragments (seven) or sidescrapers (five). We cannot exclude that these items reflect loss or discard with no onsite use, as seems to b e the case with the K19 14 core . The two documented instances of on site re sharpening and/or re use suggest , however, that this is unlikely. Even though t he lithics have yet to be analyzed for microscopic wear, these data imply that tasks involving the use of stone tools were carried out in the context of the layer IIl occupation(s), no matter how sporadic and transient they must have been . Given the apparent absence of anthropogenic faunal remains, such tasks may well have been limited to repair and replacement — of clothing, hunting equipment and /or travel gear. Even i f no supporting dating evidence had been available , there c ould be no doubt that, from both the technological and the typological standpoints, this assemblage, small as it is , belongs fully in the Middle Paleolithic. I n the absence of refits or characteristic by products, abandoned cores such as K19 14 can represent the end point of either discoid or Levallois recurrent centripetal reductions . Both are diagnostically Middle Paleolithic blank production methods. Moreover, in Western Eu rope, no assemblages of similar size consisting entirely of sidescrapers, notches and unretouched flake blanks have ever been found with in Upper Paleo lithic stratigraphic sequences, or radiometrically dated to th at period . In short, (a) the small size of t he layer II l assemblage must relate to the constraints on human use imposed by the site’s then extant spatial configuration , and (b) such a small size i s no impediment for the unambiguous manifestation of the technological ly diagnostic concepts underpinni ng raw material economy and blank production. 2. 3 . THE ARCHEO LOGY OF LAYER I -k Excluding the bones of birds, micromammals and rabbits accumulated by the eagleowl, layer I k yielded 34 faunal remains. O nly a cervid, probably Cervus elaphus , could be determined . One long bone splinter belongs to a larger herbivore , probably an equid ( Zilho et al., 2016: Table 5). The edges and the surfaces of the bones are fresh, but none is burnt and no cut marks have been identified . Therefore, th is small assemblage may well be unrelated to the human occupation, and reflect instead episodic use of the site by carnivores. That most of the remains come from the eastern part of the layer’s excavated area, which yielded no artefacts , supports this inference ( Fig. S2. 16) .


Of the ca.54 m excavated in the Z L columns of the grid, layer I k existed in some 36 m. To the West , the boundaries observed in plan as much as in cross section view s suggest that the layer extended into the area of the 1991 trench only marginally. It is therefore entirely plausible that the lack of finds from I-k among that year’ s faunal and stone tool collections stands for a genuine absence. However, the distribution of refitted flint items links grid units J19 and K19 , excavated in 20 08, with I20, excavated in 2011. This connection suggests that products and/or byproducts of the reduction of the same block could have been present, even though missed, in the intermediate grid unit J20, excavated in 1991 (Figs. S2. 16S2. 18) . Given these data, t he density values one can derive from the stone tool counts in Table S2. 4 are of the same order of magnitude as those for layer II -l : 0.9/m for the faunal remains, 0.6/m for the lithics. E xcluding the 1991 trench and postulating that the grid unit s east of column D, which yielded no stone tools, lied outside of the occupied area, the total surface decreases to 25 m . Using this denominator brings the density up, but not significantly : to 1.4/m and 0.8/m for fauna and lithics, respectively. Such low numbers mean that the finds made in layer Ik could well stand for a few sporadic, pass through occupations, if not a single such episode of site use — one whose remains were then syn depositional lly scattered along the E W dip of the stratification by runoff and gravity . This evidence further suggests that, like the fauna, the abundant but quite dispersed wood charcoal found in layer I -k may well be largely non anthropogenic . There can be no question, however, that its stratigraphic associat ion w ith the layer’s stone tools is reliable . Indeed, of the 90 pieces of charcoal that underwent anthracological examination, all those that could be classified belong to Pleistocene taxa ( Juniperus sp. and Pinus sylvestris/nigra) ( Zilho et al., 2016: SI Table 2). The absence of such Holocene taxa as Olea sp. or Pinus hal epensis is matched by the results obtained with the radiocarbon dating of samples selected from th e examined assemblage: of the 11 that were ABOx -processed , seven failed due to low yields , and the four successful ones yielded Pleistocene results consistent with expectations derived from the characteristic s of the associated stone tools . The absence of intrusive charcoal can also be explained by the truncation undergone by layer I -k — its uppermost reaches, those that, through long-term exposure as ground floor, would have been affected by trampling and reworking, must have been eroded away ( most recently because of river damming and the attendant submersion of the site ). Given its scarcity in archeo logical remains, layer I k was excavated as a single field unit , with no internal spit subdivision . Compared with th e values recorded for the surface and base of the layer, the finds’ elevation data, however, indicate that most if not all come fro m its lower reaches, if not the very bottom . In grid unit I20, for instance, the elevation of two of the items included in this layer’s refit (I20 6 and I20 7; Fig. S2. 18) could be precisely noted : 117 cm below datum . This is 1 cm above the base of the uni t at the cent er of the NW quadrant where th ose items came from , and 5 10 cm below the surface of the unit as estimated from the values for surrounding points . Likely , therefore, the erosion al truncation of the upper part of layer I k entailed no significant loss of this unit’s original archeo logical content .


O ne of the ABOxdated samples from layer Ik also came from grid unit I20 . I t was retrieved in the NE quadrant and at an elevation of 115 cm below datum, closer to th e surface of the deposit and above the flint material. In contrast, the other sample came from the very base of layer I k in grid unit G21, 2.3 m away ( Fig. S2. 16 ). Since there is no reason to suppose that the charcoal scatter originates in a single depositional event, or that it must be of one and the same age, this spatial information carries two implications: firstly, that the result obtained for the G21 sample best reflect s the time of occupation ; secondly, that an upper boundary for such a time is provided by the terminus ante quem represented by t he result for the I20 sample . Be it as it may, both results are consistent with the terminus post quem represented by the result obtained for a sample (E21 11) from immediately underlying layer II-a. Given these stratigraphic constraints and the samples’ ages (Table 1) , the artefacts contained in layer Ik must have been used and/or discarded at the site no earlier than 37. 1 ka . This late date has led some to question the h omogeneity and definition of layer Ik (Wood et al., 2013 ). However, the refitting evidence ( Fig. S2. 18A C) concurs with the absence of intrusive charcoal in supporting the layer’s stratigraphic integrity, and n o Upper Paleolithic stone tool types are present in the lithic assemblage, in which diagnostically Upper Paleo lithic blank production methods are not represented either. No indication exists, therefore, that we might be dealing with a palimpsest containing both Middle and Upper Paleolithic material and that the dated charcoal might relate to the latter instead of the former . T he Middle Paleo lithic affinities of the stone tool assemblage are , in turn, undeniable : e ach one of th os e methods that are documented would suffice, in and of itself, to support assignment to the Middle Paleolithic , and the more so since they co occur . Such methods are: Centripetal (Discoid or Levallois) , represented by the J19 4 core and its products and byproducts, most of which could be refitted back ( Fig. S2. 1 7, no. 1; Fig. S2. 1 8A -C); Discoid, represented by the blank for the D20 2 notched piece, which is made on a recycled , discoid core trim ming byproduct ( Fig. S2. 17, no. 5 ); Levallois, represented by the H21 8 laminar flake ( Fig. S2. 18D E, Fig. 19, no. 1 ); Kombewa, represented by the F18 1 core ( Fig. S2. 17, no. 4 ); the product ion of naturally backed, orange segment like flakes, probably deliberately overshot discoid blanks, repre sented by the E19 7 denticulate ( Fig. S2. 17, no. 3) ; f or flint, th is method is unknown in the region al Upper Paleo lithic but well documented in its Mousterian (see Chapter 3 ). The other core is a “splintered piece” documenting the use of the bipolar technique ( Fig. S2. 17, no. 2), also represented by the useworn, unretouched blank F19 1 ( Fig. S2. 19, no. 2 ) . Bipolar reduction can also be found in the regional Upper Paleo lithic (see Chapter 4 ) but , given its ubiquity across the culture stratigraphic sequence of Southwestern Europe’s Upper Pleistocene (Aubry et al., 1997), it remains entirely consistent with the Middle Paleo lithic affinities of everything else .


I n a context of such short, sporadic visits to the place ( by small groups and/or isolated individuals ) as those recorded in layer I k, one would expect the stone tool s used at the site ( as opposed to simpl y therein lost or discarded ) to have been mostly involved in repairing and re tooling tasks involving the processing of wood . Th e macro or microscopic use wear evidence is consistent with this notion ( Table S2.5 ). Contact with bone or antler is documented in addition to wood in the case of the H21 8 Levallois blank ( Fig. S2. 19, no. 1) . However, given that the layer’s animal bone is likely to be non anthropogenic, that wear may well represent prior, offsite use of one of this imported object’s cutting edges. As with layer II l, the small size of the layer I k assemblage must relate to the limited accessibility of the site and its general unsuitability to serve as a shelter for a significant number of people , or a significant length of time — the more so for layer I k, as it formed after valley incision had left the cave hanging above the streambed at the base of a 20 m high, vertical cliff-face. We could reconstruct layer II l as a crescent shaped 50 m beach tucked under the cave’s roof that, most of the time, would have had no dryland access. From Figs. S2. 10 and S2. 16 we can reconstruct layer I k as a narrow, sloping tongue of angular limestone gravel extending across a similar surface — a band of terrain no more than 5 m wide running along the back wall of the cave, access to which would have been subject to the sam e constraints as during layer IIl times . No wonder, therefore, that, functionally, despite being 35,000 years apart, the two occupations are strikingly similar. Such a similarity is also apparent in the realm of raw material economy, as the I k refit unit matches the II l episode of limestone knapping in documenting the association of imported items with the exploitation of locally available stuff. The flint variety represented in the I k refit has primary sources a few hundred meters away, in the valley’s Cretaceous limestone bedrock, where it occurs as narrow, tectoniz ed bands featuring much cleavage. Such structure is well apparent in the block that was reduced at Cueva Antn, and is one that renders this raw material unsuitable to produce large blanks u sing complex reduction methods. These characteristics explain well why the knapping episode documented in layer I k apparently produced a single usable piece ( the flake extracted at the end of the sequence), which we could not find and would seem to have b een exported. The above has implications for the interpretation of the K19 3 perforated, ochre stain ed halfvalve of Pecten maximus ( Fig. S2.20 ) found in close spatial association with the J19 4 core and some of its products and byproducts ( Fig. S2. 16). Ochre may have utilitarian functions , namely in the preparation of hides, but no wear related to such functions has been identified in the lithics from layer I -k , and none is ochre stained. This evidence rejects interpretations of the shell ’s pigmentation as accidental, i.e., as resulting from post depositional accumulation of iron oxides brought in for hide processing tasks or locally p roduced by diagenetic processes. The additional evidence acquired since publication of the shell’s age, association, dating and symbolic significance (Zilho et al., 2010) therefore supports the original argument: the parsimonious interpretation of the Cueva Antn Pecten remains that it enter ed the site as a painted/perforated item of body decoration therein discarded after breaking.


2.4. CONCLUSION At the site, the layer I k assemblage is by no means unique . Its characteristics relate to the exceptional ly high resolution of the Cueva Antn archeo logical record, which enabled the preservation as separate stratigraphic units of contexts that either result from single, sporadic occupation events leaving a limited number of remains , or stand for the accumulated result of a small number of such eve nts occurring within a very short time span. In contrast, the site formation frameworks most commonly encountered when dealing with cave and rock shelter sites are characteriz ed by lower sedimentation rates , entailing the formation of stratigraphic sequences of less er resol ution . In such “normal” frameworks, substantial assemblages can b uild up even if each of the individual occupation events subsumed therein is of the same kind as those recorded in layers Ik and II l of Cueva Antn . T he reason is simple: in such frameworks, the time span represented by the accumulation is much longer . At times of cultur al or technological transition, low sedimentation rates may end up in the formation of stratigraphic units containing assemblages that feature the different terms, o r steps , of such a transition . In that situation , disentangling the different component s and the meaning of their co occurrence can only be made against an external frame of re ference. Put another way, when sedimentation is slow, establishing whether the co occurrence of different technological systems stand s for their true integration , for their true coexistence, or for a palimpsest effect can only be done post hoc, against previously acquired knowledge. A t Cueva Antn, rapid build up is documented by the radiocarbon dating of sub complex AS1 and the sedimentological characteristics of last interglacial sub complexes AS2 to AS5 (Zilho et al., 2016) . In such a context, it is entirely to be expected that, large or small, the artefact assemblages retrieved from individual stratigraphic units will reflect use and discard within a restricted time window and, hence, that, in culture stratigraphic terms, their content will be homogeneous . Thus, questioning the Middle Paleo lithic nature of layer I k would be no more warranted than doing it for layer II l, which yielded an assemblage of similar size and bearing the same kinds of technological diagnostics. When the two assemblages are compar ed, t he only difference resides in that th e retouched tools made on flint are all sidescrapers in II-l and all notches, denticulates , or use worn, unretouched blanks in Ik. Given the small size of the assemblages, the significance of this difference, which may well be of a functional nature, is d ifficult to assess. I n this respect, i t is nevertheless worth noting that, elsewhere in Iberia, the few secure, very late Middle Paleo lithic contexts known also feature very low percentages of retouched tools . T h os e that we do find in such contexts are , a s in layer Ik of Cueva Antn, only or mostly notches and denticulates. Layer 8 of Gruta da Oliveira (Almonda karst sys tem , Torres Novas, Portugal), placed by radiocarbon in the 35.6 38.6 ka interval, is a case in point . H ere , t he age of the assemblage is independently supported by U series results for the same layer ( Hoffmann et al., 2013) and, among a total of 84 blanks greater than 2.5 cm , only five — all of which are discontinuously notched/denticulated pieces — were retouched (Marks et al., 2001).


A similar pattern is apparent at the open air site of Foz do Enxarrique (Vila Velha de Rdo, Portugal), where the archeo logical context is buried in rapidly accumulated inundation silts belonging to the 510 m alluvial terrace of the Tagus , dated by OSL t o the 31 40 ka interval (Cunha et al., 200 8, 2012). N early 10,000 artefacts were recovered at this site, where, per Brugal and Raposo (1999: 369 370), “the industry is characterized by numerous discoid and Levallois recurrent centripetal cores ” and “retouc hed tools are dominated by notches and denticulates, sidescrapers are rare and tools of ‘Upper Paleo lithic’ type are virtually absent.” Layer I k therefore does not “walk alone.” Dating error and/or problems of assemblage definition ha ve been shown to underpin the attribution of some Iberian occurrences to a late persisting Middle Paleo lithic (Zilho, 2006; Wood e t al., 2013; Kehl et al., 2013). However, t he high resolution of the Cueva Antn stratigraphic sequence and the quality and accuracy of the radiocarbon dating of sub complex AS1 and, within it, layer I-k, confirm that, at least in South east Spain, the Middle Paleo lithic did not disappear — and, hence, neither did its Neandertal makers — until after some 37,100 years ago. 2. 5 . REFERENCES ANGELUCCI, D.; ANESIN, D.; SUSINI, D.; VILLAVERDE, V.; ZAPATA, J.; ZILHO, J. (2013) — Formation processes at a high resolution Middle Paleolithic site: Cueva Antn (Murcia, Spain) . Quaternary International, 315, p. 24 41. AUBRY, Th.; ZILHO, J.; ALMEIDA, F.; FONTUGNE, M. (1997) — Production d’armatures microlithiques pendant le Palolithique suprieur et le Msolithique du Portugal , in BALBN, R.; BUENO, P. (eds.) — II Congreso de Arqueologa Peninsular. Paleoltico y Epip aleoltico, Zamora, Fundacin Rei Afonso Henriques, p. 259 272. BRUGAL, J. P.; RAPOSO, L. (1999) — Foz do Enxarrique (Rdo, Portugal): first results of the analysis of a bone assemblage from a Middle Paleo lithic open site , in The Role of Early Humans i n the Accumulation of European Lower and Middle Paleolithic Bone Assemblages, Mainz, Monographien des Rmisch Germanischen Zentralmuseums 42, p. 367 379. BUROW, C.; KEHL, M.; HILGERS, A.; WENIGER, G.C.; ANGELUCCI, D. E.; VILLAVERDE, V.; ZAPATA, J.; ZILHO, J. ( 2015 ) — Luminescence Dating of Fluvial Deposits in the Rock Shelter of Cueva Antn, Spain. Geochronometria, 42, p. 107 125. CUNHA, P. P.; MARTINS, A. A.; MURRAY, A. S.; HUOT, S.; MURRAY, A.; RAPOSO, L. (2008) — Dating the Tejo river lower terraces in the Rdo area (Portugal) to assess the role of tectonics and uplift . Geomorphology, 102, p. 43 54. CUNHA, P. P.; ALMEIDA, N. A. C.; AUBRY, Th; MARTINS, A. A.; MURRAY, A. S.; BUYLAERT, J.P.; SOHBATI, R.; RAPOSO, L.; ROCHA, L. (2012) — Records of human occupation from Pleistocene river terrace and aeolian sediments in the Arneiro depression (Lower Tejo River, central eastern Portugal) . Geomorphology, 165166, p. 78 90.


HOFFMANN, D. L.; PIKE, A. W. G.; WAINER, K.; ZILHO, J. (2013) — New U series results for the speleogenesis and the Paleo lithic archeo logy of the Almonda karstic system (Torres Novas, Portugal) . Quaternary International, 294, p. 168 182. KEHL, M.; BUROW, C.; HILGERS, A.; NAVAZO, M.; PASTOORS, A.; WENIGER, G.C.; WOOD, R.; JORD P ARDO, J. F. (2013) — Late Neanderthals at Jarama VI (Central Iberia)? Quaternary Research, 80, p. 218 234. MARKS, A.; MONIGAL, K.; ZILHO, J. (2001) — The lithic assemblages of the Late Mousterian at Gruta da Oliveira, Almonda, Portugal , in ZILHO, J.; AUBRY, Th.; CARVALHO, A. F. (eds.) — Les premiers hommes modernes de la Pninsule Ibrique. Actes du Colloque de la Comission VIII de l’UISPP, Vila Nova de Foz Ca, Octobre 1998, Trabalhos de Arqueologia 17, Lisboa, Instituto Portugus de Arqueologia, p. 145 154. MARTNEZ SNCHEZ, C. (1997) — El yacimiento musteriense de Cueva Antn (Mula, Murcia) . Memorias de Arqueologa de la Regin de Murcia, 6, p. 18 47. RASMUSSEN, S. O.; BIGLER, M.; BLOCKLEY, S. P.; BLUNIER, TH.; BUCHARDT, S. L.; CLAUSEN, H. B.; CVIJANOVIC, I.; DAHL JENSEN, D.; JOHNSEN, S. J.; FISCHER, H.; GKINIS, V.; GUILLEVIC, M.; HOEK, W. Z.; LOWE, J. J.; PEDRO, J. B.; POPP, T.; SEIERSTAD, I. K.; STEFFENSEN, J. P. ; SVENSSON, A. M.; VALLELONGA, P.; VINTHER, B. M.; WALKER, M. J. C.; WHEATLEY, J. J.; WINSTRUP, M. (2014) — A stratigraphic framework for abrupt climatic changes during the Last Glacial period based on three synchronized Greenland ice core records: refining and extending the INTIMATE event stratigraphy . Quaternary Science Reviews, 106, p. 14 28. REIMER, P. J.; BARD, E.; BAYLISS, A . ; BECK, J. W.; BLACKWELL, P. G.; BRONK RAMSEY, C.; BUCK, C. E.; CHENG, H.; EDWARDS, R. L.; FRIEDRICH, M.; GROOTES, P. M.; GUIL DERSON, T. P.; HAFLIDASON, H.; HAJDAS, I.; HATT, C.; HEATON, T.J.; HOFFMANN, D. L.; HOGG, A. G.; HUGHEN, K. A.; KAISER, K.F.; KROMER, B.; MANNING, S. W.; NIU, M.; REIMER, R. W.; RICHARDS, D. A.; SCOTT, E. M.; SOUTHON, J. R.; STAFF, R. A.; TURNEY, C. S. M.; VAN DER PLICHT , J. (2013) — IntCal13 and MARINE13 radiocarbon age calibration curves 0 50000 years calBP . Radiocarbon, 55 (4), p. 1869 1887. SNCHEZ GOI, M. F.; BARD, E.; LANDAIS, A.; ROSSIGNOL, L.; d’ERRICO, F. (2013) — Air sea temperature decoupling in western Europe during the last interglacial glacial transition. Nature Geoscience, 6, p. 837 841. SNCHEZ GOI, M. F.; LANDAIS, A.; FLETCHER, W. J.; NAUGHTON, F.; DESPRAT, S.; DUPRAT, J. (2008) — Contrasting impacts of DansgaardOeschger events over a western European latitudinal transect modulated by orbital parameters . Quaternary Science Reviews, 27, p. 11361151.


SNCHEZ GOI, M. F.; BARD, E.; LANDAIS, A.; ROSSIGNOL, L.; d’ERRICO, F. (2013) — Air sea temperature decoupling in western Europe during the last interglacial glacial transition. Nature Geosci ence, 6, p. 837 841. STUIVER, M.; REIMER, P. J. (1993) — Extended 14C Data Base and Revised CALIB 3.0 14C Age Calibration Program . Radiocarbon, 35 (1), p. 215 230. WOOD, R.; BARROSO-RUZ, C.; CAPARR”S, M.; JORD PARDO, J. F.; GALVN SANTOS, B.; HIGHAM, T. F. G. (2013) — Radiocarbon dating casts doubt on the late chronology of the Middle to Upper Paleo lithic transition in southern Iberia. Proceedings of the National Academy of Sciences USA, 110 (8), p. 2781 2786. ZILHO, J. (2006) — Chronostratigraphy of the Middle to Upper Paleolithic Transition in the Iberian Peninsula. Pyrenae, 37, p. 7 84. ZILHO, J.; VILLAVERDE, V. (2008) — The Middle Paleolithic of Murcia . Treballs d’Arqueologia, 14, p. 229 248. ZILHO, J.; ANGELUCCI, D.; BADAL GARCA, E.; d’ERRICO, F.; DANIEL, F.; DAYET, L.; DOUKA, K.; HIGHAM, T. F. G.; MARTNEZ SNCHEZ, M. J.; MONTES BERNRDEZ, R.; MURCIA MASCAR”S, S.; PREZ SIRVENT, C.; ROLDN GARCA, C.; VANHAEREN, M.; VILLAVERDE, V.; WOOD, R.; ZAPATA, J. (2010) — Symbolic Use of Marine Shells and Mineral Pigments by Iberian Neandertals . Proceedings of the National Academy of Sciences USA, 107 (3), p. 1023 1028. ZILHO, J; AJAS, A.; BADAL, E.; BUROW, Ch.; KEHL, M.; L”PEZ SEZ, J. A.; PIMENTA, C.; PREECE, R. C.; SANCHIS, A.; SANZ, M.; WENIGER, G. C.; WHITE, D.; WOOD, R.; ANGELUCCI, D. A.; VILLAVERDE, V.; ZAPATA, J. (2016) — Cueva Antn: A multi proxy MIS 3 to MIS 5a paleoenvironmental record for SE Iberia. Quaternary Science Review s, 146, p. 251 273.


