Paleolithic occupations of the Göllü Dağ, Central Anatolia, Turkey

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=yjfa20

Journal of Field Archaeology

ISSN: 0093-4690 (Print) 2042-4582 (Online) Journal homepage: https://www.tandfonline.com/loi/yjfa20

Paleolithic occupations of the Göllü Dağ, Central

Anatolia, Turkey

Steven L. Kuhn, Berkay Dinçer, Nur Balkan-Atlı & Mehmet Korhan Erturaç

To link to this article: https://doi.org/10.1179/2042458215Y.0000000020

© Trustees of Boston University 2015

Published online: 06 Aug 2015.

Submit your article to this journal

Article views: 1955

View related articles

View Crossmark data

Paleolithic occupations of the Go

¨llu

¨ Dag˘,

Central Anatolia, Turkey

Steven L. Kuhn

, Berkay Dinc

er

, Nur Balkan-Atlı

, Mehmet Korhan Erturac



1

University of Arizona, School of Anthropology,2Ardahan University, Faculty of Social Sciences and Humanities, Prehistory Department,3Istanbul University, Faculty of Letters, Prehistory Department,4Sakarya University, Faculty of Arts and Sciences, Geography Department

Systematic archaeological surface reconnaissance of the Go¨llu¨ Dag˘ volcanic complex from 2007 to 2012 documented more than 230 findspots with Paleolithic artifacts, ranging from isolated finds to extensive and dense scatters of artifacts. Most of the activities represented relate to exploitation of the rich obsidian resources in the region. Paleolithic artifacts are attributed mainly to the Middle Paleolithic based on the presence of Levallois technology but there is a substantial Lower Paleolithic component represented by handaxes and other large bifacial tools. Upper and Epipaleolithic sites and artifacts are scarce or absent in the survey sample. The distributions of handaxes and Levallois elements differ substantially, reflecting differences in site preservation and exposure as well as organ-ization of prehistoric activities. Multiple variants of Levallois are represented but centripetal preferential and unipolar flake production dominate. The frequent co-occurrence of different Levallois forms suggests flexible reduction strategies. Distributions of different classes of artifact across the survey area indicate that the Middle Paleolithic occupations of Go¨llu¨ Dag˘ were not entirely oriented toward workshop activities.

Keywords: Anatolia, archaeological survey, Paleolithic, obsidian, Levallois

Introduction

Anatolia, the Asian part of Turkey, plays a unique role in accounts of hominin dispersals out of Africa and across Eurasia. On a map of the world Anatolia seems to form an obvious land-bridge connecting the Levant with the Balkans, making it the most direct route between Africa, western Asia and Europe. We know that Anatolia was a highway for multiple incursions of peoples from both east and west over the past few thousand years. It is not surprising then that representations of hominin dispersals at various times during the Pleistocene typically feature bold arrows drawn across Anatolia indicating hypothetical movements of people and/or culture (e.g., Mellars 2011: fig. 1; Oppenheimer 2012: fig. 1). Predictably, the situation on the ground is very different. Anatolia exists in more than two dimen-sions. Most of the central plateau ranges upwards from 1000 masl. The plateau is also surrounded by a rim of high, steep mountains, the Taurus range on the south and the Pontic range on the north, which restrict the movement of warm, moist air

masses from the Mediterranean and Black Seas into the interior. Today the central plateau is character-ized by dry hot summers and cold, snowy, windy winters, and it supports arid or semi-arid steppe biomes. Because of the elevation the Anatolian pla-teau sees winter temperatures and growing seasons similar to parts of England and northern Germany, although with much less precipitation. The climate would of course have been much harsher during glacial periods within the Pleistocene. Consequently, at times during the Pleistocene the central Anatolian plateau could have been an obstacle to hominin colonization rather than a conduit for dispersal. Of course, the low-lying Mediterranean and Black Sea coasts would have remained conducive to settlement even during cold glacials. Consequently, understand-ing the history of occupation of the high Anatolian plateau during the Pleistocene may tell us as much about hominins’ proclivity to occupy cold, dry environments as it does about the timing of move-ments out of Africa.

The other factor undermining easy assumptions about Anatolia as a dispersal route is the scarcity of direct evidence for hominin presence during the Pleistocene. There has been surprisingly little Paleo-lithic research in Turkey compared to surrounding

Correspondence to: Steven L. Kuhn, School of Anthropology, University of Arizona, 1099 E. South Campus Drive, Tuscon, AZ, 85719. Email: skuhn@email.arizona.edu

areas (Kuhn 2002; Dinc¸er in press). In all of Turkey, an area of more than 780,000 sq km, only four Lower and Middle Paleolithic sites have been investigated systematically. Surface finds of handaxes, Levallois flakes and other typical Paleolithic artifacts have been reported from throughout south-central Anato-lia (Aksaray, Nig˘de, Nevs¸ehir, and Kayseri pro-vinces), demonstrating that hominins were present at least sporadically, but few sites have been investi-gated further. As a consequence we have little infor-mation about hominin presence in Anatolia during the key periods for geographic expansion.

This paper reports results from a systematic survey of prehistoric archaeological occurrences within and around the Go¨llu¨ Dag˘ volcanic complex in central Anatolia. The survey, conducted by a Turkish/ French/American team was initiated in order to document changing use of the area, and of its rich obsidian sources, from the earliest occupations through the Chalcolithic. Go¨llu¨ Dag˘ offers two unique advantages with respect to investigating hominin presence in central Anatolia during the Pleistocene. First, the area contains extensive exposures of actively eroding Pleistocene volcanic sediments, increasing the visibility of early archaeolo-gical sites. Second, it abounds in high-quality volca-nic rocks for tool making—basalt, andesite and obsidian. If Pleistocene populations were present in the neighboring areas we expect that the abundance of toolstone would have brought then to Go¨llu¨ Dag˘.

Previous Research on the Paleolithic of Central

Anatolia

Prior to excavations and surveys in the Go¨llu¨ Dag˘ area, Paleolithic research in south-central Anatolia was restricted to non-systematic studies. Sog˘anlı, near Kayseri, was discovered as early as 1910, making it one of the first reported Paleolithic sites in Turkey (Campbell-Thompson 1910). Middle Paleolithic obsidian scrapers were reported by the site’s discoverer but no Paleolithic finds were observed during later visits to the site in 1945 (Ko¨kten 1947; Inan 1947). The only two Lower Paleolithic findspots reported in Central Anatolia prior to 1990 are Barsık, near Kayseri and Avladag˘, near Nevs¸ehir. At Barsık two pebble tools made of limestone and one disk-like core made of chert were found on one of the terraces of Kızılırmak River (Tomsky 1982). At Avladag˘ a basalt biface and some obsidian Levallois products were recovered from the terraces of the Damsa River (Todd and Pas-quare 1965). Reported Middle Paleolithic sites are somewhat more frequent than Lower Paleolithic ones. Obsidian artifacts described as ‘‘Levallois tools’’ were collected at Acıgo¨l and Suvermez (Kansu 1945). Two chert sources with associated

Middle Paleolithic tools have also been identified in the region. At the village of Karain near U¨ rgu¨p, two Mousterian tools made of obsidian and a chert source were recorded (Inan 1947). A large chert source has also been identified at C¸ akmaktepe, nearly 70 km south of Go¨llu¨ Dag˘. A total of 96 arti-facts attributed to Middle Paleolithic were collected from around the outcrops: the assemblage includes discoidal cores and many large flakes (Minzoni-De´roche, 1993). Another limitation of earlier research on the Paleolithic in south-central Anatolia is the lack of systematic and multidisciplinary studies. Most pub-lications regarding the Paleolithic in this region are little more than summary descriptions of a non-systematic sample of surface finds. However the earlier research at least shows that in the Paleolithic the distri-bution of artifacts was not limited to areas immedi-ately surrounding raw material sources.

