Hayonim Cave: a TL-based chronology for this Levantine Mousterian sequence N. Mercier a, * , H. Valladas a , L. Froget a , J.-L. Joron b , J.-L. Reyss a , S. Weiner c , P. Goldberg d , L. Meignen e , O. Bar-Yosef f , A. Belfer-Cohen g , M. Chech h , S.L. Kuhn i , M.C. Stiner i , A.-M. Tillier j , B. Arensburg k , B. Vandermeersch j a Laboratoire des Sciences du Climat et de l’Environnement-IPSL, UMR CEA-CNRS-UVSQ, Domaine du CNRS, avenue de la Terrasse, 91198 Gif-sur-Yvette, France b Laboratoire Pierre Su ¨e, Groupe des Sciences de la Terre, CEN Saclay, 91191 Gif-sur-Yvette, France c Weizmann Institute of Science, Structural Biology Department, 76100 Rehovot, Israel d Boston University, Department of Archaeology, 675 Commonwealth Avenue, Boston, MA 02215, USA e Centre d’Etudes, Pre ´histoire, Antiquite ´, Moyen Age, UMR 6031 du CNRS, 250 rue Albert Einstein, 06560 Valbonne, France f Harvard University, Peabody Museum, 11 Divinity Avenue, Cambridge, MA 02138, USA g Institute of Archaeology, The Hebrew University, Mount Scopus, Jerusalem, Israel h MNHN, Laboratoire d’Anthropologie, Muse ´e de l’Homme, 17 place du Trocade ´ro, 75116 Paris, France i University of Arizona, Department of Anthropology, Building 30, Tucson, AZ 85721-0030, USA j Laboratoire d’Anthropologie des Populations du Passe ´, UMR5199-PACEA, Universite ´ Bordeaux 1, avenue des Faculte ´s, 33405 Talence, France k Department of Anatomy and Anthropology, Sackler School of Medicine, Tel Aviv University, Ramat Aviv 61108, Israel Received 1 April 2005; received in revised form 12 December 2005; accepted 27 September 2006 Abstract The thermoluminescence dating method was applied to 77 heated flints from the Mousterian layers of Hayonim Cave in order to provide a precise TL-based chronology for this important Levantine sequence. A detailed dosimetric study was performed by using 76 dosimeter cap- sules and revealed strong spatial dose-rate variations. In parallel, Fourier transform infrared spectrometry enabled the identification of various mineral assemblages in the sediments of the cave and to localize the boundaries of these assemblages. By comparing these two data sets, it is shown that low dose-rate values (w500 mGy/a) are systematically recorded in areas where the calcite-dahllite (CD) assemblage is preserved, whereas higher values (up to 1300 mGy/a) are associated with the leucophosphite, montgomeryite, variscite and siliceous aggregates (LMVS) assemblage. The dosimetric and mineralogical information was combined in order to assess, where possible, the dose-rate experienced by each flint during its burial. Some of the flint samples analyzed were too close to mineral assemblage boundaries and were therefore discarded. This rigorous selection led to TL ages ranging from 230 to 140ka for the lower part of the Mousterian sequence (layers F and Lower E), which contains lithic industries characterized by blade production using the laminar method. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: Thermoluminescence dating; Mousterian; Chronology; Near-East; Hayonim Cave 1. Introduction Hayonim Cave is located in the western Galilee, Israel, on the right bank of Nahal Meged (Fig. 1). The cave was formed by karstic activities in an Upper Cretaceous limestone reef and is characterized by a huge arch providing access to several chambers, one of which is open to the outside by means of a 25-m-high chimney. The sedimentary infilling is at least 6 m thick and is mainly of anthropic origin (Goldberg and Bar-Yosef, 1998); the sediments contain remains of combus- tion products (charcoal, ash remnants and phytoliths .), lithic * Corresponding author. E-mail address: [email protected](N. Mercier). 0305-4403/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.jas.2006.09.021 Journal of Archaeological Science 34 (2007) 1064e1077 http://www.elsevier.com/locate/jas
Transcript
Journal of Archaeological Science 34 (2007) 1064e1077http://www.elsevier.com/locate/jas
Hayonim Cave: a TL-based chronology forthis Levantine Mousterian sequence
N. Mercier a,*, H. Valladas a, L. Froget a, J.-L. Joron b, J.-L. Reyss a, S. Weiner c, P. Goldberg d,L. Meignen e, O. Bar-Yosef f, A. Belfer-Cohen g, M. Chech h, S.L. Kuhn i, M.C. Stiner i,
A.-M. Tillier j, B. Arensburg k, B. Vandermeersch j
a Laboratoire des Sciences du Climat et de l’Environnement-IPSL, UMR CEA-CNRS-UVSQ,
Domaine du CNRS, avenue de la Terrasse, 91198 Gif-sur-Yvette, Franceb Laboratoire Pierre Sue, Groupe des Sciences de la Terre, CEN Saclay, 91191 Gif-sur-Yvette, France
c Weizmann Institute of Science, Structural Biology Department, 76100 Rehovot, Israeld Boston University, Department of Archaeology, 675 Commonwealth Avenue, Boston, MA 02215, USA
e Centre d’Etudes, Prehistoire, Antiquite, Moyen Age, UMR 6031 du CNRS, 250 rue Albert Einstein, 06560 Valbonne, Francef Harvard University, Peabody Museum, 11 Divinity Avenue, Cambridge, MA 02138, USA
g Institute of Archaeology, The Hebrew University, Mount Scopus, Jerusalem, Israelh MNHN, Laboratoire d’Anthropologie, Musee de l’Homme, 17 place du Trocadero, 75116 Paris, France
i University of Arizona, Department of Anthropology, Building 30, Tucson, AZ 85721-0030, USAj Laboratoire d’Anthropologie des Populations du Passe, UMR5199-PACEA, Universite Bordeaux 1, avenue des Facultes, 33405 Talence, France
k Department of Anatomy and Anthropology, Sackler School of Medicine, Tel Aviv University, Ramat Aviv 61108, Israel
Received 1 April 2005; received in revised form 12 December 2005; accepted 27 September 2006
Abstract
The thermoluminescence dating method was applied to 77 heated flints from the Mousterian layers of Hayonim Cave in order to providea precise TL-based chronology for this important Levantine sequence. A detailed dosimetric study was performed by using 76 dosimeter cap-sules and revealed strong spatial dose-rate variations. In parallel, Fourier transform infrared spectrometry enabled the identification of variousmineral assemblages in the sediments of the cave and to localize the boundaries of these assemblages. By comparing these two data sets, it isshown that low dose-rate values (w500 mGy/a) are systematically recorded in areas where the calcite-dahllite (CD) assemblage is preserved,whereas higher values (up to 1300 mGy/a) are associated with the leucophosphite, montgomeryite, variscite and siliceous aggregates (LMVS)assemblage. The dosimetric and mineralogical information was combined in order to assess, where possible, the dose-rate experienced byeach flint during its burial. Some of the flint samples analyzed were too close to mineral assemblage boundaries and were therefore discarded.This rigorous selection led to TL ages ranging from 230 to 140 ka for the lower part of the Mousterian sequence (layers F and Lower E), whichcontains lithic industries characterized by blade production using the laminar method.� 2006 Elsevier Ltd. All rights reserved.