Table S2. 1. Cueva Antn stratigraphy. Depositional environments recorded in the succession. The solid lines indicate major erosive surfaces, the broken lines indicate minor discontinuities (after Angelucci et al., 2013 and Zilho et al., 2016) Complex Unit(s) Depositional environment DD twentieth century artificial reservoir TL exposed surface AS1 Ig, I h, I -k alluvial (floodplain plus bar/levee intercalation and one lacustrine event) alternating to (and ending with) wall degradation and runoff I-i Ij, II -a II-c II-b AS2 II d, II e alluvial bar/levee alternating to wall degradation and runoff II f II g fining upward alluvial sequence (channel, bar and floodplain), with intercalated events of wall degradation II h, II i II k II l top IIl, II -m AS3 II, II z, II -o fining upward alluvial sequence (bar and floodplain) capped by lacustrine event II-p wall degradation followed by alluvial floodplain II-q alluvial sequence (channel, bar and floodplain) followed by wall degradation II-s II-t AS4 IIalluvial bar/levee AS5 II-u alluvial bar/levee with events of wall degradation and slope outwash II-w IIt, II -y III-a IIIb, III -c alluvial bar/levee III-d alluvial bar III e, III f, III g III i wall degradation and alluvial bar III j III k III l III m, III n alluvial bar FP IV 'lacustrine' Table S2. 2 . Cueva Antn layer II l. Stone tool technological categories. T wo small flint nodules (possibly manuports, or else, given the fluvial nature of the accumulation, part of the natural geological background) are not included. N = number, M = mass in grams R AW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 1 49.3 3 17.7 4 6.6 2 * 1.5 – – – – 6 1.4 – – 6 102.3 22 178.9 Quartzite – – – – – – – – – – – – – – – – – – – – Limestone 1 148.0 2 17.2 6 37.9 – – – – – – 2 1.0 – – 1 17.3 12 221.4 Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 2 197.3 5 34.9 10 44.5 2 1.5 – – – – 8 2.4 – – 7 119.6 3 4 400.3 *The two small flint flakes are by products of a single sidescraper re sharpening event


Table S2. 3 . Cueva Antn layer II l. Classification of cores and retouched tools. Includes only formal, retouch modified items Cores N Formal retouched tools N chopper 1 notched piece 1 centripetal (Levallois or discoid ) 1 sidescraper TOTAL 2 simple 4 transversal 1 convergent 1 TOTAL 7 Table S2. 4 . Cueva Antn layer Ik. Stone tool technological categories. Two small, unmodified limestone cobbles (possibly manuports, or else, given the fluvial nature of the accumulation, part of the natural geological background) are not included . N = number , M = mass in grams RAW MATERIAL CORE S (a) FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOL S (b) TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 3 110.3 3 21.8 4 5.9 2 3.7 – – – – 5 0.7 1 3.3 2 21.3 20 167.0 Quartzite – – – – – – – – – – – – – – – – – – – – Limestone – – – – – – – – – – – – – – – – – – – – Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 3 110.3 3 21.8 4 5.9 2 3.7 – – – – 5 0.7 1 3.3 2 21.3 20 167.0 (a) 1 Kombewa, 1 discoid and 1 bipolar (b) 1 notched piece and 1 denticulate Table S2. 5 . Cueva Antn layer I -k. U se wear evidence (flint) . Material used on, inferred function , and traces of residue present Illegible None Wood Hide Meat Bone Projectile Ochred Total notches – – 1 – – – – – 1 denticulates – – 1 – – – – – 1 unmodified blank – – 2 – – 1 – – 3 TOTAL – – 4 – – 1 – – 5


Chapter 3. The rock shelter of Finca Doa Martina 3. 1. DESCRIPTION Finca Doa Martina ( FDM ; 3804’43” N, 0129’25” W ) is located at the base of an overhanging rock face rising above the left slope of the Rambla P erea valley, part of a large cliff generated by the incision of the watercourse. The bedrock belongs to an upper Miocene, poorly deformed sedimentary formation . A t the site, it is formed of limestone : a fine to very fine calcarenite with poorly visible bedding. The back wall is the interior side of a joint exposed by the massive loss of a triangular prism of bedrock . This slabbingoff event created an extensive, dihedral, wedge like recess in the cliff face that , today , de spite subsequent events of overhang collapse, still features a sheltered band of terrain extending up to 5 m behind the drip line ( Fig. S1. 2 ; Fig. S3. 1). The archeo logically fertile sediment ary fill rest s on and against bedrock . Its accumulation was made p ossible by the presence of two stepped, largely flat bedrock platforms exposed by archeo logical excavation in the western half of the recess ( Figs. S3. 2S3. 3) . The upper platform, narrow and featuring a varied micro relief, runs along t he back wall. It covers an area roughly corresponding to rows 4 -7 of the excavation grid , decreasing in width from ca.3 m , at the western end of the rock face, to ca.1 m in column O, at the eastern end of the trench . Longitudinally, its surface is near ly horizontal; in row 6, it rises from 40 2.75 m, in column O, to 402. 93 m, in column D . Sagittally, it presents a slight outward dip; along the separation between columns J and K , it goes down from 403. 13 m against the back wall t o 402.6 4 m at the edge of the ca.1.5 m escarpment separating it from the lower platform. The latter runs parallel to the back wall, is almost 50% wider than the upper platform ( ca.5 m in columns F and G), and dips slightly to the E ast (from 401.65 m in column E to 401.20 m in column L). A second, signifi cantly higher (in column G, ca. 4 mhigh ) escarpment marks the outward boundary of the site . Along this boundary , the sediment ary facies change s from a rock shelter accumulation formed behind a drip line to a slope d eposit devoid of archeo logical content . 3.2. EXCAVATION APPROACH AND METHOD S At the time of discovery, the site’s ground surface featured a significant dip to the West in the eastern half ( Figs. S3. 1D , S3. 4A) . The area where it flattened out — approximately corresponding to column L of the excavation grid — was also where the band of sheltered terrain was wide st ( Fig. S1. 2) . These factors underpinned the decision to place here the initial testing trench, open in 2007 in grid units L/5 -7 ; a perpe ndicular trench was started at the same time in grid units N O/5 ( Fig. S3. 4B) . In subsequent seasons, these trenches were extended to grid units L/ 89, L4 and N O/4 , and taken down to bedrock . This work establish ed the site’s basic stratigraphic outline an d was followed , between 2008 and 2010, by the open area excavation of grid units D K/4 -7 . In this phase, the different strata recognized were exposed across the entire surface of the trench, layer by layer, until bedrock was reached ( Fig. S3. 4C E).


Eventually , stratigraphic units that did not exist elsewhere or had gone unrecognized until then were identified at the base of the deposit i n the western half of the DK/4 7 trench . Thus, i n 2012 and 2013, to gain a better understanding of the new units a nd their archeological content, th e trench was extended south ward to grid units E G/8 12 ( Fig. S3. 4F -G). The outward escarpment was identified at this stage . G iven the inconclusive results of geophysical surveys carried out in the interim, a trench ( grid u nits F G/13 14) was open ed in 2014 to test the possibility that a third bedrock platform existed at lower elevation ( Fig. S3. 4H) . This 5 m deep sounding cut through slope deposits until it hit an accumulation of huge boulders beyond which no further progress was possible . Whether a third platform cum shelter exists at FDM remains therefore uncertain, even though, w hen the general dip of the Miocene stratification is considered , the bottom elevation reached at FDM roughly corresponds to that at which b edrock was found 50 m downstream at La Boja (ADB) . Work at FDM concluded in 2016, when grid units M7 and N O/6 7 were excavated to verify the geometry and boundaries of the basal Middle Paleo lithic deposit in the eastern part of the site ( Fig. S3. 2) . The s tratigraphic observations made are accounted for in the following, but not so the additional stone tool f inds , which remain unprocessed but support the chrono stratigraphic framework derived from the 2007 2014 material. In the end, the excavation’s planar surface and volume totaled 59 .9 m and 9 0. 3 m, respectively. During the initial testing phase, the excavation proceeded via slicing the sedimentary fill into spits that respected observed stratigraphic boundaries whenever possible. If the thickness and heterogeneity of the deposit made the recognition of such boundaries difficult, or if these where gradual and hard to follow, the testing operations switched to horizontal spits of arbitrary thickness ( 5 or 10 cm ) . A good grip on the cha racteristics of the succession was thus acquired , and produced the guidelines and criteria required to undertake the open area excavation of the different strata recognized. This work proceeded entirely along natural stratigraphic boundaries and, when nece ssary, via the subdivision of the different units into spits that followed the dip of the stratification ( Fig. S3. 5). The finds from the arbitrary spits of the testing phase were retrospectively assigned via comparison of their x,y,z information with the v olume try of stratigraphic units as reconstructed from cross sections . W here ambiguity remained, no assignment was made and such finds were excluded from consideration. Post depositional disturbance features of two kinds were encountered. P atches of grey colo red sediment extending down from the top soil (layer 2) affected layers 3 and 4 without physically altering the general texture and structure of the deposit. These were diagnosed as accumulations of organic matter related to the vegetation growing on the surface and treated as lateral variation. Westward of the 2007 trench, a large disturbance feature, denoted by a much looser sediment fill, the near absence of finds, and the pr esence of large vertical blocks, affected a significant portion of laye rs 4 and 5 ( Fig. S3. 5F). Diagnosed as a rabbit warren, th is feature was entirely emptied prior to excavation of the rump of th o se units . A similar procedure was followed in the case of a tunnel identified in grid units E G/8 along the edge of the upper pla tform ( at the elevation of layers 8 and 9 , and partly filled with layer 7b material) .


All dcapage surfaces and stratigraphic cross sections were drawn and/or photographed. T he DStrectch plugin for ImageJ was used to highlight colo r contrasts and produce prints used in the field to help with the dcapage of stratigraphic interfaces ( Fig. S3. 5J) . Photo mosaics were assembled using PT GUI or Microsoft ICE , and orthorectified with the University of Venice’s RDF software. Elevation maps, 3D models and area/volume estimations were produced with Surfer. Elevations and finds were recorded and/or piece plotted with the help of a laser level, to the nearest centimet er , against the grid and the site datum, which was placed on the back wall of the shelter at t he elevation of 405.85 m. Cores and core fragments, complete and proximal blades and bladelets, retouched tools and fragments thereof were systematically piece plotted , and , in the lowermost Middle Paleolithic , the same for complete flakes . The rest were b agged together with sieve finds from the same spit and grid unit . T he charcoal for anthracological study was collected in similar manner, but the fragments spotted during excavation as suitable for radiocarbon dating were all piece plotted . In addition, m u ltikg samples of carbon or charcoalrich sediment associated with hearths or their remnants were saved for subsequent flotation in the laboratory . In the 20072010 field seasons, finds were numbered sequentially, 1 to n , per grid unit (e.g., L5 1 to L5n, H7 1 to H7 n , etc.); in 20122016 , they were numbered sequentially, 1 to n , per year o f excavation (e.g., 2012 1 to 2 012n , 20131 to 201 3n ). The sandy, dry nature of the deposit dispensed wet sieving, so the excavated sediment was dry sieved in its entirety through two sieve sta cks with meshes of 2 and 1 mm ( Fig. S3. 5A) . Undisturbed s amples for soil micromorphological analysis of the sediment s and of the few fire features identified were collected throughout. Despite the non acidic nature of the d eposit , bones were seldom present. Even if no more than very small splinters, all were therefore pieceplotted. Mollus k shell preserved well. L and snail remains were particularly abundan t in disturbed or organic matter-rich areas of the sedimentary fill, and the dominant taxon ( Iberus alonensis ) is well known for its burrowing behavio r and preference for sheltering in rocky fissures (Moreno Rueda, 2006) . For these reasons, land snail shell was considered as post depositionally intrusive if not natural background noise (the latter otherwise corroborated by the present day ubiquity of I. alonensis shell in the slopes of the gorge). Given the inland location of the site, marine and fluviatile mollus k shells were , however, clearly anthropogenic . Therefore, such shells , often pierced for use in body ornamentation, were piece plotted when seen at excavation, even when they were no more than small valve fragments, while land snail remains were discarded. Surveys of the region were carried out to identify the provenience of the diverse flint varieties represented . Several potential sources located within a radius of 20 km have been identified but this aspect of the project remains in a preliminary stage. The study of l ithics used traditional typological classification but otherwise followed the “economics of stone” approach outlined in Zilho (1997) : in the analysis of reduction sequences and site function, “ splintered pieces ” were treated as bipolar cores, while “ burins ” and thick (carinated and nosed) “scraper” forms were defined as blanks for the extraction of bladelet products.


All c ores, retouched tools and unretouched blanks (complete flakes in the Middle Paleo lithic assemblage ; blades and bladelets , including fragments, in the Upper Paleolithic ) were individually measured, weighed and recorded for a set of technological attributes. To enable comparison across the Middle/Upper Paleo lithic technological divide, “small flakes” (<2.5 cm) were counted separately. Usewear analysis of selected samples was based on differential interference contrast microscopy, carried out with a BHMJ Olympus model (at 200 or 400 magnification) , and follow ed standard recommendations for the cleaning and preparation of the material (Plisson, 1985). 3. 3 . STRATIGRAPH IC OUTLINE All units contain local limestone fragments, variable in shape, size and abundance, and are enriched of post depositional calcium carbonate or gypsum. Due to the complex geometry and variable extent of the different stratigraphic units , none of the reference cross sections and associated dcapage plans ( Figs. S3. 6S3. 13 ) features the complete succession, which , from top to bottom, is made up of the units described below , designated in the field as “ layers ” and primarily defined after the L>M cross section ( Fig. S3. 8) (a ll Munsell colo r descriptions are on moist sediment ). Unit 1 T hin surface layer disturbed by trampling, exposed only in the inner parts of the rock shelter, along the back wall. The texture i s silt, with few stones (often with a sub horizontal orientation plane), a weakly developed “ platy ” structure, and a clear boundary. Unit 2 P oorly developed A horizon a top the succession. It is a 10YR4/3 soft silt with very weak crumb structure, some roots, common dispersed organic matter, and a sharp linear boundary. Stones of distinct types, randomly distributed and oriented, are common . In areas of the excavation extending beyond the drip line , layers 4, 5 or 7 , bioturbated in a manner resembling unit 2 , are exposed as ground surface. The horizontal transition from a “buriedunder the overlying strata” to an “exposed at ground surface” condition being gradual, the bioturbated, surface exposed, devoid of finds reaches of those layers were not differentiat ed during excavation, as reflected in t he captions to the cross sections illustrated here . Unit 3 Possibly an E horizon poorly developed from slope sediment. It is a 10YR5/4 sandy silt , massive, slightly firm, with no organic matter and a sharp boundary. S tones of distinct types, with random distribution and orientation, are common . I n section view , i t is discontinuous , weakly visible. At dcapage, it consisted of a variably mottled lens of light, fine, largely loose sands rich in angular debris and contain ing numerous fragments featuring in situ shattering (frost weathering?) , comprised between a dark Unit 2 and a somewhat cemented, salmon or light grey colored Unit 4 ( Fig. S3. 5 B, S3. 5 F) . As is also the case with Unit 2, the stone tools retrieved in Unit 3 are of Epimagdalenian affinities.


Unit 4 L aterally variable unit in which two facies were recognized . Unit 4a is a 10YR4/4, massive, slightly firm sandy silt with scarce stones ( the tabul ar ones often being parallel to the lower interface), scarce , dispersed organic matter (increasing toward the back wall), and a clear boundary. Unit 4b is the same without the organic matter and of color 10YR6/6. Unit 4a was restricted to column L of the excavation grid and adjacent areas of columns K and M -O . It remains to be clarified whether this subunit’s organic matter content is anthropogenic or reflects bioturbation and soil formation during a depositional hiatus. The lithics retrieved in Unit 4 are of Upper Solutrean affinities. Unit 5 Thick layer of breccia/ diamict , formed of cross bedded or festooned , poorly recognizable lenses , with stone line intercalations. It is a 10YR6/6 s andy silt with no organic matter, firm, weakly cemented by calcium carbonate, and a clear boundary . Stones of different types are common . It yielded Upper Solutrean diagnostics. Unit 6 Organic matterenriched lens , wedging out in all directions ( Figs. S3. 5C, S3. 8, S3. 14) . It is a 10YR2/1 s andy silt with some clay, few sto nes, frequent , finely dispersed organic matter, massive, firm, and with a clear boundary. It remains to be clarified whether this facies stands for a remnant, poorly developed A horizon , or for a localized accumulation of micro charcoal particles related t o human occupation . Unit 7 Layer of breccia/diamict material. It is a 10YR5/4 s andy silt, massive, firm, with some calcified roots penetrating from unit 5, weak carbonation and a sharp boundary. Stones are scarce (near wall) to abundant (outwards) . The presence of large, darker mottles denotes bioturbation ( few , but relatively large burrows filled with organic sediment ). At excavation, layer 6 was treated as lateral variation of layer 7 and , archeologically, they have been considered as a single unit, layer 6/7 , within which the boundaries of unit 6 broadly coincide with the area featuring the highest density of stone tool finds. In rows 4 6 of the western part of the D K/4 7 trench this deposit was shallow and clearly post depositionally disturbed. Along row 7 and the outward half of row 6, its surface was scoured by erosion , leaving a steep micro ravine atop and against which unit 5 came to lay ( Fig. S3. 14). The stone tools from layer 6/7 are of Gravettian affinities. Unit 7b S imilar to unit 7 , whic h it underlies in the western part of the site. T he excavation trench in which this unit was recognized (E-G/8 12) extends out ward of the drip line, explaining why the sequence is enriched in large horizontal stones , some of which are >1 m long, derived from the degradation of the shelter’s overhang ( Figs. S3. 10S3. 12) . As with overlying Unit 7, this layer yielded diagnostic Gravettian lithics .


Unit 8 Clast supported fine to medium breccia filling channels and scours cut in underlying unit 9 (erosive fe atures oriented E to W, with platy stones sometimes imbricated and often weathered) . The matrix is a fine material as in unit 7b, local ly enrich ed in sandy silt derived from unit 9 . T he boundary is erosive and the mass of the deposit presents horizontal di scontinuities making for situations of lateral contact of its edges with the interface between layers 7 b and 9 ( Figs. S3. 12S3. 13). The stone tools retrieved in the body of unit 8 are of Aurignacian affinities. Unit 9 Relatively homogeneous , 10YR6/4 silt wit h fine sand and scarce stones (plus occasional fragments of reworked clay at the base ), massive, firm, poorly cemented by calcium carbonate, with discrete packing, no organic matter, and a gradual, poorly distinct boundary. Some darker, large mottles de note bioturbation. During fieldwork, further subunits (9b and 9c) were distinguished, based on local variation in texture and amount of stones. Inward, this unit lean s against the edge of the upper bedrock platform; outward, it becomes much thicker (ca. 1 m ) and dips S/15 as it extends beyond the boundary of the lower platform ( Figs. S3. 8, S3. 10) . In the N-O /4 -5 grid units and along the southern edge of row 7 of the D K/4 7 trench, layer 9 outcropped as narrow bands of yellow sandy sediment bounded by the 5 >6 and 7>8 cross sections , outward, and by the upper bedrock platform, inward. A t the time of excavation, these bands of sediment, which featured significant bioturbation, w ere thought to represent the basal aspect of layer 7. They w ere eventually recognized for what they were, and the corresponding excavating spits and finds retrospectively assigned, based on analysis of the cross sections and on lateral continuity with the typical aspect of the unit — first defined in L/7 -8 and subsequently found to extend to surrounding grid units with the 2010 excavation of the M column, the 2012 excavation of E G/8 9, and the 20 16 excavation of M7 and N O/6 -7. All st one tools retrieved in Unit 9 are of unambiguous Middle Paleo lithic (Mousterian) affinities. Unit 10 Also known as “ pisto ,” this archeo logically sterile unit was first recognized in the L/7 8 deep sounding, where it corresponded to a 7.5YR4/4 massive, weakly porous s ilty loam with common , mm to 2 cm long fragments of pink / red laminated cl ay /marl, scarce fragments of local rock , occasional fragments of calcite /aragonite crust s, common microcrystalline gypsum nodules, scarce carbonate and Mn Ox nodules , no organic matter, and a sharp boundary to bedrock. This unit was also found in E G/8 10. Here, and especially so towards the back wall of the lower bedrock platform, it was rich in large, angular, non weathered blocks, and heavily cemented. In this area, a <5 cm wide fissure , internally padded with a continuous layer of calcite/aragonite crystals , separated the sediment from the escarpment leading to the upper bedrock platform against which it leaned.


Unit 11 T hick, archeo logically sterile breccia making up the bulk of the FDM talus scree . Formed of poorly stratified alternations of a cl ast supported breccia of local rock (locally partially open work) with in a 10YR5/6 silty sand with common to frequent stone fragments , i t dips S/15 and was recognized over a thickness of ca . 3 m in the F G/13 14 sounding ( Fig. S3. 10). Unit 80 S aprolite developed from bedrock. It is a fine open work breccia made of cm siz ed platy and tabular bedrock fragments oriented parallel to bedrock. Prior to recognition of the basal horizons of the succession lying beyond the edge of the upper platform , this unit separated layer 7 from bedrock and, as such, was initially designated as “layer 8” (in e.g. , annual excavation reports and in Zilho et al., 2010) , subsequently changed to unit 80. 3.4. RADIOMETRIC DATING The provenience, composition and nature of the samples submitted for radiocarbon dating are given in Table S3. 1, and the results obtained are listed in Table S3.2. No charcoal was found in layers 3, 5, 8 and 9 ; h ence , the lack of results for these units. In 2012, sediment samples were taken in the E G/8>9 cross section for the luminescence dating of layers 7b and 9. The tests carried out in the luminescence l aboratory of the Institute of Geography, University of Cologne , indicated , however, that the material was unsuitable for dating. The cause probably lies in the sediment being freshly derived from the weathering of the local bedrock and, thus , not having undergone the number of transport deposition cycles required to develop favorable luminescence characteristics. Radiocarbon dating was carried out on samples submitted to the Vienna laboratory (VERA) . Despite reasonable size and good preservation of tissue structure, enabling ready taxonomic identification, the standard Acid -BaseAcid (ABA) pre treatment proved to be too aggressive for the site’s charcoal. The cause probably lies in leaching and other geochemical processes related to site formation mechanisms. Regardless of this problem’s exact nature, the consequence was that dating had to be carried out on the humic acid s fraction or after a mild form of ABA. Incomplete decontamination , or contamination by younger humic acids present in the humic acids fraction being an issue, t he results must be treated as mini mum ages only. At nearby La Boja, better charcoal preservation enabled measurement of the ABA treated sample material (and obtained humic fraction), and, i n some cases, the more aggressive ABOx SC extraction procedure was also applied (see Chapter 4). No difference was observed between the ABA and ABOx SC results for a single sample . S ome of the humic acid results were up to two millennia younger , while others fell in the same statistical ball park . The latter was the case for results no older than about 21 23 ka (uncal ibrated ), but also for some samples in the 40 45 ka (uncal ibrated ) range. These comparative data, the lack of evidence for soil formation in the areas of FDM where the dated samples came from, and the stratigraphic consistency of the results obtained ( Figs. S3. 15 S3. 16 ) suggest that, even though almost certainly underestimated, the results reported in Table S3. 2 probably are only moderately so.