The only excavated Lower or Middle Paleolithic site in central Anatolia is Kaletepe Deresi 3 (KD3). The site is situated within the area covered by this study, on the lower slopes of the Go¨llu¨ Dag˘ volcano, close to the well-known Kaletepe/Ko¨mu¨rcu¨ obsidian workshops (Slimak et al. 2007, 2008). The 6-meter-deep stratigraphic sequence at KD3 is made up of a series of geogenic deposits, mainly colluvial and alluvial accumulations of reworked volcanic tephra and larger clasts of rhyolite and andesite. The 14 archaeological horizons represent periods of more or less intense frequentation of the area by Paleo-lithic groups rather than discrete anthropogenic deposits or living floors. The lower part of the sequence (levels XII through IV) contains a series of Acheulean assemblages. The richest of these, from layer V, includes bifacial and unifacial han-daxes of obsidian as well as flake cleavers of obsidian and andesite. The upper part of the sequence (levels II-I) contains Mousterian assemblages with Levallois technology. Assemblages from the two middle levels (II9 and II) contain too few artifacts to characterize. There is very limited chronological control over the Paleolithic sequence at KD3. A series of volcanic tephras within the Middle Paleolithic deposits come from an eruption of the Acıgo¨l volcano, which has recently been redated to roughly 190–200 ka (Schmitt et al. 2011). The base of the archaeological sequence is formed by rhyolitic lava associated with the Kabak Tepe dome, dated to ca. 1.1 ma.

Evidence for an Upper and Epi-Paleolithic pre-sence is even scarcer in Central Anatolia than are traces of the Lower or Middle Paleolithic. There is a single excavated late Epi-Paleolithic site, Pınarbas¸ı, near the Konya plain (Baird et al. 2013). Other reported Upper Paleolithic surface materials have yet to verified, and could very well be early Neolithic in age (Baird et al. 2013: 175–176).

Description of Go

¨llu

¨ Dag˘ and the Survey Project

The study area (FIG. 1) is situated within the Central Anatolian Volcanic Province (CAVP). The CAVP was subject to more-or-less continuous volcanic activity from the Neogene through the late Pleisto-cene and early HoloPleisto-cene. Go¨llu¨ Dag˘ itself was formed during a comparatively recent phase of activity. The Bu¨yu¨k (‘‘big’’) Go¨llu¨ Dag˘ volcanic cone dominates the visual landscape but the Go¨llu¨ Dag˘ is better characterized as a rhyolitic volcanic complex. At least ten distinct rhyolitic domes are identifiable, and several of these preserve traces of earlier domes within them (Binder et al. 2011) (FIG. 2). The entire complex formed during the Lower and Middle Pleistocene. The oldest volcanic rocks within the complex are dated to 1.95 Ma+0.06 (Ar/Ar, Cauvin and Chataigner 1998) and 1.71+0.03 Ma (K/Ar, Mouralis 2003). The domes most visible today refer to later eruptive phases, ranging from 1.48+0.09 Ma for the Kayırlı area at the north end of the study area to

0.44+0.007 Ma on Ku¨c¸u¨k (‘‘small’’) Go¨llu¨ Dag˘ at the south end (Bigazzi et al. 1993; Mouralis 2003).

Go¨llu¨ Dag˘ is well known for the extensive exposures of obsidian and associated workshops. Obsidian from these sources was widely circulated throughout Anatolia, Mesopotamia, the Levant and Cyprus, particularly during the aceramic Neolithic (Binder et al. 2011; Cauvin and Chataigner 1998). The sources on Go¨llu¨ Dag˘ were recognized in the first sourcing studies of Mediterranean obsidian (Renfrew et al. 1966, 1968) although their exact locations were unknown. In fact, there are at least seven chemically differentiable varieties of obsidian within the Go¨llu¨ Dag˘ complex (Binder et al. 2011), each of which outcrops in several locations (FIG. 2).

Many of the obsidian deposits were formed as intru-sive ring-dykes around collapsed calderas (Mouralis 2002, 2003), but recent field study by one of the authors (KE) demonstrates the presence of extrusive obsidian flows as well. As might be expected from the diversity of sources and formation contexts,

Figure 1 Map of area surrounding Gollu Dag, showing major mountains and cities (black squares). The survey area is indicated by broken lines.

Go¨llu¨ Dag˘ obsidians occur in a variety of forms, ranging from massive outcrops to tabular/bedded exposures to isolated ‘‘bombs’’ of various sizes. The quality of obsidian also varies widely. Glass from outcrops in the Kaletepe/Ko¨mu¨rcu¨ and Bitlikiler/ Ekinlik areas is of remarkably high quality, but in many other exposures vesicular texture or numerous perlite or spherulite inclusions compromise the obsi-dian’s workability. Tool makers showed distinct pre-ference for different forms and qualities of obsidian at different times in the past.

The Go¨llu¨ Dag˘ today is a very dynamic landscape. Most slopes support little vegetation, a legacy of centuries of intensive grazing and wood-cutting activities. Powerful winter and late summer storms can cause rapid erosion and reworking of poorly-consolidated volcanic soils and tephra deposits. Parts of the Go¨llu¨ Dag˘ landscape appear very differ-ent from one year to the next, especially if rains have been heavy. On the other hand, other recent and

historic anthropogenic impacts have been compara-tively minor. Three villages, Ko¨mu¨rcu¨, Kayırlı and Pınarcık, are situated at the base of mountain along the survey area’s northern boundary. Contemporary land use is mainly limited to agriculture and pastoral activities, the latter being especially common at higher elevations (w1600 masl). Abandoned terraces and evidence for field clearance in areas no longer cul-tivated indicate that small-scale agriculture was much more widespread in the past. Currently there are some small, informal quarries for pumice and perlite as well as borrow pits for fill used in construction. However, the area may be opened up to large-scale per-lite mining in the future. This very destructive activity would result in serious damage to an important and irreplaceable archaeological resource.

The largest and best-known ancient site within the survey area is the fortified Neo-Hittite site situated adjacent to the precipitation-fed lake within the main Go¨llu¨ Dag˘ volcanic crater (Schirmer

Figure 2 Major domes in Gollu Dag volcanic complex, indicated by medium gray shading. Crosses mark Paleolithic find-spots. Cross-hatching indicates extent of major obsidian-bearing formations. Dark gray polygons are contemporary villages.

1993).The ancient road leading to the mountaintop site is still visible in places, and at least one site ident-ified in the survey (n. 572) may be associated with the road near the bottom of the mountain. Aside from this, however, evidence for human presence on Go¨llu¨ Dag˘ after the Chalcolithic is remarkably sparse. Only a handful of sites located during the survey con-tained sherds predating the 19th or 20th centuries. Apparently resources available in the area were not sufficient to attract much human settlement once obsi-dian was no longer an important commodity.

Survey Methods

Intensive pedestrian survey coverage focused on areas of low to moderate slope, in many cases with active erosion. For both practical and scientific reasons we attempted only minimum coverage of the steepest slopes. We also avoided areas blanketed by Holocene alluvium, especially at the foot of the mountain, unless they were erosionally dissected. We also intentionally minimized survey within the well-known Kaletepe/Ko¨mu¨rcu source and work-shop zone. This area has been under intensive study since the late 1990s by a team of French and Turkish researchers (Balkan-Atlı et al. 1999), and a compre-hensive publication on the Neolithic and Chalcolithic workshops is being prepared.

Systematic surface survey is the most appropriate way to approach the archaeological record of a place such as Go¨llu¨ Dag˘. Although it contains abun-dant cultural material on the surface and in buried contexts, except for the single neo-Hittite fortified settlement the area lacks the sort of large mounded village sites or limestone caves that have been the tar-gets of research throughout much of the history of archaeological research in Anatolia. In such situ-ations, survey becomes the primary tool for research-ing human land-use and ecology, rather than simply being a means of discovering sites to excavate. Sys-tematic surface reconnaissance is growing in import-ance as a research tool in Turkey (e.g., Baird 1996; Garrard et al. 2004; Luke and Roosevelt 2009; Mat-thews 2007; Rauh 2012) as in the rest of the world.

We employed a conventional means of recording finds, based on individual ‘‘sites’’ with limited spatial boundaries. Although a ‘‘non-site,’’ object-based approach to recording might have been more fitting, especially for the earlier periods and some of the more extensive workshop deposits, the sheer density of material made this impractical. In places the ground surface is carpeted with obsidian artifacts from different periods. In many respects the entire Go¨llu¨ Dag˘ study area should be treated as a single archaeological ‘‘district.’’ It contains a very high aver-age density of archaeological materials compared with surrounding areas. Moreover, almost all of the

archaeological occurrences noted reflect activities that were either oriented toward, or were strongly affected by, a single, abundant natural resource, obsi-dian. While the use of obsidian and the roles of the Go¨llu¨ Dag˘ sites in regional land use systems certainly changed over time, the presence of obsidian was always important to the people using the area in prehistory.

