Introduction

The Balkans and Middle Paleolithic Occupation Intensity

While Neanderthals inhabited large portions of Eurasia, including most of Europe, southwest Asia and eastward into Asia as far as the Altai, the archaeological record of their presence is not evenly distributed across this region. Certain regions, such as southwest France, exhibit a rich and continuous presence of Neanderthals. The Balkan Peninsula, like several other regions in Europe, lacks evidence of their continuous presence, either spatially or temporally. The reasons behind the disproportionate density of the record across Eurasia may lie in a less intensive research history compared to other areas, or in differences in occupational intensity in the Pleistocene, or some combination of the two, although most probably reasons vary between regions. For instance, it has been suggested that the Pleistocene record may no longer be visible, in some parts of eastern Europe where thick loess deposits cover potential Lower Paleolithic sites (Iovita et al., 2012; Romanowska, 2012).

Aside from factors that can alter the preservation through time of human occupations (Surovell et al., 2009), the incomplete and mosaic-like presence of Neanderthal occupation throughout Eurasia may have several explanations. Variations in the biogeography of a population can arise in relation to climate, environment, topography, settlement preference, existence of migration routes, etc. These, in turn, affect the demography of a population, which may go through local extinctions or withdrawals and create specific spatial patterns across large regions with continuities and gaps in the record of human settlements. The combination of these factors, on a larger scale, creates the patchy character of the occupation record. The varied presence of Neanderthals has been more intensively investigated in central, northern, and north-western Europe (Gaudzinski-Windheuser & Roebroeks, 2011; Hublin & Roebroeks, 2009; J. Richter, 2016; Roebroeks et al., 2011; Wenzel, 2007), but some have made a Europe-wide assessment (Davies et al., 2003a, b; J. Richter, 2005, 2006). Occupational discontinuities are more distinct in the northern areas, which suggests that during climatic deteriorations hominins migrated from the north to southern areas and then re-colonized the north once the climate ameliorated. This 'ebb and flow' model implies that population fluxes, in response to deteriorating climatic conditions, leave a distinct spatio-temporal pattern in the archeological record. However, it is alternatively possible that Neanderthals experienced a series of regional extinctions (Hublin & Roebroeks, 2009; Roebroeks et al., 2011), while other populations would have survived in the southern regions and persisted throughout the glacial periods. In either case, in many models parts of southern Europe, including the Balkans, figure as favorable regions for habitation during climatic fluctuations and are expected to exhibit a rather continuous record (Dennell et al., 2011; Finlayson, 2008; Hublin & Roebroeks, 2009; Roebroeks et al., 2011; Serangeli & Bolus, 2008; J. Stewart, 2005).

For a large region, the Balkan peninsula has a low number of Middle Paleolithic locations, particularly sites with long and securely dated sequences, aside from a couple of known sites (e.g. Krapina, Vindija, Lakonis). Regardless of the reasons for the sparse Pleistocene record, the region has insufficient information for understanding the adaptations of Pleistocene hominins. This appears paradoxical: if the Balkans are one of the southern refugia, and if certain areas offer topographically and environmentally favorable conditions for habitation, why is the record so sparse? The Balkans certainly merit further research, particularly if we consider their large area and particular ecological and geographical position within Europe. In this context, several questions arise. Were the Balkans occupied by ‘core’ populations, particularly during the harsher climatic conditions, thus evidencing a continuous human presence, or were they host to low population densities and often devoid of human presence? Furthermore, given the climatic and geographical diversity of the peninsula, are there any correlations between these factors and human presence? Were Neanderthal adaptations in this region, away from the ‘heartland’ of Europe in the west, somewhat different, relative to the Balkans’ geography, varied topography, climate, and paleoenvironment?

The Balkans and Middle Paleolithic Industrial Variability

Typological classification systems were the initial tool for understanding variability in Middle Paleolithic material culture, and they gave rise to the early definitions of separate Mousterian varieties, mainly in southwest France (Bordes, 1953; Bordes & Bourgon, 1951; Mellars, 1965). For other regions within the geographic range of the Middle Paleolithic, the established nomenclature was accepted and sometimes new groups were defined. While Balkan Paleolithic researchers have generally adopted the Western classification, there has also been a concern about whether these groups can be easily and appropriately applied to the regional record.

The past several decades of research into Middle Paleolithic lithic technologies have shown that typology is not an adequate reflection of different technological traditions, and rather indicated instead differential tool reduction practices (Dibble, 1987, 1995). In this view, a large part of typological variability is related to how intensely sites have been occupied and resources utilized, as a response to changes in climate and environment (Rolland, 1977, 1981; Rolland & Dibble, 1990). This work, along with the work of the French technological school, helped shift the field away from tool types to a greater emphasis on the technology of blank production in assemblage interpretations. The technological repertoire of Middle Paleolithic material culture consists of diverse methods of blank production, for example, Levallois, discoidal, bifacial, Quina, and various core-on-flake methods (Boëda, 1993, 1994; Bourguignon, 1997; Delagnes et al., 2007; Peresani, 2003; Tixier & Turq, 1999; Turq, 1989), and the predominance of one over others led researchers to group them into Mousterian variants (e.g., Delagnes et al., 2007). This research continues today with the re-definition of groups and the addition of new groups, this time based on predominant technological methods. Balkan Paleolithic archaeology, however, lagged behind these interpretative changes. It did not fully adopt the research transformation from a typological to a technological basis for the definition of technocomplexes or the recognition of other factors affecting assemblage variability. When Paleolithic research in Europe almost abandoned the terms Typical and Charentian Mousterian, Balkan archaeology was contending with the applicability of these terms to the regional record. Further work is still needed to deconstruct the typological groups and factor in various aspects that influence assemblage variability.

Setting aside the question of how the different Middle Paleolithic techno-complexes are defined, variation in Neanderthal technological behavior is potentially patterned at both temporal and regional scales (Discamps et al., 2011; Jaubert, 2010; Monnier & Missal, 2014; Pettitt, 2003; J. Richter, 2000; Ruebens, 2013; Thiébaut et al., 2014). For instance, prismatic blade assemblages are particularly common in MIS 5 in northern Europe (Delagnes, 2000; Koehler, 2011; Révillion, 1995; Tuffreau & Révillion, 1995), while on the other hand, there is still some debate about how chronologically constrained technologies like Quina (Discamps et al., 2011; Guérin et al., 2016; Guibert et al., 2006; Le Tensorer, 1978; Morin et al., 2014; D. Richter et al., 2013a, 2013b) or bifacial (Monnier & Missal, 2014; J. Richter, 2016; Ruebens, 2013) are. Some have argued that the patterning in lithic industries in southwest France is responding to changes in climate and environment. Whether the industrial variability in the Balkans is similarly patterned is a question that can be answered only after the record has been updated.

Besides being a potential refugium for Neanderthal populations, the Balkans are noteworthy for their position at the entrance to Europe from southwest Asia and from the Eurasian steppes to the north and east. The region represents a transit zone and can provide a valuable record of patterns of dispersals, timing, routes, and possible hominin interactions (Fu et al., 2015). In this regard, an open question is Neanderthals’ longer persistence in this region than in other areas of Europe, and whether their demise is associated with climatic factors or with incoming modern humans. Regarding the latter, what scenarios can be envisioned for population replacement in this region, which modern humans entered from southwest Asia prior to reaching central Europe?

Given these many questions, this review attempts to present the current Middle Paleolithic (MP) record from the Balkans and our understanding of Pleistocene hominin adaptations, with a particular emphasis on the variability of their technological behavior and population history apparent from the available record. While recent and ongoing research in the Balkans is conducted in many countries, an overview of the available record is so far missing. The patchiness of the Paleolithic record of Europe, and certainly of the Balkans, is to a certain extent a result of variations in the level of attention paid to science at the political level, and of the international connectedness of any given country (Gamble, 1986). Most reviews tend to cover the record within the political borders of a particular country (Darlas, 2007; Ivanova & Sirakova, 1995; Karavanić, 2004, 2007; Tourloukis & Harvati, 2017). Here instead we try to adhere to geographical parameters and to incorporate the region of the Balkan peninsula.