0305-4403/$ - see front matter � 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jas.2006.09.021
by karstic activities in an Upper Cretaceous limestone reef andis characterized by a huge arch providing access to severalchambers, one of which is open to the outside by means ofa 25-m-high chimney. The sedimentary infilling is at least6 m thick and is mainly of anthropic origin (Goldberg andBar-Yosef, 1998); the sediments contain remains of combus-tion products (charcoal, ash remnants and phytoliths .), lithic
1065N. Mercier et al. / Journal of Archaeological Science 34 (2007) 1064e1077
materials and bone residues from both large (gazelles, fallowdeer, aurochs) and small (tortoises, lizards and rabbits) animals(Stiner, 2006; Stiner et al., 2000). The sequence is dividedinto several layers, from G (bottom) to A (top), containinga succession of occupations representing Middle Palaeolithic,Upper Palaeolithic and Natufian periods (Bar-Yosef, 1991;Bar-Yosef et al., 2006).
The excavations carried out between 1992 and 1999 wereconfined to two distinct locations within the site, the CentralArea and the Deep Sounding, and mainly focused on the Mous-terian sequence. During these excavations, the otherwise homo-geneous Mousterian deposits of Layer E could be subdivided onthe basis of lithic densities in order to evaluate chronologicalchanges in the history of human occupation of the cave.
In Layer F and in the lower part of Layer E (units 6e4), theassociation of Levallois debitage with true blade technology(Laminar debitage) provides a specific aspect to this industry(Meignen, 1998, 2000). Characterized by the presence of elon-gated products (blades and points, often retouched) togetherwith short often triangular Levallois blanks, the lithic assem-blages closely resemble the material found in Abou Sif(Neuville, 1951). Generally speaking, they show global affi-nities in their core reduction strategies with the assemblagesclassically labelled as ‘‘Tabun D type’’ or Early LevantineMousterian found in several sites throughout the Near East,even if the tool-kits show some variations. In contrast, in theUpper E assemblages (units 1e3), Levallois technology isclearly oriented towards the production of short and thin flakes(Meignen, 1998). Core management is evidenced by the
Fig. 1. Location of some Near-Eastern Middle Palaeolithic sites, including
Hayonim Cave.
presence of centripetal and unidirectionnal removals resultingin oval, quadrangular and subtriangular blanks. Retouchedtools include mainly side-scrapers on large Levallois blanks.By all these criteria, the Upper E assemblages (especially units1 and 2, with centripetal core preparation) show some affinitieswith the so-called ‘‘Tabun C type’’ industries found, for exam-ple, in Tabun unit I (Jelinek, 1981), in most layers from Qafzeh(Hovers, 1997) and Ksar Akil XXVI (Marks and Volkman,1986). In fact, the preliminary technological study of theseUpper E assemblages once more illustrates the variabilityand flexibility of the Levallois technology, resulting in an arrayof products of controlled morphology (Meignen, 1998).
Zooarchaeological analyses of the ungulate remains inMousterian Layer E indicate that large animals were obtainedby hunting, and prey biomass comparisons demonstrate thatlarge game were the principal sources of meat, supplementedby modest quantities of tortoises and other slow-moving smallanimals (Stiner, 2006). The results also indicate significantexpansion in human dietary breadth immediately followingthe Mousterian, and predator pressure on tortoise populationsimplies subtle increases in human population densities aroundthe time of the Middle to Upper Paleolithic transition (Stineret al., 2000). Human populations of the early Middle Paleolithicappear to have been small and widely dispersed, accounting forthe relatively ephemeral nature of the Mousterian occupations atHayonim Cave in contrast to those documented for the late Mid-dle Paleolithic at Kebara Cave (Bar-Yosef, 1998; Stiner, 2006).
The few Middle Paleolithic hominid remains recovered atHayonim Cave were concentrated in the central area of thecave (except for one specimen discovered in the entrancearea) and the majority of them originated from upper strati-graphic units of layer E. To date, the fossil hominid recordfrom other Mediterranean Levantine sites (i.e. Amud, Geula,Kebara, Qafzeh, Skhul, and Tabun) provides evidence thatMousterian assemblages can be associated with distinct humangroups. In the past decade, there have been contradictory opin-ions expressed on the interpretation of anatomical differencesbetween these fossil hominids and especially on the taxonomicdesignation of the Amud, Kebara and Tabun individuals (e.g.Arensburg, 2002; Arensburg and Belfer-Cohen, 1998; Hoverset al., 1995; Mann, 1995; Rak, 1998; Tillier, 1998; Tillieret al., 2003; Trinkaus, 1995; Vandermeersch, 1995). A majordifference between the fossil record from the above-mentionedsites and that from Hayonim site lies in the fact that the latterhas a low density of hominid remains and is dominated by iso-lated specimens (fragmentary skull remains, isolated teeth, cer-vical vertebra, hand and foot bones, parts of long bones) thatare not suitable for precise taxonomic assignment (Arensburget al., 1980; Arensburg et al., 1990). Moreover, in contrastwith other Levantine sites, Hayonim Cave provides no evi-dence for mortuary practices within the Mousterian sequence.