The significance of ‘‘sites’’ as bounded distributions also varies over time. For the Lower and Middle Paleolithic, geomorphology and surface processes had a very strong influence on where we found things and where we did not. Some of the variation in the frequency and nature of finds is certainly due to the ways Pleistocene hominins used the Go¨llu¨ Dag˘ landscape, but processes of erosion and depo-sition have had an extremely pronounced influence on the structure of the Paleolithic record. Holocene (Neolithic and Chalcolithic) sites were certainly affected by geological processes but less so than ear-lier ones. Consequently they present a more complete representation of the distribution of human activities on the mountain than are the earlier sites.

The archaeological record documented through the survey is not highly varied. The finds noted and collected consist overwhelmingly of flaked obsidian artifacts, along with occasional ground-stone and pottery. Given the limited array of materials recorded, as well as the difficulties of establishing site boundaries in this highly dynamic landscape, there is little utility in constructing an elaborate typology of sites. A few kinds of localities stand out: workshop areas; sites with a range of artifacts besides production debris, which probably encom-pass a residential component; a few small Chalco-lithic and NeoChalco-lithic living sites; and a number of small rockshelters.

Lithic raw material was easily accessed throughout the study area. No recorded find spot was more than five km from an outcropping of obsidian, and most are within one km. Because of the ready access to raw material in most of the survey area, stone arti-facts are locally abundant to superabundant: in places dense scatters of obsidian flaking debris extend over thousands of square meters. From a logistical and practical perspective it was impractical to conduct systematic surface collection at all localities. Field collection and documentation focused on a limited range of artifacts likely to yield the greatest information about chronology, technology, and range of activities conducted at sites. Collection protocols emphasized cores and core rough-outs, reduction products such as blades, Levallois flakes, ‘‘technical pieces’’ (byproducts of core shaping and rejuvenation), and of course shaped tools. Future phases of study will entail

more systematic in field documentation of specific sites, especially workshops.

As a result of this intentionally biased collection strategy, the survey database provides systematic information on certain features of the archaeological record but not on others. Our database reliably records the presence/absence of shaped artifacts and cores from different time periods, as well as the most common technological procedures carried out at various sites. We can also extract comparative information about assemblage content and the con-dition of artifacts useful at a coarse grained, ordinal level (absent, low frequency, high frequency). How-ever, the data generated during this first stage of research do not supply reliable quantitative infor-mation about artifact density or the frequencies of different procedures or reduction stages.

Survey results

A total of 333 archaeological loci were identified in the 2007–2012 survey of Go¨llu¨ Dag˘. These loci range from isolated artifacts to scatters of debris from multi-period workshops covering thousands of square meters. The total includes 100 sites with only Neolithic and/or Chalcolithic artifacts, 111 with just Paleolithic artifacts, and 122 yielding evi-dence for both Pleistocene and Holocene occupations (FIG. 3). We collected just over 2,000 artifacts of

obvious Paleolithic age, mainly cores, shaped tools and Levallois flakes. Given the limited contextual information and absence of reliable radiometric dates, we can only describe the Paleolithic finds by general period. Lower and Middle Paleolithic arti-facts are well represented in the survey area. In con-trast, evidence for later Pleistocene Upper and Epi-Paleolithic occupations is notably scarce.

Lower Paleolithic

The artifacts we can attribute with most confidence to the Lower Paleolithic are handaxes, cleavers and certain other core tools. A proportion of less-diagnostic artifacts collected may well date to the Lower Paleolithic but it is difficult to identify them with certainty. A total of 176 handaxes and cleavers (Large Cutting Tools or LCTs) and other core tools were collected from between 2007 and 2012. Although some of these may pertain to the later Middle Paleolithic occupations, based on mor-phology and technological characteristics the majority are Lower Paleolithic/ Acheulean in age.

The large core tools recovered during the survey are made using a range of raw materials (TABLE 1).

Despite the abundance of obsidian in the project area, just over 1/3 of the LCTs are made of basalt, andesite, or trachyte. Similar raw material prefer-ences were noted in some of the Acheulean levels at

KD3 (Slimak et al. 2007). Apparently the greater durability of basalt or andesite was sometimes favored over the workability of obsidian in making heavy-duty tools.

Handaxes are the most abundant general class of large core tool in the survey collections. Most of the specimens collected are generally amygdaloid in shape, but ovate and triangular forms are also pre-sent (FIGS. 4, 5). Rarer forms include a few bifacial

cleavers, backed handaxes and discs. There are also a range of artifacts intermediate between handaxes and cores, scrapers or other forms. For the most part these are broken or only partially worked, although a few have clearly been reused, judging from differing levels of patina. Cleaver flakes are pre-sent in the excavated Acheulean layers at KD3, especially in level 5, and we expected to find them during the survey, but only one possible example was collected. This may be a function of the difficulty of recognizing simple cleaver flakes as isolated sur-face artifacts. One special class of artifacts consists of small bifacial handaxes less than 80 mm in maxi-mum dimension. Based on their size and morphology these artifacts could date to either the Middle Paleo-lithic or to the late Acheulean: they are similar to small bifaces, termed ‘‘Micoquian,’’ that were col-lected on the surface around the Kaletepe/Ko¨mu¨rcu¨ obsidian source (Slimak 2004). The small bifaces are made of obsidian more frequently than the larger handaxe or cleaver forms (chisq.54.68, df51, p50.030).

We could determine the form of the original blanks for 107 large tools, handaxes and bifaces: the others are either too fragmentary or all traces of the original blank morphology have been elimi-nated. Of the pieces for which the original form can be discerned, 68 (63.6%) were made on large flakes. The remaining specimens were manufactured from nodules or tabular pieces. It is likely that a large proportion of indeterminate pieces, most of which lack cortex, were made from large flakes or blocks. The frequent selection of flakes as blanks for heavy-duty tools is consistent with some of the Acheulean assemblages from KD3, levels 5–12, which represent variants of the ‘‘Large Flake Acheu-lean’’ (Sharon 2007, 2009).

The distribution of handaxes and cleavers differs from most other classes of artifact. Although they occur in the same parts of the landscape as other arti-facts (FIG. 3), ‘‘large cutting tools’’ tend to occur as

isolates or as single specimens in scatters of artifacts from other periods. More than two or three speci-mens are seldom recorded together except for the instances of general collections made over broad areas containing lag deposits. Other classes of Paleo-lithic artifact, such as Levallois cores or flake tools,

tend to occur in larger concentrations. Similar con-trasts in the distributions of Acheulean and Mouster-ian artifacts have also been observed in the course of systematic survey in desert basins in Syria (Conard

et al. 2010). The dispersed distribution of handaxes undoubtedly reflects the life histories of these particu-lar artifact forms. Handaxes and cleavers may have been widely transported as generalized aids to

Table 1 Large core tools from survey collections.

Form Obsidian Basalt Other Volcanic Total

Handaxe 43 35 2 80 Handaxe preform? 7 0 0 7 Bifacial cleaver 1 1 0 2 Backed handaxe 2 0 0 2 Bifacial scraper 5 0 0 5 Point/handaxe 0 1 0 1 Small biface/point (_{,80 mm)} 19 4 0 23 Discoid 1 2 0 3 Discoid/handaxe 0 2 0 2 Cleaver flake 0 0 1 1 Flake chopper 1 0 0 1 Uniface/core 2 2 0 4 Core/tool 12 1 0 13 Miscl. biface 18 2 0 20

Other large tools 4 2 1 7

Biface fragments 1 3 1 5

Total 116 55 5 176

Figure 3 Major domes in Gollu Dag volcanic complex with Paleolithic findspots. Circles mark locations with Levallois cores, triangles are locations with handaxes or other large cutting tools.

foraging and other activities, and discarded individu-ally (e.g., Soressi 2002, 2004).