In our review of the MP record, we follow a geographic rather than political definition of the region and consider topographic and environmental parameters of relevance to Pleistocene human occupations. In our case, however, this approach would oblige us to review the record north of the Danube and Sava, that is, Romania, Hungary, and northern parts of Serbia and Croatia usually considered part of the Balkans, because this constitutes the same environmental and topographic zone as the northern parts of the Balkan Peninsula, including the wide Danube and Sava valleys. While we will refer to the MP record of these areas, nonetheless we will not thoroughly present this area in our review and keep largely within the accepted limits of the peninsula. This might, for instance, leave out the important site of Vindija. Wide river valleys in the north are relevant for human migrations, and the record from this region will be accounted for in the discussions of early modern human dispersal into Europe.

We begin with a geographical, paleoecological, and paleoclimatic review of the region that sets the scene for examining human occupations and adaptations in the sections that follow. With the vast region of the Balkan peninsula, subregional variation in landscapes and climate may be crucial for understanding the adaptations and particularly the migrations of past populations. To get a better picture of the variability of MP industries and how they change through time, we chose to review the record chronologically rather than by techno-complex. While we are fully aware of the palimpsest problem in the archaeological record (G. Bailey, 2007; Holdaway & Wandsneider, 2008; Stern, 1994), and finer resolution is seldom realistic (Aldeias et al., 2012; but see S. J. P. McPherron et al., 2005; Rezek, 2015), we are still interested in observing the variability within periods of more or less similar climatic conditions, that is, Marine Isotope Stages (MIS).

While a chronological approach to the regional record is preferred in this case, it is also true that this approach is challenging, given the relative lack of chronological information and absolute dating for many key sites. Chronological framing of assemblages has often been based on stratigraphy and geology with the help of climatically sensitive faunal species. Therefore, this review will be primarily based on sites with reliable chronological estimates, and to a somewhat lesser extent on stratified sites where some temporal trends in technology can be observed but assemblages cannot be chronologically constrained. A more detailed review of these sites, their sequences, and lithic assemblages, together with sites where either no chronological estimates can be made or insufficient data is available for assessing temporal patterns in technology (e.g., non-stratified sites) is presented in the Online Supplementary Materials.

It should also be mentioned that many assemblages come from collections excavated during the early twentieth century and analyzed with outdated methods. Lithic data was collected with variable methods, and with incomplete data, for instance, usually only information relevant to the F. Bordes method is available. A technological analysis, aside from the Bordian calculations of Levallois index (e.g., Ivanova, 1979) is often missing. There are some early attribute analyses (e.g., Sirakov, 1983), though many assemblages have not been subjected to recent re-analysis. The many inconsistencies in the way industries are reviewed here are a result of the non-standardized way these assemblages have been analyzed and reported. A final caveat is that a consistent pattern for MP sites from the Balkans is that they represent evidence of ephemeral occupations, with assemblages too small for clear patterns in lithic technology to be expressed.

Geography and Paleoenvironment of the Balkans

Geography of the Balkan Peninsula

The Balkan peninsula is often thought to coincide with a geographical region of Southeastern (hereafter SE) Europe. The two terms are used interchangeably, and the latter has been widely used over the former in recent decades, to avoid historical, political, and cultural connotations (Todorova, 1997). The Balkan peninsula is, nevertheless, a part of SE Europe. SE Europe, as a political region of Europe, consists of the following countries whose territories are fully or partially included here: Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Greece, Moldova, North Macedonia, Montenegro, Romania, Serbia, Slovenia, and Turkey. The Balkan peninsula, on the other hand, can be defined by its topographic boundaries: the rivers Danube, Sava, and Krka (Cvijić, 1922) to the north, and the Adriatic, Ionian, and Black Seas as its marine boundaries in the west, south, and east (Reed et al., 2004) (Fig. 1).

Fig. 1
figure 1

Map of the Balkan peninsula with sites considered in this review (site categories 1–3, see SM1), main geographical features (mountain ranges and rivers), and country borders. Digital elevation model and coastline data obtained through European Environmental Agency (http://www.eea.europa.eu). Data on rivers and country borders from http://www.naturalearthdata.com. Sea level as reconstructed for the MIS 4 (Dutton & Lambeck, 2012; Lambeck, 1995, 1996). Map created using the Free and Open Source QGIS. Sites: (1) Campanož, (2) Krapina, (3) Veternica, (4) Vindija, (5) Mujina pećina, (6) Crvena stijena, (7) Zobište, (8) Pecina pod Lipom, (9) Bioče, (10) Petrovaradin, (11) Šalitrena pećina, (12) Hadži Prodanova pećina, (13) Risovača, (14) Asprohaliko, (15) Theopetra, (16) Pešturina, (17) Balanica Cave Complex, (18) Kalamakia, (19) Lakonis, (20) Kozarnika, (21) Klissoura, (22) Samuilitza, (23) Temnata Dupka, (24) Shyroka polyana, (25) Mouselievo, (26) Bacho Kiro

Full size image

Characterized by varied topography and climate, the Balkan Peninsula is broadly divided into three main geographical zones: the Mediterranean zone; mountain chains; and northern lowlands (Furlan, 1977; Reed et al., 2004). Almost 70% of this territory is a high relief of the large mountain chains of the Dinarides and Balkan-Carpathian mountains stretching from the east to the west of the peninsula, then the Rhodopes in the east and Pindus in the southern Balkans. These mountain chains run from the coasts of the Adriatic, Ionian, and Aegean Seas and separate the Mediterranean geographic and climatic region from the northern plains (Fig. 1). As such, they represent natural barriers and are rarely intersected by broad river valleys. Exceptions to this are the Maritza river basin in the southeast of the peninsula (European part of Turkey and Bulgaria), running from the very entrance to the Balkans from Anatolia towards the central areas of the peninsula, and the Vardar river valley, leading from northern Greece to the Morava river valley in Serbia. Several potential passageways from the Sava valley southwards to the northern Adriatic coast may have been possible. These represent south-to-north routes across the peninsula, though the high mountains might have impeded migrations in this direction. The wide valleys of the Danube and Sava rivers represent a major east–west route across the peninsula, along its northern borders, connecting Eastern Europe with the Central European plain.

In the north of the peninsula, the southern edges of the great Pannonian Plain meet the gently sloping mountains in an area of low hills. Major wide river valleys running to the south represent the extension of the Pannonian Plain and its southernmost parts, lying at less than 200 masl (Čalić et al., 2012). The lowlands include the Middle and Lower Danube Basin, their limit being the Iron Gates gorges on the Serbia–Romania border. This is the topographic and environmental transitional zone between the mountain ranges and the Pannonian basin and is comparable to other 'overlap' zones in Europe between northern and southern regions (Davies, Valdes, Ross, & van Andel, 2003a, b; J. Stewart, 2005), all characterized by greater ecological diversity. The widest lowland area is the large Danube valley and its main tributaries, Velika Morava, and Sava with its tributaries in Bosnia and Croatia. Along the Mediterranean coast, by contrast, mountains rise more steeply than in the north, leaving no wide valleys towards inland, at least not in the present day. The coast, nevertheless, has had a more fluctuating coastline throughout the Pleistocene, and even more dramatically along the northern Adriatic. Additional migration corridors along the Mediterranean coast were created once the landscape changed with the formation of a large land mass connecting Asia Minor and Greece (Tourloukis, 2010).

The Balkans as a Refugium

The Balkan peninsula, along with two other southern European peninsulas, the Iberian and the Apennine, is a refugial region where flora and fauna, particularly temperate deciduous forest and its biota, existed during the full glacial conditions (Bennett et al., 1991; Eastwood, 2004; G. Hewitt, 2000; Miracle et al., 2010; O’Regan et al., 2002; Stewart et al., 2010; Tzedakis & Bennett, 1995; Tzedakis, 2004; Weiss & Ferrand, 2007; Willis, 1994). The southern refugia were a source of expansion and re-colonization for many species of plants, insects, and vertebrates in central and northern Europe (Hewitt, 1999, 2000). Several phylogeographic studies have shown that species expanded their ranges from the southern peninsulas of Europe at the end of the ice ages when more favorable climatic conditions commenced, and that northern populations derive from these Mediterranean 'refugial' populations (Hewitt, 1999, 2000).

