Revisiting Kokkinopilos: Middle Pleistocene radiometric dates for stratified archaeological remains in Greece

lable at ScienceDirect

Journal of Archaeological Science 57 (2015) 355e369

Contents lists avai

Journal of Archaeological Science

journal homepage: http : / /www.elsevier .com/locate/ jas

Revisiting Kokkinopilos: Middle Pleistocene radiometric dates forstratified archaeological remains in Greece

V. Tourloukis a, b, *, P. Karkanas c, J. Wallinga d

a Palaeoanthropology, Senckenberg Center for Human Evolution and Palaeoenvironment, Eberhard Karls Universit€at Tübingen, Ruemelinstr. 23, 72070Tübingen, Germanyb Human Origins Department, Faculty of Archaeology, Leiden University, 2300 RA Leiden, The Netherlandsc The Malcolm H. Wiener Laboratory for Archaeological Science, American School of Classical Studies, Souidias 54, Athens 10676, Greeced Soil Geography and Landscape Group, and Netherlands Centre for Luminescence Dating, Wageningen University, PO Box 47, 6700 AA Wageningen,The Netherlands

a r t i c l e i n f o

Article history:Received 12 November 2014Received in revised form4 March 2015Accepted 5 March 2015Available online 17 March 2015

Keywords:Middle PleistoceneLower PalaeolithicMiddle PalaeolithicLuminescencepIRIRGreeceMicromorphologyGeoarchaeological study

* Corresponding author. Palaeoanthropology, SencEvolution and Palaeoenvironment, Eberhard Karls Ulinstr. 23, 72070 Tübingen, Germany. Tel.: þ49 (0)707

E-mail addresses: [email protected]@ascsa.edu.gr (P. Karkanas), jakob.wallinga@

http://dx.doi.org/10.1016/j.jas.2015.03.0120305-4403/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

The red-bed site of Kokkinopilos is an emblematic and yet also most enigmatic open-air Palaeolithic sitein Greece, stimulating controversy ever since its discovery in 1962. While early research raised claims forstratigraphically in situ artifacts, later scholars considered the material reworked and of low archaeo-logical value, a theory that was soon to be challenged again by the discovery of in situ lithics, includinghandaxes. Here we present results of a latest and long-term research that includes geoarchaeologicalassessments, geomorphological mapping and luminescence dating. We show that the site preserves anoverall undisturbed sedimentary sequence related to an ephemeral lake, marked by palaeosols andstratigraphic units with Palaeolithic material that is geologically in situ and hence datable. Our studyresolves the issues that have been the source of controversy: the depositional environment, stratigraphicintegrity, chronological placement and archaeological potential of the site. Moreover, the minimum agesobtained through luminescence dating demonstrate that the lithic component with bifacial specimensconsiderably pre-dates the last interglacial and therefore comprises the earliest stratigraphically definedand radiometrically-assessed archaeological material in Greece. Kokkinopilos has served as a referencesite for the interpretation of all other red-bed sites in north-west Greece, therefore our results havesignificantly wider implications: by analogy to Kokkinopilos, the open-air sites of Epirus should notanymore be considered ‘by default’ as inscrutable palimpsests with limited archaeological potential;rather, these sites can be excavated and chronologically constrained. This realization opens up newprospects for future research in Epirus, an area that is the most prolific in Palaeolithic remains in Greece.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Middle Pleistocene archaeological evidence in Greece is sparseand the Greek Lower Palaeolithic involves lithic material that lacksradiometric dates and has been chronologically bracketed only inthe broadest of terms, based mainly on the inferred archaicmorphology of the artifacts and/or on usually insufficient strati-graphic correlations (Tourloukis, 2010; Tourloukis and Karkanas,

kenberg Center for Humanniversit€at Tübingen, Rüme-1 2974071.ebingen.de (V. Tourloukis),wur.nl (J. Wallinga).

2012). The Late Pleistocene record of Greece is also largelycomposed of undated, surface lithic assemblages from open-airsites with inadequate contextual information. From a total of ca.two hundred open-air Middle Palaeolithic sites and findspots only ahandful have been excavated (Harvati et al., 2009). In most cases,the material has been assigned to the Middle Palaeolithic on thebasis of typo-technological rather than chronostratigraphic criteria,whereas sites with radiometric dates are even fewer (but see Popeet al., 1984; Runnels and van Andel, 2003). As a consequence, thebackbone of the Greek Middle Palaeolithic is essentially restrictedto only five excavated cave sequences, dated with chronometricmethods (for reviews see Darlas, 2007; Runnels, 1995). On currentevidence, the earliest-dated stratified Middle Palaeolithic site andthe earliest-dated appearance of the Levallois technique in Greece

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369356

is placed around ca. 130 ka, most probably within the time-span ofthe transition from the penultimate glacial to the last interglacial(Theopetra Cave; Karkanas et al., in press; Valladas et al., 2007).Besides Theopetra, the rest of the sites are variously and/or tenta-tively dated to between ca. 120 and 40 ka, due to the paucity ofradiometric dates (Darlas, 2007; Elefanti et al., 2008; Kuhn et al.,2010; Harvati et al., 2013). In sum, apart from five excavated Mid-dle Palaeolithic cave sites, the bulk of the Lower and MiddlePalaeolithic records of Greece essentially lack supporting chro-nostratigraphic data.

The province of Epirus (north-west Greece) has so far yieldedthe largest Greek Palaeolithic database (Papagianni, 2000 andreferences therein). Furthermore, it is three sites from Epirus,namely the rockshelter of Asprochaliko, the cave of Kastritsa andthe open-air site of Kokkinopilos, which provided for the first timea framework for a stratified Palaeolithic sequence in Greece (Baileyet al., 1992). Most of the open-air Palaeolithic sites in Epirus areassociated with red sediments ethe so-called ‘red-beds’. One such‘red-bed site’, Kokkinopilos (Fig. 1), is the best-studied and mostimportant open-air site for the understanding of the early Palae-olithic of Greece, and yet its interpretation is surrounded by along-lasting controversy (Dakaris et al., 1964; Higgs and Vita-Finzi, 1966; Bailey et al., 1992; Runnels and van Andel, 1993;King et al., 1997; Papakonstantinou and Vassilopoulou, 1997;Papagianni, 2000; Runnels and van Andel, 2003; van Andel andRunnels, 2005). Kokkinopilos holds a central position in theongoing debate over the chronological and depositional rela-tionship between artifacts and sediments at all red-bed sites ofEpirus, because it has been used as a reference-site in all theproposed models (Papagianni, 2000). Moreover, it has yielded alarge collection of artifacts, which typologically span the entirePalaeolithic period and overall indicate an intensive and/or‘persistent’ occupation by human groups. Last but not least, thissite provided for the first time in Greece convincing arguments forin situ lithic occurrences, which can be attributed to the LowerPalaeolithic on the basis of chronostratigraphy -and not merely onlithic typology (Runnels and van Andel, 1993, 2003; Tourloukis,2009, 2010).

Our research at Kokkinopilos aimed to assess previous con-trasting interpretations on the stratigraphic integrity of the site,evaluate the controversy about the ‘in situ’ versus ‘reworked’ sta-tus of the finds and refine the local chronostratigraphy. Anothergoal was to evaluate the prospects of finding stratified lithic ma-terial, which is so rare in Greek early Palaeolithic open-air sites.Earlier discoveries of stratified artifacts (including bifaces;Tourloukis, 2009, 2010) served as a further reason that promptedus to carry out extensive fieldwork at this important and yetenigmatic site.

On one hand, our study confirms the existence of a palimpsestcharacter in the distribution of the surface finds (Bailey et al., 1992),but on the other hand it demonstrates the presence of in situ arti-fact concentrations, which can be resolved into a geochronologicalsuccession, albeit with a relatively low temporal resolution. Muchof the disagreement in the conclusions of previous investigations(e.g. Bailey et al., 1992 versus Runnels and van Andel, 1993) arosefrom inadequate understanding of the overall geometry of thepedo-sedimentary units and their boundaries. As shown below,geomorphological mapping of the stratigraphic contacts, combinedwith micromorphological analyses and a re-assessment of thegeological sequence by focusing on stratigraphic markers, revealsthe presence of stratigraphically defined, potentially datable lithicassemblages. In effect, this approach allowed us to acquire reliableradiometric minimum ages, which are thus far the earliest to bepublished for Middle Pleistocene archaeological remains fromGreece.