Distributions of handaxes and similar artifacts docu-mented during the survey are strongly biased by geologi-cal factors. In the field it was observed that handaxes were often collected from comparatively flat, apparently deflating land surfaces, whereas Middle Paleolithic arti-facts came from a wider variety of situations, including steep, actively eroding slopes. This general observation is confirmed by analysis of locational data. For this comparison, Levallois cores are used as an indicator of the presence of Middle Paleolithic occupations. There are no significant differences in the mean elevations at which handaxes and Levallois cores were collected (t50.902, df51049, p50.367). However, han-daxes and other LCTs do tend to occur on less steep slopes than Levallois cores (t54.547, df51047, pv0.001) (Supplemental material 1; http://www.maney online.com/doi/suppl/10.1179/2042458215Y.0000000 020). In fact, the majority (*56%) of handaxes were collected from surfaces with slopes of 10uu or less, whereas only about one third of Levallois cores (*35%) came from such level surfaces. We note that these differences in topographic situation are not vis-ible on the large-scale distribution maps for different

artifact classes (FIG. 3). The local relief in parts of Go¨llu¨ Dag˘ is quite steep, so that stable or deflating surfaces may be bounded by actively eroding slopes.

These differences in situations of discovery do not indicate that the Acheulean record of Go¨llu¨ Dag˘ is lim-ited to deflated land surfaces. The stratified site of KD3 contains several meters of sediments with in situ Acheu-lean assemblages (archaeological levels 5–12). How-ever, the results do suggest that the surface record is very incomplete. Lower Paleolithic deposits at KD3 are at least three to four meters below the surface. At the time of the site’s discovery the Acheulean layers were covered with a mantle of colluvium, and they were only recognized after being exposed during exca-vation. Thus, a large part of the sedimentary record corresponding to the earlier Acheulean occupations of Go¨llu¨ Dag˘ may be buried deeply and/or obscured by recent colluvial deposits, leaving us with a view biased toward stable surfaces or lag deposits.

Middle Paleolithic

Evidence for a Middle Paleolithic presence is abun-dant and widespread on Go¨llu¨ Dag˘. Middle Paleo-lithic artifacts were collected in almost every part of the study area. They also occur in a wide range of

geological and topographic situations, from deflated lag deposits with heavily weathered artifacts to buried, actively- eroding deposits containing very fresh-looking finds. The range of products and byproducts encountered as well as the conditions of artifacts indicate that Middle Paleolithic hominins did more on Go¨llu¨ Dag˘ than simply collect and process obsidian. The Middle Paleolithic assemblages vary somewhat in their composition and may well document a range of activity types ranging from blank production to tool use and discard. This

section provides summary descriptions of the princi-pal classes of Middle Paleolithic artifacts. The sub-sequent section discusses variation in artifact samples within the survey area.

Because of collection strategy the most abundant evidence for the Middle Paleolithic consists of Leval-lois cores and products. Approximately 1,220 cores of likely pre-Neolithic age were collected during the 2007–2012 surveys (FIGS. 6, 7). More than 90% of these cores are Levallois or ‘‘para-Levallois’’ forms. ‘‘Para-Levallois’’ cores possess most of the key

Figure 5 Small obsidian bifacial handaxe and larger basalt handaxe on flake from Gollu Dag survey. Note different levels of patina on two faces of obsidian handaxe.

characteristics of Levallois as defined by Boe¨da (1995), but one or more character is either missing or poorly expressed. Other core forms, such as discoid or polyhedral, are comparatively rare in the study sample. The great majority of these cores, especially the Levallois and ‘‘para-Levallois’’ forms almost certainly correspond to the Middle Paleo-lithic. Although a certain number may also come from the earlier Acheulean occupation(s) it is difficult to sort these out based on topographic or geological contexts or archaeological associations. In contrast to handaxes, Paleolithic cores from the Go¨llu¨ Dag˘ survey are almost exclusively (98.2%) of obsidian. This certainly reflects a strong preference for using the finer-grained material to produce small blanks throughout the occupational history of the region,

and is consistent with findings from the upper levels at KD3 (Slimak et al. 2007, 2008).

While they preferred obsidian as a raw material, Middle Paleolithic tool makers seem to have been pre-pared to use a wider range of obsidians than later people. They sometimes made artifacts of vesicular or perlite-rich obsidians that later knappers tended to ignore. Middle Paleolithic hominins were not unselec-tive however. They did prefer particular forms of obsidian, tending to select flattish nodules or tabular pieces as blanks for Levallois cores, even where larger blocks of obsidian were also available. This preference is likely related to the shapes of the intended products: the kinds of flat pieces selected could be trans-formed into Levallois cores with a minimum of prep-aration. On the other hand, maximizing the sizes of

products does not seem to have been a dominant concern for Middle Paleolithic tool makers. Some very large cores and Levallois flakes were collected, but most of the products are of moderate size. Levallois products average 61.6 mm in maximum dimension (sd517.8), while Levallois and para-Levallois cores average around 74 mm in largest dimension. Artifact sizes also vary across the study area according to the proximity to outcrops of obsidian (see below).

We might hypothesize that the Levallois cores were, as a group, the most recent of the Paleolithic forms, and that the less formal ‘‘para-Levallois’’ forms and the non-Levallois cores more often represented earlier hominin occupations. However, differences in weath-ering suggest that this is not the case—as a class, the ‘‘classic’’ Levallois cores show the highest overall level of surface alteration (Supplemental Material 2, 3; http://www.maneyonline.com/doi/suppl/10.1179/2042 458215Y.0000000020). Thus, there is no reason to believe that the ‘‘para-Levallois’’ or non-Levallois cores as groups date to an earlier occupation. The distinction between the ‘‘para-Levallois’’ and Levallois groups instead represents different levels of core

preparation and exploitation at the time of discard. Cores in the ‘‘para-Levallois’’ group tend to be both larger and thicker than more classic Levallois cores (Supplemental material 4; http://www.maneyonline. com/doi/suppl/10.1179/2042458215Y.0000000020). The difference in thickness is particularly substantial, amounting to about 20%. Cores in the ‘‘para-Levallois’’ group also preserve cortex on their upper (Levallois) surfaces more often (Supplemental material 5; http://www.maneyonline.com/doi/suppl/10.1179/204 2458215Y.0000000020). Together, these results indicate that the ‘‘para-Levallois’’ cores as a group tend to be less extensively exploited and were aban-doned at slightly earlier stages in their reduction. These differences are subtle, and the artifacts in the ‘‘para-Levallois’’ group are not simply preforms or failures. Instead, they represent part of the broad range of variation in the execution of Levallois production within the Go¨llu¨ Dag˘ study area. The two groups of cores will consequently be combined in analyses that follow under the heading ‘‘Levallois.’’ Middle Paleolithic tool makers in the Go¨llu¨ Dag˘ area employed diverse systems for preparation and

Figure 7 Levallois cores from the Gollu Dag survey. Note different levels of patina and surface erosion on the two specimens.

exploitation of Levallois cores. Cores were classified based on the orientation and arrangements of scars on the most recent (final) Levallois surface. Preferen-tial Levallois surfaces, with traces of a single large central flake removal, are the most common (TABLE 2) (FIG. 6: 1, 3). The second most abundant

group of cores preserves traces of unidirectional (or more rarely, bidirectional) recurrent exploitation (FIG. 6: 2, 4). While unidirectional Levallois

pro-duction is often associated with blade manufacture that is not the case on Go¨llu¨ Dag˘, where most of the products of the unidirectional Levallois cores would have been elongated flakes rather than true blades. Centripetal recurrent cores (FIG. 6: 5; FIG. 7) are the third most abundant, followed by a range of rarer morphologies. It is notable that evidence of point production is very scarce: the cores with convergent flake scars produced mainly elongated flakes and blades. There is a single example of a highly atypical Levallois point core in the survey collections.

We had expected that the Levallois exploitation sys-tems would exhibit independent spatial distributions, perhaps indicative of temporal sequencing as occurs in the Mediterranean Levant (Bar Yosef 1998; Copeland 1975), but this does not appear the case on Go¨llu¨ Dag˘. All large samples of Levallois and para-Levallois cores include specimens showing several different systems of exploitation. Every collection with seven or more cores contains examples showing both preferential and unidirectional exploitation, and often other schemes as well. There are a few exceptions to this general ten-dency. Locality 392, for example, yielded a series of large, very well executed centripetal recurrent Levallois cores that are unlike artifacts found in other parts of the survey area. The sample from Locality 575, a buried workshop site adjacent to an obsidian outcrop, consists mostly of products, byproducts and cores from uni-directional recurrent production, including some unique examples of Levallois blades. However, the vast majority of the samples contain evidence of diverse systems of core preparation and exploitation.