The layer 4 samples all come from o ne and the same hearth feature and ought to have returned the sa me date. That such was not the case corroborates that the results are affected by incomplete decontamination or the presence of younger humic acids in the humic acids fraction , and implies that the older result , placing the feature in the 22.8 23.4 ka cal BP range, best approximates its real age. This interval is likely to still represent something of an underestimation but, be it as it may , the Upper Solutrean affinities of the layer 4 stone tools constrain its accumulation to no earlier than the beginnings of the 24th millennium cal BP. One of the layer 6/7 samples (H6 51) comes from the periphery of a h earth, while the other (H6 63) was collected a few cm underneath, near the contact with layer 80. Assuming that the oldest result (VERA 5367d HS_2) is least affected by residual contamination and, hence, provide s the best age estimat e , layer 6/7 would date to the first half of the 31st millennium cal BP. However, this sample was made up of scattered charcoal and, given the gradual nature of the stratigraphic boundary with layer 7b, it cannot be excluded that some relates to the latter. Indeed, if layer 7b extend ed toward the back wall as no more than a thin lens atop the basal saprolite, it could well have gone unrecognized beyond the boundaries recorded during the dcapage of the basal spits of row 6 of the E K/4 7 trench. Bearing this possibility in mind and knowing that the VERA 6170HS result comes from the sample most closely associated with the layer’s single preserved hearth feature (H6 51; Fig. S3. 15), assigning to this occupation an age within the 29th or the 30th millennium cal BP would seem more reasonable . The result for layer 7b (VERA5368HS) is consistent with those for overlying layer 6/7. At the adjacent site of La Boja, the correspo nding stra tigraphic slot (horizon OH14) is undated and was poor . However, it yielded Gravettian diagnostics , and t he age of the horizons that sandwich it is reliable: 27,260230 BP (VERA 5789) for OH13 ; 30,548/+363/347 BP (VERA 6153) for OH15 (see Chapter 4 ). The humic acids result for VERA 5789 is in agreement with the ABA result. Indeed, s tatistically, the three Early Gravettian results for La Boja OH14 and FDM layer 7b are indistinguishable, while broad synchronicity in the phasing of the tw o sites’ sedimentary accumulation is to be expected. Thus, we can conclude that FDM layer 7b likely dat es to the end of the 3 2n d millennium cal BP. 3. 5 . SITE FORMATION A 3D model of the extent, geometry and stratigraphic relationships of the different unit s, based on the combination of the cross section and dcapage records , is given in Fig. S3. 17. As the basal deposits (layers 10, 11 and 80) do not overlap ( Fig. S3. 17 AD), the sequence in which they formed remains to be clarified and can only be addressed in indirect manner . With current evidence, layer 10 would seem to be the olde st . The lack of finds an d organic matter, combined with the non weathered condition of the angular detrital fragments it contains , suggest that layer 10 could well correspond to t he interior deposit that once filled the joint along which the rock shelter formed and that the slabbingoff event triggering the site formation process eventually exposed . If so, the geometry of this unit’s erosional surface ( Figs. S3. 8, S3. 10 S3. 11) woul d result from the initial phase of configuration of a slope linking the streambed below with the recess thus newly formed in the valley’s left bank cliff face .


In Fig. S3. 17C, l ayer 80 appears to cover bedrock in the interior grid units of the western part of the trench in discontinuous manner only ; this is due to the fact that saprolite was often too thin there for differentiation at dcapage to be possible . Beyond the drip line , atop the largely horizontal lower platform , the absence is, however, genuine. Layer 80 therefore corresponds to the degradation of the local bedrock under sheltered conditions , prior to a time when the formation of slope deposits outside and lower down generated a topographic baseline e nabling the accumulation of a stable rock shel ter fill. The formation of the saprolite and the accumulation of the slope deposit are probably coeval, but all that can be said with certainty is that they predate the accumulation of layer 9, which leans against or overlies layer 80 in ward of the drip -l ine and, outward, overlies layers 10 and 11 ( Fig. S3. 17E ). Layer 9 of grid units L9 and E -G/914 featured tabular carbonate crusts of variable thickness and extension, and many of the stone tools therein presented thick carbonate coatings ( Fig. S3. 18S3. 19). No coatings of similar thickness and extent were observed among the lithics from overlying units but, at comparable elevation (>65 75 cm below ground surface), those units only existed behind the drip line. Outward, only the portions of layers 7b and 8 extending to rows 911 of the grid were buried so deeply , and some of their lithics also feature carbonate coatings (even though thinner and less extensive). T he distribution of these precipitates closely follow ing the shelter’s extant drip line , they ca nnot be relicts of a mid Upper Pleistocene paleosoil. Instead, they must relate to the calcic horizon of a Holocene Mediterranean soil, denoting how far down water percolated from the ground surface during its formation. A massive erosional event removed most of layer 9 in the central and western parts of the upper platform , where bedrock was left exposed, bare or covered by a thin layer 80. The geometry of the erosive scar suggests that the implicated agent was water running along the back wall of the site and cascading to S and SW from an area of the upper platform ’s edge located in columns JK of the grid . The origin of that water probably lies in the fissure that prolongs the back wall to the East, reactivated as a karst ou tlet draining the plateau abov e . As a consequence of these processes , the baseline for subsequent depositional events featured a prominence of the upper bedrock platform that effectively divided the site into two different parts that, for a while, underwent separate depositional histories: FDMWest, where, after a hiatus of unknown duration, layer 9 was overlain by layers 8 and 7b; and FDM East, where layer 9 was eventually covered by layers 7 and 5 ( Fig. S3. 17F -I). The geometry of these stratigraphic interfaces explains why the basal reaches of layers 7b and 6/7 yielded a couple of items that, from technological and typological standpoint s, are of Middle Paleolithic affinities ( Table S3.3 ; Fig. S3. 20) . One ( 2012 159 ) is a flint denticulate made on a cortically backed, orange segmentlike flake, probably a deliberately overshot discoid blank . It was found in grid unit E9, in an area where layer 7b was in lateral contact with layers 8 and 9 ( Figs. S3. 12S3. 13). There can be little doubt that this is a reworked object. The other (J6 63) is a slightly patinated sidescraper fragment found at the base of layer 6/7 in a place that , at the time of formation, lied at the foot of a steeply sloping layer 9 long exposed as FDM -East’s ground surface ( Fig. S3.17E H). It is clearly an inherited item incorporated by progradation.


Blanks such as that used for the 2 012 159 denticulate are well represented among the complete, modified and unmodified flake s from layer 9 — by four items ( three sidescrapers and one naturally backed knife ) out of 26, i.e., 15% — but absent from the equivalent ensemble of flint blanks from layers 8, 7b and 6/7 (N=122) . Th e J6 sidescraper is m ade on quartzite, a raw material rarely found in the site’s Upper Paleo lithic levels , in which it wa s exploited in expedient manner , to produce blanks that remained unretouched . That only two such “contaminants” were found among the 210 formal retouched tools found in layers 8, 7b and 6/7 (see below) suggests that the presence of reworked , undiagnostic Middle Paleolithic material in these units is anecdotal and of no consequence for the study of spatial distributions and the lithic econom y and lithic technology of the site’s Upper Paleo lithic occupations. T he subsequent accumulation of layer 6/7 testifies to a change in the source of the sediment, now coming mostly from the opposite side of the recess, and to the onset of the constitution of the site’s still extant Westdipping ground surface ( Fig. S3. 17 H) . From this time onwards, FDM West behaved as the brim of a depositional cone in which most sedimentary accumulation occurred eastward of column H . The thickness of layer 5 in the central part of FDM East ( Fig. S3. 17 I) , especially apparent in the L>M cross section (Fig. S3. 8), is a byproduct of th is change in the dynamics of the accumulation , combined with the presence of a depression caused by the scouring of the previously formed deposit ( Fig. S3. 17H) — probably due to a new episode of reactivation of the fissure prolonging the site’s back wall t o the East. Once ground surface again reached a stab le level, the dynamics of the accumulation was reduced, leading to the formation of layer 4 ( Fig. S3. 17 J) , and eventually came to a halt. After the hiatus — manifested in the incipient soil formation probably present in unit 4a and during which rabbits extensively burrowed the site — came the last stage of the accumulation , represented by layers 3 and 2 ( Fig. S3. 17 KL) . This phase corresponds to the eventual filling-up of the slightly inward sloping, b asin shaped space comprised between the back wall and the drip line extant since layer 4 times . A number of expectations for the conservation of an archeo logical record can be derived f rom this formation process. Firstly, that sedimentation will have been rapid and the remains abandoned on site quickly buried in packages of significant thickness . Secondly, bearing in mind the average duration of Middle and Upper Paleo lithic technocomplexes, that individual layers will correspond to separate phases of the re gional chrono stratigraphic sequence , with minimal palimpsest effects. Thirdly, not withstanding the localized impact of bioturbation and the potential presence of inherited material , that the level of assemblage integrity will be, in general, high . Fourthly , that the nature of the sedimentary accumulation combined with the site’s significant exposure to the elements (despite the presence of an overhang, FDM is more a piedde falaise than a true rock shelter) will cause significant syn depositional scattering of the remains of individual occupation episodes by gravity and water driven surface dynamics, generating relatively low density, spatially homogeneous find distributions. Fif thly , that the preservation of features (e.g., hearths) will be exceptional a nd limited to those areas most protected from such surface dynamics — close to the back wall and where the slope extant at the ti me of occupation flattened out .


These expectations, which we developed at the end of the initial testing phase , guided the excavation approach and the methodology subsequently used . Realizing that the potential of the site principally lied i n providing a framework for the region’s Upper Pleistocene c ultural stratigraph y, emphasis was put in the acquisition of lith ic assemblages fit for purpose, i.e. , sufficiently large (in composition) and complete (in the representation of the different stages of the reduction sequence). For that potential to be realized , however, a large area and a significant volume of sediment would have to be excavated . Hence the adopt ion of an optimized strategy for the recording of finds, features, surfaces and sections , and the decision to prioritize stratigraphic control over resolution in the spatial distributions ( implying, for instance, that the sediment was excavated and siev ed by grid units, not by m quadrants) . 3. 6. SPATIAL DISTRIBUTIONS The spatial distribution of the lithics from layers 9, 8, 7b and 6/7 retrieved between the 2007 and 2014 field seasons is illustrated in Figs. S3. 21 S3. 24. T o avoid distortions caused by different excavators potentially using different cut off criteria for the recovery of chippage in the sieving process ( e.g. whe n sorting the smaller mesh size), the distributions are assessed by mass instead of numb er of items . Consequently , for layer 6/7, only flint was considered, as the presence of a few large cores and manuports made on locally available raw materials (quartzite and limestone) creates distributional peaks that are more apparent than real. In agre ement with site formation expectations, the distributions are fairly homogeneous and, when apparent exceptions are considered in their detail, unimodal. The sorting and classification of sieve finds from layers 5, 4 and 3 has yet to be carried out in full. Based on field observations and the distribution of piece plotted items, there is no reason, however, to suspect that they will pattern differently . In layer 9, the higher values tend to be found outward. Grid unit F8 is exceptional because of a large (32 .2 g) limestone flake, a raw material otherwise represented in this layer by a small flake fragment and nothing else. The absence of finds in E L/4 7 is explained: (a) close to the back wall, by the surface of the site corresponding, at th e time , to the upper platform’s rock floor; (b) away from the back wall, by erosional loss of the deposit , le ad ing to exposure of the sagi t tally oriented bedrock prominence found in c olumns I K of the grid ( Fig. S3. 9) . Loss to erosion also explain s well the marked differ ence between the total for the grid units where layer 9 remained thick (e.g., L/8 9) and those where erosion truncated its upper reaches. The layer 8 stone tool distribution is biased by two factors: the low number of finds, which means that peaks will be present in those squares that yielded larger, heavier pieces, namely cores (e.g. , F11, where two of them come from); and the highly irregular geometry of the deposit, which implies that mass /area ratios are in this instance a poor indicator of the density of finds per volume unit of excavated sediment. The impact of these factors is clearly apparent when the layer 8 artefact distribution is contrasted with those of layers 7b and 6/7, which yielded larger assemblages and featured a more regular geometry.


In deed , despite post depositional, surface dynamics induced homogenization, the original spatial structure of the Gravettian occupation s has remained relatively well preserved. The layer 7b distribution is clearly unimodal and features higher values coincident with the location of the hearth features excavated therein (Figs. S3. 25 S3. 26). The layer 6/7 distribution is only apparently bimodal, as the trough in HJ/6 7 correlates with (a) the loss to erosion , along row 7, of a significant part of the de posit ( Fig. S3. 14) , and (b) the decreased thickness of the layer in row 6 , due to the inward rising of the bedrock ( Fig. S3. 15) . If we filter out the impact of these factors we see a clear concentration in K M/4 6, coincident with the extent of the black facies ( “layer 6” ) . Th e concentration is surrounded by a scatter of gradually decreasing density , t he apparent (and minor) anomaly represented by O5 being due to the bias introduced by the presence of a large core. 3. 7. FEATURES Figs. S3. 25 S3. 28 illustrate the few occupation features that could be identified during excavation. All correspond to hearths that underwent erosion to a varying degree. Hearth 1, at the base of layer 4, is the single instance in which a remnant of the original succession of ash, charcoal and burnt sand remained in situ ( Fig. S3. 28BE). The reason lies in the cementation of parts of the ash component, within which there were large chunks and even short fragments of carbonized juniper twigs. Hearth 2, in the same stratigraphic position, had been almost completely eroded, and only the presence of a few burnt stones associated with a dark stain , denoting a concentration of micro charcoal particles , betrayed the presence of a fire feature ( Fig. S3. 28 A) ; even so, the evidence is weak, and the identification of this feature is tentative. Hearth 3, at the top of layer 6/7, corresponded to a hard patch of cemented ash associated with a scatter of charcoal and reddened stones ( Figs. S3. 14, S3. 27) . This c ementation explains why Hearth 3 could survive at the edge of the ravine created in the G K/7 grid units by the washing away of the upper part of layer 6/7 . Conversely, the burrowing of the rabbit warren occupying grid units I J/5 6 and their periphery exp lains why the preservation of Hearth 1 and Hearth 2 was partial or poor. While Hearths 1, 2 and 3 correspond to fires lit on a bare ground, Hearths 4 and 5, at the top of layer 7b ( Figs. S3. 25S3. 26) , are of a different type, one that is represented by sev eral intact examples at the nearby La Boja site (see Chapter 4) . From the se examples we can infer that Hearths 4 and 5 of FDM are the subsurface , stone filled component of complex features whose above surface parts — the ash and charcoal produced by the bu rn t fuel — were lost to erosion . T he dark staining of the sediment downslope from , and inwards of Hearth 4 testifies to the process, the retention of micro charcoal particles being due to the barrier s represented by (a) the outcropping of stones that already belong in underlying layer 8 , (b ) the bedrock prominence that at this stage separated FDM West from FDM East, and (c ) the escarpment leading to the site’s upper platform . T he fact that all the charcoal from layer 7b was collected in row 7 (F7, G7 a nd H7), and most in G7 ( i.e., in the grid unit placed in the exact direction leading downslope to NE from the middle of the E G/89 trench ), is consistent with this interpretation .


3. 8 . THE MIDDLE PALEO LITHIC (LAYER 9) Tables S3. 4S3. 5 summarize t he composition of the layer 9 stone tool assemblage , and a representative sample is illustrated in Fig. S3. 29. Flint accounts for 9 6 % of the finds (8 7 % by mass ), and 100% of the cores. The latter’s small number ( nine ) and average mass at discard (1 8.6 g; 15 .0 g considering only the five unbroken ones) imply a type of site occupancy characteriz ed by intensive consumption of mostly imported blanks , as otherwise suggested by the high toolblank ratio (0. 6 by mass ). Supporting this inference , the raw material available was reduced for the extraction of even minimal amounts of cutting edge, as indicated by (a) the size of the cores’ l ongest measurable removal scars (between 9.7 and 16. 4 mm), and (b) small flakes representing 67 % of all the unretouched, complete flint blanks. This economic pattern explains the prepondera nce of the Kombewa technique ( Fig. S3. 2 9, no. 1), well suited for the recycling of flake blanks as cores for the production of smaller flakes . The Levallois and Discoid methods are represented by a few diagnostic blanks ( Fig. S3. 29, no. 2) and preparation products, as well as by one core. The latter is a recurrent centripetal Levallois core for which , given the known difficulty in distinguishing the two techniques when dealing with exhausted specime ns , a discoid classification would also have been reasonable . Out of the 3 6 tools , five are typologically unclassifiable, broken pieces. Excluding the se , sidescrapers represent 68 % of the total . The use wear evidence ( Table S3.6, Fig. S3. 30 ) indicate s that the y were used indistinctively to work on wood, hide , and hard animal tissue , as well as for cutting meat. This evidence is consistent with the notion that their retouch respond s to the need to extend the use life of a blank ’s worn out edges rather than to normative considerations related to intended design. That such edge rejuvenation tasks were carried out on site is documented by re sharpening debris and by extensions of well defined use wear polish cut by retouch removals ( Fig. S3. 30 , no. 2). The other tools are one denticulate, one notch ed piece, three backed knives (e.g., L9 6; Fig. S3. 19), one atypically retouched piece and two Mousterian points. T he latter were indeed used for the stonetipping of projectiles , a s shown by their diagnostic wear. D istally , the larger ( Fig. S3. 29 , no. 4; Fig. S3. 30, no. 6) features an impact fracture and, proximally, bears characteristic microscopic impact striations . I n addition, the blank’s bu t t had been retouched to facilitate hafting through alternating removals that thinned the dorsal edge of the platform and, ventrally, eliminated the bulb. T he very tip of t he smaller point ( Fig. S3. 2 9, no. 3; Fig. S3. 30, no. 7) — made on an older, patinated, recycled blank — is missing (possibly due to impact), but i ts classification as a point is supported by (a) the regularly retouched edges distally converging towards a 45 angle, (b) the overall symmetry of its triangular outline , and (c) the microscopic impact striations present proximally on the ventral side. Th e small size of the assemblage suggest s infrequent and short lived stays, while the fact that tools related to on site, domestic tasks make up most of the assemblage means that such stays were not purely logistical in nature. A residential signal can also be seen in the fact that most phases of the reduction sequence — namely, core preparation and core trimming, blank production, tool consumption, re sharpening — are represented .


Neither unworked nodules nor cores discarded during the initial configuration phase have been recovered . The rawmaterial acquisition stage of the reduction sequence is therefore missing . This absence concurs with the intensi ve reduction of flint volumes to suggest that (a) raw material sources were for the most part located at a significant distance and (b) raw material procurement was embedded in daily subsistence tasks and/or routine travelling (e.g., when relocating settlement) . 3. 9 . THE EARLY UPPER PALEO LITHIC 3. 9. 1. Aurignacian (Layer 8) The inventory of the lithics recovered in layer 8 is given in Table S3. 7, and details on the kinds of cores, bladelet blanks and retouched tools are given in Table S3. 8. A representative sample is illustrated in Figs. S3. 31 S3. 32 . All cores, blanks and tools are on flint , but q uartz and quartzite are represented in the debris category . As identical recovery techniques were used in the excavation of both layers 8 and 9 , t he much higher percentage of the assemblage represented by debris in layer 8 — 77% of the total, up from 5 4 % in layer 9 , the percentages by mass being 30% and 15%, respectively — might indicate that on site knapping was more frequent than before . Note, however, that the retrospective assignment to layer 9 of the finds made at the base of the MO/4 6 trench could only be made for the diagnostic material, not for the potentially associated debris . Given the low numbers involved, the bias introduced by this factor is , however, marginal. Therefore, the lower lithics per square meter ratio of layer 9 probably reflects a higher level of syn depositional scattering , including the downslope washing-away of a larger proportion of the lighterweight debris component. In this context, the lower overall number of blanks, cores and tools in layer 8 (75, dow n from 176 in layer 9) could result from the smaller size of the area in which it was found. T he average mass of (unbroken) discarded cores is the same as in the Mousterian, even though all we re now used for the extraction of bladelets, mostly using the c arinated/nosed “scraper” reduction method ( Fig. S3. 32, nos. 1 2). Th is technological change is reflected in the average mass of complete , retouched and unretouched flint blanks (flakes, small flakes, blades and bladelets) , which decreas es more than t wo fold , from 5.0 g in the Mousterian to 1. 9 g in the Aurignacian. There are also three proximal, unretouched blade blanks and one endscraper made on the proximal end of an overshot laminar blank ( Fig. S3. 32, no. 3). These data suggest a flint economy no different from that obtaining during the Mousterian, i.e., characterized by intensive consumption of mostly imported blanks with on site recycling of exhausted volumes (broken or worn out blades and flakes) to extract small amounts of cutting edge used to retool o r replenish the tool kit . The fact that the toolblank ratio (0.8 by mass ) i s in layer 8 even high er than in layer 9 further supports this conclusion , but there is a significant difference: in layer 8 , no cores, core preparation or core trimming by products document the on site production of non microlithic blanks .


Functionally, both occupations yielded evidence for the carrying out of the same kinds of tasks, as t he activities represented by the layer 8 retouched tools are the same as in layer 9 : usewear document s the mounting of bladelets, including unretouched ones, as elements of composite projectiles ( Fig. S3. 32, no. 6) , while wood and hide processing tasks were carried out with scrapers and atypically retouched flakes ( Fig. S3. 32, no. 3; n ote how this item’s polished surface is abruptly cut by the retouch, documenting on site re sharpening of a previously used tool) . The differences are technological, relate to the methods used in the reduction of available volumes — now geared towards the production of bladelet s — and result in a more efficient exploitation of the raw material. Typologically , the se differences are expressed in the replacement of sidescrapers by endscrapers , and in weapon tips being now armed with series of barbs weighing in the range of 0.2 g each instead of a single, axially fitted, triangular point weighing in the range of 5 to 30 g. However, the spatial segmentation of the reduction sequence, as best seen in the lack of evidence for the on site production of blades, sugge sts occupations whose logistic, hunting component overshadowed the residential one. Despite the small size of the assemblage, there can be no doubt as to its affinities with the Aurignacian ; i ts stratigraphic position — sandwiched between Mousterian and Ea rly Gravettian layers — matches its combination of carinated or nosed “scrapers” with very small bladelets bearing a marginal, mostly inverse or alternate retouch ( i.e., of the Dufour type ) . Specifically, the presence of one nosed “scraper ,” and the fact t hat 40% of the 15 bladelet blanks complete or sufficiently complete to assess torsion are twisted (e.g. , the retouched specimen of Roc de Combe subtype illustrated in Fig. S3. 31, or the unretouched bladelet illustrated in Fig. S3. 32, no. 4 ) support assignment to the Evolved (Aurignacian II) rather than the Early (Aurignacian I) phase of the technocomplex. 3. 9 .2 . Early Gravettian (Layer 7b) The Early Gravettian stone tool inventory (Tables S3. 9S3. 10) nearly quadruples that of the Aurignacian in both number and mass . This increase is due neither to a large r amount of debris ( 73% of the total , about the same as in layer 8 ) nor to the marginally larger size of the area of the site in which layer 7b was present (see Figs. S3. 22S3. 23). All other things having remain ed equal, these numbers suggest that either layer 7b spans a longer time interval or FDM was now being more frequently occupied . As before, flint is overwhelmingly dominant (97% by mass), with quartzite being represented by a handful of flakes and some chippage. By mass, however, the average size of unbroken cores (6.1 g) is halved by comparison with both the Mousterian and the Aurignacian , even though the average mass of complete, retouched and unretouched flint blanks remain s broadly similar (2. 4 g in layer 7b, 1.9 g in layer 8). This pattern correlates with a change from carinated/nosed “scrapers” to “burins” as the preferred type of bladelet core . It may well be the case, therefore, that the emergence of this preference ref lects a continuing trend towards increased efficiency in the exploitation of immediately available volumes , in the context of a flint econom y that remained characteriz ed by intensive consumption of imported blanks coupled with on site recycling of exhausted material.