2. Chronometric dating at Hayonim Cave
The chronology of the Hayonim Mousterian sequencehas attracted much attention during the last two decades.Comparisons of micro-faunal remains with other Levantine
1066 N. Mercier et al. / Journal of Archaeological Science 34 (2007) 1064e1077
archaeological sequences (Tchernov, 1988; Tchernov, 1998)indicated that the faunal assemblages from Upper E in Hay-onim Cave predated those discovered in Qafzeh Cave. Abso-lute dating methods have also been applied extensively: theU-series method was applied to a speleothem unearthedfrom the Mousterian deposits and provided the first reliabledate (Schwarcz et al., 1980). Optical dating of sedimentswas attempted (Godfrey-Smith, personal communication)and preliminary results were obtained from the Thermolumi-nescence (TL) study of burnt lithics (Valladas et al., 1988).In parallel, the Electron Spin Resonance (ESR) datingmethod was applied to tooth enamel (Schwarcz and Rink,1998). More recently, combined ESR/U-series analysesyielded detailed radiometric information indicating that theage of the Upper E Mousterian layer is around 180 ka, andnot younger than 155 ka (Rink et al., 2004).
In this paper, we report the results of a study initiated10 years ago whose aim was to provide a TL-based chronologyfor the Hayonim Mousterian sequence. A total of 77 age esti-mates obtained on heated flints is presented with associatedradiometric information. Particular attention is given to envi-ronmental dose-rate variations through space and time, and totheir possible correlations with the distribution of authigenicminerals formed as a result of diagenetic processes (Weineret al., 2002). It is shown how these data can be combined inorder to discard some TL results and to increase the reliabilityof the remaining TL ages.
3. Samples and methods
During the last 30 years, the TL method has been extensivelyused for the dating of burned flints discovered in archaeologicalsites, in particular those sites beyond the range of radiocarbon(Huxtable and Aitken, 1988; Mercier et al., 1991, 1995a,b,c,2000; Valladas et al., 1986, 1987, 1988, 1998). The utility ofthe method is based on the fact that flints behave as excellentnatural dosimeters which record doses delivered by the differ-ent types of radiation coming from the decay of radioisotopes,such as the U- and Th-series and potassium. The TL method is
based on dosimetric measurements of these ionizing radiations.Short-ranged a and b particles (w20 mm and w2 mm, respec-tively) emitted from radioisotopes within the flint itself, providean ‘‘internal dose’’. An ‘‘external dose’’ from long range(w40 cm max.) g-rays (as well as from cosmic radiations)necessitate consideration of the immediate environment of theflints.
3.1. Burnt flint specimens and dosimetric measurements
More than 100 flint artefacts showing signs of heating wereselected during the excavation campaigns at Hayonim Cave.The sample preparation followed the protocol defined byValladas (1992). After the outermost 2 mm of each flintwere removed, the remaining core was crushed and subjectedto chemical treatment and its TL emission was analyzed. Only77 flints had been heated at temperatures sufficient to ensurea complete resetting of the TL signal, and only these were se-lected for further analyses. Fig. 2 shows the TL emissions ofsample (HAY 7) as well as the equivalent dose (ED). The nat-ural and regenerated growth curves are typical of the sub-linear behavior observed for most of the Hayonim samples.These samples come from the Deep Sounding (squares D-E-F-28-27-26), and from the southern (squares I-J-K-24-23-22)and northern parts (squares G-H-I-J-20-19-18) of the CentralArea of the excavations (Fig. 3).
As the excavations covered a relatively large area(w23 m2), great effort was made to obtain detailed informationon spatial variations in environmental dosimetry. A total of 76calcium sulfate dosimeters were placed in different sections ofthe cave and remained in place for approximately 1 year. Mostof them were inserted horizontally in the newly exposed pro-files close to squares containing numerous burnt flints: G18,H19 and I20 e North section; K23-24 and I24 e South sectionand, in the Deep Sounding: squares D-E-F27 and F28. A fewdosimeters were also inserted vertically at the end of the fieldseason in squares where excavation was scheduled in subsequentyears (see Fig. 3). Additional dosimetric countings were carriedout on-site using a portable g-detector inserted in natural holes
HAY 7
0
5000
10000
15000
20000
25000
30000
180 240 300 360 420Temperature (°C)
TL
(a.u
.)
TL
(a.u
.)
-50
0
50
100
150
200
250
300
ED
(G
y)
(a)
0
50000
100000
150000
200000
250000
300000
0 500 1000 1500Dose (Gy)
1st growth curve2nd growth curve
(b)
Fig. 2. (a) TL emission of sample HAY 7 for doses: 0, 281, 562, 843 Gy. The equivalent dose (ED) calculated as a function of temperature (‘‘plateau test’’) is
indicated. (b) Growth curves for sample HAY 7 (natural þ given doses (1) and regenerated (2)) are typical of the sub-linear behavior observed for most of the
Hayonim samples.
1067N. Mercier et al. / Journal of Archaeological Science 34 (2007) 1064e1077
Fig. 3. Schematic map of Hayonim Cave showing boundaries of excavations, the Deep Sounding and the Central Area. The location of the dosimeters and flints are
indicated with crosses and circles, respectively.
(burrows) present in some profiles, a few other measurementswere made in the laboratory with a high purity germaniumdetector. In this case, about 100 g of sediment were analyzedby g-spectrometry for obtaining information on possible dise-quilibria in the radioactive decay chains and investigating therepartition of the g dose-rate between the U- and Th-seriesand potassium.