It could be the case that different Levallois exploi-tation systems were applied sequentially to the same core, and that the variety we see is the result of cores

abandoned at different stages of reduction. This proposition can be evaluated simply by comparing sizes of discarded cores. Supplemental material 6; (http://www.maneyonline.com/doi/suppl/10.1179/204 2458215Y.0000000020) shows the lengths and thick-nesses of unbroken cores exhibiting different final Levallois surfaces. There are some obvious differ-ences in mean values, which are confirmed by ANOVA analyses that show significant differences in length (F53.745, df56/670, p50.001) and thick-ness (F54.490, df56/670, pv0.001) among some categories of Levallois core. Centripetal recurrent and rare convergent cores tend to be the largest, while the means for the other groups cluster closely together. Cores with a large preferential removal tend to be thinner than others, which is not surpris-ing: they are identified based on having been aban-doned after a large flake was detached. However, as Figure 8 shows, the distributions of dimensions overlap extensively among all categories: each group contains both very large and very small specimens. It is very likely that the ways Levallois surfaces were exploited changed over the life of a core. However, it does not appear that the sequencing of exploitation strategies was highly redundant, such that one strategy was always applied when cores were new (and large) and another when cores were small and worn out. Thus, the diversity of Levallois methods in evidence at the Go¨ llu¨ Dag˘ sites is not likely to reflect the presence of different reduction stages alone.

Two other scenarios may be proposed to explain the co-occurrence of different forms of Levallois core in the sites on Go¨llu¨ Dag˘. One is that Middle Paleolithic toolmakers in the Go¨llu¨ Dag˘ area employed multiple Levallois systems simultaneously, and there is no strongly-expressed temporal sequencing. A second is that nearly every findspot samples multiple occu-pations spanning a long period of time. Given the range of site contexts, which include rapidly buried, in situ deposits as well as surface lag deposits, the second option seems unlikely. We can also conduct a partial test of the second scenario. If the different var-ieties of Levallois core were produced at different times we might also expect them to show different degrees of weathering based on the potential time of exposure on the surface. Of course weathering will also relate to the geomorphological context but this should be independent of the age of Middle Paleolithic artifacts. However there are no significant differences in levels of patina among cores representing the var-ious Levallois systems (Supplemental material 7; http://www.maneyonline.com/doi/suppl/10.1179/2042 458215Y.0000000020). These results weigh in favor of the first scenario, that the diversity of core forms reflects flexible aims and production tactics on the part of Mousterian toolmakers on Go¨llu¨ Dag˘.

Table 2 Frequency of Levallois system on last exploited surface of cores from survey collections. Indeterminate specimens not included.

Levallois system Total

Preferential 335 Unidirectional recurrent 259 Bidirectional recurrent 44 Convergent 30 Orthogonal 40 Centripetal recurrent 147 Opposed preferential 42 Total 897

In addition to the large number of cores, samples of nearly 400 retouched tools (TABLE 3) (FIGS. 9, 10) and 147 unmodified Levallois flakes (FIG. 11) were also

col-lected during pedestrian surveys of Go¨llu¨ Dag˘. The dominance of cores in collections is certainly a partial result of collection strategies. It should be noted that whereas flake tools and Levallois products are prob-ably under-represented within the survey collection as a whole, the bias should be consistent across sites and collections. Moreover, in a dynamic landscape, surface assemblages are likely biased towards the

largest, least mobile artifacts. As fine-grained sedi-ments were removed by gravitational or aeolian forces, cores probably stayed put while flakes and chips moved. Sites with very fresh, recently-exposed artifacts typically show fairly broad range of products as well as byproducts. Nonetheless, the very low abun-dance of Levallois flakes and retouched tools in many Paleolithic assemblages almost certainly reflects the behavior of Middle Paleolithic hominins as well.

As with cores the great majority of retouched tools and Levallois specimens are made of obsidian.

Figure 8 Dot density graphs showing distributions of length and thickness measurements for different varieties of Levallois core (whole specimens only).

Table 3 Retouched tools, typological groups.

Retouched tool class Obsidan Basalt Other volcanic Flint Total Proportion

Mousterian point 15 0 0 0 15 0.054 Limace 5 0 0 0 5 0.015 Simple sidescraper 82 3 1 0 86 0.26 Double sidescraper 27 0 0 0 27 0.082 Cvgt. sidescraper 7 1 0 0 8 0.024 Dejete scraper 20 0 0 1 21 0.060 Transverse scraper 36 2 0 0 38 0.115 “Other” sidescraper 10 0 1 0 11 0.033 Endscraper 13 0 0 0 13 0.039 Burin 1 0 0 0 1 0.003 Percoir 1 0 0 0 1 0.003 Backed knife 5 0 0 0 5 0.015 Truncation 6 0 0 0 6 0.018 Notch 23 1 0 1 25 0.085 Denticulate 27 1 0 0 28 0.085

Piece w/ bifacial retouch 6 0 0 0 6 0.018

Tayac point 1 0 0 0 1 0.003

Rabot 1 0 0 0 1 0.003

Diverse 8 0 1 0 9 0.027

Multiple tools 6 0 0 0 6 0.018

Truncated-faceted piece 11 2 0 0 13 0.039

Flake with partial retouch 20 0 1 0 21

Intdet. Fragment 43 1 0 0 44

The sample of retouched tools is dominated by side-scrapers and retouched Mousterian points (TABLE 3).

The ratio of sidescrapers to denticulates and notches is certainly biased by our collection strategies and analytical protocols. Obsidian is a comparatively brittle raw material and flake edges are easily damaged by fluvial action and by trampling by ani-mals and people. This unintentional damage gener-ally produces irregularly denticulated edges, often with scars on both faces. In order to avoid including large numbers of naturally damaged pieces in our counts we were extremely critical in classifying arti-facts as denticulates or notches. As a consequence these artifact classes are probably significantly under-counted in the sample.

Although evidence of Levallois production is in evidence everywhere on Go¨llu¨ Dag˘, Levallois blanks are not especially common as tool blanks (TABLE 1) (Supplemental material 8; http://www.

maneyonline.com/doi/suppl/10.1179/2042458215Y.00 00000020) (FIG. 9). Instead, many tools are made

from flakes produced early in core reduction. The majority of retouched tools are made on ‘‘plain’’ flakes or blades. Nearly 55% of retouched pieces have some primary cortex (the natural surface of the original nodule) and about 16% have cortex covering more than half of the dorsal surface. A very large proportion also have secondary cortex (weathering). Most Levallois blanks and most unmo-dified Levallois products are flakes. There are some

Figure 9 Retouched flake tools from the Gollu Dag survey. 1) simple sidescraper; 2, 7) transverse scrapers; 3) retouched point with thinned butt; 4) truncated/faceted piece; 5) double sidescraper; 6) dejete scraper; 8) denticulate/perc_{oir. All} obsi-dian except n. 5, which is basalt.

blades or elongated flakes, while true Levallois points are essentially absent. Counting both modified and unmodified Levallois blanks, the survey collections contain about four times as many Levallois and proto-Levallois cores as it does obvious products. Collection biases notwithstanding, this highly imbal-anced ratio certainly indicates that Middle Paleo-lithic people exported Levallois products from sites located near to the obsidian for use elsewhere.

There is also evidence, albeit limited, for movement of artifacts into the Go¨llu¨ Dag˘ Middle Paleolithic sites. As might be expected in an area with so much high-quality raw material, many flake tools show only minimal levels of retouch and resharpening. However, there are also some very heavily reworked pieces, including several limaces and even a few scra-pers with Quina retouch. This shows that not all arti-facts were made locally and discarded quickly on the spot: some seem to have experienced long use lives before being abandoned. Two retouched tools made of flint which is not of local origin were also collected. Flint source attribution in Anatolia is still in its infancy (e.g., Nazaroff et al. 2013) and we cannot say where these raw materials originated.

Variation in use of the Go

¨llu

¨ Dag˘ landscape

Given the abundance of obsidian it is not surprising that raw material procurement and artifact pro-duction are the best-represented activities on Go¨llu¨ Dag˘ from the Mousterian through the Chalcolithic. In the later periods it appears that people visited the mountain mainly if not exclusively to get obsi-dian. However, the same is not true of the

Paleolithic. The range of stone artifacts present and their conditions suggest that Paleolithic hominins pursued a range of activities on Go¨llu¨ Dag˘, using and discarding tools as well as making them. The fol-lowing analyses examine evidence for variation in activities across the survey area.