The toolblank ratio of layer 7b (0.05 by mass) is more than one order of magnitude below the ratio seen in both the Mousterian and the Aurig na cian . This is due to the fact that, three notches and three atypically retouched piece s excepted, this layer’s retouched tools are all small, light weight projectile points: G ravette (see Fig. S3. 33 ) or microgravette points, and backed bladelets (all broken, probably fragments of microgravettes, given the sur enclume retouch and impact fract ures seen on most). Sampling bias cannot explain the lack of endscrapers because the y are present in the layer 8 assemblage, which is significantly smaller. In addition, the functionally equivalent sidescrapers dominate in layer 9. The refore, even though t he economy of flint remained the same, a significant change in functionality would seem to have occurred at the onset of the Gravettian : t he shelter would have becom e a place used in more specializ ed manner , in relation to hunting activities only and with few to no domestic tasks ( e.g. , hide work ing) being carried out on site. This inference, which remains to be tested via usewear analysis of the unretouched material, is consistent with two other observations: the absence of blade cores, core trimming or core preparation elements, and the fact that all bladelet cores derive from the recycling of dbitage blanks. The fact that >25% of all bladelet blanks (and 33% of the retouched ones , all transformed in backed micro points ) bear diagnostic marks of having been removed from “burins” corroborates the techno economic classification of the se items as cores. Typologically, the significant feature of layer 7b is the disappearance of the Dufour bladelet , replaced by ch aracteristically Gravettian backed elements. There are two marginally backed bladelets, but one comes from around a small burrow at the base of layer 7b of grid unit E9 that disturbed the underlying Aurignacian in layer 8. This item could well be, therefor e, in derived position. However, this microlith type was found through the site’s Upper Paleo lithic sequence, so no a priori reason exists to question its presence in the layer 7b bladelet tool assemblage. 3. 9 .3 . Middle Gravettian (Layer 6/7) Tables S3. 11S3. 12 provide stone tool counts for layer 6/7 . As the percentage of debris is roughly the same as in layers 8 and 7b, including or excluding them does not change the fact that the total number of finds more than triples relative to layer 7b. This increase is broadly proportional to the difference in the size of excavated areas (see Figs. S3. 22 S3. 23) . W hen only flint is considered, the m ass data replicate the pattern, even though including two large quartzite cores biases the mass per square met er ratio . As is also the case with limestone, t he use of this locally available raw material was , however, anecdo t al. The logic of flint economics therefore suggests that not much changed in the frequency of site visits. T he average mass (12.4 g) of the unbroken, disc arded flint cores increases in layer 6/7 to about the same as in the Mousterian and the Aurignacian . However, the average mass ( 1.8 g) of complete, retouched and unretouched flint blanks is in line with the values for the earlier Upper Paleo lithic occup a tions of the site , while t he toolblank ratio (0.1 by mass as well as by number ) , even though doubling the Early Gravettian value , remains more in line with the latter than with the much higher values seen earlier on.


This aspect of continuity with the Ear ly Gravettian is insufficiently explained by the importance of the microlithic component of the toolkit, which represents only 3 3 % of the total in layer 6/7, down from 57 % in layer 7b. In addition, e ndscrapers , absent in the latter, are now well represented: among the tools that could be typologically classified they are 14 . Adding seven broken ones that were counted as part of the “retouched piece fragment” category , endscrapers amount to nearly a quarter of the retouched tools found in layer 6/7. Knowing that the average mass of cores is higher than in the Early Gravettian while that of the blanks remained broadly the same, the mass and typology of layer 6/7’s r etouched tool s would seem to point out to one of the following: (a) a type of site occupancy th at resulted in more waste and was therefore less stressed by rawmaterial procurement concerns; or (b) a n underestimati on of the amount of cutting edge actually use d resulting from significant functi onal use of unretouched blanks . Bearing in mind the amount of debris, by products , and lesserquality blanks discarded in the course of the reduction of a single large volume (for an archeo logical example of such a complete sequence, see the refitted mottled flint core reduced in Terminal Gravettian layer 2 of Lapa do Anecrial, in Portugal; Zilho, 1997 ; Almeida et al., 200 7 ), the former would seem to be the parsimonious hypothesis. The notion that flint was being less intensively reduced and that more on site knapping of flint volumes introduced as raw or te sted nodules was taking place in layer 6/7 times is consistent with two other observations: (a) the presence of a significant number (27) of core preparation and core trimming elements, inc l uding platform rejuvenation tablettes and crested flakes, blades and bladelets; and (b) the fact that, excluding fragments, prismatic types bearing blade and bladelet removal scars, none of which were found in layer 7b, represent a quarter of all cores retrieved in layer 6/7 and amount to a total mass of 231.0 g ( almost as much as the Aurignacian assemblage as a whole). Taken together, these lines of evidence suggest continuity with the Early Gravettian and preceding periods in the low frequency of site visit s, coupled with a shift back to the Mousterian pattern of occupations being of a more residential nature — hence the increase in the number and mass of tool kit components related to domestic and on site production tasks , and the representation of all stages of the stone tool reduction sequence. Residency, even i f still for short durations , is additionally hinted at by the anthropogenic imprint possibly represented by the black, micro charcoal staining of the sediment in the area featuring the higher concentration of artefact remains (grid units K M/4 -6 , across wh ich the “layer 6” aspect of layer 6/7 could be observed ; see Figs. S3. 5C, S3. 8, S3. 14, S3. 2 4, S3. 27 ). It was adjacent to this area of layer 6/7, in grid units I J/5 6, that the only substantial faunal remains recovered at the site were found — a cluster of Equus sp. teeth , a couple complete, the rest fragmentary . E xcluding the few Holocene intrusions present in layer s 2 and 3, these teeth amount to 87%, by mass, of the site’s 128.7 gstrong assemblage of faunal remains . Their preservation must be related to highly localized chemical conditions : micro sheltering from runoff coupled with incipient , syn depositional cementation , the latter due to increased carbonate precipitation in the presence of ash and in proximity to both back wall and bedrock .


Th e sh ift in emphasis from the essentially logistic occupations of layer 7b times back to ones of a more residential nature probably explains as well the number of body ornaments found in layer 6/7 ( Table S3. 13; Fig. S3. 34). Indeed, layer 7b yielded no more than two small, sieve retrieved marine or fluviatile shell artefacts, while a third comes from a large burrow cutting through layers 7, 7b and 8 . In contrast, layer 6/7 yielded 16 marine or fluviatile shell finds . The perforations borne by some (mostl y gastropods and scaphopods, but also an ochred halfvalve of Pecten jacobaeus ; Fig. S3. 34 , no. 1) suggest they were used, combined or in isolation, as beads, pendants, necklaces or other items of personal ornamentation requiring suspension. A number of va lve fragments from different taxa may correspond to fragments of the same, to cutting tools ( as in e.g. the case of Mytilus sherds), or to containers. In agreement with the fact that diagnostic “burin spalls” are one third of all bladelet blanks, either retouched or unretouched (and 57% of the total among backed microliths) , “b urins” remain the most common type of bladelet core ; a s in layer 7b ( Fig. S3. 33) , on truncation types dominate , and Gravette points (represented in layer 6/7 by a single apical fragment with diagnostic impact stri ations ; Fig. S3. 35 ) are outnumbered by microgravettes ( Fig. S3. 36 , no. 3) . The domestic tool kit is mostly made up of endscrapers, two of which (L6 38 and K4 24) have been examined for usewear and present typical hide working polishes ( Fig. S3. 35). Notched pieces come next, and three continuously retouched blades complete the domestic tool component of the lithic assemblage. 3.10 . THE LATER UPPER PALEO LITHIC 3. 10.1. Site function The sieve collected fraction of the assemblages from layers 2 to 5 remain ing to be sorted and analy z ed, patterns of site occupancy in post Gravettian times can only be derived , for the time being, from field observations and typological considerations ( Table S3.14 ) . This evidence suggest s a pattern of site use similar to that inferred for layer 6/7 . For layer 3, however, the completeness of reduction sequences, the number of blade and bladelet cores and of core preparation and core trimming elements implies more frequent and lengthier vis its . Th ese inferences are supported by the presenc e of hearth features in layer 4. No such features were found in the other units of the site’s later Upper Paleolithic sequence , in which , however, the absence probably relates to preservation issues : i n layer 5, it can be explained by the higher energy of the syn depositional environment; i n layer 3, by the homogeni z ation of the deposit necessarily deriving from a more intensive use of a more restricted space. Such preservation factors, however, fail to explain why , above layer 6/7, marine or fluviatile mollus k shell, and hence items of personal ornamentation, are represented by only a single specimen (a scaphopod retrieved in the upper reaches of layer 5 during the 2016 field season) . One possibility is that the phenomenon relate s to issues of territoriality, e.g., the geographical extent of these periods’ subsistence and exchange catchments. This is most certainly the case with Epimagdalenian layer 3, as a similar absence seems to characterize the inten sive Upper Magdalenian occupation of the nearby La Boja rock shelter.


3. 10.2. Upper Solutrean (Layers 5 and 4) The index fossils leave no doubt as to the Solutrean affinities of layer 5. Laurel leaf foliates are represented by a broken piece recycled into an endscraper ( Fig. S3. 36, no. 2), manufacture errors ( e.g. , the overshot flake that eliminated the opposite edge of a large preform; Fig. S3. 36, no. 1), several fragments of unfinished pieces, and a few, characteristic bifacial thinning flakes . The barbedand tanged, socalled Parpall point is also present ( Fig. S3. 36 , no. 4) . This combination unambiguously places the assemblage in the Upper Solutrean. In the eastern part of the excavated area , layer 5 accumulated atop and against the markedly Westdipping slope of layer 6/7 . Therefore , in this area, the artefact content of layer 5 could hav e been impacted by progradation effects . In the ce ntral and we stern parts of the site , the interface with underlying layer 6/7 corresponded to a major erosional hiatus that , in grid units H K/7 and L/7 9, entailed deep scour ing ( Figs. S3. 14, S3. 27) . This process left the walls of the thusly formed rill exposed to erosion and, hence, available to contribute laterally derived material to the deposit that eventually filled it up, i.e., to layer 5. The homogeneity of the layer 5 stone tool assemblage therefore needs to be critically examined. Given the primarily gravity and/or water driven nature of the formation processes invol ved, their impact, if any, on the layer 5 lithics ought to be most apparent among the smallersize d components : the bladelet tool class , which includes six backed items . A ll come from columns L, M and N of the grid, i.e., from those areas of the excavation in which the Westdipping of the layer’s lower boundary was more pronounced . The “burins” on truncation, which form the overwhelming majority of this core category in Gravettian layers 7b and 6/7 ( Tables S3. 10, S3. 12) and represent >50% of the corresponding typological tool class in layer 5 ( Table S3. 14), represent a different, potentially technologically problematic component . O f the ten found in layer 5, four come from grid unit L8, 15-5 5 cm below the elevation of the uneroded edge of layer 6/ 7 as exposed in grid units K L/5 , and eight come from columns L and M of the grid , e ast of where the interface between layers 5 and 6/7 flattened out. If, in order to filter out such potential sources of contamina tion , we restrict the analysis of the layer 5 assemblage to finds that (a) come from a 2 mwide band of columns D L running along the back wall and (b) were made at elevations above the edge of the rill, we end up with no more than 20 items, all from grid units HI/5 6, J5 and K L/4 -5 . This stratigraphically secure assemblage includes five of the seven Solutrean diagnostics in Table S3. 14, including the Parpall point, but only one burin on truncation and one backed bladelet fragment (with sur enclume retouch) . T he se two items , however, come from the last spit (A7) into which, in the D K/4 7 trench, layer 5 was subdivided ( the base of this spit therefore corresponding to the dcapage of the surface of layer 6/7 ). Outside the black stained “layer 6” areas , such dc a page was fraught with the difficulties arising out of the gradual nature of the stratigraphic interface . I ndeed, the field notes contain warnings to the effect that, in some grid units (namely K5), the exposed surface had undercut the interface by some 2 -3 cm . Given this evidence, whether a component of Gravettian tradition was indeed present in the Upper Solutrean must remain an open issue — at least based on the information provided by the excavation of FDM.


Layer 4 yielded no bifacial thinning byproducts, but contained a small number of impact fract ured shouldered point fragments. Three more were recovered in layers 2 and 3 : they reflect the significant bioturbation undergone by the upper reaches of the stratigraphic sequence, not a persistence of the type into later periods. Th e impact fractured tang of a Parpall point (J7 18; Fig. S3. 36 , no. 5) was also recovered in this layer. Interpreting the latter’s significance is difficult for the same site formation related reasons that complicate the interpretation of the layer 5 assem blage : the possibility, in fact the necessity, that progradation processes will have introduced in the sediments making up layer 4 material derived from areas of layer 5 located upslope to the East. This is especially so for the smallsized items , the impl ication being that we cannot exclude that J7 18 derives from layer 5 . W hether Parpall and shouldered points were used in association during the interval represented by layer 4 is therefore an issue that , at FDM, cannot be resolved with present evidence . 3. 10.3. Epimagdalenian (Layers 3 and 2) The Epimagdalenian affinities of the assemblages from layers 3 and 2 were established by Romn et al. (2013), who already provided a summar y of the data available until 2010, the last field season during which these units were extensively excavated. The counts in Table S3. 14 differ slightly from those given in Romn et al. ’s Table 1 , basically because stricter criteria were followed in separating simply edge worn from atypically retouched items. These differences have no impact on the fundamentals of Romn et al.’s conclusions , which remain unchanged: even though the assemblages could not be dated, the ir typological structure follows the Spanish Mediterranean wide pattern and allow assignment of layer 3 to an early phase of the Epimagdalenian, characteri z ed by the significant number of Malaurie points, and of layer 2 to a more advanced phase of the technocomplex, characterized by the introduction of lunates. 3. 11 . CONCLUSIONS The excavation of the Finca Doa Martina rock shelter advanced our knowledge of the Paleolithic archeo logy of E astern Spain, with implications beyond regional boundaries . The following is a nonexhaustive selection focus ing on key points of ongoing scientific debate: The excavated d eposit contains a record of human occupation ranging from the Mousterian to the Epimagdalenian, showing that the Mula basin was settled during glacial maxima (the Solutrean) as much as during periods of milder, close to present, transitional climatic conditions (the early Epimagdalenian, which radiocarbon dates elsewhere place in the Allerd , and the Mousterian, which, by stratigraphic pos ition, probably dates to the second half of Marine Isotope Stage 3 ). Given the chronological and paleoclimatic range of the site’s culturestratigraphic sequence, the parsimonious interpretation of its hiatuses is that they relate to geological processes, i.e., that such hiatuses are depositional or erosional in nature, not a reflection of human demography ( specifically , that they cannot be interpreted as humans having been driven away from the region for extended periods because of environmental change s d etermined by the rapidly oscillating climates of the Late Pleistocene).


When analyzed from the perspective of flint economics, and considering as well the spatial segmentation (or lack thereof) revealed by the representation of the different phases of reduction sequence s, the stone tool assemblages suggest that the site was mostly used in residential manner ; a focus on huntingrelated as opposed to domestic tasks is , however, apparent in the Early Upper Paleo lithic (and especially so in Early Gravettian lay er 7b) , while more frequent visits , and a more intensive use of the space available , can be inferred for the Epimagdalenian. Contra recent claims that the Gravettian represents the earliest phase of E astern and S outhern Spain’s Upper Paleo lithic sequence (de la Pea, 2013), the stratigraphic position, level of assemblage integrity , and techno typological features of the stone tools recovered in layer 8 corroborate the evidence for the regional presence of the Aurignacian derived from a number of sites in the Pas Valenciano (e.g., Cova Beneito, Mallaetes) and Andalusia (e.g., Cueva Bajondillo) ( Zilho , 2006). T he finding in basal layer 5, during the 2016 field season, of a Mediterranean type backed and shouldered blade ( i.e., an unfinished point) , document s the association of this type with the barbed and tanged, Parpall point. The hypothesis that the emergence of the latter predates the first appearance of shouldered points and defines an initial phase of the Upper Solutrean in the region , as suggested by Tiffagom (2006) and accords with the sequence seen in Portugal, e.g. at Gruta do Caldeiro (Zilho, 1997), is therefore not supported; but neither can it be excluded, given the significant hiatus that separates layer 5 from the underlying Gravettian depos it. If the area excavation of FDM had been limited to 10 15 m against the back wall and around the initial test trench we would have failed to detect the Early Gravettian, Aurignacian , and Mousterian occupations, and we would have been unable to carry out a proper analysis of formation process ; our results show how, at sites exposed to syn depositional surface dynamics, bioturbation, an d post depositional disturbance — as is the rule with rock shelters — the interpretation of stone tool assemblages require s prior assessment of their homogeneity , even though the assemblages are recovered in and/or labelled according to geologically defined units of provenience ; proper evaluation of their homogeneity will in turn contribute to a fuller understanding of lateral variation and accumulation dynamics , both of which depend on the excavation of sufficiently extensive areas . 3.12. REFERENCES ALMEIDA, F.; BRUGAL, J. Ph.; ZILHO, J.; PLISSON, H. (2007) — An Upper Paleolithic Pompeii: Technology, Subsistence and Paleoethnography at Lapa do Anecrial , in BICHO, N. (ed.) — From the Mediterranean basin to the Portuguese Atlantic Shore: Papers in Honor of Anthony Marks. Actas do IV Congresso de Arqueologia Peninsular, Faro, Promontoria Monogrfica 07, Universidade do Algarve, p. 119 139. BRONK RAMSEY, C. (2009) — Bayesian analysis of radiocarbon dates . Radiocarbon, 51(1), p. 337 360.


MORENORUEDA, G. (2006) — Seleccin de hbitat por dos subespecies de Iberus gualtieranus (Gastropoda, Helicidae) en Sierra Elvira (SE de Espaa) . Zoologica Baetica , 17, p. 4758. PEA, P. de la (2013) — The beginning of the Upper Paleolithic in the Baetic Mountain area (Spain) . Quaternary International, 318, p. 69 89. PLISSON, H. (1985) — tude fonctionnelle d’outillages lithiques prhistoriques par l'analyse des micro usures: recherche mthodologique et archologique . Ph.D. dissertation, Universit de Paris I (Panthon Sorbonne). REIMER, P. J.; BARD, E.; BAYLISS, A.; BECK, J. W.; BLACKWE LL, P. G.; BRONK RAMSEY, C.; BUCK, C. E.; CHENG, H.; EDWARDS, R. L.; FRIEDRICH, M.; GROOTES, P. M.; GUILDERSON, T. P.; HAFLIDASON, H.; HAJDAS, I.; HATT, C.; HEATON, T.J.; HOFFMANN, D. L.; HOGG, A. G.; HUGHEN, K. A.; KAISER, K.F.; KROMER, B.; MANNING, S. W .; NIU, M.; REIMER, R. W.; RICHARDS, D. A.; SCOTT, E. M.; SOUTHON, J. R.; STAFF, R. A.; TURNEY, C. S. M.; VAN DER PLICHT, J. (2013) — IntCal13 and MARINE13 radiocarbon age calibration curves 0 50000 years calBP . Radiocarbon, 55 (4), p. 1869 1887. ROMN, D.; ZILHO, J.; MARTN LERMA, I.; VILLAVERDE, V. (2013) — La ocupacin epimagdaleniense del abrigo de la Finca de Doa Martina (Mula, Murcia) , in DE LA RASILLA, M. (ed.) — F. Javier Fortea Prez. Universitatis Ovetensis Magister . Estudios en homena je, Universidad de Oviedo/Mnsula Ediciones, p. 167 178. SONNEVILLE BORDES, ; PERROT, J. (1954 1956) — Lexique typologique du Palolithique suprieur . Bulletin de la Socit Prhistorique Franaise, 51, p. 327 335; 52, p. 7679; 53, p. 408412, 547 559. STUIVER, M.; REIMER, P. J. (1993) — Extended 14C Data Base and Revised CALIB 3.0 14C Age Calibration Program . Radiocarbon, 35 (1), p. 215 230. TIFFAGOM, M. (2006) — De la Pierre L’Homme. Essai sur une paloanthropologie solutrenne , Lige, tudes et Recherches Archologiques de l'Universit de Lige 113. ZILHO, J. (1997) — O Paleoltico Superior da Estremadura portuguesa, 2 vols., Lisboa, Colibri. ZILHO, J. (2006) — Chronostratigraphy of the Middle to Upper Paleolithic Transition in the I berian Peninsula. Pyrenae, 37, p. 7 84. ZILHO, J.; ANGELUCCI, D.; BADAL, E.; LUCENA, A.; MARTN, I.; MARTNEZ, S.; VILLAVERDE, V.; ZAPATA, J. (2010) — Dos abrigos del Paleoltico superior en Rambla Perea (Mula, Murcia) , in MANGADO, X. (ed.) — El Paleoltico superior peninsular. Novedades del siglo XXI, Barcelona, Universidad de Barcelona, p. 97 108.