3.2. Fourier Transform Infrared analyses of sediments
Mineralogical analyses were carried out on-site during theexcavation of Hayonim Cave using Fourier transform in-frared (FTIR) spectroscopy in order to better understandthe archaeology via sediment diagenesis and site formationprocesses (Weiner et al., 2002). This method is capable ofidentifying mineral assemblages using only a few tens of mi-crograms of sample (Weiner et al., 1993). More than 2100samples were analyzed. It was shown that three major min-eral assemblages are present in the archaeological assem-blages at Hayonim: the first, termed CD (Weiner et al.,2002), is composed mainly of calcite and secondarily ofdahllite. Calcite is the main residue of the fresh ash whichoriginally made up the bulk of the sediments (Schieglet al., 1994). Dahllite (also known as carbonated hydroxyap-atite) results from interaction of calcite with phosphate-richsolutions. The second assemblage, termed LMVS, representssediments that are more diagenetically altered. In the courseof the diagenetic process, the dahllite dissolves and is replac-ed by other more stable phosphate minerals such asleucophosphite, montgomeryite and variscite. Due to thedissolution and transport of other minerals, the relative pro-portions of so-called siliceous aggregates and phytoliths, mi-nor components in the original ash, increase with ongoing
diagenesis. Note that the siliceous aggregates themselves,as well as one of the common authigenic phosphate minerals(leucophosphite) contain potassium (Schiegl et al., 1996). Asnoted above, the formation of the LMVS assemblage also re-sults in the dissolution of bones which are composed of dahl-lite, a relatively unstable mineral with respect to the LMVSgroup. The third identified mineral assemblage consists ofthe most highly altered sediments, in which clays have bro-ken down and silica has been released.
The extensive study of sediment mineralogy in the fieldalso provided a detailed map of the spatial distribution ofthe three mineral assemblages. This map showed sharp varia-tion in sediment mineralogy over distances as short as a fewcentimeters (Weiner et al., 2002; see their Figs. 10 and 11,for instance).
At this stage, it is interesting to notice that the local differ-ences in sediment chemistry and rates of diagenetic activity,based on a combination of taphonomic methods and Fouriertransform infrared (FTIR) spectroscopic analyses of sedimentand bones (Stiner et al., 2001; Weiner et al., 2002), correlateclosely with the distribution of bones in the sediments at Hay-onim Cave. The quality of bone preservation follows an ‘‘all-or-nothing’’ pattern in the central area, with relatively abruptchanges in bone abundance rather than gradual transitions be-tween good to poor preservation. The areas of good bonepreservation are extensive and associate consistently withCD mineral assemblages dominated by calcite, dahllite orboth. Bone preservation in Layer E was best within 2e3 mof the cave walls in the central area, especially between420 and 500 cm below datum. Very few bones were preservedanywhere in Layer F, where secondary mineral assemblages(LMVS and altered clays) predominate in the sediments(Weiner et al., 2002).
1068 N. Mercier et al. / Journal of Archaeological Science 34 (2007) 1064e1077
Fig. 4. Environmental dose-rates (mGy/a) measured with calcium sulfate dosimeters as a function of depth Z, for squares: (a) H19-G18; (b) L22-23-24; (c) I25-H23-
24; (d) K-J25; (e) F27-28-29-E28-29 and (f) E26-D26-27-28. Note that for (a) and (b), the values are high (w1100 mGy/a) and low (w500 mGy/a), respectively, but
do not vary significantly with depth. In contrast, for (c) and (d), the recorded dose-rates vary significantly, even on a relatively small scale (10e20 cm). In the Deep
Sounding the values are around 1000e1200 mGy/a (e) or range from 500 to 800 mGy/a (f), but some are higher by w30%.
4. Dosimetry results
4.1. Present-day environmental dose-rates
In Fig. 4 the environmental dose-rates recorded by the do-simeters are plotted as a function of depth. For simplification,they were grouped into nearby squares (a, H19-G18; b, L22-23-24; c, I25-H23-24; d, K25-J25 for the Central Area; e,F27-28-29-E28-29; and f, E26-D26-27-28 for the DeepSounding). In squares H19-G18, the values are homogeneous,ranging from 900 to 1100 mGy/a (not including one value at
w1300 mGy/a) with 10% dispersion around the mean, anddo not vary significantly with depth from Z ¼ 250 toZ ¼ 550 cm (Fig. 4a). The same trend with depth is observedfor squares L22-23-24 (Fig. 4b), although the values are muchlower (400e550 mGy/a). The range of variation in both pro-files is about the same (w10%). On the other hand, thedose-rates can vary even over small differences of elevationin squares I25-H23-24 (Fig. 4c) and K25-J25 (Fig. 4d). Inthe Deep Sounding, significant variation also exist in squaresE26-D26-27-28 (Fig. 4f) where dose-rates range from 500 to1000 mGy/a. One can then conclude that in certain sections
1069N. Mercier et al. / Journal of Archaeological Science 34 (2007) 1064e1077
of the cave, the environmental dose-rates do not vary signifi-cantly with depth over 1m or more, while in other parts, thereare substantial variations at even small scales (<10 cm), hori-zontally and/or vertically.
These observations have direct implications for dating whendosimetric methods are used. We note that the range of themost energetic g-rays (w40 cm) is greater than the small-scaleheterogeneity. Moreover present-day dose-rate measurementsare seldom recorded at the precise locations where flint sam-ples were collected, but at distances up to 1 m or more. Thisis not a problem in a homogeneous mineralogical environmentwhere the dose-rate calculated by averaging the doses recordedin neighboring dosimeters should be representative of the dose-rate received by a sample in the same environment; in a hetero-geneous mineralogical environment, however, the dose-ratevalues measured at different locations may differ significantlyfrom the dose experienced by the sample. This fact clearly hasto be taken into account when determining the environmentaldose of a given sample.