We begin by subdividing space only in terms of dis-tance from obsidian outcrops. Table 4 shows the breakdown of Paleolithic artifact inventories accord-ing to binned distances from the nearest documented obsidian outcrop. The intervals are not even, but rep-resent natural breaks in the distribution of finds and sites. The great majority of sites and findspots are within two km of an obsidian source, and some are much closer. Distance effects are detectable but are not very pronounced. At all distances except the largest (w2200 m), cores make up more than 50% of the col-lections, reflecting the importance of artifact pro-duction within the Go¨llu¨ Dag˘ survey area. Only beyond 2,200 m do ‘‘end products’’ (flakes and retouched tools) predominate, and at such distances from obsidian the artifact inventories are small. Whereas the overall frequency of cores reflects collec-tion bias, the same colleccollec-tion procedures were applied to all sites, so the pattern of frequency fall-off is likely genuine even if the actual numbers are inflated. More-over, retouched flake tools make up more than 16% of survey collections even in locales within 100 m of an obsidian source. This suggests that even where artifact production was most common it was not the only activity carried out. It is also a marked contrast to later periods, when dedicated workshops dominate the record. Levallois products are most common

either close to obsidian or far from it, which could reflect a dichotomy between pieces discarded at the production site and pieces selected for transport. Note that unretouched pieces other than Levallois or technical elements were not collected systematically. The category ‘‘other flake/blade’’ consists mainly of unretouched ‘‘technical’’ pieces such as crested

blades, eclats de´bordants and other products derived from preparation and renewal of cores.

In order to further examine variation in activities across the Go¨llu¨ Dag˘ landscape we compare three parts of the study area with different attributes that might influence how humans used them. The slopes of Bitlikeler ridge, in the north-central part of the

Table 4 Inventories of Paleolithic survey collections according to distance from obsidian outcrops. Distance from obsidian Levallois cores Other cores Levallois

product Flake tool

Handaxe/ biface

Other flake/ blade

Other

core tool Total , ¼ 100 m 373 (53.8%) 67 (9.7%) 68 (9.8%) 113 (16.3%) 31 (5.3%) 37 (5.3%) 4 (0.6%) 693 100 – 200 m 139 (53.9) 30 (11.6) 13 (5.0) 48 (18.6) 22 (8.5) 2 (0.8) 4 (1.6) 258 200 – 850 m 330 (54.1) 68 (11.1) 34 (5.6) 112 (18.4) 53 (8.7) 3 (0.5) 10 (1.6) 610 850 – 1250 m 107 (41.0) 26 (10.0) 16 (6.1) 77 (29.5) 29 (11.1) 0 (—) 6 (2.2) 261 1250 – 2200 m 38 (40.0) 12 (12.6) 10 (10.5) 28 (29.5) 5 (5.3) 0 (—) 2 (2.1) 95 .2200 m 5 (21.7) 4 (17.4) 1 (4.3) 13 (56.5) 0 (—) 0 (—) 0 (—) 23 Total 992 (51.1) 207 (10.7) 142 (7.3) 391 (20.2) 140 (7.2) 42 (2.2) 26 (1.3) 1,940 Figure 11 Levallois flakes from Gollu Dag survey. Specimens 1 – 3 are from same location. Note heavier level of surface ero-sion and recent (unpatinated) damage on the 4th specimen.

study area, contain extensive outcrops of high-quality obsidian. Extensive Neolithic and Chalcolithic work-shops associated with these outcrops have been known for decades. Ballıkaya, in the southern part of the survey area, also contains obsidian outcrops of moderate quality. The soft rhyolite cliffs also contain small rockshelters suitable for habitation. Ku¨c¸u¨k Go¨llu¨ Dag˘, also at the southern end of the study area, is the collapsed caldera of a small volcanic dome. Today there is a shallow, marshy, precipi-tation-fed seasonal lake at one end of the caldera. Ku¨c¸u¨k Go¨llu¨ Dag˘ contains many cliffs and rock walls with shallow rockshelters around the summit but no usable obsidian deposits, although rich sources of the material can be found within a kilometer or two. Simply based on the local affordances (obsidian, rock-shelters) we might expect that activities on Bitlikeler would be connected mainly with exploitation of obsi-dian, whereas human presence on Ku¨c¸u¨k Go¨llu¨ Dag˘ would be a response to other resources or needs. Ballı-kaya, which contains both obsidian outcrops and natu-ral shelters, might be expected to show an intermediate mix of activities, with a combination of both work-shops and habitation sites.

The general compositions of Paleolithic assemblages from these three areas, as well as the conditions of arti-facts at discard, summarized in Supplemental Material 9 and 10 (http://www.maneyonline.com/doi/suppl/ 10.1179/2042458215Y.0000000020), fit fairly well with the expectations outlined above. Assemblages from Ballıkaya and Ku¨c¸u¨k Go¨llu¨ Dag˘ contain much higher proportions of reduction products (flake tools and Levallois flakes) (39–40%) than does Bitlikeler (14%). This is consistent with the expectation that Bitlikeler should be more dominated by workshops, whereas the other two areas should have supported a wider range of activities. The elevated frequency of unmodified Levallois flakes at Ballıkaya may reflect the fact that the area also has a significant workshop component, as most of the Levallois flakes in this collection are broken.

Although cores comprise between 52% and 63% of the collections from all three areas, their conditions vary. On average, the cores collected from Bitlikeler are both larger and thicker than specimens from Ku¨c¸u¨k Go¨llu¨ Dag˘, while the sample from Ballıkaya shows intermediate means for these variables. These differences are entirely consistent with the availability of obsidian and evidence for other activities. The areas with the most abundant and highest-quality obsidian, Bitlikeler, yields the larger and presumably less extensively reduced cores, whereas the area with-out obsidian with-outcrops, Ku¨c¸u¨k Go¨llu¨ Dag˘, contains smaller, more reduced cores.

We emphasize that the differences between the three areas are quantitative rather than categorical.

Both the compositions of the collections and the con-ditions of artifacts overlap across the three areas. Flake tools were discarded more frequently in Ku¨c¸u¨k Go¨llu¨ Dag˘ than in other areas, but still more than half of the survey collection from the area consists of cores. Conversely, although work-shop activities may have been dominant at sites on the Bitlikeler ridge, there is still evidence for other activities involving production and discard of shaped tools. A few Paleolithic sites on Bitlikeler, such as Locality 280, contain comparatively high frequencies of retouched pieces and Levallois pro-ducts, whereas others yielded mainly cores. Likewise, the distributions of core sizes in the three areas over-lap substantially. The differences described here likely reflect shifting frequencies of different activities across space rather than a strict geographic compart-mentalization of activities and site functions.

Discussion and Conclusions

The Go¨llu¨ Dag˘ area contains a robust record of Lower and Middle Paleolithic occupations. The Lower Paleo-lithic evidence is limited mainly to handaxes and other LCTs because of collection biases and problem of site recognition. The Middle Paleolithic record is more plentiful, more broadly distributed and more varied. Middle Paleolithic sites from within about 2 km of an obsidian source usually include a strong component of debris from in situ manufacture, and it is likely that some products, especially Levallois flakes, were removed to other locations. However, many findspots, even ones located very close to obsidian outcrops, also contain retouched tools and other ‘‘products.’’ This indicates that Middle Paleolithic hominins did not visit Go¨llu¨ Dag˘ for the sole purpose of collecting and working raw materials, but rather that they lived in the area, at least temporarily.

Although we do not have direct dates any of the Paleolithic localities at present, the geographic pos-ition of these sites is itself significant. The presence of numerous Paleolithic artifacts on Go¨llu¨ Dag˘ pro-vides evidence of hominin adaptations to moderately high elevations and continental climatic conditions. The basal level for the survey area is approximately 1400 m, and the average elevation for Paleolithic sites recorded during the survey is just over 1600 m. Paleolithic artifacts were collected from elevations as high as 2000 m. What remains to be seen is whether the early Paleolithic occupations were con-fined to interglacial periods sites (e.g., Ghukasyan et al. 2011) or whether hominins continued to use the area during harsher glacial periods.