Table S3. 1. FDM . Radiocarbon dated samples . All samples are trench collected fragments of Juniperus sp. charcoal Sample Spit x y z Subsample Lab # Observations Layer 4 H6 37 A5base Hearth 1 Anthr ac o #51 VERA 6171HS in the H6NW cemented ash remnant H6 38 A5base Hearth 1 Vial 1 VERA 5101a in the H6NW cemented ash remnant; single 20 mg fragment, normal size branch Vial 4 VERA 5101bHS in the H6NW cemented ash remnant; single 10 mg fragment, normal size branch Layer 6/7 H6 51 A9 42 57 262 Anthr ac o #1 VERA 6170HS parts of a single charcoal piece broken for analysis H6 63 A10 75 50 265 Vial 3 VERA 5367cHS single , excavation broken fragment, normal size branch Vial 2 VERA 5367bHS single fragment, twig size branch Vial 4, A VERA 5367dHS in 290 mg batch of small fragments, normal size branches Vial 4, B VERA 5367dHS_2 Layer 7b G7 43 A13base 36 54 304 Vial 1, A VERA 5368HS in 80 mg batch of fragments possibly from a single branch Table S3. 2. FDM . Radiocarbon dating results. The ages have been calibrated against IntCal13 (Reimer et al., 2013) in Calib 7.0.4 (Stuiver and Reimer, 1993); the calibrated ages are given as 95.4% probability intervals Sample Lab # Age BP Age cal BP 13 C Observations Layer 4 H6 37 VERA 6171HS 18320100 21906 22417 19.41.8 humic acids H6 38 VERA 5101a 1869080 22375 22780 20.22.2 mild ABA (a) VERA 5101bHS 1918090 22842 23435 23.52.0 humic acids Layer 6/7 H6 51 VERA 6170HS 24450170 28058 28837 23.01.3 humic acids H6 63 VERA 5367cHS 23480150 27408 27864 23.40.6 humic acids VERA 5367bHS 23890170 27654 28354 humic acids VERA 5367dHS 24730170 28378 29188 19.80.6 humic acids VERA 5367dHS_2 26610210 30472 31137 humic acids Layer 7b G7 43 VERA 5368HS 26990220 30765 31312 23.61.4 humic acids (a) alkaline step (0.01M NaOH) ceased earlier due to risk of complete dissolution; the residue was used for the dating Table S3. 3. FDM Mousterian. Lithics found (in derived or inherited position) within overlying stratigraphic units Inventory # Raw material Condition Blank technology Classification In layer 6/ 7 J6 63 quartzite ridge worn, slightly patinated discoid flake (distal) sidescraper fragment In layer 7b 2012 159 [E9] flint retouch exhausted naturally backed flake denticulate


Table S3. 4 . FDM Mousterian (layer 9; 20072014 field seasons ). Stone tool technological categories. T he diagnostic Mousterian material found in derived/inherited position within overlying stratigraphic units and a quartzite manuport are not included . N = number , M = mass in grams RAW MATERIAL CORE S FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOL S TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 9 167.6 15 148 76 154.0 31 39.0 1 2.4 – – 174 50.8 31 93.0 34 206.2 371 861.0 Quartzite – – 3 27.5 3 8.8 – – – – – – 6 6.2 – – 2 53.5 14 96.0 Limestone – – 1 32.2 1 1.7 – – – – – – – – – – – – 2 33.8 Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 9 167.6 19 207.7 80 164.5 31 39.0 1 2.4 – – 180 57.0 31 93.0 36 259.7 387 990.8 Table S3. 5 . FDM Mousterian (layer 9; 20072014 field seasons ). Classification of cores and retouched tools . T he diagnostic Mousterian material found in derived/inherited position within overlying stratigraphic units is not included Cores N Formal r etouched tools N polyhedral 1 Mousterian point 2 Kombewa 2 notched piece 2 Levallois 2 denticulate 1 d iscoid (broken) 1 sidescraper fragment 3 unilateral 10 TOTAL 9 transversal 1 convergent 1 denticulated 3 double 2 fragment 4 atypically retouched piece 2 naturally backed knife 3 retouched piece fragment 5 TOTAL 36 Table S3. 6. FDM . Mousterian and Aurignacian use wear evidence (2007 2014 field seasons; flint ) . Material used on, inferred function , and traces of residue present Illegible None Wood Hide Meat Bone Projectile Ochred Total MOUSTERIAN sidescrapers 4 3 2 2 3 1 – 1 16 denticulates ( a ) 2 – – – – – – – 2 notches – 1 – – – – – – 1 points – – – – – – 2 – 2 TOTAL 6 4 2 2 3 1 2 1 21 AURIGNACIAN endscrapers 2 – – 2 – – – – 4 notches 1 1 – – – – – – 2 atypically retouched or broken 2 1 1 – – – – – 4 bladelet tools – 2 – – – – 2 – 4 TOTAL 5 4 1 2 – – 2 – 14 (a) includes one found in derived position at the base of layer 7b


Table S3. 7. FDM Aurignacian (layer 8 ; 20072014 field seasons ). Stone tool technological categor ies . N = number , M = mass in grams RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 7 69.0 4 18. 1 16 23.8 15 12.1 3 2.8 15 3.6 246 53.4 4 22.5 15 49.1 32 5 254.3 Quartzite – – – – – – – – – – – – 2 0.4 – – – – 2 0.4 Limestone – – – – – – – – – – – – – – – – – – – – Quartz – – – – – – – – – – – – 1 0.3 – – – – 1 0.3 TOTAL 7 69.0 4 18. 1 16 23.8 15 12.1 3 2.8 15 3.6 249 54. 0 4 22.5 15 49.1 328 255.0 Table S3. 8 . FDM Aurignacian (layer 8; 20072014 field seasons ) . Classification of cores, retouched tools and bladelets. Bladelet counts include both retouched and unretouched blanks Cores N Bladelets extracted from N R etouched tools N carinated “scraper” 2 carinated/nosed “scraper” 9 endscraper on retouched piece 1 nosed “scraper” 1 “burin” 4 endscraper on flake 1 prismatic for blades 1 splintered piece/bipolar core 1 broken endscraper 2 prismatic for bladelets 3 other 7 notched piece 1 TOTAL 7 TOTAL 21 Dufour bladelet 4 marginally backed bladelet 1 irregularly retouched bladelet 1 atypically retouched piece 2 retouched piece fragment 2 TOTAL 15 Table S3. 9 . FDM Early Gravettian (layer 7b ; 20072014 field seasons ). S tone tool technological categories . N = number , M = mass in grams) . Counts include t he diagnostic items found in the E G/8 bioturbation feature and exclude a Solutrean piece from a smaller burrow and an inherited Mousterian piece from the base of the layer RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 9 46.1 18 95.1 146 174.3 100 109.1 5 6.1 39 14.3 877 200.6 63 157.0 21 21. 9 127 8 824.4 Quartzite – – 3 13.2 4 9.1 2 2.4 – – – – 4 2.3 – – – – 13 26.8 Limestone – – – – – – – – – – – – – – – – – – – – Quartz – – – – – – – – – – – – 2 0.1 – – – – 2 0.1 TOTAL 9 46.1 21 108.2 150 183.4 102 111.4 5 6.1 39 14.3 883 203.1 63 157.0 21 21.9 1293 851.4 Table S3. 10 . FDM Early Gravettian (l ayer 7b ; 20072014 field seasons ). Classification of cores, retouched tools and bladelets. Bladelet counts include both retouched and unretouched blanks Cores N Bladelets extracted from N Burin types N R etouched tools N “burin” 7 “burin” 13 on truncation Gravette point 2 splintered piece/bipolar core 1 other 38 oblique 2 microgravette 3 fragment 1 TOTAL 51 concave 1 notched piece 3 TOTAL 9 convex 1 backed bladelet 7 multiple 1 marginally backed bladelet 2 transversal on retouch 1 atypically retouched piece 2 Noailles 1 retouched piece fragment 1 TOTAL 7 pointed blade 1 TOTAL 21


Table S3. 11 . FDM Middle Gravettian (layer 6/7 ; 2007 2014 field seasons ) . Stone tool technological categories. N = number , M = mass in grams . T he content of the heavily bioturbated grid units E F/5 -6 is excluded; a limestone cobble used as a hammerstone and four quartzite thermoclasts are not counted RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 75 824 . 3 62 345.1 33 7 6660.0 241 284.8 40 119 . 3 223 170 . 0 2495 686.6 170 397.7 91 214.5 3734 3702.2 Quartzite 2 661 . 0 7 111 . 2 22 10 1.2 4 7 . 8 – – – – 12 5.8 2 28 . 1 – – 49 915.1 Limestone – – 1 4 . 6 3 12 . 1 – – – – – – – – – – – – 4 16 . 6 Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 77 1485 . 3 7 0 460.8 362 7 73.2 245 292.6 40 119 . 3 223 170 . 0 2507 692.4 172 425.8 91 214.5 378 7 4633.9 Table S3. 12 . FDM Middle Gravettian (l ayer 6/7 ; 2007 2014 field seasons ) . Classification of cores, retouched tools and bladelets . Bladelet counts include both retouched and unretouched blanks ; one diagnostic item found in derived position in layer 3 is included Cores N Bladelets extracted from N Burin types N R etouched tools N “burin” 31 carinated/nosed “scraper” (?) 2 endscraper burin 1 simple endscraper 6 splintered piece/bipolar core 7 “burin” 86 dihedral atypical endscraper 3 nodule 1 splintered piece/bipolar core 1 djet 1 ogival endscraper 2 polyhedral 1 other 170 on angle 2 endscraper on retouched blank 2 centripetal 1 TOTAL 269 on break 2 endscraper on flake 1 Kombewa 2 on truncation Gravette /Vachons point s 2 prismatic for blades 2 straight 3 microgravette 7 prismatic for bladelets 13 oblique 7 truncated piece 2 prismatic for flakes 6 concave 2 continuously retouched blade 3 core fragments 13 multiple 4 notched piece 13 TOTAL 77 multiple mixed 1 trapeze 1 Noailles 6 truncated bladelet 3 plan 2 backed bladelet 7 TOTAL 31 notched bladelet 7 marginally backed bladelet 3 atypically retouched piece 11 retouched piece fragment 15 pointed blade 2 pointed bladelet 2 TOTAL 9 2 Table S3.13. FDM Gravettian. Marine and fluviatile shell finds (2007 2014 field seasons) . In all of the perforated gastropods, the perforation is located on the body whorl of the dorsum (a) Ornaments (perforated) Tools, containers , manuports ? (unperforated and/or fragments ) # Taxon Perforation origin # Taxon Description G6 27 Dentalium vulgare natural (tube segment) I4 7 Pecten maximus small rib fragment of large valve G7 23 Nassarius incrassatus percussion I7 34 Cerastoderma sp. small, ochred valve fragment I5 30 Littorina obtusata percussion J5 48 undetermined small fragment of nacre J5 62 Littorina obtusata percussion J5 49/K5 71 Mytilus sp. two small, conjoining valve fragments J7 27 Pecten jacobaeus undetermined K6 42 Acteon tornatilis unperforated M6 21 Cyclope neritea percussion L5 40 Pecten maximus ventral edge fragment of right valve O4 6 Trivia sp. percussion L5 72 Pecten maximus large ventral edge fragment of right valve F7 13 Dentalium vulgare natural (tube segment ) L6 53 Theodoxus fluviatilis unperforated (a) F7 13 and 2012 111 are from layer 7b , and 2012 148 comes from a large burrow cutting through layers 7, 7b and 8; all the others are from layer 6/7 O4 8 Pecten sp. small fragment of rib 2012 111 Glycymeris insubrica ventral margin fragment of <25 mm long valve 2012 148 Pecten maximus ventral edge fragment of right valve


Table S3. 14 . FDM Later Upper Paleo lithic. Standard typological classification of stone tools (2007 2014 field seasons). Following the type list of Sonneville Bordes and Perrot ( 1954 56 ) with the modifications introduced by Zilho (1997) Layers Layers # Type 2 3 4 5 # Type 2 3 4 5 ENDSCRAPERS RETOUCHED BLADES 1a simple on blade 4 8 2 3 65 unilaterally retouched blade – – 1 – 1b simple on flake 1 1 – 2 66 bilaterally retouched blade 1 – 1 – 2a atypical , s i mple on blade – 1 1 – SOLUTREAN TOOLS 2b atypical , simple on flake 1 2 3 – 70n fragment of bifacial foliate – – – 6 3 double – 4 – – 72a fragment of shouldered point – – 7 – 4 ogival 3 – – 3 72b Parpall point – – 1 1 5a on retouched blade 2 4 4 2 SUBSTRATE 5b on retouched flake – 2 1 – 74 notched piece 1 4 4 8 6b on laurel leaf fragment – – – 1 75 denticulate – 1 – – 8 on flake 1 1 – – 76 splintered piece – 2 1 3 10 thumbnail 5 6 – – BLADELET TOOLS 13 nosed – – – 1 83 segment 2 – – – 14a flat nosed – 1 – – 84 truncated – – 2 2 COMPOSITE TOOLS 88 denticulated – – 1 – 17 endscraper burin 1 2 – – 89 notched – 2 – 1 18 endscraper truncation – 1 – – 85a backed 3 10 2 1 PERFORATORS 85c partially backed – – 1 – 23 perforator – 1 – – 85f backed , fragment 29 60 19 5 BURINS 86a truncated backed – – 1 – 27 dihedral straight – – – 1 87a denticulated backed – – 1 – 28 dihedral djet 2 1 1 3 90a Dufour – 1 4 – 29 dihedral on angle – 2 1 4 90b Areeiro 1 1 – – 30a angle on break 1 1 1 – 90c marginally backed – – 9 6 31 multiple dihedral – 2 1 – 91b Malaurie point 2 9 – – 34 on straight truncation – – – 2 91d fusiform point – – 1 – 35 on oblique truncation 1 1 – – VARIA 36 on concave truncation – – – 4 92a atypically retouched piece – 10 – 1 37 on convex truncation – 1 – 1 92b retouched piece fragment 5 7 13 8 38 transverse – 1 – 1 92c pointed blade – – 1 1 40 multiple on truncation – 1 2 1 92d pointed bladelet – – – – 41 multiple mixed – – 1 1 TOTAL 71 157 90 7 5 44a plan – 1 – – the layer 2 counts subsume the s urface and layer 1 material type 70n includes one layer 5 item found in a burrow in layer 7b type 72a includes three layer 4 items found in derived position in overlying layers 2 and 3 type 91b includes one layer 3 item found in the large rabbit warren of grid units I J/5 6 and their periphery BACKED TOOLS 51a microgravette – – – 1 51b unilateral micro point – 3 1 – 51c unilateral micro point fragment 2 1 – – 51d fragment of retouched base micro point 1 1 – 1 51e bilateral micro point 1 – – – 58 backed blade fragment 1 – – –


Chapter 4. The rock shelter of La Boja 4. 1. DESCRIPTION The Abrigo de La Boja (ADB ; 3804’43” N, 0129’23” W ) is located 50 m downstream from Finca Doa Martina (FDM) , on the same calcarenite rock face that structures the left bank of the middle section of the Rambla Perea gorge ( Figs. S1. 1S1. 2 ). When seen from the SE, the cliff displays here a wedge like recess akin to FDM ’s , even i f smaller and less pronounced. T his c ommonality of pattern suggest ed the potential existence, buried under slope deposits , of an archeo logical site formed in the context of a “back wallrising above horizontalplatform” morphology created by the same processes that operated at FDM : tectonic (joint formation) , and erosional (slabbing off of multi-ton prisms) . These expectations were confirmed . The excavation exposed a sub vertical, nearly plane , inwardly slant , several meter shigh back wall, and , opposi te, an outer wall formed by the NE SW ori ented , flat face of a large rock mass ( Figs. S4. 1S4. 5). Bedrock could be reached over no more than 2 m adjacent to the back wall, but the existence of an extensive basal platform supporting the stratification is suggested by th e fact that the deeper occupation horizons tend to be more substantial and richer as one moves outward ( Figs. S4. 6S4. 7) . The fill of ADB spans the same time interval as FDM’s but the band of terrain behind the former’s current drip line is only ca.4 m at its widest and no more than ca. 10 m long in total — i.e., it is both smaller and narrower. Compared to FDM’s , however, t he p ost depositional alteration of the succession is negligible , impl ying little or no exposure to the elements through the sedimentary accumulation process . From t his apparent contradiction, we can infer that significant change occurred in the morphology of ADB since it was first used by humans . Indeed, t he multi ton slabs interstratified in the deposit, which document tw o major episodes of roof fall ( Figs. S4. 6S4. 7) , bear witness to a process of change in the geometry of this locality . In addition, the many large boulders strewn downslope of the site corroborate that the cliff face receded significantly over time. Determining how the site originally looked like requires additional excavation . W ith current evidence, two hypotheses can be entertained. One is that the site once was a true cave and became a rock shelter only after the breaking off of the rock mass curr ently bounding the sedimentary fill to the South . In this scenario, that mass would represent the upper portion, now loose and inward tilted , of the cave’s original external wall. The parsimonious hypothesis , however, is that the site always was a rock she lter and that the rock face presently making for an external wall correspond s to the flat bottom of a massive chunk of the overhang — which, formerly , would therefore have protected a much more extensive area . In this scenario , the space between the back w all and the inward face of the tilted collapse would nonetheless have also functioned as a kind of cave — closed to N and E, with an entry facing S , and naturally roofed by the substantial overlap of the two drip lines (external, at higher elevation, the shelter’s; internal, at lower elevation, the collapsed overhang’s).


The pristine preservation of fireplaces with thick ash deposits across the Aurignacian to Solutrean sequence suggests the persistence of cavelike conditions through out . Conversely, the lesser integrity characteriz ing the fire features of the Upper Magdalenian could relate at least in part to the site having by then already acquired its extant form. 4.2. EXCAVATION APPROACH AND METHOD S To verify whether an archeo logical fill did exist subsurface , a n initial 2 m test trench , p laced where the band of terrain behind the extant drip line i s widest (grid units S/4 5) , was open ed in the S pring of 2008 ( Fig. S4. 1A). It was then enlarged over a couple of annual, one month long field seasons . As the excavation went deeper into the deposit , the trench was stepped to keep the height of its walls within safety parameters ( Fig. S4. 1B D). Between 2008 and 2010 , a n area of ca.6 m was thus taken down to ca.2.5 m below surface, from where a deep sounding (grid units T/5 6) , launched in 2012 , eventually reached bedrock ( Fig. S4. 1E F). Between August 2013 and November 2014 , the gridded area (squares SX/2 6) was open area excavated over a total of seven months of field work ( Figs. S4. 2S4. 3). Eventually, t his wide r trench also had to be stepped . T he size of excavated surface s thus decreased with stratigraphic depth : from 20 m in the Magdalenian to 14 m in the Solutrean and Gravettian, 8 12 m in the Aurignacian, and only 2 m in the Mousterian ( Fig. S4. 4). In 2016, the excavation was extended outward ; at the end of the field season, it had reached the Middle Solutrean ( Fig. S4. 5). The yellow sand y deposit encountered by the excavation a t the base of the dark, Holocene soil extended all the way down to bedrock ; only the signature left by the site’s successive human occupations allowed stratigraphic subdivision ( Figs. S4. 6S4. 7 ). At dcapage, t his arc heo logical stratigraphy defined by hearth features, soil burning, and/or artefact scatters coul d be followed with sub centimeter precision over extensive surfaces . Away from the area in which the anthropogenic signature was strongest , the dcapage w ould be extended laterally along the same plane, using stone lines as an auxiliary instrument . Each of these true paleo surface s, plus the thickness of enveloping sediment above , was defined as an “Occupation Horizon” (OH). Intervening deposits found to be sterile or containing scarce, intrusive remains were designated as “Intermediate Levels” (IL). When OHs were thick or featured multiple, stacked up occupation lenses , they were subdivided to the extent possible. In the field, each excavated slice (“spit”) was numbered sequentially (in the format “A1” to “An”) . Even though , i n some cases , a one to one , spit to OH correspondence exists, often each OH is an analytical unit made up of the addition of two or more spits. Even though the same general approach was followed in the initial phase , the limited size of the trench implied that some OHs could not be recogni z ed, while the delimitation of some that were only marginally represented was imprecise. The integ ration of the two phases of excavation into a single archeo stratigraphic scheme used a 3D model in which the paleo surface s of 20132014 were projected o nto the testexcavated grid units using boundary information derived from cross section record s. In th e few cases where ambiguity remained as to the assignment of a given test excavation unit to one of the 2013 2014 OHs , the corresponding finds were removed from consideration.


Otherwise , given broadly similar levels of bone and shell preservation, work fol lowed , with a few differences , the methodology described in Chapter 3 for the excavation of FDM. Finds were piece plotted with the same cutoff criteria, against the site grid and a datum placed on the back wall at the elevation of ca. 399 m; from the begin ning, however, finds were numbered sequentially, 1 to n, per year of excavation (e.g., 2008 1 to 2008n, etc. ) , and the sediment for sieving was s eparated per m units . In the open area excavation phase, the <2 mm fraction of two such units was entirely saved for subsequent flotation. Sediment samples associated with the fire features were taken for phytolith analysis , and charcoal collected in and around them was sampled for biomolecular analysis . C ontrol samples from non anthropi z ed areas of the same pa leo surface s were taken alongside. 4. 3 . STRATIGRAPH IC OUTLINE 4.3.1. Archeo stratigraphic units Figs. S4. 6S4. 7 illustrate the main cross sections recorded at the end of the 2012 and 2014 field seasons, respectively. C ombined, they represent the horizontal and vertical variation observed across the excavated area of the site. The radiometric dating results available for the archeo stratigraphic sequence and the corresponding provenience information and analytical data are given in Tables 23 and S4. 1, and in Figs. S4. 8S4. 9. F rom top to bottom, the sequence can be subdivided into the ten main blocks described in the following . Holocene soil Gr ey, pulverulent sediment with abundant vegetal detritus and rare artefacts (wheeled pottery, metal, a few prehistoric sherds). It corresponds to the field unit “layer A.” OH0 B and of brown grey sediments (field unit “layer B” ) overlying brown yellow sands (field unit “layer C”) containing small, slablike clasts . Layers B and C contain scarce stone tools of Ep imagdalenian affinities, akin to those retrieved in layer 3 of FDM . Due to animal burrowing and the presence of anthropogenic , pit and channel like disturbance features penetrating from the ground surface deeply into the deposit , this horizon’s stratigraphically intact areas are of limited extent. OH1 Band of yellow sands, heavily anthropi z ed, with abundant charcoa l and ash , which give it a grey colo r over extensive surfaces (field unit “layer D”). It contains diagnostic Upper Magdalenian stone tools. Negative features filled with a dark, organic matterrich sediment containing abundant land snail shell and charcoal traverse the horizon in its entirety . These features correspond to the base of the network of large animal burrow s penetrat ing into the Pleistocene deposit from the overlying Holocene levels ( Fig. S4. 10 ).