The success of the re-colonization, nevertheless, depended on the extent to which the mountain ranges represented barriers for the species concerned (Hewitt, 1996, 1999). The high-altitude mountain chains of the Alps and the Pyrenees, above 2000 m, could have impeded the migrations northwards from the southern refugia. As for the Balkans, temperate species existed in the forested areas along the Mediterranean coast (the Adriatic coast of Greece and European Turkey), and the movement towards the north required passing mountains over 1000 m, less of an impediment than the Alps or the Pyrenees (Hewitt, 1996). Genetic studies of several species of small mammals demonstrated that these populations failed to expand and remained in geographical isolation in the Mediterranean peninsulas. As such, these Mediterranean regions were, for some species, 'hot spots' for endemism rather than a source of their re-expansion; this is the case for small mammals and numerous plant species (Bilton et al., 1998; Stewart & Lister, 2001; Stewart et al., 2003). Species capable of crossing the mountain range, such as bears, could have easily migrated out of southern refugia. Compared to northern Europe, southern areas, including the Balkans, are characterized by greater biodiversity and endemism for both plant and animal species (Griffiths et al., 2004).

This may have been true for human populations as well, where 'core' populations did not necessarily retract but survived in southern refugia during the glacial periods (Dennell et al., 2011; Roksandic, Mihailović, et al., 2011a, b; Serangeli & Bolus, 2008). As populations outside of this region went through local extinctions (Hublin & Roebroeks, 2009; Roebroeks et al., 2011), the survival of the species would have depended on their persistence in South and Southeast Europe, in a manner similar to the case of plant species (Dennell et al., 2011). Potentially, the Balkans may have been one of the major core areas of settlement for hominin populations (Dennell et al., 2011, p. 1522). In terms of the archaeological record, if this region was a glacial refugium, it should be demonstrated that it was populated during cold, glacial phases, otherwise, it has to be considered as a region of peripheral settlement (Dennell et al., 2011; Stewart & Stringer, 2012; Stewart et al., 2010). This said, the Balkan peninsula, with its diverse topography, refugial character, and as a transit zone between eastern and central Europe, potentially acted both as a migration corridor and cul-de-sac (Kozlowski, 1992) for hominin populations.

Pleistocene Environment and Climate

The peninsula falls in the temperate zone between 35 and 48N latitude. With three geographical zones, it exhibits different climatic features across its territory, but overall it is broadly continental. With mountain ranges running W–E, warm air masses cannot reach the northern lowland area. Likewise, the Mediterranean part is sheltered from continental winters from the north. European continental, Mediterranean, Atlantic, and Central Asian climatic systems all affect variation in the climate of SE Europe (Ducić & Radovanović, 2005; Furlan, 1977). Hence, the climatic and environmental history of the region was susceptible to the relative influence of these competing systems (Stevens et al., 2011). In general, the mountain region is characterized by an Alpine climate, the northern plains by a continental climate similar to central Europe, and the southern coast by a Mediterranean climate (Furlan, 1977).

Pleistocene climate and environmental reconstructions of the Balkans are possible, given the extensive existing research in several fields:

  1. a.

    The Southern Balkans provided exceptional records of well-known pollen sequences in lacrustine contexts that span several glacial cycles: Tenaghi Philippon (Tzedakis et al., 2006), Pamvotis/Ioannina (Lawson et al., 2002), and Kopais (P. C. Tzedakis, 1999) in Greece; Lake Ohrid on the North Macedonia–Albania border (Lézine et al., 2010; Sadori et al., 2016; Vogel et al., 2010; Wagner et al., 2010); and Lake Prespa on the North Macedonia–Albania–Greece border (Leng et al., 2013; Panagiotopoulos et al., 2014). The abundance of plant taxa and their geographical and chronological changes are a response to climatic changes and hence represent the essential evidence for paleoclimate and paleoenvironment in the Mediterranean parts of the Balkan peninsula, its role as a refugium, etcetera.

  2. b.

    Northern regions of the peninsula are part of the Danube river drainage system, and contain thick loess-paleosol deposits that preserve a valuable record of past environments (Marković et al., 2012; Marković, Fitzsimmons, et al., 2015a, b; Marković-Marjanović, 1968; Smalley & Leach, 1978; Vasiljević et al., 2011). The importance of the Danube loess belt was recognized early (Marsigli, 1792), and intensive research has provided one of the most complete European terrestrial paleoclimatic records, enabling the construction of a regional stratigraphic model as well as comparisons with models coming from other European and Chinese loess-paleosol sequences (Marković et al., 2012; Marković, Stevens, et al., 2015). A continuous record of alternating aeolian and pedogenetic depositions has an excellent preservation potential due to a rather stable deposition on loess plateaus. These are numerous in the Serbian part of the Panonian basin, where some loess deposits reach as much as 40 m in thickness, such as at Stari Slankamen (Marković et al., 2011) and Batajnica (Buggle et al., 2008, 2009; Marković et al., 2009). Loess deposits even extend south into the inland of the peninsula at the sites of Belotinac (Basarin et al., 2011) and Stalać (Obreht et al., 2016), along the Morava river valley in Serbia. In the Dobrogea region of southeast Romania, the Danube River delta loess-paleosol deposits reach 30 m or more, for instance, Mircea Voda (Buggle et al., 2008, 2009; Fitzsimmons et al., 2012). The same depositional thickness is encountered at several locations in Croatia (Galović et al., 2009; Wacha et al., 2013) and Bulgaria (Jordanova & Petersen, 1999; Jordanova et al., 2007). Studies in paleopedology, malacology, magnetic susceptibility, geochemistry, and luminescence dating have made a significant contribution to paleoenvironmental reconstruction.

  3. c.

    Mountains in Croatia (Lj. Marjanac & Marjanac, 2004), Montenegro (Adamson et al., 2014, 2016; Djurović, 2009; Hughes et al., 2010, 2011; Petrović, 2014), Greece (Adamson et al., 2014; Hughes et al., 2006a, 2006b, 2007; Woodward et al., 2004), eastern Bosnia (Milivojević, 2007), the border between Serbia and Albania (Menkovic et al., 2004; Milivojević et al., 2008), and Slovenia (Bavec et al., 2004; Ferk et al., 2015) bear evidence of glaciations during multiple glacial cycles and are another valuable paleoenvironmental record, while the highest Bulgarian mountain range holds evidence only of the LGM glaciation (Kuhlemann et al., 2013). The most comprehensive data come from the mountains of NW Greece, where reliable geochronological data have been obtained (Mount Tymphi, Olympus, Pindus), which allow for a correlation with the nearby pollen records (Hughes, Woodward, & Gibbard,2006; Hughes, Woodward, Gibbard, et al., 2006; Hughes & Woodward, 2006). Even though the pioneering work on glaciation and glacio-karstic landscape was initiated in the Balkans (Cvijic, 1900, 1917; Cvijić, 1899, 1903a, b), in many areas of the peninsula, in fact outside of Greece, this research is still not at an advanced stage and lacks well-dated glacial and periglacial sequences. This hampers attempts to obtain a clearer picture of the history of glaciation in the central and southern Balkans (Hughes et al., 2006a, 2006b).

Paleoclimatic Reconstructions

Late Middle Pleistocene

Glacials: MIS 12 (478–424 ka BP) marked a major glaciation in northwestern Europe and the Alps and was one of the coldest stages in the Balkans as well. In fact, the most extensive glaciation recorded in the Mediterranean region of the peninsula comes from MIS 12 (Adamson et al., 2014, 2016; Hughes & Woodward, 2008; Hughes et al., 2007, 2010, 2011). During this time, all of the conjoining ice-caps in central Montenegro were spread over an area of 1500 km2, thus representing one of the largest and lowest ice caps in the Mediterranean (Hughes et al., 2010). Some of the lowest glaciers in the Balkans were located on the Dalmatian coast in Croatia, some of them as low as 90 masl and probably even reaching sea level or below (L. Marjanac & T. Marjanac, 2004; T. Marjanac & L. Marjanac, 2016). Major glaciations in the Dinarid mountains have been recorded for MIS 12, and MIS 6, though there is possible evidence of another, smaller, glaciation in central Montenegro prior to MIS 7, namely in MIS 8 and/or MIS 10 (Hughes et al., 2011). In the north of the peninsula, starting from MIS 8 (~ 300 ka), loess accumulation rates increased in southern Pannonia, and a more steppic and relatively more arid environment developed there (Buggle et al., 2009; Fitzsimmons et al., 2012; Marković et al., 2011).