4.2. Basis for the dosimetric heterogeneities
In the course of the field seasons, a general overview of thedistribution of mineral assemblages in the cave and dose-ratevalues showed a strong correlation, as did the distributions ofmineral assemblages and bone abundance. In sections wherecalcite and dahllite were abundant (CD assemblage), dose-ratesaround 500 mGy/a were recorded, whereas in sections rich inleucophosphite, montgomeryite, variscite and siliceous aggre-gates (LMVS assemblage), the dose-rates were significantlyhigher (up to 1300 mGy/a). In order to understand the basisfor this correlation, sediment samples (w300 mg) were takenfrom various sections, homogenized and the major mineralsidentified by FTIR. In parallel, neutron activation analysis(NAA) was performed on the remaining powder (w250 mg)and the radioisotope contents (U, Th and K) were measured(Mercier et al., 1995). Fig. 5 gives the g dose-rates calculatedby assuming that the identified minerals were the only compo-nents in the environment (within a radius of at least 50 cm, i.e.,greater than the typical range of the g-rays). It is worth notingthat the results basically confirm at a very small scale (the‘‘scale’’ of a 300 mg sample) what was observed at a largescale: low dose-rates are associated with the CD assemblage(calcite and dahllite), and high dose-rates are associated withthe assemblage made up of siliceous aggregates with varyingquantities of leucophosphite and montgomeryite. The on-sitesurvey with a gamma spectrometer also revealed that the lowestdose-rates (w300 mGy/a) were measured from sediments dom-inated by calcite. Note too that Fig. 5 shows some exceptions,the most glaring being that one sample dominated by dahllitewhich has the third highest dose-rate. It should be noted thatcarbonated apatite has a tendency to readily occlude a varietyof different ions, including apparently potassium.
These mineralogical data can also explain the variation indose-rate observed along different profiles (Fig. 4). In mineral-ogically homogeneous sections, dose-rate values are fairly ho-mogeneous: either low (w500 mGy/a) as in squares L22-23-24
or high (w1000 mGy/a) as in squares H19-G18. In areas wheresharp mineralogical boundaries exist, the measured dose-rateshave intermediate values depending on the relative distanceof the dosimeters from these boundaries (see Weiner et al.(2002) for details on mineral assemblage distributions,and Aitken (1985) for dose-rate variations as a function ofthe relative distance). The implication of these observationsto TL dating is that it is difficult, if not impossible, to reliablyestimate a representative dose-rate for any burned flint sampleslocated within 30 cm of a boundary between mineral assem-blages. These samples cannot therefore be used for establishinga chronology.
4.3. Dose-rate changes with time
The NAA data can provide some insight into the basis of thedose-rate changes. Fig. 6a shows the distribution of the doseinto the different components (the U- and Th-series and K-40). There is no obvious correlation between major mineraltypes and the source of radiation. As the samples are listed ac-cording to increasing radiation dose, one can however noticethat high dose-rates are correlated with samples dominatedby siliceous aggregates, which have high potassium contents(Schiegl et al., 1994, 1996). This is clearly demonstrated inFig. 6b, which shows that the total g dose-rate increases line-arly as a function of the potassium content of the samples. Con-sequently, the contribution of the U- and Th-series to the annualdose-rate does not exceed 30e40%, except for a few samples.
As siliceous aggregates are only a minor component offresh ash, samples dominated by siliceous aggregates musthave undergone severe diagenesis resulting in the loss of theother components. This raises the important question ofwhen this diagenesis occurred, as the timing could clearly af-fect the mean dose-rates experienced by the samples and con-sequently, the calculated ages. If, for example, the diagenesis
0 500 1000 1500 2000Infinite dose-rate (µGy/a)
Apatite
Apatite
Variscite
Apatite
Apatite
Clay
Apatite
Apatite
Apatite
Apatite (S.A.)
Leucophosphite (Montg.)
Leucophosphite (S.A.)
Leucophosphite (S.A.)
Leucophosphite (S.A.)
Crandalite
Crandalite
Montgomeryite (S.A.)
S.A.
S.A. (Leucophosphite)
S.A. (Leucophosphite)
S.A.
S.A. (Montgomeryite)
S.A. (Crandalite)
Fig. 5. Infinite g dose-rates (mGy/a) for the different minerals identified by
FTIR. For comparison, each value was computed by taking into account the
U, Th and K contents determined by NAA and assuming that the identified
mineral was the only component in the environment. The arrow indicates
the exception mentioned in the text. (Carbonated hydroxyapatite (or Apatite)
is used instead of Dahllite.)
1070 N. Mercier et al. / Journal of Archaeological Science 34 (2007) 1064e1077
was a slow steady process that took place during most of theperiod of time that elapsed after the flint was deposited inthe sediments, then using the present-day relatively highdose-rate would result in TL ages that are too young. If, onthe other hand, the diagenesis took place soon after burial,
(b)
(a) 0% 20% 40% 60% 80% 100%
Apatite (S.A.)
Apatite
Variscite
Apatite
Clay
Apatite
Apatite
Apatite
Apatite
Leucophosphite (Montg.)
S.A. (Montgomeryite)
S.A. (Leucophosphite)
Montgomeryite (S.A.)
S.A.
Leucophosphite (S.A.)
S.A. (Leucophosphite)
Crandalite
S.A.
Leucophosphite (S.A.)
S.A. (Crandalite)
Apatite
Crandalite
Leucophosphite (S.A.)
U - fraction Th - fractionK - fraction
0
500
1000
1500
2000
2500
0 1 2 3 4 5 6 7 8K content (%)
To
tal d
ose-rate (µ
Gy/a)
Fig. 6. (a) Contribution to dose-rate of the different components (the U- and
Th-series and K-40). Note that the samples are listed from top to bottom with
increasing total g dose-rate. As no significant disequilibria have been detected
in the 100 g sediment samples analyzed by g-spectrometry in the laboratory,
the listed g dose-rates were calculated by taking into account all the g-rays of
the U- and Th-series. The second most abundant constituents are shown in
brackets. S.A. refers to siliceous aggregates. (b) A plot of the total g dose-rate
as a function of the K content of the samples shows a linear correlation.
this would not affect the TL ages as the dose-rate measure-ments today would be representative of the dose-rates experi-enced by the flints over most of the time since deposition.
Karkanas et al. (2000) and Shahack-Gross et al. (2004)have presented several lines of evidence all pointing to theconclusion that diagenesis in prehistoric caves in the Levantand in Greece, took place soon after burial, possibly evenwithin hundreds or tens of years. One reason is that the diagen-esis is driven mainly by the breakdown of organic materialfrom bat and bird guano. This produces acids that in turn causethe dissolution of less stable minerals and the formation ofmore stable ones. So we assumed that the diagenetic historyof the mineralogical assemblages in Hayonim Cave is nota major factor which induced changes in the dose-rates withtime and so we used the present-day dose-rates to calculatethe TL ages.