OH2 to IL1b Uppermost part of the yellow sands with clasts of different sizes that, variably anthropiz ed, extend all the way down to bedrock (field unit “layer E”) . OH2 is a lithic artefac t scatter found at the interface between layers D and E in rows 23 of the 2013 2014 open area excavation . Sample 2008 774 , collected in the initial test trench, has the same interface provenience and may well relate to this OH2 occupation. Immediately b elow OH2, artefacts are still found, downwardly displaced, in bioturbation features . In the excavated area, however, the deposit then becomes virtually sterile until a paleo surface defined by an eroded hearth feature is reached. Denoted by its imprint (su bsurface sediment reddening; Fig. S4. 1 1) , this feature and the few associated , undiagnostic artefacts form the OH3 context. The deposit above, between OH2 and OH3, is IL1a (ca. 20 cm thick) , the deposit below, between OH3 and OH4 (ca.10 cm thick), is IL1b . The dating evidence shows that the OH2 IL1b package, even though 35 cm thick , accumulated very rapidly , in less than one millennium, and, as with OH1 above, entirely within the time frame of the regional Upper Magdalenian. OH4 to OH12 Extensively anthropi z ed yellow sands featuring a complex stack of hearth defined paleo surfaces ( Figs. S4. 12S4. 20) that associated artefact s and radiocarbon dating place in the Last Glacial Maximum (LGM). From top to bottom, the assemblages are Early Magdalenian, Solutreo grave ttian, Solutrean and Gravettian. At this time, the site seem s to have been visited regularly , with stratigraphic segregation of the different horizons rendered possible by a steady rate of sedimentary accumulation . An erosional hiatus between OH7 and OH8 is suggested by a change in the dip of the paleo surfaces and by their being in direct contact , with no intermediate sterile deposit, despite more than a thousand years apart . Another erosional episode exists at the interface between OH11 and OH12, which , in the excavated area, are separated by two millennia. Adjacent to the back wall, the existence of this latter hiatus could be suspected based on the amount of syn depositional, small-mammal burrowing observed at the corresponding elevation (F igs. S4. 17S4. 18). Outward , the hiatus manifests itself in the rill that cuts the OH12 deposit and is filled with a dark, organic matter rich sediment whose nature and origin remain to be established ( Fig. S4. 19 ). IL2 to OH14 Large collapsed slab, on average ca.50 cm thick, and sediment fill of the empty space extending to the back wall behind . Based on a microgravette found just below the slab’s underside and the radiocarbon date for the small hearth lit against the base of its inward sloping upside ( Fig. S4. 21), t he time of the collapse can be constrained to between the Late Aurignacian deposit it rests upon and the Early Gravettian. T h at hearth defines an exceedingly poor OH13 occupation . The thickness of deposit between OH12 and OH13 is IL2 . It was sterile , except at the very top, where the finds represent downward migration from OH12. The thin deposit comprised between OH13 and the topographic base of the collapse is IL3 . OH14 is the wedge of deposit filling the latter’s raised butt .


OH1 5to -O H20 This package corresponds to ca.75 cm of “layer E” sandwiched between two episodes of major roof collapse . The uppermost horizon, OH15, underlies the a rtefact poor and undated OH14 , which yielded a microgravette and, therefore, given its position in the stratigraphic sandwich, belongs in the Early Gravettian. Lithic assemblage composition and radiocarbon dating place in the Aurignacian the six horizons discriminated within the OH15 OH20 package, which feature s well preserved, spatially extensive fire features ( Figs. S4. 22S4. 28) . Numerous s hell beads and , in OH17 , even a bone needle , were also found . The dating suggest s a hiatus between OH1 8 and OH1 7 , which accords well with the fact that the OH18 hearths are eroded ( Fig. S4. 25 ). OH20 represents the earliest manifestation of the Aurignacian at the site but, in the excavated area, yielded no fire features ( Fig. S4.29 ) . Th is occupation took place once sedimentary accumulation had filled the space behind the collapsed slab atop which its remains are found — levelling the ground surface , that accumulation eventually provide d a sub horizontal platform amenable to settlement . IL4 Large slab, with an average thickness of ca. 90 cm , plus the sedimentary fill of the space extending behind . The slab represents a roof-collapse episode post dating the last Mousterian occupation recorded in grid units T U/6 7. Through the accumulation of IL4, t he narrow space ( Figs. S4.29 S4. 30 ) that remained between the back wall of the shelter and the inward facing upside of the roof collapsed slab was not amenable to human use . Thus, a s with the few items retrieved in the other ILs, the IL4 lithics must reflect downward migration from the overlying occupation — as indeed otherwise proven by t he refits linking them to the OH20 stone tool assemblage. OH21 to OH23 Basal occupation levels , spanning a thickness of ca.75 cm , comprised between bedrock and the base of the IL4 collapse ( Fig. S4. 31) . T he fire reddened lenses observed during excavation allow subdivision of this deposit into three horizons. Th e subdivision remains tentative and is bound to be revised once the excavation is able to proceed outward and explore the sediments under the collapse. Saprolite Thin level of clast supported, angul ar breccia resulting from the degradation of bedrock under sheltered conditions ( Fig. S4. 31 ; see Chapter 3 for a similar unit at the base of FDM). 4. 3.2. Radiometric d ating The provenience and composition of the samples submitted for radiocarbon dating, all of charcoal, is given in Table S4. 1. T he results obtained are listed in Table 2 . The Olea sample com es from one of the burrows penetrating layer D . It was submitted to test whether the taxon was regionally present in Tardiglacial times or, as suspected , represent ed post depositional intrusion . This sample was dated at the Oxford laboratory (OxA). The other results are all AMS dates run at the VERA facility, University of Vienna.


All but three of the submitted samples were successful. The exceptions are 20121518, 20143129, and 20143421, for which the ABA treatment proved too aggressive. The y concern charcoal taken from the core of fire features, which may well explain the failure , as samples of similar age collected in the periphery of the same or identical features turned out to be wholly unproblematic . Nevertheless, the humic acids result for sample 2014 3421 seems to represent a reliable age estimate, as it is statistically indistinguishable from the results obtained for the other, successful OH1 9 sample (2014 3348) . T he other two unsuccessful samples’ humic acids results are clear underestimations, a s shown by their comparison with the date obtained on the successful sample for the same OH17 unit (20143184). Similar underestimations are apparent when comparing pairs of carbon and humic acids results obtained on other samples (e.g., from OH12, OH20, IL4) . Therefore, all humic acids results have been excluded from consideration and are not included in the Fig. S4. 8 plot. The fact that the magnitude of the underestimation falls within the uncertainty of the results obtained for the Mousterian occupation horizon s probably explains why the effect is not apparent in th ose cases . The reliability of the successful ABA dates is supported by their agreement with the ABOx results for subsamples of the same charcoal ( Table 2) . Optically Stimulated Luminescence (OSL) dating of the associated sediment corroborates the robust ness of the si te’s radiocarbon chronology ( Table 3 ; Figs. 6 -8 , S4. 9) . For the Mousterian, the OSL date s from immediately under the base of the overlying collapse , in OH21, and for OH23, 30 cm above bedrock, constrain the accumulation to the ca. 51. 557 .7 ka (thousands of years) interval — in agreement with the terminus ante quem provided by the radiocarbon dating to 40.8 42.4 ka of a charcoal sample taken in the overlying IL4 deposit. Likewise, radiocarbon places the OH17 and OH18 horizons in the 34.2 35.6 and 34.9 37.0 ka intervals , which are encompassed by those obtained via OSL for the same units ( 28.8-36.4 and 30. 2-41.4 ka, respectively). 4. 3.3. Culture stratigraphic sequence Data on the stratigraphic distribution of personal ornaments and stone tool types found across the ADB stratigraphic sequence, grouped by major occupation phases, are provided in Tables S4. 2S4. 3 . The vertical variation in the presence of the mollus k taxa used in the manufacture of shell beads is plotted in Fig. S4. 32. Representative sample s of the stone tools from the mid and later Upper Paleo lithic occupation horizons and of the ochred and perforated shell finds made at the site are illustrated in Fig. S4. 33 and Fig. S4. 34, respectively . Preparation and analysis of the large artefac t collections being a work in progress, the counts in Table s S4. 2S4. 3 are bound to be modified, for instance via the inclusion of the sieve fraction . So far, this fraction has been processed for the basal levels (Early Gravettian, Aurignacian, and Mousterian) only. For Upper Magdalenian horizon OH1 ( Fig. S4. 33 , nos. 1 4), counts based on the 122 formal tools collected 2008 2009 have already been published (Zilho et al., 2010 ). T he finds made since multiply that total but do not change the structure of the assemblage . Marine shell beads are present in OH2 but , so far, absent from OH1, where the meaning of an ensemble of spatially clustered, unperforated Melanopsis river shells from grid units X/5 6 remains elusive.


The Early Magdalenian is represented in OH4, dated to ca.18.6 ka on a hearth associated sample ( Fig. S4. 12). The assemblage is small and almost exclusively made up of backed and very small, marginally backed bladelets ( Fig. S4. 33 , nos. 5 6). Lucena et al. (2012) provide preliminary data on the Solutrean sequence. The Solutreo gravettian ( Figs. S4. 13S4. 1 4; Fig. S4. 33 , nos. 7 8) is represented in horizons OH5 and OH6 , dated to the 19.8 21.3 ka interval . The dominant tool type s are backed and/or pointed microliths ; e xcluding retouched piece fragments, such items represent 46% of the total. A n unifacial foliate was recovered in two fragments, distal and proximal, that join minimally at the break . S o far, this is the only recovered item evok ing the Solutrean ancestry of the technocomplex. The U pper Solutrean is represented in horizon OH7 ( Fig. S4. 1 5; Fig. S4. 33 , nos. 9 -13) , dated on a hearth -a ssociated sample to ca.23.2 ka. It yielded a small number of shouldered points associated with backed and marginally backed bladelets in an industrial cont ext dominated by endscrapers. No evidence of bifacial thinning was found . A small (basal?) fragment of an unifacial foliate point dorsally featuring the covering, flat re touch typical of the Solutrean is tentatively classified here as a willow leaf. The Middle Solutrean is represented in horizons OH8 and OH9, dated on hearth associated samples to 24. 3-25.2 ka ( Fig. S4. 16) . In agreement with their position in the sequence and regional chrono stratigraphic patterns, these horizons yielded a few bifacial thinning flakes and a bifacial foliate fragment. Lower Solutrean assemblages featuring a total of 2 4 unifacial points — 21 % of these occupations’ retouched piece s ( fragments excluded ) — were recovered in OH10 and OH11, dated on hearth associated samples t o 25.025. 5 ka ( Figs. S4. 17S4.19; Fig. S4. 33 , nos. 1 417) . Aside from the Upper Magdalenian, these occupation s w ere the site’s densest and yielded the most retouched tools ( despite the material from the rill in row 3, where obviously reworked Gravettian items were mixed in , not having been count ed). The few backed bladelets and burins on truncation may also derive from the underlying Gravettian, as they all come from peripheral areas at the interface between OH11 and OH12 that had been disturbed by small mammal activity . O therwise , such activity also resulted in Middle Solutrean charcoal — sample 2012178 — being present in the fill of an OH12 burrow ( see above and Tables 2 and S4. 1 ). Combined, the Solutrean and Solutreo -g ravettian yielded most of the sequ ence’s shell beads, including all the Tritia neritea and Acteon tornatilis specimens, a few of which were unperforated ( Figs. S4. 32 , S4. 34 ). A piece of red coral, Dentalium tubes , and o ther small, marine and fluviatile gastropods round up this assemblage, which contrasts with the underlying Gravettian in the absence of perforated bivalves (a small fragment tentatively assigned to Glycymeris being the single possible exception). The Middl e Gravettian ( Fig. S4. 33 , nos. 18 19) is contained in OH12, dated on a hearth associated sample to 27. 427.9 ka ( Fig. S4. 20) . Even though endscrapers made on rather large blanks dominate, the stone tool assemblage features the characteris tic association of backed or marginally backed bladelets with ( numerous ) burins on truncation .


The Early Gravettian, Aurignacian and Mousterian occupations of ADB date beyond 31 ka . They are spread across eleven Occupation Horizons (OH13 to OH23) situated beyond the base of the uppermost of the two large, collapsed roofslabs inter stratified in the “layer E” sands. The se horizons substantiate the transition from the Middle to the Upper Paleo lithic in the region . T heir artefact assemblages ar e presented and discussed below in greater detail. 4. 3.4. Site formation The preservation of undisturbed f ire features bespeaks of the exceptional integrity of most occupation horizons within “layer E.” Small scale bioturbation affected limited areas of the excavated surfaces, and can be expected to have caused some degree of post depositional displacement of finds — mostly of small items brought up by burrowing that ended up mixed on extant ground surface s with the remains of later, coeval with the time of burrowing, occupatio ns. Otherwise, when assessing the potential presen ce of intrusive or inherited material among the remains excavated from any given horizon , the only processes one need s to bear in mind are “wall effect,” dcapage error, and palimpsest formation (the latter, when dealing with material either side of the few hiatuses of erosion and/or sedimentation identified in the sequence ) . The distribution of phytoliths across the different lenses that make up a given OH provides independent corroboration tha t, where the artefact content of the deposit is concerned, an exceptionally high degree of assemblage integrity ought to be expected. The results obtained for samples taken in exposed cross sections and spanning the micro stratigraphy of each OH over thick nesses of 2 to 5 cm reveal a very clear pattern : (a) high concentrations in the white lenses ; (b) low concentrations in the underlying black lenses ; and (c) zero or near zero values in the reddened sediment below as much as in the yellow matrix sandwiching the fire altered deposit ( Fig. S4. 35) . The FTIR (Fourier Transform Infrared) spectroscopic analysis of the white lenses shows that they are formed of ash entirely generated by the burning of wood (not bone or grass). Conversely, t he absence of such a whit e lens (and, hence, of phytoliths), as in Column F ( Fig. S4. 35) , coincides, in plan view (Fig. S4 .24 ) , with unstructured surfaces whose above ground fire features we re not intact. These phytolith data fully support the geo archeo logical interpretation of the micro stratigraphic organiz ation of the different OH lenses developed in the course of excavation. The black lenses represent the ground fire was lit on . T heir geometry is therefore a proxy for the topography of the surface at the time of occupation — that upon which the associated remains were discarded. The red lenses represent the immediate subsurface sediment — the color change result s from chemical alteration of the iron component , brought about by the high temperatures reached in the fireplaces . T hese lenses may contain some trampled in material related to the occupation above. The white lenses are an anthropogenic deposit consisting of the ash formed by the burning of the fuel used in the fireplaces. Finally, the yellow sediment covering each of t he white black red ensembles of lenses corresponds to the material, principally derived from the degradation of the shelter’s walls, that eventually buried the occupation.


When the preservation of the fire lenses is pristine or near pristine , it can be safely inferred that burial occurred very rapidly . Thus, in such cases, the artefact component of the yellow deposit capping a given package of fire altered lenses can be considered of the same general archeological age, if not necessarily pertaining to the s ame occupation event . That component could represent, for instance, a scatter of finds produced during subsequent visits, ones during which the fire related activity would have taken place in parts of the site located o utside of the excavation trench . Even so, from a technological and culturalstratigraphic perspective, no issues of assemblage heterogeneity arise from the conflation of such material with that found on the exact paleo surface defined by a hearth or an extensive black lens. In plan view, the fire altered areas of the surface of a given OH are sometimes reduced to no more than a homogeneous red stain with a spatter of associated charcoal, or to a mottled mix of red, black and white . This lesser degree of preservation bespeaks of significant exp osure to erosional agents, resulting in smallscale redistribution of both sediments and artefacts due to wind, surface dynamics , and/or animal activity. Except for the rill at the interface between OH11 and OH12, no evidence was found that run off, or oth er types of water related modifications, acted on the deposit in any significant manner. It is therefore likely that the eroded fire lenses reflect longer periods of exposure rather than exposure to stronger agency. Radiocarbon dating further shows that OH s featuring some post depositional disturbance of the original micro stratigraphic structure tend to be separated from overlying, better preserved ones by measurable amounts of time. This is consistent with a reduction in the rate of sediment accumulation, representing (or including) actual hiatuses during which little or no deposition occurred. Two good examples of this pattern are OH18 ( Fig. S4. 25 ) and OH6 ( Fig. S4. 14 ). OH18 corresponds to a thin lens of eroded , fire related sediment separated from the base of overlying OH17 by <5 cm of sediment but one millennium of time. OH6 is separated from overlying, statistically younger OH5 , by an equally thin slice of deposit but itself yielded two statistically distinct dates . This dating evidence suggest s that OH6 is a palimpsest corresponding to the horizontal redistribution and vertical compression of at least two different occupation events separated by as much as five centuries. In contrast, note how the dating of th e ca.20 cm thick package of wellpreserved, stacked up fire lenses making up OH15 and OH16 yielded statistically indistinguishable results ( Table 2 ). 4. 4. THE HORIZONS OF THE TRANSITION 4. 4.1. Lithic taphonomy Systematic intraand inter level refitting was carried out for the finds made in the OH18 IL4 sequence, i n order to (a) validate the expectation that the stone tool assemblages recovered within OH19 and OH20 ought to be characterized by an exceptionally high degree of stratigraphic integrity , and (b) test the interpretation of the artefact component of IL4 as representing downward migration from OH20. The analysis focused on a bioclastic flint variety that, even though present through the sequence, is especially abundant in th e se horizons and could be traced to a primary source located 1 2.7 km to the East .


Sixteen “nodule” units were discriminated, of which M and N almost certainly correspond to a single initial volume while L and Z almost certainly do not ( Table S4.4) . Nodule M did no t refit , but six of the ten items assigned to it came from OH20 and four from IL4. The successful refits further substantiate the connection between th es e two archeo stratigraphic units : o f the 20 items that went into the five refitting units obtained, 16 came from OH20 and three from IL4 . One, however, came from OH18 . It belongs to Refit 5 in nodule N, which joins 2014 3596a, a Siret accident from OH20 , with 20121303, a flake from OH18 ( Fig. S4. 36 ). A non refitted small flake fragment is the other nodule N item , and it also came from OH20. The OH18 OH20 connection documented by Refit 5 is at odds with the integrity of the intervening OH19 horizon, documented by the pristine preservation of its fire features ( Figs. S4. 25S4. 28 ). Note, h owever, that all the items in Refit 5 are sieve finds, which are not processed and logged until after the end of the corresponding season’s field work (and whose coordinates are approximate — to the center of the quadrant of provenience and the mid point of the spit’s thickness) . The 2012 -1 303 item was retrieved in the field season of August September 2012, during which the T/5 6 test trench reached IL4 to a depth of ca.525 cm below datum . Therefore, i t cannot be excluded that its assignment to OH18 derives from labelling error, for instance, a misreading of the spit information given in the accompanying label as “a45” (OH18) instead of “a55” (IL4 ) . Alternatively, if the provenience data associated with the inventory number are correct, th e object may have been brought up by the kinds of “wall effects” that one can expect to have been in operation in the NW quadrant of grid unit T6 (cf. Fig. S4. 5 for the vertical reach of such effects along the site’s external wall). Excluding the OH18 item in Refit 5, only nine other pieces (4% of the total, but only 0.1% by mass) were found above OH20 ( Tables S4. 5S4. 6 ). These counts exclude the items indicated as from “OH19/20” because they correspond to sieve material from grid unit T6 , excavated to this depth as part of the 2012 deep sounding. Likely, this material derive s from OH20 but its precise correlation with the OH stratigraphy of 20132014 could not be carried out ; h ence, the uncertainty expressed by their provenience designation . Of the nine other bioclastic flint items analyzed whose stratigraphic position we can be confident about, seven are from those “nodules” (L and Z; see above) that are almost certainly made up of material from different initial volumes . A s to issues of pote ntial post depositional displacement , t heir OH provenience is therefore uninformative . The other two items are from nodule J, which yielded 50% of the refits, all concerning material from OH20 and IL4 . The se two items are therefore likely to represent indeed upward displacement into : OH18 , for a 0.5 g sieve bladelet from the southern half of grid unit U3 ; and OH19 , for a 0.5 g chip from the NE quadrant of T6. Given their position in the trench , the causes underpinning the displacement must lie in (a) the same kinds of wall effects possibly responsible for the Refit 5 anomaly, in the case of the piece from T6 , and (b) the small-scale burrowing identified in grid units T U/3 between the base of OH17 and the base of OH20 ( Figs. S4. 24S4. 25 , S4. 28S4. 29 ), in the case of the piece from U3 .


Among the flint artefact s retrieved in OH20 and IL4, t he bioclastic variety used in the lithic taphonomy study represents 43 % (1 92 out of 451 ), by number , and 80 % ( 858 out of 1075 g), by mass. Given the representativeness of the sample and the insignificance of the anomalies , we can therefore derive from the data concerning the basal Aurignacian levels analy z ed a robust infer ence for the whole of the sequence : that an assumption of assemblage integrity is, more than warranted, mandatory . B earing in mind the phytolith data, this is even more so when fire features are well preserved. T he fact that all other 15 refit units obtained so far — in OH17, OH16 and OH15 — link items from the same horizon, often from the same excavation spit , further strengthens the inference. 4. 4.2. Mousterian The basal Mousterian in OH21, OH22 and OH23 dates beyond 44 ka. Judging from the Aurignacian pattern, the main are a of habitation must have been located 2 -3 m outward of the grid units located against the back wall in which bedrock was reached and the Mousterian could be excavated. The limited size (2 m) and marginal position of this excavation explains the small size of the assemblage s ( Tables S4. 7S4. 8) and preclude s a firm conclu sion on the nature of th e occupations’ stone tool economics. A few preliminary observations can nevertheless be made. Cortex was present in 3 0 % of the flints (2 5 out of 8 3, debris and unretouched flake fragments excluded ), suggesting the in troduction of unmodified nodules . T hat some knapping did occur at the site is otherwise attested by two coretrimming elements and a Kombewa core . The latter is a cortical blank whose ventral surface was centripetally exploited to produc e small flakes ; the largest scar is no more than 18.5 mm long . T hree sidescraper resharpening elements document a local consumption history for these kinds of items. The coretrimming elements concern the correction of hinges and/or platform angles in the course of core red uction using the Discoid method. Of the 38 items of dbitage and formal retouched tools for which a technological reading was possible, most (ten) also reflected use of that method . The Kombewa and Levallois methods are represented by five specimens each. Among the latter there is a double denticulate on quartzite discarded after breakage that , given this raw material’s poor representation in the assemblage, is mostly certainly an imported item ( Fig. S4. 37 , no. 1) . The range of formal retouched tools is lim ited to notches, denticulates (Fig. S4. 37 , nos. 1, 6) and sidescrapers ( Fig. S4. 37 , nos. 2 5). A cortical, orange segment flake might be added to this ensemble as a possible naturally backed knife but it bears no evidence, macro or microscopic, of having been used . D itto for three unmodified Levallois products — two small points and a typical preferential flake. The use wear analysis, however, showed that some unretouched or atypically retouched pieces had in fact been functional tools ; most were used on wood, but defleshing is also documented ( Table S4. 9 ; Fig. S4. 38).