The frequent use of flakes as blanks for handaxes and other LCTs in the survey collection suggest that we are dealing with a local variant of the ‘‘Large Flake Acheulean.’’ Evidence from the

excavated site KD3 suggests a mid-Middle Pleisto-cene age for the Acheulean assemblages with large flakes, although other earlier or later assemblages may be present as well. In the Jordan River valley the ‘‘Large Flake Acheulean’’ seems to begin around 780 ka (Goren-Inbar et al. 2011) but we do not know when the production of similar artifacts began in Central Anatolia or how long it lasted. Although the potential duration of the Acheulean is much greater than that of the Mousterian, the earlier period is represented by many fewer diagnostic arti-facts in the survey collections. If not a function of recognition biases, this fact could be telling us that the Lower Paleolithic occupation of Go¨llu¨ Dag˘ was less intense and continuous than the Middle Paleo-lithic occupation, perhaps confined to warm intergla-cial periods. However, experience at KD3 suggests that in situ Acheulean-age artifact-bearing sediments are likely buried deeply and/or covered with recent colluvium and will only be recognized through excavation.

The evidence from Go¨llu¨ Dag˘ helps fill in the map of Lower Paleolithic industries in Turkey. Numerous finds of handaxes in open-air contexts, especially on the terraces of the Euphrates river, have confirmed the presence of Acheulean in southeast Anatolia (Albrecht and Mu¨ller-Beck 1988; Bostancı 1962; Dinc¸er in press; Garrard et al. 2004; Minzoni-De´roche and Sanlaville 1988; Tas¸kıran 1988, 2002). Although there have been occasional reports of handaxes from the central Anatolian plateau for decades, the better controlled evidence from KD3 and this survey con-firmed that there was also a substantial Acheulean occupation at higher elevations. The western edge of the Acheulean distribution in Turkey remains an open question. A few handaxes have been collected from the eastern shores of the Bosporus (Jelinek 1980), again from surface contexts. So far, surveys in Thrace (European Turkey) have identified likely Lower Paleolithic assemblages with choppers, cores and flake tools, but no clear Acheulean (Dinc¸er 2011; Dinc¸er and Slimak 2007; Runnels and O¨ zdog˘an 2001). Material traces of Middle Paleolithic occupation on Go¨llu¨ Dag˘ are both widespread and abundant. Levallois cores, Levallois products, and typical Mousterian retouched flake tools are virtually ubi-quitous within the survey area. Like all cultural remains, they are concentrated in certain areas and deposits, and are particularly common in close association with outcrops of obsidian. The largest group of Middle Paleolithic artifacts collected con-sists of Levallois cores. The sample of cores shows a range of different production systems on the final Levallois surface. Centripetal preferential Levallois surfaces are most common but most other variants on the Levallois theme are present as well. Cores

showing different systems of preparation and exploi-tation tend to co-occur in the same localities, and based on the admittedly limited surface evidence, there is no evidence of separation or temporal sequencing in Levallois variants. At KD3, techno-logically similar Mousterian assemblage surface materials occur both immediately above and below a tephra from the Acıgo¨l volcano which erupted 190–200 ka (Schmitt et al. 2011). However, we cur-rently have no narrower time constraints on the Middle Paleolithic occupations of Go¨llu¨ Dag˘.

There are no local comparators to the Middle Paleolithic assemblages of Go¨llu¨ Dag˘: in fact, there are no other well-documented Mousterian assem-blages from within a 300 km radius of the mountain. Owing to the dominance of Levallois technology, the Middle Paleolithic collections from Go¨llu¨ Dag˘ as a group bear a general resemblance to Levallois Mous-terian assemblages from the eastern Mediterranean Levant. Because of the presence of varied production systems, mainly centripetal and preferential, they are most similar to the ‘‘middle-Middle Paleolithic’’ from sites such as Qafzeh (Hovers 2009) and Tabun (Jelinek 1981). Somewhat similar Middle Paleolithic assemblages, characterized by centripetal and uni-polar Levallois production have been reported from the Mediterranean coast south of the Taurus Moun-tains (Kuhn, 2002). On the other hand, the material from Go¨llu¨ Dag˘ is very different from both the Taurus/Zagros Mousterian and ‘‘proto-Charentian’’ assemblages of Karain cave (near Antalya to the southwest), the most extensively-excavated Paleo-lithic site in Turkey (Otte et al. 1998, 1999). However, any comparison must be made with caution due to differences in the nature of the sites and raw materials. At Karain most artifacts were manufac-tured using small nodules of radiolarite. Moreover, the site was probably used as a habitation for most of its history. In contrast, the Go¨llu¨ Dag˘ sites are adjacent to abundant, large size raw materials, and many have a substantial component of in situ reduction. Thus, the small size and consistently advanced reduction of tools which are hallmarks of the Karain Mousterian are unlikely to have been expressed in the Go¨llu¨ Dag˘ sites, regardless of the other technological choices made by Middle Paleo-lithic groups.

As discussed above, the diversity of Levallois var-iants within sites on Go¨llu¨ Dag˘ and the similarities across sites suggest a number of possible scenarios. It could be that we are dealing with a single local tra-dition characterized by diverse Levallois strategies and lacking the kinds of temporal trends in Levallois found further south in the Levant. An alternative possibility is that the sites all pertain to a single period of occupation, perhaps related to the

‘‘C-type’’ or ‘‘middle’’ Levantine Mousterian. The middle Mousterian at Qafzeh dates to MIS 5 (Schwarcz et al., 1988; Valladas et al. 1988) although the thermoluminescence results from Tabun place it earlier in time, ca. 165–200 ka (Mercier and Valladas 2003). Based on its stratigraphic relationship to the Acıgo¨l tephra, dated to 190–200 Ka, some if not all the Mousterian at the excavated site KD3 predates this interval by tens of millennia. However we cur-rently have no basis for determining the span of time that the Mousterian assemblages in the Go¨llu¨ Dag˘ survey area represent. Consequently the Levan-tine sequence is not an appropriate model for assign-ing different kinds of Levallois technology in Anatolia to different periods or hominin taxa.

Although there are clear similarities, the collec-tions from Go¨llu¨ Dag˘ also differ in several ways from the Levantine Levallois-Mousterian. The most important is the rarity on Go¨llu¨ Dag˘ of both Leval-lois points and blades, typical of the Levantine sites at various periods. The scarcity of these artifacts within the study area does not appear to be a result of using obsidian as a raw material. Assemblages from the Middle Paleolithic in Armenia (Chataigner et al. 2003: 13; Ghukasyan et al. 2001) and the Middle Stone Age in Ethiopia (Wendorf and Schild 1974; Sahle et al. 2014) show that it is possible to make Levallois blades and points from obsidian. Instead, the predominance of flake production, whether by centripetal, unidirectional or bidirec-tional technique, does appear to have been the result of choices by prehistoric artisans. Whether those choices reflect functional or logistical consider-ations or simple cultural biases remains to be determined.

Given the potential role of Anatolia as a pathway for dispersal of anatomically modern Homo sapiens from the Near East into Europe, the possible pre-sence of Upper Paleolithic was of particular interest at the outset of the project. In situ Upper Paleolithic deposits are very rare in Turkey generally, and almost unknown in central Anatolia. The large exposures of Pleistocene sediments and ready avail-ability of high-quality raw material makes the Go¨llu¨ Dag˘ area an ideal place to prospect for Upper Paleolithic sites. However, the survey resulted in the collection of just a handful of possible Upper Paleolithic artifacts, and no obvious Upper or Epipa-leolithic sites were identified.

There are a number of potential explanations for the scarcity of Upper or Epipaleolithic materials on and around Go¨llu¨ Dag˘. It could simply be a problem of recognition. Whereas Middle and Lower Paleo-lithic artifacts stand out clearly from later ones, Upper Paleolithic peoples generally used prismatic blade technologies similar (if not identical) to those

of later Neolithic groups. As researchers in Armenia have found (Gasparyan et al. 2014: 109) it may be difficult to recognize an Upper Paleolithic site, par-ticularly a workshop, when it is blanketed in debris from Neolithic and Chalcolithic blade production. If this is the case Upper Paleolithic deposits will only come to light after stratigraphic excavation. On the other hand, we did not encounter obvious Upper Paleolithic sites or artifacts even in areas more distant from the obsidian where they would not be obscured by material from later periods.

A second possibility is that the scarcity of Upper Paleolithic remains is simply a function of a narrower time window. The Eurasian Upper Paleolithic lasted at most 35,000 years. The duration of the Middle Paleolithic, sensu lato, is at least six times greater, and the potential duration of the Lower Paleolithic longer still. In this case Upper Paleolithic sites and artifacts are simply rare and may come to light only after more intensive reconnaissance.