5. Thermoluminescence age estimates
5.1. Sample selection and results
As noted previously, the primary difficulty in evaluating thedose-rate for a given burned flint is related to its distance froma boundary between two mineral assemblages. For a sampleunearthed from one of the two distinct environments (CDand LMVS assemblages) at a distance greater than 30 cmfrom an identified mineralogical boundary, the dose-rate wascalculated by averaging the individual dose-rates recordedby the neighboring dosimeters in the same mineralogical envi-ronment. However, every sample located at a distance lessthan 30 cm from a boundary, as defined using the FTIRdata, was noted and is identified with an asterisk (last columnof Table 1), since the present-day environmental dose-rate itreceived just before it was unearthed cannot be determinedwith any reliability. The marked samples were then excludedfrom the discussion on the chronostratigraphy.
Table 1 lists 77 burnt flint samples from Hayonim Cavewith their laboratory numbers, the layers and units in whichthey were discovered, their locations and relevant dosimetricdata, as well as the estimated TL ages and associated errors.Details concerning the radiometric data and luminescencemeasurements are given in the legend of Table 1. All theseTL ages are plotted in Fig. 7, as a function of stratigraphy.In general, it appears that the rejected samples (marked byan asterisk in Table 1 and open symbols in Fig. 7) have ageswhich are often scattered and in disagreement (older or youn-ger) with those obtained for samples located in homogeneousdosimetric environments. Consequently, our methodology forselecting samples appears reliable and gives us confidence inthe average ages listed below. However, this approach madeus exclude several samples whose ages were in the same rangeas the others for the same layer.
5.2. Average ages
The selected samples (without an asterisk) were groupedinto three sets depending on their location and their identified
Table 1
Analytical data and TL age estimates (all errors are given at one sigma level)
Equivalent dose Age estimates
Gy � ka �
149 6 126 12
183 16 127 14
217 21 155 19
149 9 119 12
147 15 114 15
204 10 146 13
156 7 124 12
161 11 125 13
169 6 139 13
266 14 129 11
147 6 125 12
153 4 119 10
181 13 144 16
202 7 140 11
153 11 148 18
216 10 146 13
189 8 142 13
189 10 143 14
170 6 169 19*
221 15 123 14*
102 4 168 29*
213 10 221 27*
139 4 157 19
133 7 208 35
122 4 163 23
128 9 140 22*
173 18 163 26*
160 17 188 35*
105 6 128 22*
141 20 141 27*
140 10 159 26 *
246 6 105 9
213 6 190 15*
161 14 213 26*
(continued on next page)
10
71
N.
Mercier
etal.
/Journal
ofA
rchaeologicalScience
34(2007)
1064e
1077
Sample
lab. no.
Square,
sub-square
Layer/unit Location: (x,y,z)
or depth (Z )
U (ppm) Th (ppm) K (%) S-a mGy/103
a/cm2Dose rates
Internal External Total
mGy/a � mGy/a � mGy/a �G18 Northern part of Central Area
51 d E base (51,92,539) 1.83 0.12 0.04 31.0 1294 109 974 101 2267
E27
58 b E base 520-525 0.94 0.08 0.07 20.6 536 39 876 102 1412
50 b F top (45,62,545) 0.40 0.10 0.06 15.2 228 14 851 102 1078
60 b F top 555-560 1.26 0.24 0.12 24.3 836 65 901 102 1737
57 c F top (76,14,578) 0.81 0.24 0.03 21.7 476 38 723 101 1199
E28
404 d E base (87,55,543) 0.48 0.12 0.14 20.0 366 27 915 101 1280
409 d E base (80,70,558) 0.74 0.16 0.11 23.2 514 42 920 101 1434
F28
406 c E base (75,35,520) 0.80 0.07 0.08 28.1 573 47 710 121 1283
516 b F base (10,75,728) 1.43 0.08 0.07 22.9 844 47 1136 121 1980
522 a F base (23,5,719) 1.16 0.05 0.06 19.7 622 47 890 121 1513
523 d F base (67,60,726) 3.28 0.04 0.02 14.9 1354 47 1077 121 2431
1073N. Mercier et al. / Journal of Archaeological Science 34 (2007) 1064e1077
5
24
aF
bas
e(2
6,1
4,7
15
)0
.67
0.1
80
.07
26
.54
89
47
90
81
21
13
97
13
02
55
80
18
36
0
52
5b
Fb
ase
(43
,98,
70
3)
1.1
40
.05
0.0
51
7.8
55
84
71
16
71
21
17
25
13
04
33
13
25
12
0
52
6b
Fb
ase
(4,9
5,7
15
)0
.83
0.0
70
.07
20
.64
80
47
11
43
12
11
62
31
30
30
32
11
87
20
52
7b
Fb
ase
(19
,87,
71
1)
0.6
10
.19
0.0
81
8.3
37
15
71
14
31
31
15
14
14
33
34
11
22
12
2
F2
7
51
7d
Fb
ase
(92
,89,
72
8)
1.6
00
.07
0.0
72
3.0
93
54
71
17
31
21
21
08
13
04
62
62
19
14
*
51
8d
Fb
ase
(54
,60,
73
3)
0.7
40
.12
0.0
52
1.2
42
74
71
16
71
21
15
94
13
03
71
12
23
32
0
51
9b
Fb
ase
(47
,77,
71
8)
0.3
70
.22
0.0
31
0.3
16
34
71
13
21
21
12
95
13
02
45
81
89
20
52
0d
Fb
ase
(74
,98,
71
8)
0.9
90
.15
0.1
12
3.3
65
34
71
16
21
21
18
15
13
03
18
32
17
52
2
52
1b
Fb
ase
(21
,99,
73
6)
1.3
50
.12
0.1
02
6.5
90
94
71
10
41
21
20
13
13
02
44
18
12
11
2*
Th
eU
,T
han
dK
con
ten
tso
ffl
ints
wer
em
easu
red
by
neu
tro
nac
tiva
tio
nan
aly
sis
(NA
A)
atth
eP
ierr
eS
ue
Lab
ora
tory
,S
acla
y(J
oro
n,
19
74)
and
hav
eea
chan
erro
ro
f1
0%
.T
he
S-a
sen
siti
vit
y(V
alla
das
and
Val
lad
as,
19
82)
was
det
erm
ined
by
com
par
ing
TL
sign
als
indu
ced
resp
ecti
vely
by
ab
do
se(S
r-9
0so
urc
e:d
ose
-rat
e8
.2G
y/m
in)
and
by
a-r
ays
fro
ma
Pu
-23
8so
urc
e(fl
ux
2.3
53
10
6a
/cm
2p
ers)
.T
he
inte
rnal
do
sera
tew
asco
mp
ute
dfr
om
the
U,
Th
and
Kco
nte
nts
usi
ng
pu
bli
shed
valu
eso
fsp
ecifi
cd
ose
-rat
es(L
irit
zis
and
Ko
kk
ori
s,1
99
2;V
alla
das
,1
98
8)
and
the
S-a
sen
siti
vit
yd
eter
min
edex
per
imen
tall
y.T
he
exte
rnal
do
se-r
ate
incl
ud
esa
cosm
icco
ntr
ibu
tio
no
f6
0m
Gy
/a(P
resc
ott
and
Hu
tto
n,
19
88)
.T
he
TL
signal
sw
ere
reco
rded
wit
han
auto
mat
icre
ader
(Val
ladas
etal
.,1994)
:th
eb
lue
com
po
nen
to
fth
e3
80� C
TL
pea
kw
as
sele
cted
by
aM
TO
380
nm
filt
eran
ddet
ecte
dw
ith
aE
MI
9635Q
Bphoto
mult
ipli
ertu
be.