4.4. 3 . Aurignacian Like those of Solutrean and Gravettian age , ADB’s Aurignacian occupation horizons yielded many shell beads, but the range of taxa is broader and includ es species so far un known in the later occupations of the site (Table S4. 2 ). Theodoxus fluviatilis ( Fig. S4. 34 ) is the most abundant taxon. A small Gibbula, possibly G. rarilineata, is re presen ted in OH20/IL4 by unperforated specimens that are heavily ochre stained, as are the Theodoxus (whether perforated or not). All the Striarca lactea are also from OH20 and some are ochre stained, so they may have been co llected for ornamental purposes. D itto for the ochrestained Mimachlamys fragments found in OH17 ; despite the lack of a perforation in the shell , or part thereof, that we have, these valves may have been used as pendants . The function of the few fragments of Pecten and other bivalves remains elusive; as wit h the numerous such remains found in the Gravettian and Solutrean, it is inferred that they were introduced to the site as complete shells, possibly used as containers. The s tone tool assemblages ( Tables S4.10S4. 21 ; Figs. S4 . 39S4. 43 ) ar e d iagnostic, e ven when t heir s ize i s small, as in t he cas e o f OH19 ’s , w hich consist s o f only 146 pieces, 7 6% of which are debris (60%) or flake fragments (16%). Among the flints (debris excluded), blades and bladelets, both retouched and unretouched, account for as much as 3% of the assemblage in O H20, 1 5 22% i n O H19 OH16, a nd 4 8% in O H15; they a re virtually n on existent in t he u nderlyin g Mousterian, w hich y ielded a single bladelet in O H22 , possibly p ost depositionally displaced fr om I L4 v ia “wall e ffect” p rocesses ( Table S 4.7 ). C onversely, a mong cores, n o L evallois, discoid or Ko mbewa ty pes — represented i n th e Mousterian deposit by discarded volumes , p reparation b y products , and diagnostic debitag e — were f ound i n t he A urignacian horizons. In the latter , all cores ar e e ither o f the re gular, p rismatic kind o r correspond t o s peciali z ed t ypes ( splintered p ieces, “ burins,” a nd c arinated or n osed “scrapers”) that wer e set up , o r expediently u sed for t he e xtraction of bladelet s o nly. Th e technological information encoded in the refits, cores and d bitage from OH20/IL4 allow s us to reconstruct two reduction strategies ( Figs. 11, S4. 39 S4. 40). One targeted the production of blade size blanks from elongated nodules set up as single platform cores; the platform is prepared by abrasion, the opposite end of the prism is configured to maintain the convexity of the dbitage surface, and the desired end product is a 5 10 cm long, 1 3 cm wide blade. The second strategy is designed for the extraction of bladelets from carinated/nosed “scrapers ,” and is carried out in two steps . The first step is the extraction of the blank s — thick, elongated , “naturally carinated” b lade s. T he second step is the production of bladelets from the “scraper fronts” configured on the blanks obtained in the first step. The evidence at hand indicates that the two strategies were used separately, not sequentially . In the examples that we have, t he production of “scraper” b lanks used dedicated nodules/volumes that were routinely discarded once found unfit for purpose. It cannot be excluded, however, that t he production of such blanks could also have been carried out on by products of the initial stages of configuration and preparation of regular blade cores.


From a functional standpoint, the basal occupation i n OH20/I L4 is significantly distinct from those that follow. This is well apparent in the total mass of the flint assemblage retrieved: 107 g, of which 510 g (47%) correspond to cores. The total flint mass in OH16 (1310 g) is higher, and that i n OH17 (830 g) approaches the OH20/IL4 values. H owever, the cores r etrieved in OH16 and OH17 represent no more than 13% and 19%, respectively, of the total flint mass therein. Core size and core type also differ markedly: prismatic cores for blades are 4 out 7 in OH20/IL4, but 2 out 24 i n OH16 and non-e xistent in the other Aurignacian horizons; and the average mass of cores is 73 g in OH20/IL4 but ranges between 3.6 and 7.6 g only in the other horizons (excluding OH15, i n which the average of 14.8 g is not significant because it corresponds t o only t wo specimens ). These d ifferences are the more significant b ecause a successi on o f dense occupation l enses featuring m ultiple, stacked up fire f eatures are subsumed i n O H17 a nd O H16. In the OH19 OH15 sequence, therefore, available flin t volumes would seem to have been much more intensive ly reduced . This inference is corroborated by the values concerning the proportion of each flint assemblage represented by chippage and chunks: by mass, 6% in OH20 /IL4 , contra between 17% (in OH19) and 35% ( in OH17 ) for the overlying horizons . Conversely, these horizons yielded significant volumes of locally available raw materials (cf. their large flakes and cores made on limestone ) , while only three limestone items, in total 1.1 gworth, were retrieved in OH20/IL4. Thes e stone economic s data suggest that the OH20 /IL4 occupation(s) were of short duration . Likely, we are dealing with transient stays during which flint volumes collected en route were processed for the extraction of quality blanks taken awa y by the individual(s) doing the knapping, and whereby volumes found to be unsuitable or whose set up failed were discarded with no further labor investment. This interpretation accords well with the lack of fire features in OH20, which is in stark contras t with the number and preservation of those found in overlying horizons . These comparisons suggest that OH19 OH15 were characteriz ed by lengthier stays during which flint volumes brought in as ready made blanks or finished tools were systematically recycled prior to discard. The decrease in size seen in similar core types as one moves up in the sequence further supports this interpretation. For instance, while the two carinated “scrapers” in OH20 weighed between 13 and 48 g at discard, the correspon ding values for the carinated and nosed “scrapers” in OH17, OH18 and OH19 range between 2 and 7 g (e.g. , item 2014 3424, Fig. S4. 42 , no. 2 , the mass of which is 3.2 g). The fact that the core assemblages in OH17 and OH16 are dominated by splintered pieces/ bipolar cores with masses comprised between 0.4 g (item 201424811, Fig. S4. 43 , no. 1) and 7 g is also consistent with a flint economy whereby extracting as much cutting edge as possible from any given volume of raw material was routinely attempted prior to discard . Indeed, out of 108 bladelet like blanks in these two horizons, no less than 28 bore diagnostic technological attributes of extraction from splintered pieces/bipolar cores (Aubry et al. , 1997) and one such blank, badly burnt, had been retouched into a Dufour bladelet ( Fig. S4. 43 , no. 8).


Reductionto exhaustion of imported blanks and tools must underpin two other aspects of La Boja’s Aurignacian assemblages. The first aspect is the dearth of endscrapers. Across all horizons, and excluding the type list items that, in fact, are cores, this category is represented by only 6 out of 69 retouched tools : two small, thumbnaillike ; one simple atypical; and three fragments ( Tables S4. 10S4. 21). The second aspect i s the low success rate in the identification of use wear. T races were identified in 14 % ( 10 out of 72 ) of the retouch ed and unretouched blanks analyz ed, in contrast with the 28% (11 out of 40) success rate encountered in the Mousterian (Table S4.9 ). Both a spects probably relate to intensified reduction in that (a) endscrapers that broke during use or could no longer be re sharpened may have supplied a significant proportion of the blanks reduced as splintered pieces/bipolar cores , and (b) the recycling of tools may have entailed the loss of the active parts bearing evidence of the uses they had been put to. In this context, it is probably to be expected that three out of the six pieces on which diagnostic use wear could be identified are projectile elements (Fig. S4. 41 , nos. 1 3) . The other tasks that left identifiable traces are the processing of wood, hide , and hard animal tissue ( Fig. S4. 41 , nos. 4 5). E conomic and functional considerations — degree of spatial segmentation of the reduct ion sequences , and intensi ty of raw material recycling — therefore suffice to explain the differences in composition, structure and metric attributes observed when comparing the assemblages retrieved in OH20/IL4 with those from the overlying Aurignacian horizons. From a typological standpoint, however, OH16 seems to represent a significant divide . This horizon marks the first appearance in the sequence of backed bladelets featuring a particular mode of abrupt retouch ( Fig. S4. 43 , nos. 4 5) — short and lim ited to the very edge, not, as in the typical backed bladelets found in the overlying Gravettian, Upper Solutrean and Magdalenian , invasive and eliminating a significant portion of the blank. Documenting typological continuity with the preceding occupation s, these “shortbacked” bladelets from OH16 and OH15 are associated with the same kinds of retouched bladelets that make up the totality of the microliths in OH17, OH18, OH19 and OH20 /IL4: Dufour , and marginally backed ( Fig. S4. 4 3, nos. 69) . Based on this typological evidence, La Boja’s Aurignacian sequence can therefore be subdivided into : (a) an Evolved Aurignacian , dated to between 35 and 37 ka and represented in OH20 /IL4 , OH19, OH18 and OH17 ; and (b) a Late Aurignacian, dated to around 35 ka and represented in OH1 6 and OH1 5 . The number of bladelet blanks bearing diagnostic features of extract ion from a “burin” type of core that we see in OH15 may stand for another element of technological innovation. We cannot exclude that assemblage size and sam pling bias are involved in the pattern to some extent. However, such “burin spall” blanks , non existent in OH20/IL4 and in OH19, account for as much as 15% of all bladelets in OH15. Conversely , OH15 yielded no bladelets bearing diagnostic features of extra ct ion from a carinated/nosed “scraper, ” whereas such blanks amount to as much as 24% of all bladelets in OH20/IL4 and to as much as 38% in OH19.


These Late Aurignacian assemblages may well represent, therefore, the beginnings of a trend that will eventually give rise to the Gravettian, one whereby (a) “burins” replace carinated/nosed “scrapers” as the primary type of specialized bladelet core , and (b) backed bladelets replace Dufours as the preferred type of lithic component arming composite projectile poin ts made of wood or antler. The last steps of such an Aurignacianto Gravettian transition, however, would seem to be unrepresented at the site. This is probably due to the existence of a hiatus of deposition or occupation , if not erosion, at the interface between OH15 and OH14 — as indicated by the three millennia that , despite a stratigraphic distance of no more than ca.15 cm, minimally separate OH15 from OH13 ( Fig. S4. 7). 4.4. 4 . Early Gravettian The Early Gravettian in OH13/IL3/OH 14 ( Tables S4. 22S4. 25) date s to 30.931. 5 ka and is very poor. The scarcity of finds clearly relates to the site (or at least this part of the site) having become unsuitable for habitation. Indeed, at that time, m uch of the area sampled by our trench was occupied by the major roof collapsed slab that fell on top of the OH1 5 ground surface , with attendant implications for human use of the place . Yet, it remains significant that even such a small assemblage as that retrieved in OH13/IL3 (35 items in total, 14 only excluding debris) yielded sufficient evidence — a burin, and three bladelets , two of which refit — to allow cultural-stratigraphic assignment in general technological terms (to the Upper Paleo lithic). Specific technocomplex attribution (to the Gravettian) is made possible by the substantial basal fragment of a microgravette point present in the similarly small assemblage from OH14 (82 items in total, 22 only if debris are excluded ). 4. 5 . CONCLUSIONS The excavation of the La Boja rock shelter prov ided replication, with much enhanced precision and resolution, of the main conclusions derived from the study of the adjacent site of Finca Doa Martina. The following is a summary of the key points . Continued occupation of the region through the Upper Pleistocene, irrespective of climatic oscillations, is implied by a radiocarbon dating record whose gaps would seem to correspond , with present evidence, to (a) sedimentary hiatuses (as indicated by direct stratig raphic contact between occupation horizons separated by up to three millennia ), or (b) lateral variation (of occupation emplacements relative to the excavation trench) . The classical tripartite subdivision of the Solutrean, derived from the pattern of inde x fossil succession observed in such thick , stratified sequences as La ugerie Haute, in France, or Parpall, in Spain, questioned by some since Straus’s (1983) proposition of a functional interpretation of that variability, is fully supported by the La Boja evidence . In particular, the real entity of a Lower Solutrean phase defined by the presence of unifacial foliate points and the absence of laurelleaves is confirmed. There can be no question that the assemblages retrieved in the OH15 OH20 horizons belong in the Aurignacian technocomplex, with associated dates placing its first presence at the site in the middle of the 37th millennium BP, at the latest.

PAGE 100

The Aurignacian in OH20 is in total technological and typological discontinuity with the underlying Mousterian in OH21 OH23. As a minimum of six millennia separate OH20 from OH21, this evidence does not suffice to exclude the possibility that truly “transitional” systems ( in the sense of Kuhn, 2003) may have existed in the region at some point during that mi ssing interval. However , a similar discontinuity is observed at Finca Doa Martina, and elsewhere in Western Europe the Middle and the Upper Paleo lithic are discrete technical packages with no overlap . This pattern makes it possible to securely diagnose even very small assemblages on the basis of tool typology or the presence of technological traits that have true index fossil value . For instance , th e evidence shows that the presence of sidescrapers , of discoid/Kombewa/Levallois cores and dbitage produc ts or by products, or of denticulates made on such blanks , suffic es to attribute an assemblage to the Middle Paleo lithic . Conversely , the presence of blades, bladelets or prismatic/”burin”/carinated “scraper” types of cores suffices to attribute an assemblage to the Upper Paleolithic. Indeed, based on such criteria , under blind testing conditions (i.e., devoid of any contextual information on dating or position in a stratigraphic sequence ), an expert lithic analyst would have correctly assigned to either the Middle or the Upper Paleo lithic even the smallest of the assemblages retrieved in La Boja’s occupation horizons . The use wear evidence shows that the same types of tasks were being carried out at the two Rambla P erea sites across the period of the transition. Therefore, putative differences in site function cannot explain the sharp technological differences underpinning the assignment of assemblages to either the Middle or the Upper Paleolithic . Instead, it is in the changes that occurred in the technical system that may reside the explanation for why a given activity, e.g., wood working, was executed with sidescrapers and denticulates in the Mousterian but with only perforators, notches and atypically edge retouch ed blanks in the Aurignacian. Only such techn olog ical changes can explain why the set s of tools one could choose from when assessing suitability for a given task were so conspicuously different . 4.6 REFERENCES AUBRY, Th.; ZILHO, J.; ALMEIDA, F.; FONTUGNE, M. (1997) — Production d’armatures microlithiques pendant le Palolithique suprieur et le Msolithique du Portugal , in BALBN, R.; BUENO, P. (eds.) — II Congreso de Arqueologa Peninsular. Paleoltico y Epip aleoltico, Zamora, Fundacin Rei Afonso Henriques, p. 259 272. KUHN, S. L. (2003) — In what sense is the Levantine Initial Upper Paleolithic a “transitional” industry? in ZILHO, J.; D’ERRICO F. (eds.) — The Chronology of the Aurignacian and of the Transitional Technocomplexes. Dating, Stratigraphies, Cultural Implications, Lisboa, Trabalhos de Arqueologia 33, Instituto Portugus de Arqueologia, p. 61 69.

PAGE 101

LUCENA, A.; MARTNEZ, S.; ANGELUCCI, D. E.; BADAL, E.; VILLAVERDE, V.; ZAPATA. J.; ZILHO, J. (201 2) — La ocupacin solutrense del abrigo de La Boja (Mula, Murcia, Espaa). Espacio, Tiempo y Forma. Serie I, Nueva poca. Prehistoria y Arqueologa, 5, p. 447 454. SONNEVILLE BORDES, D. de; PERROT, J. (1954 1956) — Lexique typologique du Palolithique suprieur . Bulletin de la Socit Prhistorique Franaise, 51, p. 327 335; 52, p. 7679; 53, p. 408412, 547 559. STRAUS, L. G. (1983) — El Solutrense vasco cantabrico: una nueva perspectiva, Madrid, Centro de Investigacin y Museo de Altamira. WoRMS Editorial Board (2016) — World Register of Marine Species . Available from at VLIZ. Accessed 2016 0725. doi:10.14284/170 ZILHO, J. (1997) — O Paleoltico Superior da Estremadura portuguesa, 2 vols., Lisboa, Colibri. ZILHO, J.; ANGELUCCI, D.; BADAL, E.; LUCENA, A.; MARTN, I.; MARTNEZ, S.; VILLAVERDE, V.; ZAPATA, J. (2010) — Dos abrigos del Paleoltico superior en Rambla Perea (Mula, Murcia) , in MANGADO, X. (ed.) — El Paleoltico superior peninsular. Novedades del siglo XXI, Barcelona, Universidad de Barcelona, p. 97 108.

PAGE 102

Table S4. 1 . ADB . Provenience of the r adiocarbon dated samples . A ll samples are wood charcoal fragments ( trench collected , unless otherwise stated). Coordinates are in cm (elevations are below datum) . For the intrusive charcoal collected in burrow features, the chronological horizon given reflects the age obtained for the sample , not the stratigraphic position of the surrounding, intact deposit Horizon Sample G rid unit x y z Taxon Sample o bservations Lab # Holocene intrusion in OH1 burrow 2008 775 T5 SE 75 25 205 Olea europaea 0.09 g; highly fused and crystalized; xyz approximate OxA 20116 MAGDALENIAN OH1 2010 27 U5 SE 75 25 208 Juniperus sp. single, large charcoal fragment ; xyz approximate VERA 5363 OH1 /OH2 2008 774 T5 SE 75 25 211 Pinus nigra 0.15 g subsample dated ; xyz approximate VERA 5212a OH3 2013 868 S4 NE 73 55 227 Pinus nigra/sylvestris 1.47 g; single, very large charcoal fragment VERA 5937 OH4 2014 846 T4 NW hearth 16 69 254 Juniperus sp. single, large charcoal fragment VERA 6080 SOLUTREO GRAVETTIAN OH5 2012 385 U5 SW hearth 13 10 273 Juniperus sp. single, large charcoal fragment VERA 5788 OH6 2010 183 T5 SE 75 10 274 Juniperus sp. vial 1; 0.07g ; xyz approximate VERA 5364a Juniperus sp. vial 2 ; xyz approximate VERA 5364b SOLUTREAN OH7 2010 225 T5 SW 3 46 295 Juniperus sp. large fragments, can be combined VERA 5365 OH 9 2014 1270 S4 SE hearth 85 36 306 Juniperus sp. 13 growth rings VERA 6081 2012 1522 S5 W hearth 2 7 300 Juniperus sp. 31 straight, wide growth rings VERA 5850 OH10 2010 316 T5 SE hearth 84 17 316 Juniperus sp. large fragments, can be combined VERA 5366 OH11 2008 760 S5 SE 75 25 323 Juniperus sp. another subsample failed ABOx at Ox A; xyz approximate VERA 5213 2014 2578 U4 NE hearth 53 94 328 Juniperus sp. single, large charcoal fragment VERA 6152 Solutrean intrusion in OH12 burrow 2012 178 T5 SW 25 25 331 Juniperus sp. 20 straight, wide growth rings; xyz approximate VERA 5851 GRAVETTIAN OH12 2012 175 T5 SE hearth 56 18 335 Juniperus sp. 18 straight, wide growth rings VERA 5852 OH13 2012 622 T5 NE hearth 99 54 384 Juniperus sp. single, large charcoal fragment VERA 5789 (a) AURIGNACIAN OH15 2014 2903 U5 SE hearth 92 27 420 Juniperus sp. single, large charcoal fragment VERA 6153 OH16 2014 3046 U4 NW hearth 13 90 440 Juniperus sp. single, large charcoal fragment VERA 6154 OH17 2012 1518 T5 S 36 40 436 Juniperus sp. wood knot; small, clean, well preserved VERA 5853 2014 3129 U6 SW hearth 10 10 426 Juniperus sp. single, large charcoal fragment VERA 6155 2014 3184 U5 SE hearth 71 3 450 Juniperus sp. single, large charcoal fragment VERA 6156 OH18 2012 1352 T6 SW 45 25 453 Juniperus sp. large branch or trunk; 20 straight, wide growth rings VERA 5854 OH19 2014 3348 U4 SW hearth 20 45 470 Juniperus sp. single, large charcoal fragment VERA 6157 2014 3421 T3 NW hearth 45 55 484 Juniperus sp. single, large charcoal fragment VERA 6158 OH20 2012 1382 T5 NE 80 88 470 Juniperus sp. 15 straight, wide growth rings VERA 5855 Hiatus IL4 2012 1481 T6 SW 35 15 503 Juniperus sp. 14 straight, wide growth rings VERA 5856 MOUSTERIAN OH22 2013 384 T6 E 100 18 593 Pinus nigra/sylvestris large, clean VERA 5899 2013 330 T6 SE 70 41 602 Pinus nigra/sylvestris 6 growth rings VERA 5900 OH23 2013 258 T7 SE 85 3 606 Juniperus sp. 19 growth rings VERA 5901 2013 361 T6 SE 75 25 618 Pinus nigra/sylvestris from sieve; xyz approximate VERA 5902 (a) The p reviously published r esult for this sample (Lucena et al., 2012: Table 1) is affected by a typo that went uncorrected at the time of proof revision. The sample’s correct age is 27260230 BP, as given in Table 2, not 27620230 BP, as printed in Lucena et al. (2012)

PAGE 103

Table S4. 2 . ADB. Stratigraphic distribution of marine and fluviatile shell finds (a) . Taxonomy follows WoRMS Editorial Board (2016). MT = Mousterian; EA = Evolved Aurignacian (Aurignacian II) ; LA = Late Aurignacian (Aurignacian IIIIV); EG = Early Gravettian; MG = Middle Gravettian; PS = ProtoSolutrean (?); LS = Lower Solutrean; MS = Middle Solutrean; US = Upper Solutrean; S G = Solutreo gravettian; EM = Early Magdalenian; UM = Upper Magdalenian; EpM = Epimagdalenian PERFORATED MT EA LA EG MG P S LS MS US S G EM UM EpM TOTAL Gastropoda Acteon tornatilis – – – – – – – – 4 1 – – – 5 Conus ventricosus – 1 – – – – – – – – – – – 1 Littorina obtusata – – – – 1 – 3 2 2 – – 1 – 9 Smaragdia viridis – – – – 1 – 2 – – – – – – 3 Theodoxus fluviatilis – 3 – – – – 6 2 2 2 – – – 15 Tritia incrassata – 1 – – – – – – – – – – – 1 Tritia mutabilis – – – – – – – 1 – – – – – 1 Tritia neritea – – – – – – 16 – 6 1 – – – 23 Trivia sp. – 1 – – – – – – – – – – – 1 Bivalvia Glycymeris insubrica (with umbo) – – – – 4 – – – – – – – – 4 Striarca lactea – 1 – – – – – – – – – – – 1 Scaphopoda Dentalium (Antalis) vulgare – – – – 2 – 5 3 4 – – – – 1 4 UNPERFORATED MT EA LA EG MG P S LS MS US S G EM UM EpM TOTAL Gastropoda Acteon tornatilis – – – – – – – – 2 – – – – 2 Gibbula sp. – 3 – – – – – – – – – – – 3 Littorina obtusata – – 1 – – – 1 2 – – – – – 4 Melanopsis sp. – – – – – – – – – – – 14 – 14 Nucella lapillus – – – – – – – 3 – – – – – 3 Theodoxus fluviatilis – 8 – – 1 1 11 1 8 3 1 – – 34 Tritia neritea – – – – – – 2 – 4 – – – – 6 Trivia sp. – – 1 – – – – – 1 – – 1 – 3 Turritela sp. – – – – – – – – 1 – – – – 1 u nidentified fragment – 1 – – 1 – 1 1 – 1 – – – 5 Bivalvia Acanthocardia sp. fragment – – – – 4 – 7 1 10 – 1 – – 23 Cerastoderma sp. fragment – – – – 6 – 1 1 4 – – – – 12 Glycymeris insubrica (fragment without umbo) – 4 – – 2 – – – 1 – – – – 7 Mimachlamys sp. fragment – 6 – – – – – – – – – – – 6 Mytilus sp. fragment – 2 – – 4 – – – – – – – – 6 Pecten sp. fragment – 3 – 1 4 1 6 9 4 – 2 – – 30 Striarca lactea – 4 – – – – – – – – – – – 4 u nidentified fragment – 4 2 1 9 1 24 9 12 10 2 6 1 81 ALL MT EA LA EG MG P S LS MS US S G EM UM EpM TOTAL TOTAL – 42 4 2 3 9 3 8 5 35 65 18 6 22 1 322 ( a) The rock shelter ’s fill contains fossil shell derived from the calcarenite bedrock. Pecten sp. is the most common. W hen sieve collected, such finds were either subsequently discarded or, if the diagnosis was uncertain, kept but not counted. For very small fragments, however, it cannot be excluded that the “bivalve fragment” and “ Pecten sp. fragment” categories are somewhat inflated by the presence of such inherited material