During the penultimate glacial, that is, MIS 6 (191–130 ka BP), the Balkans, like the rest of the Europe, witnessed very intense, unfavorable climatic conditions. Precipitation from the Mediterranean sea did not extend to the interior of the peninsula, and this lack of humidity led to extremely dry conditions (Obreht et al., 2016). This glaciation was of lesser extent and volume, however, than the most extensive glaciation the region witnessed in MIS 12, and no larger scale glaciations in the Dinarid Alps have occurred since MIS 6 (Adamson et al., 2014, 2016; Bavec et al., 2004; Hughes et al., 2007, 2010, 2011; Hughes & Woodward, 2008; Lj. Marjanac & T. Marjanac, 2004; Woodward et al., 2004).

The climate of the cold stages during the Middle Pleistocene was dry and cold, but the eastern Mediterranean still had a sustained moisture supply. This humidity, however, could not reach the inland Balkans because of the ice caps in the Dinarides, hence this area maintained dry conditions (Hughes et al., 2010) and the forests contracted (Tzedakis et al., 2006). The glacials in the southern regions were characterized by open grass vegetation and very low pollen concentration of Pinus and Juniper species (Sadori et al., 2016; Tzedakis et al., 2003). Some parts of the Balkans had higher precipitation and fewer temperature fluctuations, favoring the survival of temperate tree species (Bennett et al., 1991; van Andel & Tzedakis, 1996), but their distribution varied in a north–south direction. Coniferous trees were represented in the northern areas with deciduous trees further south and evergreen forests on the coast (Bennett et al., 1991; van Andel & Tzedakis, 1996).

The greater extent of the ice lowered the sea level during Middle Pleistocene glacials to a similar degree as during MIS 2, probably not more than 120 m (van Andel & Tzedakis, 1996), and substantially changed the coastal environment in the Mediterranean region by creating large plains and river valleys. As it has been reconstructed for the Late Glacial Maximum, water retraction in the Adriatic Sea exposed the large continental shelf and formed a wide open plain that linked the Apennine and Balkan peninsulae (Maselli et al., 2011, 2014). Rivers running from the Alps provided rich water resources and hosted abundant large migratory animals (Shackleton et al., 1984). Similarly, the Pinios and Axios Rivers in northern Greece and the plains of the western coast of Albania formed plains larger than they are today, while the retraction of at least half of the Aegean Sea caused large-scale land formation in the glacial periods, with extensive drainage systems, plains, and lakes (Lykousis, 2009).

Interglacials: The pollen record from the south shows that MIS 11 (424–374 ka BP) was a warm and extensive interglacial, clearly the warmest interglacial in the last 600 ka years (Tzedakis & Bennett, 1995). In the loess-paleosol sequences in Vojvodina, this stage is marked by an unexpectedly weak paleosol development, a surprising finding for a stage considered to be a long interglacial (Marković et al., 2011). The record, however, suggests very warm and wet conditions during MIS 9 (337–300ka BP) that did not reoccur in the subsequent interglacials (Buggle et al., 2008, 2009). Starting from this stage, the Middle and Lower Danube Basins are characterized by the progressive aridization of the interglacials during the Middle Pleistocene, most likely as an outcome of the surface uplift of the Dinarids, Carpathians, and the Alps (Buggle et al., 2013; Marković et al., 2011). Interglacials in the Middle Danube Basin in the last 350 ka have continental climate and steppe or forest-steppe environment (Marković et al., 2012). The central areas of the Balkans remained outside the influence of the Middle Danube basin and bear evidence of a more pronounced influence of the Mediterranean climate in MIS 9 and 7 (Buggle et al., 2008; Obreht et al., 2016). This influence, however, progressively diminishes: starting from MIS 7 each interglacial was less warm than the previous one, suggesting an increased influence of the continental climate, in the central areas of the peninsula at least (Buggle et al., 2008; Obreht et al., 2016). The pollen records show some fluctuations during MIS 7 and overall colder conditions than the subsequent MIS 5e interglacial (Roucoux et al., 2008).

Upper Pleistocene

The Last Interglacial, MIS 5e (130–115 ka BP, the Eemian), was less warm and humid than the previous interglacials, and, in the central Balkans, this was the first interglacial characterized by a more continental climate (Obreht et al., 2016). Overall, since the beginning of the Late Pleistocene, the influence of the Mediterranean climate in the interior of the Balkans is less pronounced and more continental conditions prevail. The climate in the northern valleys was nonetheless warm and humid, characterized by a moderate to warm steppe environment, probably forest-steppe, based on intense pedogenesis in the loess-paleosol deposits and the intensely weathered steppic soils (Fitzsimmons et al., 2012; Marković et al., 2004, 2011). The influence of Atlantic air masses brought westerly flows and wetter maritime air masses into the region (Stevens et al., 2011).

In the Mediterranean, however, it was the warmest interglacial in the last 450 ka years (Abrantes et al., 2012) but with smaller scale variations (Tzedakis et al., 2003). At the onset of the Eemian, sea levels rose, as much as 9 m higher than today (Dutton & Lambeck, 2012; Gallup et al., 1994; Kaufman, 1986), drastically changing the coastal topography in the Mediterranean. Sea level rise during warmer periods would have reduced the productive eco-zone, particularly in the Adriatic, resulting in diminishing migratory species (Shackleton et al., 1984; van Andel & Shackleton, 1982). Interglacial vegetation spread out of the refugial zones in the south across the rest of the continent. The southern regions of the Balkans would have had abundant oak (Quercus), elm (Ulmus/Zelkova), Mediterranean olive (Olea), and evergreen forests (van Andel & Tzedakis, 1996).

The interglacial climate started to deteriorate with the onset of the Early Glacial, MIS 5d–5a (115–71 ka BP). In the northern Balkans, accumulation rates of loess deposits of the last glacial cycle were more substantial than in earlier periods (Fitzsimmons et al., 2012; Marković et al., 2008; Stevens et al., 2011). In the central Balkans, the shift from Last Interglacial to Last Glacial was nevertheless mild, and conditions during the early and late Last Glacial were rather mild and humid (Obreht et al., 2016). In northern Bosnia and Herzegovina, at the sites of Visoko Brdo and Kadar, in deposits associated with Mousterian artifacts, pollen samples show a dominant tundra-grassland environment with sporadic dwarf birch, juniper, and pine (Gigov, 1973; Montet-White & Johnson, 1976). In the vicinity of springs and rivers, there may have been small stands of dwarf willow and alder. In the absence of radiometric dates at sites with pollen records, it is not possible to make a correlation with the MIS stages, but sites with corresponding industries are dated to MIS 5d–a.

In the Dinarid mountains, however, only smaller glaciers developed during MIS 5d to MIS 2, which facilitated the movement of warm and humid air into the interior of the peninsula from the Mediterranean (Hughes et al., 2011). The Last Glacial in general had warmer summer temperatures compared to Middle Pleistocene glacials, higher annual precipitation, and more humid summers (Hughes et al., 2007). In the south, according to pollen records, Mediterranean parts of the peninsula were covered in closed evergreen, deciduous, and mixed forests and had warmer and wetter climates (Panagiotopoulos et al., 2014; van Andel & Tzedakis, 1996).

At the culmination of the glacial during MIS 4 (71–57 ka BP), a substantial increase in loess accumulation in the northern plains was triggered by the cold conditions. The boundary between the paleosol and loess deposition recorded at several sites is rather sharp and relates to an age of 75 ka BP (Marković et al., 2008). The northern plains and river valleys of the Danube Basin had a dry glacial climate, with alternating windy and less windy phases and with less influence from Atlantic or Mediterranean moisture (Stevens et al., 2011). According to the loess records, MIS 4 and MIS 3 environments in this region were characterized by grassland, with trees still preserved in the river valleys, and the whole Last Glacial in the south Carpathian region was dry and relatively warm (Marković et al., 2004, 2008).

In the south, the onset of MIS 4 saw a contraction of warmth-loving tree populations into the southern refugia, and most central and southern areas of the peninsula were characterized by an open landscape, wooded steppe with pine dominating and rare oak species. Temperatures and moisture availability declined, and cold and dry conditions prevailed (Panagiotopoulos et al., 2014).