2. Summary of previous research and interpretations

There have been controversial views about the origin, deposi-tional context and age of the sediments (for extensive reviews seePapagianni, 2000; Runnels and van Andel, 2003; Tourloukis, 2010).In short, Higgs and colleagues first suggested an aeolian origin ofthe deposits and postulated an accumulation during the ‘LastGlaciation’ (Dakaris et al., 1964), but afterwards they argued for analluvial context (Higgs and Vita-Finzi, 1966). King and Bailey (1985)originally accepted the alluvial origin, but later Bailey et al. (1992)proposed a colluvial deposition of the sediments and claimedthat they are at least of Middle Pleistocene age, if not much older;two TL dates of >150 ka were regarded as inconclusive, if notsuggesting that the sediments at both sampling localities (units Aand B; see below) are older than this age. Runnels and van Andel(1993, 2003; van Andel, 1998) advocated that the deposits of Kok-kinopilos -and of the other red-bed sites- consist of terra rossa,which has been re-deposited in a runoff-collecting karst depression(a polje). The latter researchers demonstrated that the sedimentsare composed of clay (50 to >90%), produced by the dissolution oflimestone and washed down from the flanks of the depression, andsilt (5e30%) of windblown origin, deriving probably from theSahara. Complete bleaching during aerial transport renders the siltcomponent suitable for TL and IRSL dating and enabled theacquirement of luminescence dates for Kokkinopilos and other sitesof Epirus (Runnels and van Andel, 2003; Zhou et al., 2000).

Kokkinopilos has three main stratigraphic units, A, B and C frombottom to top, and a palaeosol caps the entire sequence (Fig. 2).Most of the lithic material has been found on the surface of units Band C; besides a small Upper Palaeolithic component, it has beendescribed as ‘Levallois-Mousterian’ with bifacial leafpoints andpreponderance of racloirs (P. Mellars, in Dakaris et al., 1964). Theraw material of the artifacts is a bluish-gray nodular flint and thespecimens are heavily patinated but in sharp condition; in contrast,lithic finds ascribed to the Bronze Age and found on the surface‘topsoil’, are made on a different type of flint and are not patinated.

Higgs and colleagues opened test-trenches in two out of thir-teen locations, where they claimed to have identified artifacts insitu (“chipping floors”; Dakaris et al., 1964). These claims were af-terwards contested by Bailey et al. (1992, 142) who argued that“none of the artifacts recovered from Kokkinopilos can bedemonstrated to be geologically in situ”, and that “[the artifacts] donot date the accumulation of the main body of red clays at all, butpostdate them by an unknown interval”; importantly, they alsoargued that the same point applies to the other open-air red-bedsites in Epirus. Only a year later, the latter view was in turn chal-lenged by Runnels and van Andel (1993), who discovered anAcheulean handaxe as well as other patinated artifacts of non-Levallois morphology, stratified in situ within deposits of unit B(Fig. 2); the artifacts were separated by clay matrix from each otherand there was no size sorting or any mixing with unpatinatedspecimens, as it would be expected if they were included in the fillof an erosional gully. Thermoluminescence dating of the uppermostcapping palaeosol at ca. 91 ka allowed the researchers to estimatethe age of the handaxe-bearing layer at ca. 150e250 ka, byextrapolating sedimentation rates corrected by the variations of thesilt/clay ratio (Zhou et al., 2000; Runnels and van Andel, 2003).

3. Materials and methods

Our work in the field included geomorphological mapping,topographic, stratigraphic and sedimentological analyses, evalua-tion of the tectonic activity, geo-archaeological assessments of thelithic artifacts and their contexts, sedimentological sampling formicromorphological analysis and luminescence dating.

Fig. 1. Panoramic views of Kokkinopilos. a) The main part of the site is visible in the center of the picture; to the right is the limestone ridge, which runs parallel to a fault andseparates the site from Louros River. b) The badland landscape of Kokkinopilos; in the center of the picture, note the palaeosol horizon formed on unit B, close to the contact withunit C (visible in the background).

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369 357

Four intact blocks of sediment were sampled for micromor-phological analysis. The samples were oven dried at 50 �C andimpregnated with polyester resin diluted with styrene. Largeformat thin sections (75 � 50 mm) were prepared and studiedunder petrographic microscope. Descriptive terminology of thinsections follows that of Stoops (2003) and Courty et al. (1989), butstandard sedimentary terminology is also used for non-pedogenicfeatures.

Three sediment samples were analyzed at the NetherlandsCentre for Luminescence Dating (hosted by Delft University ofTechnology at the time). KPL-1 was sampled from the locationwhere a biface was found in situ, stratified in sediments of unit C(Figs. 2 and 3: I; Figs. 9 and 10b; see also below). Sample KPL-2 wastaken from the uppermost part of unit C (Fig. 3: II), directly belowthe palaeosol that caps the entire sequence and has been TL-datedto 91 ± 14 ka by Zhou et al. (2000). KPL-3 was sampled from de-posits of unit B (Fig. 3: III) and at a location that is topographicallyand stratigraphically close to the findspot of the ‘Micoquian’ han-daxe found by Runnels and van Andel (1993).

Initial grain-size choice for testing of optically stimulatedluminescence characteristics followed Runnels and van Andel(2003), who report that quartz and feldspar are most abundant inthe 10e64 mm grain-size fractions. An investigation of the bulk<63 mm fraction did not show any luminescence signal from quartz(r > 2.62) or potassium-rich feldspar (r < 2.58). Several other grain-size fractions were prepared for further testing (45e63 mm,63e90 mm and 90e180 mm), of which the quartz (90e180 mm) andfeldspar (63e90 mm) did show a luminescence response to opticalstimulation. However, the quartz optically stimulated

luminescence (OSL) signal was in saturation, and the feldsparinfrared stimulated luminescence (IRSL) signal is known to be un-stable (anomalous fading; e.g. Wallinga et al., 2007). Several addi-tional approaches were then attempted to date the samples. Theuse of the thermally stimulated luminescence (TL) signal of quartzwas investigated, but only the unstable 110 �C peak showed anyresolution above background. Despite use of several different filtersto isolate the datable 375 �C peak, it could not be resolved from apeak with broad background and blackbody irradiation occurringaround 400 �C. A recently proposed approach for extending lumi-nescence age ranges using the quartz thermally transferred-OSL(TT-OSL) signal (e.g. Wang et al., 2007; Jacobs et al., 2011; Dullerand Wintle, 2012) was also tested. Unfortunately, TT-OSL signalswere too dim to use for dating. After these failed attempts to datequartz minerals we switched attention to feldspars.

Recent research has indicated that a more stable signal can beobtained by measuring the infrared signal at elevated temperature,after a low temperature infrared exposure (post IR IRSL, or pIRIR;Thomsen et al., 2008). pIRIR methods isolate a more stable lumi-nescence signal, and have been shown to be hardly affected byanomalous fading (Buylaert et al., 2012; Thiel et al., 2011). Weadopted a protocol where the pIRIR signal was measured for100 s at 290 �C after an infrared bleach for 100 s at 50 �C. A preheatof 60 s at 320 �C was used, and each SAR cycle ended with anelevated temperature infrared bleach (40 s at 330 �C). The net signalused for analysis was obtained by subtracting the backgroundsignal (80e100s) from the initial signal (0e10s). Internal tests of theSAR procedure were used to accept only valid data (recycling ratiowithin 15% of unity, test-dose known to within 10%). As the pIRIR

Fig. 2. Geomorphological map and stratigraphic column of Kokkinopilos, the latter modified after Tourloukis, 2010. The stratified biface of Fig. 9, and samples KPL-1 and MKPL-1, arelocated at log I; KPL-2 is located at log II; KPL-3 and MKPL-2 are located at log III, which lies stratigraphically and topographically close to the findspot of the biface found by Runnelsand van Andel (1993). See also Fig. 3 for more details.

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369358

signal may be affected by incomplete resetting, we subtract a doseof 30 ± 15 Gy from the dose estimates. This estimate covers allvalues reported in the literature for modern analog samplesmeasured with pIRIR methods (Buylaert et al., 2012; Alexandersonand Murray, 2012).