PAGE 104

Table S4. 3 . ADB. Standard typological classification of stone tools . F ollowing the type list of Sonneville Bordes and Perrot (1954 56) , with the modifications introduced by Zilho (1997) . MT = Mousterian; EA = Aurignacian II (Evolved Aurignacian); L A = Aurignacian III/IV (Late Aurignacian); EG = Early Gravettian; MG = Middle Gravettian; PS = Proto -S olutrean (?); LS = Lower Solutrean; MS = Middle Solutrean; US = Upper Solutrean; S G = Solutreo gravettian; EM = Early Magdalenian # Type MT EA LA EG MG P S LS MS US S G EM ENDSCRAPERS 1a simple endscraper on blade – – – – 6 – 1 5 5 6 2 – 1b simple endscraper on flake – – – – – – 1 1 1 – – 2a atypical endscraper on blade – – – – 2 1 1 1 – – – 2b atypical endscraper on flake – – 1 – – – – – 2 – – 3 double endscraper – – – – 5 – 3 1 1 – – 4 ogival endscraper – – – – – – 1 – – 2 1 5a endscraper on retouched blade – – – – 4 – – 3 2 2 – 5b endscraper on retouched flake – – – – – – 18 1 5 1 – 7 fan endscraper – – – – 1 – 4 – – – – 8 endscraper on flake – – – – 1 – 1 – 1 – – 10 thumbnail endscraper – 2 – – – – – – 1 – – 11 carinated endscraper – 3 – – – – – – – – – 12 atypical carinated scraper – 2 – – – – – – – – – 13 nosed endscraper – 2 – – – – – – – – – 14a flat nosed endscraper – – – – 1 – – – – – – 14b flat shouldered endscraper – – – – – – 1 – – – – COMPOSITE TOOLS 17 endscraper burin – – – – 1 – 2 – 1 1 – 18 endscraper truncation – – – – – – 1 – – – – PERFORATORS 23 perforator – – 2 – – – – – – – – 24 atypical perforator – – – – – – 1 – 1 – – BURINS 27 straight dihedral burin – – – – – – – – – – 1 28 dihedral djet burin – – – 1 3 – 1 – 1 – – 29 dihedral burin on angle – – – – – – – – 1 – – 30a angle burin on natural surface or break – 2 – – 1 – 2 2 – – – 34 burin on straight truncation – – – – 1 – – – – – – 35 burin on oblique truncation – 1 – – 2 – 2 – – – – 36 burin on concave truncation – – 1 – 5 – 1 – – – – 38 transversal burin on truncation – – – – – – – – – 1 – 40 multiple burin on truncation – – – – 3 – 1 – – – – 41 multiple mixed burin – – – – 2 – 1 2 – – – 42a Noailles burin – – – – 1 – – – 1 – – 42b Vale Comprido burin – – – – – – – – 1 – – BACKED TOOLS 51d basal microgravette – – – 1 – – – – – 1 – TRUNCATIONS 61 oblique truncation on blade – – – – – – 1 – – – – 62 concave truncation – – – – – – 1 – – – – RETOUCHED BLADES 65 continuous retouch blade, unilateral – – – – 1 – 1 – 1 – – 66 continuous retouch blade, bilateral – 1 – – 1 1 – 1 1 – – SOLUTREAN TOOLS 69a unifacial point – – – – – – 24 2 – 1 – 70n laurel leaf fragment – – – – – – – 1 – – – 71 willow leaf fragment – – – – – – – – 1 – – 72a shouldered point – – – – – – – – 5 – –

PAGE 105

Table S4. 3 . ADB. Standard typological classification of stone tools (cont.) # Type MT EA LA EG MG P S LS MS US S G EM SUBSTRATE 74 notched piece 1 7 7 – 2 – 1 – 3 1 – 75a denticulate 4 – – – – – – – – – – 75b double denticulate 1 – – – – – – – – – – 76 splintered piece – 19 11 – 4 – 1 7 5 4 2 1 77a sidescraper 2 – – – – – – – – – – 77b transversal sidescraper 1 – – – – – – – – – – 77c convergent sidescraper 1 – – – – – – – – – – 77d denticulated sidescraper 1 – – – – – – – – – – 77f sidescraper fragment 1 – – – – – – – – – – 78b Vascas scraper – – – – – – – – 2 – – BLADELET TOOLS 81 trapeze – – – – – – – – 2 – – 83 segment – – – – – – – – 1 – – 84 truncated bladelet – – – – 1 – – – – – – 85a backed bladelet – – 2 – – – – – 1 – 3 85c partially backed bladelet – – – – – – 1 – – – 1 85d double back bladelet – – – – 1 – – – – – – 85f backed bladelet fragment – – 5 – 2 – 2 – 3 7 2 86a backed truncated bladelet – – – – 2 – – – 1 – – 86b backed, double truncated bladelet – – 1 – – – – – – – – 87a denticulated backed bladelet – – – – – – – – – 1 2 89 notched bladelet – 3 1 – – – 1 1 5 – – 90a Dufour bladelet – 3 5 – 1 1 1 – – – – 90b Areeiro bladelet – 1 2 – – – – – – – – 90c marginally backed bladelet – – 5 – 1 – 1 1 5 5 3 VARIA 92a atypically retouched piece 5 4 5 – 3 1 3 1 9 2 – 92b retouched piece fragment 3 8 4 1 6 1 26 12 2 5 6 – 92d pointed bladelet – – – – – – 1 – 1 2 1 92e hammerstone – 1 – – – – – – – – – TOTAL 20 5 9 5 2 3 64 6 138 40 95 3 7 15 Table S4. 4 . ADB bioclastic flint analysis. Nodule types defined in horizons OH1 8to IL4 TYPE DESCRIPTION OBSERVATIONS A yellowish brown, opaque, very thin cortex – B light brown, semi opaque, very thin iron impregnated cortex, possibly rolled – C light brown, semi opaque, very thin iron impregnated cortex, possibly rolled; subcortical white weathering – D brown, dark spotted, semi opaque, patinated – E light brown, semi opaque, patinated – F fine grained, brown, opaque, with rolled, whitish cortex – G greyish brown, opaque, thin, rough textured, iron impregnated cortex refits H brown, semi opaque, heavily iron spotted, with very thin, coarse textured beige cortex – I greyish brown, semi opa que, thin, rough textured cortex without iron impregnation refits J reddish brown, semi opaque, heavily iron spotted with thick limestone like cortex refits K reddish brown, semi translucent (cortex absent) – L beige, semi transluc ent , with limestone like cortex refits; more than one volume M yellowish brown, semi translu cent (cortex absent) likely a single volume N light reddish brown, semi opaque, iron spotted with thin limestone like, weathered cortex refits; likely a single volume O yellowish brown, opaque, white thin iron impregnated cortex with subcortical whitish weathering – Z undetermined to nodule bioclastic flint of the same type and source more than one volume

PAGE 106

Table S4. 5 . ADB bioclastic flint analysis. Vertical distribution of each n odule type in units OH18 to IL4 (number of items per archeo stratigraphic unit) Nodule OH18 OH18/19 OH19 OH19/20 OH20 OH20/IL4 IL4 TOTAL A – – – – 1 – 1 2 B – – – – 1 – – 1 C – – – 1 3 – 1 5 D – – – – 1 – – 1 E – – – – 1 – – 1 F – – – – 1 – – 1 G – – – – 2 1 1 4 H – – – 1 28 1 10 40 I – – – – 3 – – 3 J 1 – 1 3 47 2 17 71 K – – – – – – 1 1 L – 1 1 1 18 – 5 26 M – – – – 6 – 4 10 N 1 – – – 2 – – 3 O – – – 2 3 – 2 7 Z 2 2 1 – 17 – 12 34 TOTAL 4 3 3 8 134 4 54 210 Frequency 1 . 9% 1 . 4% 1 . 4% 3 . 8% 63 . 8% 1 . 9% 25 . 7% 100 . 0% Table S4. 6 . ADB bioclastic flint analysis. Vertical distribution of each nodule type in units OH18 to IL4 (total mass of items, in grams, per archeo stratigraphic unit) Nodule OH18 OH18/19 OH19 OH19/20 OH20 OH20/IL4 IL4 TOTAL A – – – – 136.6 – 1.9 138.5 B – – – – 14.2 – – 14.2 C – – – 0.1 24.2 – 0.1 24.4 D – – – – 0.9 – – 0.9 E – – – – 4.5 – – 4.5 F – – – – 7.5 – – 7.5 G – – – – 28.5 6.7 3.1 38.2 H – – – 0.2 49.1 0.2 15.2 64.7 I – – – – 145.3 – – 145.3 J 0.5 – 0.5 9.0 235.4 0.6 77.8 323.8 K – – – – – – 0.5 0.5 L – 4.4 1.5 0.5 37.1 – 7.7 51.2 M – – – – 2.7 – 2.4 5.1 N 14.2 – – – 16.5 – – 30.7 O – – – 1.2 15.2 – 8.6 25.0 Z 1.0 0.7 0.2 – 7.2 – 8.8 17.8 TOTAL 15.7 5.0 2.2 11.1 724.9 7.4 126.0 892.4 Frequency 1.8% 0.6% 0.2% 1.2% 81.2% 0.8% 14.1% 100.0%

PAGE 107

Table S4. 7 . ADB Mousterian. S tone tool technological categories . N = number , M = mass in grams . OH21, OH22 and OH23 combined RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 2 37.8 27 144,71 32 62.8 35 32.4 – – 1 0.3 153 19,76 1 10.3 19 113,89 270 422.0 Quartzite – – 1 21,27 – – 1 0.8 – – – – 3 0,73 – – 1 13,62 6 36.5 Limestone – – – – 2 6.7 1 0.4 – – – – 4 1,19 – – – – 7 8.3 Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 2 37.8 28 166.0 34 69.53 37 33.6 – – 1 0.3 160 21.7 1 10.3 20 127.5 283 466.7 Table S4. 8 . ADB Mousterian. Classification of cores and retouched tools . OH21, OH22 and OH23 combined Cores N R etouched tools N Kombewa 1 notched piece 1 discoid ? (fragment) 1 denticulate 5 TOTAL 2 sidescraper unilateral 2 transversal 1 convergent 1 denticulated 1 fragment 1 atypically retouched flake 5 retouched piece fragment 3 TOTAL 24 Table S4. 9 . ADB Mousterian and Aurignacian stone tools (flint) . Usewear evidence Illegible None Wood Hide Meat Bone Projectile Ochred Total MOUSTERIAN sidescrapers 2 3 1 – – – – – 6 denticulates 3 – 1 – – – – – 4 notches – – 1 – – – – – 1 edge retouched pieces and fragments 1 1 1 1 1 (a) – – – 5 unretouched flake – 19 5 – – – – – 24 TOTAL 6 23 9 1 1 – – – 40 AURIGNACIAN cores (b) 1 5 – – – – – – 6 endscrapers 1 5 – – – – – – 6 perforators 1 – 1 – – – – – 2 notches 2 10 1 1 – – – – 1 4 bladelet tools 4 22 – – – – 3 – 29 retouched pieces and/ or fragments 6 8 2 1 – 1 – – 1 8 unretouched blanks 1 2 – – – – – – 3 TOTAL 1 6 5 2 4 2 – 1 3 – 7 8 (a) flesh removal or skinning , but uncertain ; (b) includes carinated/nosed scrapers,” “burins , ” and splintered piece s Table S4. 10 . ADB Aurignacian (OH20) . Stone tool technological categories . The items retrieved in IL4 are included. N = number , M = mass in grams RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 7 510.5 25 190.4 56 161.0 27 23.5 18 91.2 37 11.4 26 2 43. 5 10 19.0 8 24.1 45 0 1074. 6 Quartzite – – – – – – – – – – – – – – – – – – – – Limestone – – – – 1 1.0 – – – – – – 2 0.1 – – – – 3 1.1 Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 7 510.5 25 190.4 57 162.0 27 23.5 18 91.2 37 11.4 26 4 4 3 . 6 10 19.0 8 24.1 45 3 107 5 . 8

PAGE 108

Table S4. 11 . ADB Aurignacian ( OH20 ). Classification of cores, retouched tools and bladelets . Bladelet counts include both retouched and unretouched blanks. The items retrieved in IL4 are included Cores N Bladelets extracted from N R etouched tools N prismatic for blades 4 carinated/nosed “scraper” 10 thick, atypical thumbnail endscraper 1 carinated “scraper” 2 other 31 notched bladelet 1 Kostenki truncation 1 TOTAL 41 Dufour bladelet 1 TOTAL 7 Areeiro bladelet 1 atypically retouched piece 2 retouched piece fragment 2 TOTAL 8 Table S4. 12 . ADB Aurignacian (OH19). Stone tool technological categories . N = number , M = mass in grams RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 6 36.0 4 8.0 23 23.5 12 11.0 – – 6 1.9 83 14.2 4 5.2 7 17.0 145 116.8 Quartzite – – – – 1 0.9 – – – – – – – – – – – – 1 0.9 Limestone – – – – – – – – – – – – – – – – – – – – Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 6 36.0 4 8.0 24 24.4 12 11.0 – – 6 1.9 83 14.2 4 5.2 7 17.0 146 117.7 Table S4. 13 . ADB Aurignacian (OH19) . Classification of cores, retouched tools and bladelets. Bladelet counts include both retouched and unretouched blanks Cores N Bladelets extracted from N Retouched tools N prismatic for bladelets (fragment) 2 carinated/nosed “scraper” 3 bilaterally retouch ed blade 1 carinated “scraper” 1 other 5 notched piece 2 nosed “scraper” 1 TOTAL 8 notched bladelet 2 splintered piece/bipolar core 2 retouched piece fragment 2 TOTAL 6 TOTAL 7 Table S4. 14 . ADB Aurignacian (OH18). Stone tool technological categories . N = number, M = mass in grams RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 6 21.8 7 43.8 23 29.7 18 20.3 – – 9 1. 7 127 28.8 5 15.0 4 9.6 1 99 170.7 Quartzite – – – – 2 3.3 – – – – – – – – – – – – 2 3.3 Limestone – – – – – – – – – – – – 1 0.3 – – – – 1 0.3 Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 6 21.8 7 43.8 25 33.0 18 20.3 – – 9 1.7 128 29.2 5 15.0 4 9.6 202 174.3 Table S4. 15 . ADB Aurignacian (OH18). Classification of cores, retouched tools and bladelets. Bladelet counts include both retouched and unretouched blanks Cores N Bladelets extracted from N Burin types N Retouched tools N prismatic for bladelets (fragment) 3 carinated/nosed “scraper” 2 angle on natural surface 1 thumbnail scraper 1 carinated “scraper” 1 “burin” 1 TOTAL 1 Dufour bladelet 1 “burin” 1 splintered piece/bipolar core 1 atypically retouched piece 1 splintered piece/bipolar core 1 other 6 retouched piece fragment 1 TOTAL 6 TOTAL 10 TOTAL 4

PAGE 109

Table S4. 16 . ADB Aurignacian (OH17). Stone tool technological categories . N = number, M = mass in grams; an ochre stained quartzite cobble is not counted RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 32 159.9 21 91.0 72 130.0 83 100.8 2 14.5 42 23.5 587 132.2 40 157.6 9 20.76 888 830.4 Quartzite – – – – 2 5.8 – – – – – – 1 0.3 – – – – 3 6.1 Limestone – – 8 184.9 9 35.1 2 2.3 – – – – 2 0.8 5 14.0 1 22.45 27 259.5 Quartz – – – – – – 2 2.3 – – – – 3 0.6 – – – – 5 3.0 TOTAL 32 159.9 29 275.9 83 170.9 87 105.6 2 14.5 42 23.5 593 133.9 45 171.6 10 43.2 923 1099.0 Table S4. 17 . ADB Aurignacian (OH1 7 ). Classification of cores, retouched tools and bladelets. Bladelet counts include both retouched and unretouched blanks Cores N Bladelets extracted from N Burin types N Retouched tools N prismatic for bladelets 4 carinated/nosed “scraper” 3 angle, on natural surface 1 notched piece 5 prismatic, other 2 “burin” 6 on oblique truncation 1 Dufour bladelet 1 polyhedral 1 splintered piece/bipolar core 19 TOTAL 2 atypically retouched piece 1 fragments and other 5 other 15 retouched piece fragment 3 carinated “scraper” 1 TOTAL 43 TOTAL 10 nosed “scraper” 1 “burin” 2 splintered piece/bipolar core 16 TOTAL 32 Table S4. 18 . ADB Aurignacian (OH16). Stone tool technological categories . N = number, M = mass in grams RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 22 166.6 32 227.3 117 292.4 82 114.5 7 10.6 51 17.0 1069 224.2 44 176.6 28 80.3 1452 1309. 7 Quartzite – – – – 1 17.3 – – – – – – – – – – – – 1 17.3 Limestone 2 327.3 4 80.8 20 141. 8 4 3.6 – – – – 50 20.2 9 33.3 1 8.1 90 615.1 Quartz – – – – – – – – – – – – 3 0.6 – – – – – – TOTAL 24 493. 8 36 308.1 138 451.6 86 118. 1 7 10.6 51 17.0 1119 244.4 53 210.0 29 88.4 1543 1942. 0 Table S4. 19 . ADB Aurignacian (OH16). Classific ation of cores, retouched tools and bladelets. Bladelet counts include both retouched and unretouched blanks Cores N Bladelets extracted from N Retouched tools N prismatic for blades 2 carinated/nosed “scraper” 8 simple endscraper, atypical 1 prismatic for bladelets 7 splintered piece/bipolar core 9 perforator 2 prismatic for flakes 2 other 48 notched piece 6 fragments and other 2 TOTAL 65 short backed bladelet 5 splintered piece/bipolar core 11 notched bladelet 1 TOTAL 24 Dufour bladelet 2 Areeiro bladelet 2 marginally retouched bladelet 4 atypically retouched piece 2 retouched piece fragment 4 TOTAL 29

PAGE 110

Table S4. 20 . ADB Aurignacian (OH15). Stone tool technological categories . N = number, M = mass in grams RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS (a) TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 2 29.7 9 95.1 17 31.2 11 8.4 8 9.1 18 9.6 151 39.4 3 10.4 12 13.8 231 246.7 Quartzite – – – – – – – – – – – – – – – – – – – – Limestone – – – – – – – – – – – – – – – – – – – – Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 2 29.7 9 95.1 17 31.2 11 8.4 8 9.1 18 9.6 151 39.4 3 10.4 12 13.8 231 246.7 (a ) two refitted fragments of a large, marginally backed bladelet, typologically listed as a single item, are here counted separately Table S4. 21 . ADB Aurignacian (OH15). Classification of cores, retouched tools and bladelets. Bladelet counts include both retouched and unretouched blanks Cores N Bladelets extracted from N Burin types N Retouched tools N prismatic for flakes 1 “burin” 4 on concave truncation 1 notched piece 1 “burin” 1 other 2 2 TOTAL 1 backed bladelet fragment 1 TOTAL 2 TOTAL 2 6 short backed bladelet 1 backed truncated bladelet 1 Dufour bladelet 3 marginally retouched bladelet 1 atypically retouched piece 3 TOTAL 11 Table S4. 22 . ADB Gravettian (OH14). Stone tool technological categories . N = number, M = mass in grams RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS (a) TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 2 13.6 4 46.7 6 7.6 3 2.0 – – 6 2.5 58 16.4 2 6.4 1 0.6 82 95.8 Quartzite – – – – – – – – – – – – – – – – – – – – Limestone – – – – – – – – – – – – – – – – – – – – Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 2 13.6 4 46.7 6 7.6 3 2.0 – – 6 2.5 58 16.4 2 6.4 1 0.6 82 95.8 Table S4. 23 . ADB Gravettian (OH14). Classification of cores, retouched tools and bladelets. Bladelet counts include both retouched and unretouched blanks Cores N Bladelets extracted from N Retouched tools N prismatic for bladelets 2 “burin” 4 basal microgravette 1 TOTAL 2 other 3 TOTAL 1 TOTAL 7

PAGE 111

Table S4. 24 . ADB Gravettian (OH13). Stone tool technological categories . IL 3 items included. N = number, M = mass in grams RAW MATERIAL CORES FLAKE BLANKS LAMINARY BLANKS DEBRIS TOOLS (a) TOTAL Complete Fragment Small Blade Bladelet Chippage Chunk N M N M N M N M N M N M N M N M N M N M Flint 1 4.9 2 11.8 2 1.1 6 9.1 – – 3 0.7 20 4.0 1 1.1 – – 35 32.8 Quartzite – – – – – – – – – – – – – – – – – – – – Limestone – – – – – – – – – – – – – – – – – – – – Quartz – – – – – – – – – – – – – – – – – – – – TOTAL 1 4.9 2 11.8 2 1.1 6 9.1 – – 3 0.7 20 4.0 1 1.1 – – 35 32.8 Ta ble S4. 25 . ADB Gravettian (OH1 3 ). Classification of cores and bladelets. Bladelet counts include both retouched and unretouched blanks. IL3 items included Cores N Bladelets extracted from N Burin types N “burin” 1 “burin” 2 dihedral djt 1 TOTAL 1 other 1 TOTAL 1 TOTAL 3


Download Options

No images are available for this item.
Cite this item close


Cras ut cursus ante, a fringilla nunc. Mauris lorem nunc, cursus sit amet enim ac, vehicula vestibulum mi. Mauris viverra nisl vel enim faucibus porta. Praesent sit amet ornare diam, non finibus nulla.


Cras efficitur magna et sapien varius, luctus ullamcorper dolor convallis. Orci varius natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Fusce sit amet justo ut erat laoreet congue sed a ante.


Phasellus ornare in augue eu imperdiet. Donec malesuada sapien ante, at vehicula orci tempor molestie. Proin vitae urna elit. Pellentesque vitae nisi et diam euismod malesuada aliquet non erat.


Nunc fringilla dolor ut dictum placerat. Proin ac neque rutrum, consectetur ligula id, laoreet ligula. Nulla lorem massa, consectetur vitae consequat in, lobortis at dolor. Nunc sed leo odio.