The
hea
ting
rate
was
5� /
s.E
Dva
lues
wer
ed
eter
min
edb
yco
mp
arin
gth
efi
rst
and
seco
nd
TL
gro
wth
curv
es(M
erci
eret
al.,
19
92),
sinc
efo
rm
ost
sam
ples
,th
eg
row
thcu
rve
of
the
TL
sig
nal
(TL
vers
us
acq
uir
edd
ose
)w
assu
b-l
inea
rw
her
eas
for
the
oth
ers,
the
shap
ew
asm
ore
lin
ear.
Ty
pic
alex
amp
les
of
sub
-lin
ear
gro
wth
curv
esar
e
show
nin
Fig
.2.
All
the
flin
tsli
sted
hav
ep
asse
dth
ep
late
aute
st(A
itk
en,
19
85)
.T
he
erro
rsas
soci
ated
wit
hth
eag
ees
tim
ates
wer
eco
mp
ute
dac
cord
ing
toA
itk
en(1
98
5,ap
pen
dix
B).
mineralogical assemblage: the northern part of the CentralArea includes the squares G18-G19-H19-I20-J20; the southernpart, the squares K22-K23-K24-J24-I24-I23 and the DeepSounding, the squares D27-E27-E28-F28-F27. For each layerand unit within a specific area, the average TL age value wascalculated (Table 2); since these layers do not represent singlebut successive occupations, the associated error was obtainedby averaging the individual errors; in brackets, the standarddeviation of the individuals results is also given. Note, how-ever, that samples from square D27 (Deep Sounding) wereall discarded because of the incoherence of the TL results,ranging from w272 to w35 ka (!). This may well be relatedto the presence of a 1.30-m-deep trench thought to be theproduct of water driven erosion which resulted in redepositionof younger sediments together with stone artefacts (Weineret al., 2002).
Beginning with the bottom of the sequence, samplescollected in the Deep Sounding from layer F (base and top)result in two average ages which are statistically equivalent(210 � 28 ka and 221 � 21 ka, respectively) (Table 2). Evenif these results seem coherent, one must note that layer F wassubjected to extensive chemical diagenesis that resulted inthe breakdown of the clays and the formation of authigenic sil-ica, as demonstrated by Weiner et al. (2002). As discussed bythese authors, this process might be related to the erosional un-conformity that marks the geological limit between the top oflayer F and the bottom of E, and indicates a period of abundantwater flow, especially at the entrance of the cave where theDeep Sounding is located. Indeed, the fine-bedded nature ofthe sediments exposed in the Deep Sounding, indicate deposi-tion in standing water. Because clays and other mineral compo-nents are altered to a depth of more than 2 m below theerosional unconformity in this area, one cannot rule out thepossibility of an increase of the g-dose-rate leading to anunder-estimation of the true deposition age. However, it is ofinterest that the overlaying layer (Lower E) does not show signsof such an alteration in this area and is dated to 186 � 20 ka.It can then be deduced that the alteration was limited in timeto Oxygen Isotope Stage 7 (OIS-7), from 244 to 190 ka(Martinson et al., 1987). In order to assess the magnitude ofthis possible under-estimation, one can consider as an extremehypothesis that the environmental dose-rate in the Deep Sound-ing during OIS-7 was as low as those recorded in areas wherethe CD assemblage has been identified (w500 mGy/a). Consid-ering the data of Table 1, it can then be shown that such a sce-nario would lead to average ages of 224 � 29 ka and231 � 22 ka for layer F (base and top, respectively). Thus,the age of layer F (w210e220 ka) is possibly under-estimatedbut it is likely that this under-estimation does not exceed 20 kaand is similar in magnitude to the errors associated with themean ages.
Moving up in the stratigraphic sequence, the TL age ofLower E in the Deep Sounding is 186 � 20 ka. This is statis-tically indistinguishable from the average values obtainedfrom the southern part of the Central Area for units 5 and 4(160 � 22 ka and 168 � 21 ka, respectively), and is consistentwith the fact that these two units were both identified as
1074 N. Mercier et al. / Journal of Archaeological Science 34 (2007) 1064e1077
0 50 100 150 200 250 300 350TL age (ka)
Northern part of Central AreaSouthern part of Central AreaDeep Sounding
Upper E - 1
Upper E - 2
Upper E - 3
Lower E - 4
Lower E - 5Lower E - 6
C/ D
Ebase Ftop
Fbase
Fig. 7. All the TL age estimates obtained for the different sectors of the cave (northern and southern part of Central Area, Deep Sounding) are plotted as a function
of the stratigraphy. The selected samples are represented with filled symbols, whereas open symbols indicate the discarded samples.
belonging to the Lower E complex and thus are expected tohave the same age. This result is also close to the averageage of unit 4 (176 � 26 ka) in the northern part. Lower Ewas thus deposited at the beginning of OIS-6, a period whichseems to have been much drier than the previous one accord-ing to the state of preservation of the contemporaneous units,especially in the Deep Sounding area, where these upper sed-iments are less diagenetically altered and commonly calcitic.