The natural dose rate is calculated from the radionuclide con-centration of sediments surrounding the sample, the depth of thesample below the surface and thewater and organic contents of thesample. We determine the radionuclide concentration by high-resolution gamma-ray spectroscopy (Murray et al., 1987). Severalradionuclides are measured for the Uranium and for the Thoriumseries to check whether they are in secular equilibrium. For the Thseries we measure Ac-228, Pb-212 and Bi-212. For the U series wemeasure Th-234, Pb-214, Bi-214 and Pb-210. Because the latterprobes the decay chain below the Rn step, it allows us to evaluatethe Rn escape in the natural environment. K-40 is measureddirectly. For the K-feldspar extracts internal dose rate contributionsare included, assuming a potassium concentration of 12.5%(Huntley and Baril, 1997) and an Rb concentration of 400 ppm(Huntley and Hancock, 2001). Spectral data are converted to ac-tivity concentrations and infinite matrix dose rates (Gu�erin et al.,2011). The natural dose rate was calculated from the infinite ma-trix dose rate using attenuation actors given by Mejdahl (1979). Acontribution from cosmic rayswas included based on the depth andburial history of the sample, following equations presented byPrescott and Hutton (1994). A correction was made for attenuation

of the dose rate by water using the attenuation factors given byZimmerman (1971). Present-day water contents of the sampleswere used (around 30% by weight), with a generous uncertainty of25% to take into account past variations in water content. Resultingdose rates for the feldspar extracts are around 3 Gy/ka, which issimilar to dose rates reported for similar deposits by Zhou et al.(2000).

4. Results

4.1. Stratigraphic, sedimentological, geomorphological andgeoarchaeological assessments

Research permit limitations prevented us from conducting ex-cavations or test pits and, besides two bifacial tools mentionedbelow, we were not allowed to collect any lithic material (previousresearchers faced similar permit issues: Runnels and van Andel,1993, 2003). A small-scale study of the finds collected by otherresearchers was conducted in the Archaeological Museum atIoannina.

Kokkinopilos is situated at 120e150 masl in a valley to the westof the Louros River, fromwhich it is separated by a limestone ridgethat runs parallel to a fault. Louros, which flows only 150 m to theeast of the site, is incised more than 40 m into the limestonebedrock. The Kokkinopilos deposits consist of 30e40 m-thickconsolidated red clayey silts and silty clays and are currently being

Fig. 3. Sediment logs from selected parts of the site, showing the stratigraphic positions of sedimentary units, palaeosols, artifacts, dating samples and micromorphological samples.The locations of the logs are indicated in Fig. 2.

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369 359

rapidly eroded in a network of gullies that create a badland land-scape (Fig. 1b).

Unit A rests on the rugged limestone bedrock (Fig. 4) and isuniformly deep red (Munsell color chart: 2.5YR 4/6) with very fewgray streaks. Sediments of this unit are scarcely exposed as local-ized patches that crop out in the bottom of deep gullies, wherebedrock is also visible (e.g. in the south-western part of the site).Where detectable, the contact of unit A with unit B appears to begradual and irregular. No artifacts have ever been found in this unitand our investigations confirm that it is probably archaeologicallysterile.

Unit B is ca. 15 to 25m-thick and yellowish red (7.5YR 4/4 to 5YR6/8, in places grading to 2.5 YR 3/6), displaying mottled bands andtapering down gray streaks that represent drab-haloed root traces(Fig. 5a). This unit is better exposed along a North/Northwest-South/Southeast axis in the central part of the site, where itforms an ‘even’ surface of relatively low gradients (Fig. 4). Else-where, unit B sediments crop out inside the gully network andgenerally occur in relatively low/lowermost elevations. The sedi-mentary sequence of unit B is interrupted by at least two truncatedmature palaeosol clay horizons (Bt). In the southern part, the up-permost palaeosol is heavily dissected by unit C and is character-ized by an upper light gray gleyed zone and a dense network of lightgray drab-haloed root traces, which penetrate deeply into thepurple red clay subsurface Bt horizon and continue down to the redunderlying sediment. The drab-haloed root traces often preserve acentral thin core of decayed organic matter. At a microscopic scale,

the material and texture of the drabs and red matrix differ only incolor, and this confirms that they are depletion redoximorphicfeatures (similar to those of Fig. 5a); the red matrix is silty clay,characterized by black ironemanganese concentric nodules andimpregnative features. A locally striated fabric, as well as thick,intact and assimilated clay pressure faces (slickensides) is oftenpresent. The same palaeosol locally displays alternating horizontalgray and purple layers associated with black ironemanganesestaining. An unambiguous archaeological horizon was identifiedinside the above-described palaeosol of unit B (Fig. 6; Fig. 3: X).Artifacts are embedded in a ca. 40 to 60 cm-thick gleyed yellowish-orange horizon, which, apart from lithics, lacks any other type ofcm-scale clastic material (e.g. gravels). There is no size-sorting orany orientation of the artifacts, which are matrix-supported andcan be traced laterally over several meters along the palaeosoloutcrop. The material consists of patinated large flakes, debitageproducts, and retouched tools (mainly denticulates and notchedpieces) without Levallois traits. Besides those artifacts in thepalaeosol horizon, lithics were found at various other places asisolated occurrences stratified in situ inside unit B: for instance, apatinated flake was found while cleaning a profile for sample KPL-3(Fig. 3: III).

Moreover, a biface was found lying on the surface (Tourloukis,2009), close to the place where Runnels and van Andel (1993)discovered the Acheulean handaxe and at about the same strati-graphic level. The specimen is patinated, made on bluish-gray fine-grained flint, it has a cortical base and typologically can be

Fig. 4. Cross-sections of the geological sequence.

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369360

described as “amygdaloid �a talon” ea typical Acheulean biface(Fig. 10a; Deb�enath and Dibble, 1994). It was associated withrecently reworked deposits, most probably deriving from unit Band most likely pertaining to the fill of an erosional gully. None-theless, the artifact is in mint condition, which suggests minimumtransport.

The contact between units B and C follows a line that for themost part is irregular (Fig. 4). In the central part of the site, thecontact has an average dip of ca. 5.5% and appears at an averageelevation of ca. 150e140 masl, but in the northeast area close to thelimestone ridge it ranges from ca. 145 down to less than 135 m asl.Particularly, it becomes highly undulating and it forms deeplocalized troughs. At most localities, the contact follows palaeosolsurfaces.

Fig. 5. a) Thin section from micromorphological sample MKPL-4 (stratigraphic position insediment. Note that the striated clay features of the red sediment continue uninterrupted isample MKPL-3 (cf. Fig. 3: VI): detail of the finely laminated clay sediment showing silty c

Unit C is ca. 10e13 m-thick, it has a reddish-brown color (5YR 3/6 to 10R 4/8) and is marked by thin gray layers, gray, red and orangemottles and sub-vertical gray drab haloed root traces similar tothose of unit B but more abundant, which create a color-bandingmore conspicuous than that of unit B. Often drab burrows andmore rarely soil cracks are also observed. However, remnants ofpurple-red crescent burrow infillings are locally depleted in iron.Drab halos are complex, but, as in unit B, they are of the samematerial and texture as the red matrix (Fig. 7c). They often preservea complex depletion pattern with a central depleted zone and twosurrounding halos: the inner one is diffuse with remnants ofdepleted orange red matrix, while the outer one is sharp with oc-casionally a dark iron concentration rim (Fig. 7d). Black iron-emanganese nodules and coatings were also observed (Fig. 7a). A

Figs. 2 and 3: VIII): detail of a drab halo formed by iron depletion of the red silty claynside the drab halo but they are discolored; b) Thin section from micromorphologicallay laminae capped by very thin organic-stained clay horizons.

Fig. 6. View of the upper palaeosol formed in unit B (a), showing the lithics that crop out in situ (b). In the center-left part of (a), note that the lithics (white, due to heavy patination)erode out of the palaeosol all along the outcrop.

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369 361

complex, mainly porostriated fabric with clay coatings and slick-ensides characterizes the silty clay matrix.

In the central part of the site and to the west of the main easterngully that runs parallel to the limestone ridge, there is a gray layerresting on the base of unit C; it is currently exposed horizontallyand its boundaries are defined by an iron-rich hardpan (Fig. 8a;Fig. 3: V). This layer is formed on the basal part of unit C in this areaand can be followed laterally to the deep paleo-troughs separatingunits C and B. A clear concentration of patinated artifacts wasobserved eroding out of this layer (Fig. 8b), which rests on analmost flat surface, topographically above and away from thenearest gullies. The mint preservation, surface condition, non-sorted spatial distribution and half-buried position of the artifactsinside undisturbed sediments precludes any clustering due torecent water action and instead indicates low-energy uncoveringand minimum horizontal displacements by mild erosional pro-cesses, such as deflation or winnowing. This observation wasconfirmed during repeated visits at Kokkinopilos (2007, 2008, 2009and 2011) and it can be confidently ascertained that this is ageologically in situ artifact scatter. Apart from this layer in the baseof unit C, surface artifact scatters were identified at a few otherlocalities of the same unit.