MIS 3 (57–29 ka BP) evidenced a return of the forest with a warm and humid climate. This period was characterized by climatic instability with short-term, extreme oscillations such as Heinrich events (hereafter HE) (Heinrich, 1988; Hemming, 2004) and Dansgaard-Oeschger (D-O) cycles of rapid warming and gradual cooling. These have been detected in both marine and terrestrial records reflecting millennia-scale variability in the climate of the North Atlantic (Dansgaard et al., 1993; Heinrich, 1988; Sánchez Goñi et al., 2000). Lake sediment records in the southern Balkans preserve these changes as well (Tzedakis et al., 2004), though not with the same effect and with much local variation. In areas with less precipitation (e.g., eastern Greece, Thenaghi Phillipon) these phases were more severe than in western areas (NW Greece, Ioannina) with higher precipitation. Higher tree densities existed only in these moist areas (Tzedakis et al., 2004; van Andel & Tzedakis, 1996). The woodland in the south of the peninsula was, however, of open character, with temperate trees, scattered pine, oak, and other deciduous trees (Panagiotopoulos et al., 2014). Overall, the wider region of SE Europe featured vegetation of temperate grassland, temperate deciduous forest, temperate woodland, and evergreen taiga/montane forest during MIS 3 with vegetation of evergreen taiga/montane forest and temperate woodland in the northwestern Balkans (Huntley & Allen 2003).

In the loess-paleosol sequences, however, during MIS 3, paleosols developed as single, double, or multiple pedocomplexes, with loess sublayers evidencing cooling and aridity. Overall the period was rather warm and moist with no cryogenic features recorded, unlike in sequences in central and eastern Europe (Marković et al., 2008). In the same zone, in the south of the Pannonian plain, the sites of Vindija, Velika pećina, and Veternica provide faunal evidence on mammalian species that indicate broadly temperate conditions during MIS 3 with a range of environments (open, forested, wetland, and rocky), although there was probably less forest cover than today (Miracle et al., 2010).

How extreme were these climatic fluctuations in the Balkan peninsula during MIS 3? As testified by the Prespa and Ioannina pollen records, climatic and environmental conditions during the HE 5 event (c. 48 cal BP) were less severe than in the North Atlantic (Panagiotopoulos et al., 2014; Tzedakis, 2004; Tzedakis et al., 2002, 2007; Wagner et al., 2010). It is thus possible that, at least in the south of the peninsula, conditions remained favorable during MIS 3. In the north, loess deposits in the Middle Danube region do not contain a consistent record of all Heinrich events during MIS 3 (Stevens et al., 2011), suggesting that their effect was a lot less than in western Europe (Fitzsimmons et al., 2012; Stevens et al., 2011).

Conditions, however, might have been more dramatic during HE 4. The Campanian Ignimbrite (CI) volcanic eruption in the Phlegraean fields, at 40 ka BP, resulted in the deposition of the Y5 tephra at many locations in Italy and SE Europe (Fedele et al., 2003, 2008; Giaccio et al., 2006, 2008; Pyle et al., 2006). Of all Heinrich events, HE 4, which immediately followed the CI eruption, had the greatest effect on vegetation according to the Prespa (Panagiotopoulos et al., 2014) and Ioannina (Lawson et al., 2002; Tzedakis et al., 2003, 2004) records. These illustrate the extent of the impact this volcanic eruption had on climatic changes in the following period. The proximity and the magnitude of the eruption probably had a significant detrimental effect on the ecosystem of the Balkans (Fedele et al., 2008; Fitzsimmons et al., 2013; Giaccio et al., 2008; Lowe et al., 2012). As evidenced by the thick Y5 tephra records, the lower Danube basin along with regions of the eastern and to some extent the central Balkans were all severely affected by HE 4 (Fitzsimmons et al., 2013; Obreht et al., 2016).

In sum, the Balkan peninsula is broadly divided into three main geographic zones: southern Mediterranean zone, central mountain chains, and northern lowlands, and characterized by the corresponding climatic zones: Mediterranean, continental, and temperate. Southern regions had the most favorable conditions and provided a refugium for flora, though not throughout, as some areas did evidence tree retraction in colder phases. Mountain chains went through a series of glaciations during glacial periods, with the last one happening during MIS 6 and with milder conditions characterizing the Last Glacial. The northern Balkan plains, the westernmost part of the Eurasian steppe belt, were never fully glaciated and did not experience tundra conditions but were characterized by open steppe conditions and sparse trees.

Middle Paleolithic Record and Brief Research History of Lithic Industries

The Balkan peninsula abounds with caves and rock shelters owing to the carbonate geology of the extensive mountain chains, which led to the development of karstic forms. Mountains in the Balkans, offering some of the most famous karstic regions in the world, unquestionably merit further explorations for Paleolithic sites. In addition to the sedimentary records of caves and shelters, some MP occupations have also been documented in open-air sites. The southern extensions of the Pannonian plain contain thick loess-paleosol deposits, but very few sites with remains of human occupation. To the contrary, in Northern Bosnia over a hundred Paleolithic sites have been discovered along river terraces at 100–300 masl in loess-like deposits interstratified with paleosols, though not many have been investigated archaeologically (Basler, 1963; Jovanović et al., 2014). Lithic material from these terraces, the only evidence of Paleolithic occupations, more often than not contains artifacts attributable to both Middle and Upper Paleolithic due to stratigraphic mixing of levels resulting from the extensive erosion. In Bulgaria, a small number of open-air sites is known, but noteworthy are instances of high-altitude surface assemblages in the context of raw material outcrops (Ivanova, 1994; Ivanova & Sirakova, 1995). Out of 240 MP sites in Greece, almost 90% are of open-air character, and only a few have been systematically excavated (Elefanti & Marshall, 2015). The scarce record of the Albanian Paleolithic consists primarily of surface collections (C. Runnels et al., 2009). Drastic sea-level changes in the northern Adriatic left many Pleistocene sites from glacial times submerged; in that light, a recent discovery of underwater sites with MP artifacts is exceptional (Karavanić et al., 2014).

Paleolithic research in the Balkans is characterized by long, though intermittent, work that has focused on extensive surveys, intensive interdisciplinary endeavors, and evaluative research on lithic industries. The history of research is presented in many publications (e.g., Mihailović, 2014b; Papagianni, 2000) and follows different research trajectories for each country; therefore, we will not engage in compiling an extensive history of discoveries and investigations but rather concentrate on the previous assessments of industrial variability. The sites mentioned in this paper that are deemed the most informative for this review are listed in Table 1. Online Table SM2 lists available radiometric ages for MP levels at the Balkan sites. Unfortunately, most sites in the Balkans lack absolute dating, and many have only radiocarbon ages that need to be considered as minimal ages. For most sites excavated in the mid or late twentieth century, no absolute dating is available, but the character of the sediments and an analogy with other sites could provide a tentative chronological placement of assemblages. More often than not, this assessment points to a Last Glacial age. At times, chronological indications based on fauna and/or microfauna or stratigraphic positions have been made, though these are typically without sufficient certainty. For stratified sites, however, it is at least possible to evaluate temporal trends of technological variation.

Table 1 The main technological and typological features of Middle Palaeolithic stone tool industries in the Balkan peninsula
Full size table

Early syntheses and classifications of MP industries were carried out in the 1970s by M. Gabori (Gabori, 1976), J. K. Kozłowski (Kozłowski, 1975), and S. Ivanova (Ivanova, 1979). Excavations at several significant locations were either finalized or in progress at that point (e.g. Krapina, Vindija, Crvena stijena, Asprochaliko, Londa, Kamen, Samuilitza, Mouselievo, Bacho Kiro), and assemblages from these sites were the primary axis for the evaluation and classification of Balkan MP industries. Even long after the height of the popularity of Bordes’ method of classification of industries and the establishment of Mousterian facies in southwest France, the method was still widely embraced across Europe, including these regions. Assessments of the Balkan MP were made using indices defined by Bordes (Ivanova, 1979; Ivanova & Sirakova, 1995), but also relying on a general evaluation of technological and typological features (Kozłowski, 1975) that served to make connections between sites and to build culture-historical units based on their similarities. Additionally, some early combination of indices and attribute analysis was undertaken (Sirakov, 1983), and a complete study of reduction sequences that significantly departed from the typological approach was also undertaken (Baumler, 1987). The final facies list that resulted from these initial studies overlapped between researchers, though with some variation in how sites were grouped. Needless to say, the different facies groups were conceived of as distinct culture-groups (Basler, 1975a; Kozłowski, 1975). Decades later these NASTIES (‘named stone tool industries’) (Shea, 2014) remained in use (Basler, 1983; Ivanova & Sirakova, 1995; D. Mihailovic & B. Mihailovic, 2009), mainly because a new and more substantial record that might provide an impetus for taking a fresh look at industrial variability is missing, but also because reevaluations of the collections that constituted the core of the facies definitions have not been systematically undertaken. Nevertheless, recent research has to a certain extent clarified what some of these assemblages represent technologically and typologically (Dogandžić & Đuričić, 2017; e.g., D. Mihailović & Bogićević, 2016; Sitlivy, 2016).