In the southern part of the Central Area, unit 3 is dated to156 � 19 ka, indicating that the deposition of Upper E sedi-ments followed those of Lower E without any interruption.The dating of two samples assumed in the field to be derivedfrom unit 2 in this part, yielded a mean TL age (156 � 17 ka)indistinguishable from those of unit 3. We assume thereforethat unit 2 was not present in the dated southern part of theCentral area. In reality the correlations between the top ofthe Mousterian deposits in the southern section and those ofthe northern section were difficult to establish due to their re-moval by the Natufian and Aurignacian occupations and theearlier excavations of this area (1971e1979). In addition,the dipping of the units in the northern section, possibly intoa sinkhole in the rear, was the reason for erroneous field cor-relations. Units 3 and 2 were TL-dated in the northern part andthe age estimates are 144 � 14 and 129 � 13 ka, respectively.These last two results then show a clear-cut correlation withthe stratigraphy indicating a deposition of these units in thesecond part of OIS-6.
6. Discussion and conclusions
We first compare the TL chronology with the most completeand reliable age data published to date for Hayonim Cave,namely from the combined ESR/U-series analyses (Rinket al., 2004). The mean TL dates obtained for units 3e5 inthe southern part of the Central Area, when taking into accountthe standard deviations, range from w135 ka to w185 ka. TheTL dates are therefore in fairly good agreement with the ESR
dates on teeth of the same units (Early uranium Uptake model[EU]: 177 � 12 ka and Linear uranium Uptake model [LU]:182 � 15 ka). The mean TL age of 156 � 19 ka obtained forthe uppermost unit (unit 3) also fits well with the U-seriesage of a fallen calcite speleothem (155.3 þ 2.9 �1.4 ka) foundat a higher elevation (Z ¼ 248) than all the dated samples. Itcan thus be concluded that the Mousterian sequence in thesouthern part of the Central Area, including layers Lower E(units 5 and 4) and Upper E (unit 3), covers a minimum timeinterval ranging from w185 ka to w135 ka ago. This rangeis reproduced in the northern part for units 4e3 including thebase of E in the Deep Sounding, but in the north the youngestunit (2) is also present with an age of 129 � 13 ka.
The dating of the erosional unconformity observed betweenlayer F and Lower E at Hayonim Cave is of particular interest.According to the TL results, it is dated to w200 ka and thuscorresponds to the second part of OIS-7. This event wasmarked by significant flow of water within the cave. It mightthen be contemporaneous with one of the discontinuitiesrecorded in the sequence of Tabun cave that were alreadydescribed (Jelinek, 1982; Jelinek et al., 1973; Mercier et al.,1995b). An important discontinuity between Unit I dated to165 � 16 ka and Unit II dated to 196 � 21 ka (Mercier andValladas, 2003) reflects the introduction of the terra rosasoil due to the opening of the chimney. Earlier unconformitieswere noted in the stratigraphy of the inner chamber whereunits IIeVIII accumulated following the deposition of UnitIX (Jelinek, 1982, p.1370). Units IIeIX were attributed towhat was originally called layer D by Garrod. The TL datesof these units are as follows: Unit V, 222 � 27 ka; Unit IX,256 � 26 ka (Mercier and Valladas, 2003). Hence, the gap be-tween Hayonim Lower E and F (both corresponding in theirindustries to the generalized Layer D at Tabun) would fall be-tween Unit II and Unit V and not as mentioned in a previouspaper (Weiner et al., 2002).
If these discontinuities represent the same series of climaticevents, they seem to appear at the same position within the
1075N. Mercier et al. / Journal of Archaeological Science 34 (2007) 1064e1077
Table 2
List of the selected samples with their TL dates and associated errors
The mean ages and deviations for the indicated Mousterian layers were calculated by averaging the individual dates and errors, respectively. The standard devi-
ations are also given.a Sample no. 412 gave a young date in comparison to other results, something hard to explain.
Levantine Mousterian sequence. In Hayonim Cave, the bound-ary is situated between layers F and Lower E, both of whichcontain blade-rich Middle Paleolithic industries. Conse-quently, by comparing the chronologies and industries of Hay-onim and Tabun caves, one can conclude that the formersequence is likely more continuous than that of Tabun andthat the latter site was probably unoccupied during the firstpart of OIS-6. It should be noted that subsidence is more dra-matic in Tabun which also may explain gaps in the physicalstratigraphic record.
In sum, this study shows that the age of the Mousterian se-quence in Hayonim Cave ranges from about 115 ka ago to atleast 220 ka. It also shows that the dating of samples such asflints or teeth by dosimetric methods can be severely affectedby the proximity of mineral assemblage boundaries to thesamples and/or to the dosimeters used to measure the currentradiation dose. Without three-dimensional information on thedistributions of these mineral assemblages in the sediments,the ages of the stratigraphic units in Hayonim Cave wouldhave been determined with much less precision and confidencethan those presented here, and would have certainly been lessreliable.
Acknowledgements
We thank the National Science Foundation, Washington(grants awarded to O. Bar-Yosef and M.C. Stiner), the FrenchMinistry of Foreign Affairs (grants awarded to B. Vander-meersch), the Israel Science Foundation (grants awarded toS. Weiner), the American School of Prehistoric Research, Pea-body Museum, Harvard University, for funding the project atHayonim cave and the ensuing laboratory analyses. S. Weiner(the Dr Walter and Dr Trude Borchardt Professorial Chair inStructural Biology) is grateful for the generous donation byMr George Schwartzman.