Finely stratified, horizontal to sub-horizontal laminated sedi-ments are also evident elsewhere inside this unit; they often fillpaleo-troughs and some of them are characterized by meter-scalefeatures indicative of past small-scale and localized syn-sedimentary slumping events. Under the microscope, the lami-nated sediments consist of thinly interlayered sitly clay and claycouplets (Fig. 7). The silty clay layers are generally homogeneous

and thicker, capped by very thin (a few tens of microns) organic-stained clay horizons (Fig. 5b).

In the upper parts of unit C, the sediments grade to over-thickened gleyed mottled horizons. A biface was found lying hori-zontally with half of its surface buried by the sediments, embeddedwithin non-reworked deposits, in the upper part of unit C and at analtitude of ca. 140 m asl (Fig. 9; Fig. 3: I). The artifact is made on aflake-blank and it displays a flat bifacial retouch, whereas on oneside, large parts have been left unretouched and there is a possiblebreakage on the left lateral edge; the base looks as if it has beendeliberately left unworked, or, alternatively, it broke in the processof manufacture and it was then left unretouched. The tip is trian-gular in section, one cutting edge is sinuous, while the other isessentially straight, largely because of the removal that occursthere. Metrical data classify it to Bordes' ‘thick bifaces’ with acordiform aspect (Fig. 10b). The specimen is in a fresh condition,with its cutting edges still sharp and the ridges of the flake-scarsclearly visible, albeit slightly worn locally. It is heavily patinatedand displays red stains on both surfaces due to long-time contactwith Fe- and Mg-oxides of the clay matrix. Overall, there are hardlyany signs of weathering, polishing or abrasion, and the generalappearance of the artifact disproves the case of significant rolling,neither by water nor by large-scale downslope movement. Glei-zation of the matrix occurred after the deposition of the artifact,and the drab halos wrap around the specimen ea fact that furthersupports the in situ character of the find. As noted earlier mostemphatically (Tourloukis, 2009, 2010), biface typology alone cannotbe used for chronological ascriptions: for us, whether this bifacemorphologically looks like a “Keilmesser” or not (Ligkovanlis, 2014)

Fig. 7. Micromorphological samples MKPL-3 (a, b) and MKPL-4 (c, d). a and b: Resin impregnated slab (a) and thin section scan (b) showing red, finely laminated sediment withblack manganese impregnative pedofeatures along cracks; c and d: resin impregnated slab (c) and thin section scan (d) showing red, finely laminated sediment dissected by graydrab halos. Detail of the structure of the drab halo is shown in the thin section scan where a central depleted zone is followed by a surrounding diffuse halo; the outer contact issharp. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369362

is of secondary importance; the value of this specimen lies mostlyin its chronostratigraphic significance, not its typological class.

More artifacts, all patinated and mainly unretouched flakes ordebitage products, were found stratified in situ, in taphonomic

conditions and positions similar to that of the latter biface, atvarying depths inside unit C and usually as isolated occurrences. Aswith the case of similar finds inside unit B, our permits preventedus from assessing whether those artifacts are indeed ‘isolated

Fig. 8. a) General view of the gray layer at the base of unit C; b) Artifacts exposed in situ.

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369 363

finds’, or belong instead to broader ‘archaeological horizons’ (suchas the one mentioned above for the palaeosol of unit B) or ‘clusters’(such as that exposed horizontally in the gray layer at the base ofunit C).

Another unit, which is designated here as ‘unit D’, overlies unitC. It occurs only patchily, mostly in depressions and/or lower ele-vations, such as those adjacent to the limestone ridge in the easternand south-eastern part of Kokkinopilos where it attains itsmaximum thickness (ca. 2 m; Fig. 3: VII). In these areas it is formedby finely laminated sediments (Fig. 11a). In the central-north part,the same unit is observed to form the base of the sequence-cappingpalaeosol (see below), but it is pedogenically altered and retains

Fig. 9. The biface shown in Fig. 10b as it was found upon discovery. Inset: closer view.

only relict bedding. At other localities, it also includes sub-horizontal, alternating gleyed gray and red thin layers associatedwith ironemanganese staining. It is separated from the underlyingunit C by a sharp erosional contact and is overall horizontal orgently undulating.

A highly mature palaeosol, which caps the entire sequence, hasbeen described in detail and TL-dated to 91 ± 14 ka by Runnels andvan Andel (1993, 2003). The Bt horizon is clayey and red (2.5 YR 3/6), it has a sticky, plastic consistency and a columnar structure withblocky and angular peds, covered by thick clay films. At places, itreaches a thickness of >80 cm. Overall, these characteristics suggestan old age for this palaeosol, in accordance with previous soilstudies in Epirus, which used pedogenic weathering data asrelative-age indicators (e.g. Woodward et al., 1994). What has beenleft out of account in previous reports on Kokkinopilos is that, whilethis palaeosol occurs mainly on expense of unit C (e.g. Fig. 3: IX),there are also parts of the site where it overlies unit D (e.g. in thesoutheast; Fig. 3: VII). The sequence-capping palaeosol can betraced from the central-north part of the badlands down toconstantly lower elevations towards the east/southeast, followingthe general inclination of the entire badland. In places, the Bt ho-rizon of this palaeosol is overlain by a thin, immature soil horizon(A), which forms the current ground-surface and supports themodern vegetation (Fig. 3: II, IV, VII, XI); unpatinated artifacts thatprobably belong to the Bronze Age are being found on and occa-sionally inside this ‘topsoil’.

The four depositional units, A to D, have accumulated in properstratigraphic order, forming a sequence that is bounded on top byan unconformity (the highly mature palaeosol) and is markedthroughout its depth by palaeosols and pedogenically alteredstructures, which indicate depositional breaks and subaerially

Fig. 10. a) Biface found on the surface of unit B; b) Biface found stratified in situ in unit C. Reproduced from Tourloukis 2010.

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369364

exposed and weathered surfaces (Fig. 11b). There is no indication ofany stratigraphic reversal or large-scale pedo-sedimentary distur-bance that would caution for significant reworking of the units.

Fig. 11. a) Exposure of unit D at the south-eastern part of the site. b) Part of thegeological sequence showing units B, C and palaeosols.

Unit C deposits are consistently encountered on top of unit B andusually at higher altitudes of topographically elevated places.Reworked sediments do exist, but they are spatially limited indiscernable entities associated with modern gully fills. Interestingly,these fills bear limestone clasts from the erosion of the limestoneridge and bedrock, a consequence of the relatively high-energydeposition occurring in gullies after the tilting of the area, theopening of the polje and the establishment of a new drainagesystem (see also discussion below). These fills are in markedcontrast to themain low-energywetland depositional environmentof Kokkinopilos, in which natural limestone clasts are entirely ab-sent. Consequently, the assertion that “none of the artifacts isgeologically in situ” (Bailey et al., 1992), can now be confidentlyrejected: there exist artifact concentrations stratified in situ andthese can be dated by dating their sedimentary matrix.

4.2. Luminescence dating results

Results of pIRIR analysis showed that the feldspar signal plottedhigh up the dose response curve for nearly all aliquots (Fig. 12).Many single-aliquot dose response curves show continued rise atvery high doses, and fitting thesewith a first order single saturatingexponential yielded unsatisfactory fits. For this reason, we adoptedthe recent suggestion of Guralnik et al. (in press) to fit doseresponse curves with a general order single saturating exponentialfunction. Fit parameters were determined from combined data ofall accepted aliquots for each of the samples, using an unweightedfitting procedure of the sensitivity corrected pIRIR responses fordoses up to 2088 Gy. Wintle and Murray (2006) suggested thatreliable estimation of equivalent dose is possible up to two times D0level of the saturated exponential fit equation of the laboratorydoseeresponse curve. Although this threshold is arbitrary (see also

Fig. 12. Single-aliquot sensitivity corrected pIRIR signals as a function of laboratorydose (black dot) for each of the samples. Laboratory data are fitted with a general ordersaturating exponential function following Guralnik et al. (in press); the 2*D0 level forthe dose response curve (DRC) is indicated on the graph. Note that nearly all singlealiquot natural pIRIR signals (triangles) plot close to the saturation level and beyondthe 2*D0 threshold. Insets show the natural, regenerative and test-dose pIRIR signalsfor a representative aliquot of the sample.