Charentian and Typical Mousterian are two facies of Mousterian from the Bordes classification system that have been commonly applied to industries of the Balkans. The Southeastern Charentian was recognized by Gabori, Ivanova, and Kozłowski, and refers to three assemblages in the Croatian northwest: Krapina, Veternica, and Vindija (Gabori, 1976; Ivanova, 1979; Kozłowski, 1975). Their features – high scraper index, low to zero Levallois index, simple cores sometimes resembling Pontian cores (on pebbles), discoidal cores, and scrapers with Quina and demi-Quina retouch particularly in Krapina – approximate Quina Charentian in France but are somewhat distinct and deserve their own separate group (Gabori, 1976; Kozłowski, 1975). Additionally, atypical Charentian is separated at several sites in Bulgaria (Bacho Kiro 14, Samuilitza, Temnata Dupka), with single convex and bifacial scrapers (Ivanova & Sirakova, 1995). Depending on the values of technological and typological Levallois indices, Kozłowski separated Levallois Mousterian (Bacho Kiro 12/13) and Moustero-Levalloisian (lower Levallois indices than the former and high scraper index: Bacho Kiro 13, Crvena stijena XIII–XVII); a division corresponding to a Typical Mousterian Levallois version (Bacho Kiro 12/13, Samuilitza, Temnata); and a non-Levallois version (Bacho Kiro, 13) (Ivanova & Sirakova, 1995). East Balkan Mousterian with bifacial points (Ivanova & Sirakova, 1995) or Moustero-Levallois with leaf points (Kozłowski, 1975) are recognized at the Bulgarian sites of Mouselievo-Samuilitza, and also Kokkinopilos in Greece, and, according to Kozłowski, is a Moustero-Levalloisian complemented by leaf-shaped points.

Under the name Balkan Mousterian, Kozłowski included assemblages with a high scraper index, average technical Levallois index, and a high Charentian scraper index. He merged assemblages, usually from lower levels, at several sites like Crvena stijena XXIV, Asprochaliko 18, Bacho Kiro 14, and Londza and Kamen, into this group. Gabori (1976) termed the latter two sites from Bosnia and Herzegovina, Late Mousterian with Levallois debitage (Gabori, 1976). Particularly based on the lower levels at Crvena stijena, which feature various side-scrapers, double and convergent forms, bifacially and unifacially retouched, and many scrapers with ventral thinning, Kozłowski gave a Charentoid character to the Mousterian at Crvena stijena and subsequently to all Balkan Mousterian (Kozlowski, 1992). Ivanova (1979), however, designated these assemblages as Typical Mousterian but very rich in sidescrapers and points (and named them ‘Crvena stijena a – Pinios’). Some of these assemblages (e.g. Krapina, Zobište, Asprochaliko) have also been examined free from the constraints of Bordes’ method and, as much as the collections allowed, described in terms of blank production methods and the associated tool use (Baumler, 1987; Gowlett & Carter, 1997; E. Papaconstantinou, 1989; Simek & Smith, 1997); however, their typological classification still remained in use.

Of other facies hitherto defined elsewhere, Pontinian and Denticulate Mousterian have been proposed for Balkan industries as well, and both are primarily identified at Crvena stijena (Basler, 1975b). For the Pontian, the flaking of pebbles to obtain cortical blanks and an abundance of scrapers link this eastern Adriatic assemblage with its counterpart in central Italy, and these are seen as broadly related to Quina/Charentian industries (Taschini & Bietti, 1979). Similar to what is seen in western MP, Basler held that the Denticulate facies was a phenomenon that characterized the late phases of the Mousterian (Basler, 1975b, 1983). This facies was potentially recognized at other sites, e.g., Temnata Dupka (Drobniewicz, Ginter, & Kozłowski, 2011; Kozłowski, 2002).

However, not all assemblages and industries were subjected to a typological classification; for instance, Micromousterian has been defined as a facies consisting of assemblages where the size of artifacts is comparatively small (though the threshold has never been defined). This group incorporates assemblages with a range of technological and typological characteristics and is predominantly related to the industries in the southern, Mediterranean belt (Asprochaliko, Crvena stijena, Klissoura, Bioče, etc.) (Dogandžić & Đuričić, 2017; Gowlett & Carter, 1997; Mihailović & Whallon, 2017; Papaconstantinou, 1989; Sitlivy et al., 2008). These assemblages have often been compared to the Pontinian in Italy, with respect to the small size of artifacts, and thus interpreted as a typical but diminutive version of Mousterian, primarily due to raw material constraints (Basler, 1975b; , 1983; Rink et al., 2002). Others have suggested that the smaller nodules have been purposefully selected for small flake production (Đuričić, 1997, 2006). Another deliberate small flake production is suggested for Asprochaliko through a core-on-flake method similar to Kombewa (Papaconstantinou, 1989).

Main Tendencies in the Chronological and Spatial Variation of the Middle Paleolithic Industries

The Lower Paleolithic Background

In contrast to western Europe, the ‘Levallois generalization’ (J. Richter, 2011) in SE Europe does not emerge in an Acheulean context. Isolated bifacial handaxes that could be attributed to the Lower Paleolithic have been discovered in southern Croatia (Malez, 1979), Albania (Harrold et al., 1999; Runnels et al., 2009), and the northern parts of Greece, including the islands (Dakaris et al., 1964; Galanidou et al., 2012, 2016; Reisch, 1982; Runnels, 2003; Runnels & van Andel, 1993a, 1993b; Tourloukis, 2010), and there are atypical bifacial elements in surface collections in the West Rhodopes (Bulgaria) (Ivanova, 2016) (Fig. 2). The bifacial component in the Kokkinopilos assemblage from Greece, one of the most notable Lower Paleolithic locations in the region, cannot be assigned with certainty to either late Lower or early MP without a proper sample and technological study, though its occurrence certainly predates the Last Interglacial based on luminescence dating of artifacts from two different overlying units yielding ages of 172 ± 25 ka and 206 ± 19 ka (Tourloukis et al., 2015). It seems, however, that the Lower Paleolithic technological ‘background’ preceding the appearance of Levallois in Central and Eastern Europe and the Balkans was not predominantly bifacial and consisted rather of mainly a core-and-flake assemblage.

Fig. 2
figure 2

Map of major Lower Paleolithic sites in the Balkans predating the widespread use of Levallois (MIS 11–MIS 9). Green: core and flake assemblages; blue: sites with handaxes. Handaxes in surface context are also found in central Albania (Runnels et al., 2009), Paleokastron in Greece (Dakaris et al., 1964), and Croatia (Mirko Malez, 1979). Sites: (1) Gajtan, (2) Western Morava sites, (3) Kozarnika, (4) Rodia, (5) Dealul Guran, (6) Yarımburgaz, (7) Kokkinopilos, (8) Rodafnidia (Color figure online)

Full size image

A non-Acheulean industry with the presence of preferential cores comparable with Levallois or discoidal technology is noted in Kozarnika cave (Bulgaria), Levels 13–11a, dated to > 400 ka (Guadelli et al., 2005; Sirakov et al., 2010). Industries that generally feature choppers, cores for flakes, flake tools, cores on flakes, and lack Levallois and bifacial production are present at Dealul Guran (Dobrogea, Romania) (Iovita et al., 2012, 2013), Rodia, an open-air site in Thessaly (Greece) (Runnels & van Andel, 1993a, 1993b; Tourloukis, 2010), Gajtan cave (Albania) (Darlas, 1995), Yarımburgaz cave, located on the Bosphorus (Turkey) (Arsebük & Özbaşaran, 1999; Kuhn et al., 1996), and most of these are tentatively placed around MIS 11, while the newly discovered site Marathousa (Greece) dates to MIS 16 or 14 (Panagopoulou et al., 2015, 2018).