A third possibility is that Upper Paleolithic groups simply spent little time in the area during the colder parts of MIS 3 and 2. A handful of pieces of central Anatolian obsidian have been identified in late Epi-Paleolithic (post-glacial) assemblages from O¨ ku¨zini cave near Antalia (more than 300 km away) (Carter et al. 2011), demonstrating that people did sometimes visit the Go¨llu¨ Dag˘ area during the terminal Pleisto-cene, but the human presence may have been so ephemeral that it is difficult to detect. To date only two sites with definite late Epi-Paleolithic layers have been identified and investigated in central Turkey (Arbuckle and Erek 2012; Baird et al. 2013). Globally, early Upper Paleolithic sites at least are reportedly rare above 1000 m or so in Iberia (Straus et al. 2000: 567), the French Alps (Evin et al. 1994: 352) and Armenia (Gasparyan et al. 2014; Pinhasi et al. 2008: 812), although they do occur in Italy (Mussi 2006). However, later Upper Paleolithic or Epipaleolithic sites dating to after the peak of MIS 2 are common in many areas, so their scarcity on Go¨llu¨ Dag˘ is noteworthy. A final possibility is that Upper and Epi-Paleo-lithic groups exploited the obsidian in ways that left only minimal records. Rather than living in the immediate area (as did Middle Paleolithic groups) or creating large workshops (as did Neolithic and Chalcolithic peoples), Upper Paleolithic groups could have carried minimally shaped blocks to work-shops or campsites located outside the survey area for further processing, leaving very little evidence around the obsidian sources. This is the situation in the Bronze Age, for example. More than one ton of Go¨llu¨ Dag˘ obsidian was collected at the Early Bronze Age (EB) site of Ku¨ltepe (Altınbilek-Algu¨l and Balcı 2010), but there is no large EB settlement

on Go¨llu¨ Dag˘. Unfortunately, the lack of systematic survey or excavation in surrounding areas makes it impossible to evaluate this alternative scenario for the moment.

Like all surveys, the study reported here can be considered a first stage of research. We have ident-ified areas with varying density and content of archaeological finds, documenting different intensi-ties of occupation and different kinds of land use on Go¨llu¨ Dag˘. The next stage of research must involve more intensive field studies tailored to par-ticular situations, including more continuous record-ing of surface artifacts, systematic study and documentation of large workshops, and excavation of intact localities. The timing and enivronmental contexts of early hominin presence on the Anatolia plateau are particularly important issues with respect to the Paleolithic occupation of Go¨llu¨ Dag˘, so additional geochronological research will be especially important. Studies of allochthonous micro-tephras have proven useful in providing some chronological controls (Slimak et al. 2007; Tryon et al. 2010). In some parts of the area Middle Paleo-lithic artifacts are also found within or capped by soil carbonate crusts (caliches), which can be dated by uranium-series methods (e.g., Ludwig and Paces 2002; Sharp et al. 2003).

Acknowledgments

The Go¨llu¨ Dag˘ survey could not have been carried out without the contributions of many individuals. Dr. Laurence Astruc, Dr. Semra Balcı and Dr. Nurcan Kayakan are responsible for the study of the Neolithic and Chalcolithic sites. They were in the field every season and, despite having much more material from the later periods to deal with, showed appropriate enthusiasm about documenting the Paleolithic localities. We would also like to thank Dr. Erhan Bıc¸akcı for his frequent help with equipment, supplies and expertise. This work was realized with the permission of the Department of Cultural Assets and Museums, Ministry of Tourism and Culture. It was supported financially by the Research Fund of Istanbul University and travel grants from the University of Arizona. We also thank three anonymous reviewers for critical com-ments that helped us improve the manuscript. Line drawings of artifacts are the work of Zehra Tas¸kıran. Berkay Dinc¸er is responsible for artifact photos and M. Korhan Erturac¸ produced the GIS maps. Steven Kuhn (Ph.D. 1990, University of New Mexico) is Rieker Distinguished Professor in the School of Anthropology at the University of Arizona. His inter-ests center on the behavioral evolution of Homo sapiens, Neanderthals and earlier hominins, with a

particular emphasis on stone tool technology and other forms of material culture.

Berkay Dinc¸er is a Ph.D candidate in the Department of Prehistory, Faculty of Letters, Istanbul University, and a researcher at Ardahan University, Turkey. His doctoral research concerns the Paleolithic of western Turkey and Anatolia.

Nur Balkan-Atlı (Ph.D. 1985, La Sorbonne) is Pro-fessor in the Prehistory Department, faculty of Letters, Istanbul University. She is an expert in the obsidian resources of Anatolia, and in the exploitation and exchange of obsidian across western Asia during the Neolithic.

Mehmet Korhan Erturac¸ (Ph.D. 2009, Eurasia Insti-tute of Earth Sciences, Istanbul Technical University) is an Assistant Professor in the Department of Geogra-phy at Sakarya University, Turkey. His research inter-ests include remote sensing and GIS applications in landscape research, geochemical characterization of obsidians, morphotectonics, terrestrial climate records, and scientific dating methods.

References

Albrecht, G., and H.-J. Mu¨ller-Beck. 1988. ‘‘The Palaeolithic of S¸ehremuz near Samsat on the Euphrates River. Summary of the Excavation Findings and Morphology of the Handaxes,’’ Pale´orient 14(2): 76–86.

Altınbilek-Algu¨l, C¸ ., and S. Balcı. 2010. ‘‘Obsidiyen Ticaretinin Merkezi Olarak Ku¨ltepe,’’ TEBE Haberler 30: 11–13. Arbuckle, B. A., and C. M. Erek. 2012. ‘‘Late Epipaleolithic

Hun-ters of the Central Taurus: Faunal Remains from Direkli Cave, Kahramanmaras¸, Turkey,’’ International Journal of Osteoarch-aeology 22(6): 694–707.

Baird, D. 1996. ‘‘The Konya Plain Survey: Aims and Methods,’’ in I. Hodder, ed., On the surface: C¸atalho¨yu¨k 1993–1995. Cambridge: Cambridge University Press, 41–46.

Baird, D., E. Asouti, L. Astruc, A. Baysal, E. Baysal, D. Car-ruthers, A. Fairbairn, C. Kabukcu, E. Jenkins, K. Lorentz, C. Middleton, J. Pearson, and A. Pirie. 2013. ‘‘Juniper Smoke, Skulls and Wolves’ Tails. The Epipalaeolithic of the Anatolian Plateau in its South-west Asian Context; Insights from Pınarbas¸ı,’’ Levant 45(2): 175–209.

Balkan-Atlı, N., D. Binder, and M. -C. Cauvin. 1999. ‘‘Obsidian Sources, Workshops and Trade in Central Anatolia,’’ in M. O¨ zdog˘an, and N. Bas¸gelen, eds., Neolithic in Turkey. Istanbul: Arkeoloji ve Sanat Yayınları, 133–145, figs. 104–115.

Binder, D., B. Gratuze, D. Mouralis, and N. Balkan-Atlı. 2011. ‘‘New Investigations of the Go¨llu¨dag˘ Obsidian Lava Flow System: a Multi-disciplinary Approach,’’ Journal of Archaeolo-gical Science 38: 3174–3184.

Bostancı, E. 1962. ‘‘Researches in South-east Anatolia, the Chellean and Acheulean Industry of Du¨lu¨k and Kartal,’’ Anadolu/Anatolia 6: 111–162.

Campbell-Thompson, R. 1910. ‘‘On Some Pre-historic Stone Implements from Asia Minor,’’ Man 10: 71–72.

Carter, T., F. -X. Le Bourdonnec, M. Kartal, G. Poupeau, T. Calligaro, and P. Moretto. 2011. ‘‘Marginal Perspectives: Sourcing Obsidian from the O¨ ku¨zini Cave (SW Turkey),’’ Pale´orient 37(2): 123–149.

Cauvin, M. -C., and C. Chataigner. 1998. ‘‘Distribution de l’obsi-dienne dans les sites arche´ologiques,’’ in M. -C. Cauvin, A. Gourgaud, B. Gratuze, N. Arnaud, G. Poupeau, J. -L. Poi-devin, and C. Chataigner, eds., L’Obsidienne au Proche et Moyen Orient: du Volcan a´ l’Outil, BAR International Series 738. Oxford: British Archaeological Reports, 325–350.