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369 365

Galbraith and Roberts, 2012), it provides a conservative means toavoid over-interpretation of luminescence ages with natural signalsclose to saturation. For each of the samples, average equivalent dosevalues were above this threshold (see Fig. 12), and for this reasonthe two times D0 value is used as minimum estimate of theequivalent dose (i.e. the true equivalent dose is greater than thisvalue, but cannot be adequately determined). To obtain a conser-vative estimate of the minimum burial dose, the residual dose of30 ± 15 Gy was subtracted from the minimum equivalent doseestimate. This procedure is similar to that applied by Joordens et al.(2014) in a recent study. Results are shown in Table 1, and vary from504 ± 70 Gy to 680 ± 52 Gy.

To obtain minimum age estimates, the minimum burial dosesare combined with the dose rate estimates. The determined doserates for the feldspar extracts are around 3 Gy/ka, which is similarto dose rates reported for similar deposits by Zhou et al. (2000). Thefinal minimum age estimates range from 172 ± 25 ka for the up-permost sample, to 206 ± 19 ka for the lowermost sample. Detailsare provided in Table 1. The data presented in Fig. 12 suggests thatthe deposits may in fact be much older than the minimum ages, asthe vast-majority of single-aliquot pIRIR natural signals plotsignificantly above the adopted threshold. Unfortunately, presentlyavailable methods do not allow us to make quantitative estimatesof the true burial age.

The obtained pIRIR ages are stratigraphically in agreement withthe results of Zhou et al. (2000), who present a TL age estimate of91 ± 14 ka for a sample from the capping palaeosol. The older ageobtained for sample KPL-2, taken just below this palaeosol, mayindicate a large hiatus represented by the palaeosol. Alternatively, itmay reflect methodological or geological issues. The TL sample ofZhou et al. (2000) was taken from a well-developed soil, at a depthof less than a meter below the surface. It is likely that some of thegrains in this layer were exposed to light due to bioturbation pro-cesses, hence the age may underestimate the time of deposition ofthe sediments. In addition, the previous TL date was obtained usingmethods that may have been affected by the instability of thefeldspar luminescence signal (anomalous fading). The original pa-per by Zhou and colleagues does not address this issue and it seemsthat no checks of signal stability were carried out in this otherwiserigorous study.

The pIRIR minimum age results indicate that the sediments thatare stratigraphically below sample KPL-2, i.e. immediately belowthe sequence-capping palaeosol, are older than 172 ± 25 ka.Accordingly, both of the artifact-bearing units of Kokkinopilos,namely unit C and unit B, date to the Middle Pleistocene: unit C isolder than 172 ± 25 ka and unit B is older than 206 ± 19 ka. Thisconclusion is essentially in agreement with the chronologicalbracketing of 150e250 ka, which Runnels and van Andel (1993,2003) estimated for the ‘Micoquian handaxe’ and the rest of theartifacts that they recovered in situ from unit B. An age older thanca. 172 ka agrees well also with the age of the artifact-yieldingsediments at Morfi eanother red-bed site with a sequence andsedimentary evolution similar to that of Kokkinopilos. In the poljeof Morphi, a 12-m. thick red-bed deposit, interspersed withpalaeosols, is very similar to unit B of Kokkinopilos and has yieldedthousands of artifacts, including Mousterian pieces; this depositimmediately overlies a 2-m thick tephra layer that has been AreArdated to 374 ± 7 ka (Pyle et al., 1998).

In sum, our dating assays generally concord with results fromprevious research (Runnels and van Andel, 1993, 2003; Pyle et al.,1998) and clearly reject the view that the accumulation of thegeological sequence is chronologically unrelated to the exploitationof the polje by human groups (Bailey et al., 1992). The pIRIR agesprovide a chronological evaluation for the Kokkinopilos materialand firmly demonstrate for the first time in Greece the existence of

a lithic component that includes bifacial specimens and pre-datesthe last interglacial. Along with other lithic material, there are bi-faces from unit C (such as that shown in Figs. 9 and 10b) and othersfrom unit B (such as that found by Runnels and van Andel), which,according to our pIRIR results, are in fact older than ca. 172 ka and206 ka, respectively.

Table 1Summary of feldspar pIRIR dating results.

Sample Location Burial dose estimate Dose rate estimates Aged

(ka)NCL Client X Y Deptha

(m)2*D0 level(Gy)

n Burial doseb

(Gy)External b(Gy/ka)

External g(Gy/ka)

Cosmic(Gy/ka)

Total dose ratec

(Gy/ka)

NCL-8309073 KPL-2 39_15_26 20_50_34 1.75 534 ± 68 4 >504 ± 70 1.21 ± 0.10 1.15 ± 0.06 0.20 ± 0.01 2.93 ± 0.13 >172 ± 25NCL-8309072 KPL-1 39_15_28 20_50_35 3.50 548 ± 62 6 >518 ± 64 1.44 ± 0.12 1.35 ± 0.07 0.19 ± 0.01 3.33 ± 0.14 >156 ± 21NCL-8309074 KPL-3 39_15_29 20_50_39 11 710 ± 50 6 >680 ± 52 1.49 ± 0.12 1.32 ± 0.07 0.12 ± 0.01 3.30 ± 0.15 >206 ± 19

a Burial depths were estimated by projecting the stratigraphic position of the capping palaeosol at the sampled locations and after correction of the dip.b An estimated remnant dose of 30 ± 15 Gy is subtracted from the 2*D0 dose, to estimate the minimum burial dose.c The total dose rate also includes a contribution from a particles (0.06 ± 0.03 Gy/ka) and a b dose contribution from internal 40K and Rb (0.30 ± 0.03 Gy/ka).d The age reported is the minimum age obtained from the adopted threshold level for reliable dose estimation.

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369366

5. Discussion

Kokkinopilos is an enclosed, fault-bounded tectonic depression,which used to host a relatively shallow lake that was periodicallydrying out either locally or entirely. Subaquaeous sedimentation isevidenced in the fine laminated layers that occur inside paleo-troughs in the lower part of unit C, and in unit D. Their occasion-ally rhythmic bedding is interpreted as sedimentation of suspen-sion clouds in stagnation basins (Reineck and Singh, 1980,123e129). The gray layer at the base of unit C is interpreted as theresult of groundwater gleying found in peaty waterlogged soils(Retallack, 2001, 90). The horizontal gleyed layers in parts of somepalaeosols of units B and C could also represent local ponding andanaerobic decay of organic-rich layers. Drab halo root traces, mot-tles and ironemanganese pedofeatures are redoximorphic features,which resulted from changes in the redox conditions of the soil inresponse to fluctuation in water saturation (Vepraskas et al., 1992;Kraus, 2002). Indeed, drab haloes are formed around root traces inclayey, periodically waterlogged soils (Retallack, 2001, 18). Thepresence of deep penetrating roots, drab burrows and generalbioturbation is related to surface gleying (pseudogley) and non-permanent waterlogged soils (Retallack et al., 2000; Retallack,2001, 18). However, some remnants of oxidized burrows areprobably the result of groundwater gley and permanent water-logging (PiPujol and Buurman, 1994). In general, Unit B palaeosolsare rather the product of surface gleying (pseudogley), whereasthose of unit C represent a complex result of both surface andgroundwater gleization (PiPujol and Buurman, 1994). Interestingly,Macleod and Vita-Finzi (1982) considered some of the drab halos atKokkinopilos as pseudogleying in seasonal waterlogged soils,whereas Runnels and van Andel (1993) recognized true fossil gley.A burial gley-origin is likely for some palaeosol horizons in unit B,where entire surface horizons provide evidence of drab derivingfrom dispersed organic matter and its anaerobic decay (Retallack,2001, 9). Nevertheless, we believe that burial gleying is onlyresponsible for accentuating the contrasting color of the matrix andthe drab-haloed root traces, and not for their original formation.Indeed, the depletion of the drabs in free iron oxide content asreported by Macleod and Vita-Finzi (1982) supports the identifi-cation of an open-system alteration during soil formation and not aclosed-system chemical reduction after burial (Retallack et al.,2000).