General assemblage features from these Middle Pleistocene sites display certain similarities with Middle Pleistocene assemblages from central or eastern Europe, such as Vértősszőlős (Kretzoi & Dobosi, 1990), Bilzingsleben (Haidle & Pawlik, 2010; Pasda et al., 2012), Korolevo I (Koulakovska et al., 2010), and many others (Burdukiewicz & Ronen, 2003; Doronichev, 2016; Rocca, 2016). The industries from this wide region between the Alps and the Caucasus share an absence of bifaces and Levallois production and are characterized by simple and informal production processes with cores with one or multiple platforms, simple toolkits on small blanks and the presence of choppers (Doronichev, 2016; Doronichev & Golovanova, 2010). The differences that do exist are most probably the result of local factors such as raw material properties and site function.

It was suggested long ago that the similarities in these Balkan industries allow for the definition of a pre-Mousterian industrial complex (Obermeier, 1925) that coincides with the Movius line (Movius, 1948), such that it is synchronous with the Acheulean but outside of the Acheulean’s geographical range. These industries chronologically fall at least in the period 400–300 ka, but an earlier age, c. 700 ka, is likely and some industries may have persisted longer, until c. 200 ka (Doronichev, 2016). A gradual transition from the Acheulean to Levallois production is, therefore, an unlikely scenario in this region (Doronichev, 2016). Several sites, however, may contain technological elements of a proto-Levallois production system, such as the open-air site in Velika Morava valley in Serbia (Mihailović et al., 2014a, 2014b) and Kozarnika (Sirakov et al., 2010). These sites have the potential to shed light on the early appearance of Levallois technology.

New data from current research in the Western Morava valley (Serbia) has the potential to elucidate the beginning of the MP in the region (Heffter, 2014; D. Mihailović et al., 2014; Mihailović & Bogićević, 2016). Over 30 locations with surface finds have been registered, with a few of them containing large amounts of material. Some locations have typical LP technology with many choppers and chopping tools, while assemblages from other locations contain proto-Levallois cores on pebbles, Kombewa cores, choppers, and some Levallois products. Both elements of Lower and Middle Paleolithic technology have been encountered. These assemblages could date to roughly 400–300 ka, certainly before the Last Interglacial (D. Mihailović et al., 2009, 2014), however, the chronostratigraphic context is not certain. The observation that the blank production incorporated elements of Levallois technology with the use of choppers and the Kombewa technique signifies the great potential of these assemblages in understanding the technological aspects behind the Lower to Middle Paleolithic transition in SE Europe and the appearance of Levallois technology.

Are there Middle Paleolithic Technologies before the Appearance of Levallois?

Levallois technology is conventionally taken as the hallmark of the MP (Ronen, 1982; White & Ashton, 2003) and is roughly concurrent with the development of Neanderthals as a taxon (Hublin, 2009). Aside from the regular use of Levallois, the dominance of other flaking systems that are characteristic of the MP, such as discoidal or Quina, may mark the beginnings of the MP (J. Richter, 2011).

The best evidence for early MP techno-typological elements predating the appearance of Levallois comes from Velika Balanica cave (D. Mihailović, 2009b, 2014a; D. Mihailović & Bogićević, 2016) (Fig. 3). Chronologically placed in MIS 9–7, assemblages from Level 3 of Velika Balanica are characterized by the production of asymmetrical, thick and cortical blanks, with cortical or flaked backs, sometimes with Clactonian platforms and high interior platform angles, a method that corresponds to Quina method (Bourguignon, 1997; Turq, 2000). These flakes were transformed into numerous scapers, many of which went through intense and/or repeated resharpening and exhibit a scalar, Quina or demi-Quina retouch prompting their classification as Charentian (D. Mihailović, 2014b; D. Mihailović, Kuhn, et al.,2022; D. Mihailović & Bogićević, 2016) (Fig. 4: 4–8). These assemblages predate the use of Levallois in Level 2 of Velika Balanica. A regional analogy can only be made with the very small but homogeneous assemblage at Theopetra Level II1 of at least MIS 6 age (Panagopoulou, 1999, 2000; Valladas et al., 2007b) (see online supplementary material SM2 for all radiometric ages). It exhibits a similar trend of producing thick and cortical blanks that are often transformed into Quina-like scrapers, and it predates an assemblage with Levallois use (Fig. 4: 1–3).

Fig. 3
figure 3

Distribution of sites in the early Middle Palaeolithic. a pre-Eemian site, likely belonging to MIS 9–6. Red: Levallois based assemblages; blue: non-Levallois assemblages. (1) Crvena stijena, (2) Velika Balanica, (3) Theopetra, (4) Kozarnika, (5) Temnata Dupka. b Sites from MIS 5. Green: MIS 5e; orange: MIS 5a–d; white: likely MIS 5 though not specific or reliable. (1) Krapina, (2) Crvena stijena, (3) Zobište, (4) Hadži Prodanova cave, (5) Asprochaliko, (6) Theopetra, (7) Pešturina, (8) Lakonis, (9) Kozarnika, (10) Klissoura (Color figure online)

Full size image
Fig. 4
figure 4

Artifacts from sites dating to MIS 9–6. Theopetra phase 3 (1–3), redrawn after Panagopoulou (2000), Velika Balanica layer 3 (4–8) and Mala Balanica (9, 10) redrawn after Mihailovic and Bogicevic (2016)

Full size image

Interregional comparisons take us to MIS 4 and MIS 3 in southwest France. Here the prevailing techno-typological behavior is marked by the distinct Quina technology with the associated typological signature of Quina scrapers. This particular techno-typological combination has been defined and researched intensively, along with the particular subsistence and mobility behavior related to reindeer hunting (Bourguignon, 1997; Delagnes & Rendu, 2011; Discamps et al., 2011; Faivre, 2008; Guibert et al., 2006; D. Richter, Dibble, et al., 2013; Turq, 1989, 2000). However, it is not restricted to the narrow timeframe of the late MP; at several sites in the same region (e.g., Les Tares, La Micoque 3), analogous techno-typological behaviors have been observed which date prior to MIS 6, most probably even earlier than 300 ka (Delpech et al., 1995; Geneste & Plisson, 1996; Geneste et al., 1997; Le Tensorer, 1978; Texier & Rigaud, 1981; Turq, 1992). However, Velika Balanica is rather chronologically and spatially related to similar techno-typological phenomena in the east (Mihailović, 2009b; Mihailović et al., 2009; Mihailović et al., 2022a, 2022b; Mihailović & Bogićević, 2016). In Anatolia, predominant Quina-like technological behavior (named ‘Proto-Charentian’) predates prevalent Levallois use at Karain (Turkey) around MIS 9 (Kozłowski, 2002; Otte et al., 1999; Rink et al., 1994); further to the east, Yabrudian industries (also, Amudian, Acheulo-Yabrudian or Mugharan) industries, dated to MIS 9 (Mercier, 2003; Rink et al., 2004), are characterized by production of thick and broad blanks that are often used for scrapers with a distinct scaled Quina-like retouch (Al Qadi, 2011; Gopher et al., 2010; Lorraine, 2000; Shimelmitz et al., 2014). Such technological and typological consistency between contemporaneous industries across the wide region might reflect population movements, transfer of cultural practices, or both (D. Mihailović, Kuhn, et al., 2022); in any case, it is important to stress this early appearance of MP technological components—other than Levallois—in industries in the Balkan region (a phenomenon of ‘Mousterianization’ as defined by Delagnes (1992) and their links with Southwest Asia.