In this generalized scheme of alternating wet and dry deposi-tional environments, such alternations would have had differentand varying life-histories in both space and time: while at one lo-cality a deep depression would be able to retain water for a longtime-span and/or in a large spatial extent, draining would havebeen more successful in the temporal and/or spatial dimension foranother, shallower locality; this is evident today in the existing,active poljes of Epirus (Runnels and van Andel, 2003; Tourloukis,2010). We therefore envisage Kokkinopilos as a complex setting

of marshy spots neighboring dry surfaces, with the dry/wet char-acter of its localities oscillating seasonally and/or in longer-term,climatically-driven periodicities. Coupled with erosional intervals,such a dynamic picture explains not only the presence of cumulicsoil horizons (such as these of unit C) and truncated soil facies ofvarying maturity, but also the difficulty in assessing the relativecontribution of pedogenic versus sedimentary processes in some ofthe identified geological entities, such as parts of unit D. Pedo-genesis occurred in submersed settings as well as on exposedsurfaces, and the alteration of oxidizing and reducing conditionsresulted inmottles of gleyed horizons occurring today inter-beddedwith material reddened by iron-bearing minerals.

Unit A accumulated on the corroded limestone bedrock duringthe earliest stages of the polje, when the depression was still tooshallow to retain spatially extended and/or temporally sustainedwater bodies. Sediment supply was high and sedimentation ratesexceeded those of soil formation, hence the general absence ofdiscernable or mature palaeosol horizons in this unit. On currentevidence, human groups did not exploit the polje during thedeposition of unit A.

Accumulation of unit B occurred in a periodically wet environ-ment with a strongly fluctuating water table. Two discontinuouslypreserved palaeosol horizons indicate relatively prolonged in-tervals of dry conditions and subaerially exposed surfaces. Homininpresence during this stage is attested by stratified artifact occur-rences, be they isolated or clustered, as well as by artifact concen-trations stratified inside palaeosols.

An erosional episode and sedimentary hiatus is indicated by theundulating contact between units B and C, as well as by a palaeosolhorizon that is at places preserved, separating the two units.Sedimentation resumed with unit C accumulating on this surfaceandmainly underwater, in relatively deep troughs and depressions.The gray layer exposed patchily at the base of unit C indicates apeaty water-logged area, of which the periphery is bounded by aniron-rich hardpan that has been desiccated and consolidated.Further up in unit C, there is evidence of an ephemerally water-logged environment, but periods with more prolonged water-logging probably occurred as well. The presence of humans isagain attested during this stage (unit C), in the form of bothpotentially widely spaced, individual occurrences (e.g. the biface ofFig. 9) and artifact clusters (e.g. the find scatter of the gray layermentioned above).

Another erosional interval and major hiatus is indicated by thesharp contact between units C and D (Fig. 11a), as well as the up-permost palaeosol that is formed either on expense of unit D ordirectly on unit C. Relatively deep underwater sedimentationresumed but only locally during the formation of unit D. Finally, it ismost probably before the formation of unit D that a tectonic eventaffected the polje and tilted its main part towards the southeast(Figs. 1 and 4); due to down-faulting, the deposits of units B and Ccurrently have an average dip of 5.5�. Unit D appears to have not

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369 367

been affected by this movement, but as this unit occurs only inisolated pockets, it is hard to observe the geometry of its contacts.Tectonic activity disrupted the hydrographic system and forced thestreams to incise and dissect the polje, draining the water bodiesand creating a gently inclined topography. The uppermost palae-osol has been formed at the expense of this inclined topographyand therefore postdates this event, but also demarcates a long hi-atus. Nonetheless, the accelerated erosion, which exposed deepsections and created the gully network seen at present, is mostprobably a very recent phenomenon, i.e. post-Roman and possiblyaccentuated after the 1950's (Dakaris et al., 1964, 213e214; Harrisand Vita-Finzi, 1968; Bailey et al., 1992, 143).

There are artifact assemblages of homogeneous nature in termsof raw material type, degree of patination and typo-technologicaltraits. These occur either as archaeological horizons visible inprofiles and sealed within palaeosols, which are themselves sand-wiched between non-disturbed deposits; or as clustered, surfaceoccurrences that have undergone minimum, if any, dislocations(e.g. in the gray layer of unit C). These two types of artifact con-centrations, seen in vertical or horizontal exposures, are reminis-cent of Higgs's “chipping floors”. In both cases, it can be securelyassessed that, instead of being later intrusions post-dating thesediment accumulation as argued by Bailey et al. (1992), the findsare in situ in the geological sense, i.e. they belong to the same time-span represented by the deposits/palaeosols inwhich they occur. Inthe case of artifacts associated with palaeosols, only proper exca-vation can assess whether the finds are also archaeologically in situ,namely that they have remained at or close to the original places ofdiscard. Should this be possible for formerly subaerially exposedsurfaces (palaeosols), such a ‘primary context’ is least likely for theaforementioned artifact scatter in the gray layer, as the gleyingdenotes subaquaeous sedimentation and probably reworking ofartifacts by low-energy water movements such as rill- and sheet-washes feeding the water bodies. The latter observation generallyapplies to all individual, widely spaced artifact occurrences, whichare found stratified in the mottled and gleyed sediments of units Band C. The degree in which lithic assemblages can be treated asanalytical units suitable for behavioral inferences, remains to beinvestigated; at the moment, these assemblages are best viewed ascumulative palimpsests produced by an unknown number of ac-tivity episodes. The identification of temporally and spatiallydistinct episodes of activity and/or occupation events can beassessed only by subsurface investigations combined with abroader dating program.

6. Conclusions

The geoarchaeological evidence presented here reinterprets thestatus of Kokkinopilos, arguably the most emblematic and yet mostcontroversial Palaeolithic open-air site in Greece. According to thearchaeological, sedimentological, stratigraphic and micromorpho-logical data described above, the depositional sequence of the sitecontains artifacts that are stratified in non-reworked deposits, i.e.they are geologically in situ and hence datable. This observation isstressed because it lies in the core of the long-lasting controversy.In the dynamic setting of an ephemeral lake, with fluctuating waterlevels and shorelines, and with temporally and spatially alternatingwet/dry depositional environments, the existence of a ‘primarycontext’ is probably unlikely or at least hardly verifiable for most ofthe observed archaeological horizons. Bailey et al. (1992) rightfullycontested this view and were even potentially correct in arguingthat some of Higgs' test trenches were excavated into disturbeddeposits. Reworked sediments do exist, yet it is misleading to over-generalize this observation and consider the entire site a mixture ofreworked deposits, since the latter are few and spatially restricted.

Different episodes of activity, related to repeated visits of hominins,are preserved potentially superimposed one upon the other andrecorded in a geological archive shaped by episodes of soil forma-tion, sediment accumulation and erosion. Kokkinopilos is thereforeno more and no less of a palimpsest than most other open-airPalaeolithic sites (cf. Bailey, 2007) and, in that respect, it is boundto entail time-averaged accumulations of archaeological material.However, this material is not intrusive and reworked into a mucholder sequence, as argued by Bailey et al. (1992). Rather, thearchaeological evidence is geologically in situ: pedo-stratigraphiccriteria can be used to distinguish artifact occurrences and defineartifact assemblages. We have identified three principal types ofstratified archaeological occurrences: (1) artifacts exposed in ver-tical or sub-vertical profiles of palaeosols, where the lithic materialcan be traced laterally following the palaeosol exposure; this is thecase of the archaeological horizon associated with the uppermostpalaeosol in unit B; (2) artifacts found stratified inside unit B andunit C but not associated with palaeosols, such as the biface fromunit C; these appear to be isolated finds, but without excavating wecannot rule out the possibility that they belong to broader con-centrations; (3) artifacts occurring as clusters in surface exposures,which are unearthed by low-energy erosional processes; this is thecase with the lithics eroding out of the gray layer at the base of unitC. The amount of time entailed in the accumulation of those as-semblages depends on the rates of sedimentation and soilformation.

The pIRIR signals of potassium-rich feldspar grains isolated fromthe investigated sediment samples from units B and C were foundto be close to saturation and hence only minimum ages could becalculated. The minimum ages obtained range from 156 ± 21 ka forsample KPL-1 to 206 ± 19 ka for sample KPL-3. With presentlyavailable methods we cannot determine the true depositional ageof the deposits, but the data presented in Fig. 12 suggest that it maybe considerably older than the minimum ages reported here. Takenat face value, the dating results presented here indicate that (1)nearly the entire depositional sequence at Kokkinopilos pre-datesthe last interglacial; (2) artifact-yielding levels in unit C pre-date172 ± 25 ka (age of sample KPL-2 at the top of this unit); (3)artifact-yielding sediments in unit B pre-date 206 ± 19 ka.

At Kokkinopilos there is a lithic component that includes bifacesand could be considered as either pre- or early Mousterian.Without an adequate sample of collected, stratified lithics for aproper typo-technological study we cannot ascertain at themoment whether this component should be attributed to a lateLower or to an early Middle Palaeolithic technocomplex. None-theless, this component considerably predates the last interglacialand comprises the earliest stratigraphically defined and radiometri-cally constrained archaeological material in Greece.

Ultimately, our research clarified the depositional environmentof Kokkinopilos, resolved the degree of the site's stratigraphicintegrity and provided the basis for developing a chronostrati-graphic framework. Kokkinopilos has served as a reference-site forthe interpretation of all red-bed sites of Epirus, which is the areawith the highest density of Palaeolithic sites in Greece. There is along-lasting view that the lithic material from Kokkinopilos, as wellas that from other red-bed sites of Epirus, is of low archaeologicalvalue, because all of it is reworked and cannot be dated “except byrelying on a type-fossil approach” (Bailey et al., 1992, 142; for otherred-bed sites, see also Papoulia, 2011; Ligkovanlis, 2011, 2014). Inlight of the results presented above, this view has to be revised. Weshowed that Kokkinopilos and, by implication, probably other red-bed sites as well, involve archaeological contexts that can beexcavated and assessed chronologically. This realization opens upnew prospects for renewed Palaeolithic research in Epirus, a regionwith the most extensive Palaeolithic database in Greece. In these

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369368

lines, future research will be able to rely more on chronostrati-graphic criteria and less on formal, typological comparisons. Animproved chronology at Kokkinopilos can serve as the basis for aregional chronostratigraphic framework that will enable compari-sons between the numerous open-air sites of Epirus. This will havewider implications for the Palaeolithic of Greece, as the latter is stilllacking such frameworks for early and/or pre-Mousterian archae-ological evidence.

Acknowledgments

VT is currently supported by the European Research Council(ERC STG 283503). Luminescence dating was supported by fundsfrom a Spinoza Prize awarded toWil Roebroeks (Leiden University),to whom VT is grateful also for the discussions about Kokkinopilos,both at the site and in Leiden. Thanks are extended to G. Riginosand the archaeologists of the Archaeological Service in Preveza andIgoumenitsa for their constant support. Candice Johns preparedsamples for luminescence dating analysis. We thank four anony-mous reviewers for their encouragement to continue our researchat Kokkinopilos and for their constructive comments that helped usimprove the manuscript.

References

Alexanderson, H., Murray, A.S., 2012. Luminescence signals from modern sedimentsin a glaciated bay, NW Svalbard. Quat. Geochronol. 10, 250e256.

Bailey, G., 2007. Time perspectives, palimpsests and the archaeology of time.J. Anthropol. Archaeol. 26, 198e223.

Bailey, G., Papakonstantinou, V., Sturdy, D., 1992. Asprochaliko and Kokkinopilos: TLdating and reinterpretation of Middle Palaeolithic sites in Epirus, north-westGreece. Camb. Archaeol. J. 2, 136e144.

Buylaert, J.P., Jain, M., Murray, A., Thomsen, K.J., Thiel, C., Sohbati, R., 2012. A robustfeldspar luminescence dating method for Middle and Late Pleistocene sedi-ments. Boreas 41, 435e451.

Courty, M.A., Goldberg, P., Macphail, R.I., 1989. Soil, Micromorphology andArchaeology. In: Cambridge Manuals in Archaeology. Cambridge UniversityPress, Cambridge.

Dakaris, S.I., Higgs, E.S., Hey, R.W., 1964. The climate, environment and industries ofStone Age Greece: Part I. Proc. Prehist. Soc. 30, 199e244.

Darlas, A., 2007. The Mousterian of Greece under the light of recent research.Anthropologie 111, 346e366.

Deb�enath, A., Dibble, H.L., 1994. Handbook of Paleolithic Typology. In: Lower andMiddle Paleolithic of Europe, vol. one. University of Philadelphia, Philadelphia.

Duller, G.A.T., Wintle, A.G., 2012. A review of the thermally transferred opticallystimulated luminescence signal from quartz for dating sediments. Quat. Geo-chronol. 7, 6e20.

Elefanti, P., Panagopoulou, E., Karkanas, P., 2008. The transition from the Middle tothe Upper Palaeolithic in the southern Balkans: the evidence from Lakonis Icave, Greece. Eurasian Prehist. 5, 85e96.

Galbraith, R.F., Roberts, R.G., 2012. Statistical aspects of equivalent dose and errorcalculation and display in OSL dating: an overview and some recommendations.Quat. Geochronol. 11, 1e27.

Gu�erin, G., Mercier, N., Adamiec, G., 2011. Dose-rate conversion factors: update. Anc.TL 29.1, 5e8.

Guralnik, B., Li, B., Jain, M., Chen, R., Paris, R.B., Murray, A.S., Li, S.-H., Pagonis, V.,Valla, P.G., Herman, F., 2015. Radiation-induced growth and isothermal decay ofinfrared stimulated luminescence from feldspar (in press). Radiat. Meas. http://dx.doi.org/10.1016/j.radmeas.2015.02.011.

Harris, D.R., Vita-Finzi, C., 1968. Kokkinopilos: a Greek badland. Geogr. J. 134,537e546.

Harvati, K., Darlas, A., Bailey, S.E., Rein, T.R., El Zaatari, S., Fiorenza, L., Kullmer, O.,Psathi, E., 2013. New Neanderthal remains from Mani peninsula, S. Greece: TheKalamakia Middle Palaeolithic cave site. J. Hum. Evol. 64, 486e499.

Harvati, K., Panagopoulou, E., Runnels, C., 2009. The paleoanthropology of Greece.Evol. Anthropol. 18, 131e143.

Higgs, E.S., Vita-Finzi, C., 1966. The climate, environment and industries of StoneAge Greece: Part II. Proc. Prehist. Soc. 32, 1e29.

Huntley, D.J., Baril, M.R., 1997. The K content of the K-feldspars being measured inoptical dating or in thermoluminescence dating. Anc. TL 15, 11e13.

Huntley, D.J., Hancock, R.G.V., 2001. The Rb contents of the K-feldspar grains beingmeasured in optical dating. Anc. TL 19, 43e46.

Jacobs, Z., Roberts, R.G., Lachlan, T.J., Karkanas, P., Marean, C.W., Roberts, D.L., 2011.Development of the SAR TT-OSL procedure for dating Middle Pleistocene duneand shallow marine deposits along the southern Cape coast of South Africa.Quat. Geochronol. 6, 491e513.

Joordens, J.C.A., d’Errico, F., Wesselingh, F.P., Munro, S., de Vos, J., Wallinga, J.,Ankjærgaard, C., Reimann, T., Wijbrans, J.R., Kuiper, K.F., Mücher, H.J.,Coqueugniot, H., Pri�e, V., Joosten, I., van Os, B., Schulp, A.S., Panuel, M., van derHaas, V., Lustenhouwer, W., Reijmer, J.J.G., Roebroeks, W., 2014. Homo erectusfrom Trinil used shells for tool production and engraving. Nature. http://dx.doi.org/10.1038/nature13962.

Karkanas, P., White, D., Lane, C.S., Stringer, C., Davies, W., Cullen, V.L., Smith, V.C.,Ntinou, M., Tsartsidou, G., Kyparissi-Apostolika, N., 2015. Tephra correlationsand climatic events between the MIS 6/5 transition and the beginning of MIS3in Theopetra Cave, central Greece (in press). Quat. Sci. Reviews. http://dx.doi.org/10.1016/j.quascirev.2014.05.027.

King, G., Bailey, G., 1985. The palaeoenvironment of some archaeological sites inGreece: the influence of accumulated uplift in a seismically active region. Proc.Prehist. Soc. 51, 273e282.

King, G., Sturdy, D., Bailey, G., 1997. The tectonic background to the Epirus land-scape. In: Bailey, G.N. (Ed.), Klithi: Palaeolithic Settlement and QuaternaryLandscapes in Northwest Greece, Klithi in its Local and Regional Setting, vol. 2.McDonald Institute for Archaeological Research, Cambridge, pp. 541e558.

Kraus, M.J., 2002. Basin-scale changes in floodplain paleosols: implications forinterpreting alluvial architecture. J. Sediment. Res. 72, 500e509.

Kuhn, S.L., Pigati, J., Karkanas, P., Koumouzelis, M., Kozlowski, J.K., Ntinou, M.,Stiner, M.C., 2010. Radiocarbon dating results for the Early Upper Palaeolithic ofKlissoura Cave 1. Eurasian Prehist. 7 (2), 37e46.

Ligkovanlis, S., 2011. Megalo Karvounari revisited. In: Fors�en, B., Tikkala, E. (Eds.),Thesprotia Expedition II, Environment and Settlement Patterns, Papers andMonographs of the Finnish Institute at Athens, vol. XVI, pp. 159e180.Helsinki.

Ligkovanlis, S., 2014. Human Activity and Technological Behavior during the Middleand Upper Palaeolithic in North-West Greece: the Evidence of the Lithic As-semblages from Megalo Karvounari, Molondra and Eleftherochori 7 (unpub-lished PhD thesis). University of Crete, Crete.

Macleod, D.A., Vita-Finzi, C., 1982. Environment and provenance in the develop-ment of recent alluvial deposits in Epirus, NW Greece. Earth Surf. Proc. Landf. 7,29e43.

Mejdahl, V., 1979. Thermoluminescence dating: beta-dose attenuation in quartzgrains. Archaeometry 21, 61e72.

Murray, A.S., Marten, R., Johnston, A., Marten, P., 1987. Analysis for naturallyoccurring radionuclides at environmental concentrations by gamma spec-trometry. J. Radioanal. Nucl. Chem. 115, 263e288.

Papagianni, D., 2000. Middle Palaeolithic Occupation and Technology in NorthwestGreece: the Evidence from Open-air Sites. In: British Archaeological Reports882. Oxford.

Papakonstantinou, V., Vassilopoulou, D., 1997. The Middle Palaeolithic industriesof Epirus. In: Bailey, G.N. (Ed.), Klithi: Palaeolithic Settlement and Quater-nary Landscapes in Northwest Greece, Klithi in its Local and RegionalSetting, vol. 2. McDonald Institute for Archaeological Research, Cambridge,pp. 459e480.

Papoulia, Ch, 2011. Mikro Karvounari in context: the new lithic collection and itsimplications for Middle Palaeolithic hunting activities. In: Fors�en, B., Tikkala, E.(Eds.), Thesprotia Expedition II: Environment and Settlement Patterns, Papersand Monographs of the Finnish Institute at Athens, vol. XVI, pp. 123e158.Helsinki.

PiPujol, M.D., Buurman, P., 1994. The distinction between ground-water gley andsurface-water gley phenomena in tertiary paleosols of the Ebro basin, NE Spain.Palaeogeogr. Palaeoclimatol. Palaeoecol. 110, 103e113.

Pope, K.O., Runnels, C.N., Ku, T.-L., 1984. Dating Middle Palaeolithic red beds insouthern Greece. Nature 312, 264e266.

Prescott, J.R., Hutton, J.T., 1994. Cosmic ray contributions to dose rates for lumi-nescence and ESR dating: large depths and long-term time variations. Radiat.Meas. 23, 497e500.

Pyle, D.M., van Andel, T.H., Paschos, P., van den Bogaard, P., 1998. An exceptionallythick Middle Pleistocene tephra layer from Epirus, Greece. Quat. Int. 49,280e286.

Reineck, H.-E., Singh, I.B., 1980. Depositional Sedimentary Environments, second ed.Springer-Verlag, New York.

Retallack, G.J., 2001. Soils of the Past, second ed. Blackwell Science, Oxford.Retallack, G.J., Bestland, E.A., Fremd, T.J., 2000. Eocene and Oligocene paleosols in

Central Oregon. Geol. Soc. Am. 344.Runnels, C., 1995. Review of Aegean Prehistory IV: the Stone Age of Greece from the

Palaeolithic to the advent of Neolithic. Am. J. Archaeol. 99, 699e728.Runnels, C., van Andel, T.H., 1993. A handaxe from Kokkinopilos, Epirus, and its

implications for the Palaeolithic of Greece. J. Field Archaeol. 20, 191e203.Runnels, C., van Andel, T.H., 2003. The Early Stone Age of the Nomos of Preveza:

landscape and settlement. In: Wiseman, J., Zachos, K. (Eds.), LandscapeArchaeology in Southern Epirus, Greece I. ASCSA, Athens, pp. 47e133. HesperiaSupplement 32.

Stoops, G., 2003. Guidelines for Analysis and Description of Soil and Regolith ThinSections. Soil Sci. Soc. Am, Madison, Wisconsin.

Thiel, C., Buylaert, J.-P., Murray, A., Terhorst, B., Hofer, I., Tsukamoto, S., Frechen, M.,2011. Luminescence dating of the Stratzing loess profile (Austria): testing thepotential of an elevated temperature post-IR IRSL protocol. Quat. Int. 234,23e31.

Thomsen, K.J., Murray, A.S., Jain, M., Botter-Jensen, L., 2008. Laboratory fading ratesof various luminescence signals from feldspar-rich sediment extracts. Radiat.Meas. 43, 1474e1486.

V. Tourloukis et al. / Journal of Archaeological Science 57 (2015) 355e369 369

Tourloukis, V., 2009. New bifaces from the Palaeolithic site of Kokkinopilos, Greeceand their stratigraphic significance. Antiquity 83 (320) (Project Gallery).

Tourloukis, V., 2010. The Early and Middle Pleistocene Archaeological Record ofGreece: Current Status and Future Prospects. In: Archaeological Studies LeidenUniversity 23. Leiden University Press, Leiden.

Tourloukis, V., Karkanas, P., 2012. The Middle Pleistocene archaeological record ofGreece and the role of the Aegean in hominin dispersals: new data and in-terpretations. Quat. Sci. Rev. 43, 1e15.

van Andel, T.H., 1998. Paleosols, red sediments and the Old Stone Age in Greece.Geoarchaeology 13, 361e390.

van Andel, T.H., Runnels, C., 2005. Karstic wetland dwellers of Middle PalaeolithicEpirus, Greece. J. Field Archaeol. 30, 367e384.

Valladas, H., Mercier, N., Froget, L., Joron, J.-L., Reyss, J.-L., Karkanas, P.,Panagopoulou, E., Kyparissi-Apostolika, N., 2007. TL age estimates for the MiddlePalaeolithic layers at Theopetra cave (Greece). Quat. Geochronol. 2, 303e308.

Vepraskas, M.J., Wilding, L.P., Drees, L.R., 1992. Aquic conditions for soil taxonomy:concepts, soil morphology and micromorphology. In: Ringrose-Voase, A.J.,

Humphreys, G.S. (Eds.), Soil Micromorphology: Studies in Management andGenesis, Developments in Soil Science 22. Elsevier, Amsterdam, pp. 117e131.

Wallinga, J., Bos, A.J.J., Dorenbos, P., Murray, A.S., Schokker, J., 2007. A test case foranomalous fading correction in IRSL dating. Quat. Geochronol. 2, 216e221.

Wang, X.L., Wintle, A.G., Lu, Y.C., 2007. Testing a single-aliquot protocol for recu-perated OSL dating. Radiat. Meas. 42, 380e391.

Wintle, A.G., Murray, A.S., 2006. A review of quartz optically stimulated lumines-cence characteristics and their relevance in single-aliquot regeneration datingprotocols. Radiat. Meas. 41, 369e391.

Woodward, J.C., Macklin, M.G., Lewin, J., 1994. Pedogenic weathering and relatice-age dating of Quaternary alluvial sediments in the Pindus Mountains ofNorthwest Greece. In: Robinson, D.A., Williams, R.B.G. (Eds.), Rock Weatheringand Landform Evolution. John Wiley & Sons, Chichester, pp. 259e283.

Zhou, L.P., van Andel, T.H., Lang, A., 2000. A luminescence dating study of open-airPalaeolithic sites in western Epirus, Greece. J. Archaeol. Sci. 27, 609e620.

Zimmerman, D.W., 1971. Thermoluminescence dating using fine grains from pot-tery. Archaeometry 13, 29e52.