The fossil evidence points to a rather complex demographic process in this period. Mala Balanica mandible dated to at least 400 ka does not have Neanderthal traits (Rink et al., 2013), while a later dentition from Velika Balanica Level 3 at c. 290 ka shows similarities to, and is somewhat younger than, the western European Neanderthal specimens, suggesting their movements in a west–east direction (Roksandic et al., 2022), contrary to what is seen in lithic industries (D. Mihailović, Kuhn, et al., 2022). At 210 ka, a cranium from Apidima cave in Greece (Apidima 1) shows a mixture of modern human and primitive features and likely points to early movements out of Africa of H. sapiens (Harvati et al., 2019), likely supported by the data on genetic introgression (Posth et al., 2017). Subsequently, only Neanderthal fossils are present, in Apidima 2 at 170 ka (Harvati et al., 2019) and Krapina at 130 ka (Rink et al., 1995). The intricate demographic picture in the Middle Pleistocene should be closely examined in relation to variations in stone tool technologies as this could provide insights into the dynamic interplay between cultural and biological changes.

Widespread Use of the Levallois Method in MIS 6 and MIS 5

The unambiguous appearance of Levallois technology is related to MIS 6 or even MIS 7 (Crvena stijena, Kozarnika, Velika Balanica, Temnata Dupka) (Figs. 5, 6a), when this, and—to a lesser extent—other centripetal methods come into rather frequent use. Particularly at Velika Balanica and Theopetra, the emergence of Levallois is fairly abrupt, as it arrives after MP industries with no indications of Levallois production. Typologically, these assemblages still consist of high scraper frequencies, but this time with lateral and double forms (e.g., Crvena stijena, Velika Balanica) rather than transversal ones, while tools with a modest amount of edge modification prevail at others (Kozarnika, Temnata Dupka).

Fig. 5
figure 5

Artifacts illustrating industries from MIS 5. Asprochaliko Basal level (1–5), redrawn after Gowlet and Carter (1997) and Theopetra phase 2 (6–8), redrawn after Panagopoulou (2000)

Full size image
Fig. 6
figure 6

Distribution of sites in the late Middle Palaeolithic. a MIS 4. White circles: occupations imprecisely associated with MIS 4; blue circles: radiometric age associated with MIS 4. b MIS 3. (1) Kličevica, (2) Mujina cave, (3) Crvena stijena, (4) Bioče, (5) Šalitrena cave, (6) Asprochaliko, (7) Dubočka) Kozija, (8) Pešturina, (9) Kalamakia, (10) Lakonis, (11) Klissoura, (12) Samuilitza, (13) Temnata Dupka, (14) Mouselievo, (15) Bacho Kiro (Color figure online)

Full size image

Levallois method remains a regular blank production method throughout the MIS 5 (Figs. 5, 6b) and it is prominently characterized by the production of larger, elongated blanks through uni- and bidirectional removals. This is evident at MIS 5e occupations at Theopetra II2-8 (Panagopoulou, 1999, 2000), the most likely MIS 5a Level XXIV at Crvena stijena (and potentially earlier Levels, XXVII–XXIV) (Kozlowski, 1992; D. Mihailović, 2009c, 2014b, 2017a; D. Mihailović et al., 2017), and MIS 5a–d levels at Asprochaliko 18 (Gowlett & Carter, 1997), and Zobište (Baumler, 1987, 1988) (Fig. 5). The production of elongated blanks often proceeded with a radial detachment resulting in regular flakes and with radial/discoidal remnant cores (e.g., Zobište, Asprochaliko). While no particular method for the production of elongated elements could be inferred due to the scarcity of cores and technical elements, elongated blanks and blades are present in significant numbers at Crvena stijena XX, XVIII (Dogandžić & Đuričić, 2017; D. Mihailović, 2014b; D. Mihailović & Whallon, 2017). These levels are associated with MIS 5d–a based on geology, but new radiometric dates place them post-MIS 5, likely MIS 4 (Mercier et al., 2017). Elongated Levallois blanks are present in MIS 5, Level 5 in Hadži Prodanova cave (D. Mihailović, 2008; D. Mihailović & B. Mihailović, 2004). The lower levels at Klissoura (XX–XV), most likely belonging to MIS 5 (Starkovich, 2014, 2017), contain a significant number of blades produced through a volumetric-like method, not Levallois, even though Levallois is often used here (Sitlivy et al., 2007, 2008). In addition to the use of Levallois to produce elongated blanks, notable is the presence of blade production at Theopetra II4-8, attested to by bidirectional blades, crested and overshot blades, and core tablets. This similarity to UP-like volumetric core exploitation led the authors to associate this with a transitional type of industry (Panagopoulou, 1999, 2000), which was subsequently invalidated by TL dates (Karkanas et al., 2015; Valladas et al., 2007a). There is a possibility that these artifacts come from the UP level above (which, however, is not rich in finds and show no straightforward blade production), but until this is resolved, one may entertain the idea that the existence of blade production in the interglacial context is rather genuine, especially having in mind the occurrence of blade production in the MP elsewhere in Europe during the Eemian and pre-Eemian (Bar-Yosef & Kuhn, 1999; Delagnes, 2000; Koehler, 2011; Locht et al., 2015; Révillion & Tuffreau, 1994; Tuffreau, 1995). At Temnata Dupka, more Levallois relative to other layers is noticed in early glacial levels, and evidence of volumetric blade production is observed as well (Drobniewicz, Ginter, Kozłowski, et al., 2011). While no finer analysis of the Lakonis sequence and material from MIS 5–MIS 3 that can address temporal trends is available, its industry with notable Levallois production of elongated blanks and a volumetric blade production is considered to display similarities with other MIS 5 assemblages (Papagianni, 2009).

Not only Levallois—Technological Variation in MIS 5

Several assemblages of MIS 5 age with low-to-moderate presence of Levallois blank production bear evidence of a production of naturally backed, thick blanks, selected for the manufacture of scrapers. As a reminiscence of earlier industries previously discussed (MIS 9–7?), blanks of this morphology are commonly produced on quartz pebbles with a method corresponding to the Quina flaking method in Pešturina’s MIS 5 Level 4, where more than half of scrapers have Quina or demi-Quina retouch (D. Mihailović, Milošević, et al., 2022; D. Mihailović & Milošević, 2012). Production of flakes of this morphology is present in the Crvena stijena Levels XXII, XX, XVIII (MIS 5 or MIS 4, see SI) (Fig. 7: 1–3) where they have been named Pontinian (Dogandžić & Đuričić, 2017; Mihailović, 2014b). If the latter levels can be securely attributed to MIS 4, the presence of these techno-typological features would coincide with a similar phenomenon during the same period in SW France (D. Mihailović, 2017a).

Fig. 7
figure 7

Artifact illustrating lithic industries from MIS 5 [Crvena stijena XX (1–3) after Mihailovic et al. (2017)] and MIS 3 [Crvena stijena XIV–II (4–7) from Dogandzic and Djuricic (2017), Asprochaliko Micromousterian (8–11) after Papaconstantinou (1989), and Mujina cave (12–14) from Karavanic et al. (2008)]

Full size image

North of the limits of the peninsula, north of Sava, this blank production method is evident in MIS 5e at Krapina, especially in the upper part of the sequence, with the frequent use of the core-wedge method, similar to bipolar (Simek, 1991; Simek & Smith, 1997). Veternica assemblage is probably associated with MIS 5, and while cobble-wedge method is used to exploit quartz pebbles, it is not directly comparable to Quina method (Banda & Karavanić, 2019). For this reason, assemblages from Krapina and nearby Vindija and Veternica were grouped into the Charentian-Quina type (Gabori, 1976; Ivanova, 1979; Kozłowski, 2016), but primarily on typological grounds and the heavy usage of quartz/ite pebbles. The geographical distribution shows that the industries with predominant Quina-like debitage [so-called Charentian of south-eastern Europe according to Kozłowski (2016, 1992) and Gabori (1976)] but also with specific-scraper-dominated toolkits are primarily present in the northern Balkans, or the southern edge of the Panonian basin, in Veternica (Banda & Karavanić, 2019), Krapina (Simek & Smith, 1997), Petrovaradin (D. Mihailović, 2009a), Erd (Mester & Moncel, 2006), Tata (Borel et al., 2016; Moncel, 2003) and a few Carpathian caves in Romania (Cârciumaru & Anghelinu, 2000). A rare presence in the central (Pešturina) and southern (Crvena stijena) Balkans demonstrates stronger links between the Balkans and central Europe (D. Mihailović, Milošević, et al., 2022).

When Levallois occurs with the production of naturally backed blanks in the same assemblage, it is worth examining whether these result from separate reduction trajectories. At the MIS 5d–a industries at Zobište the production of naturally backed elements, which were selected for scrapers, was part of one continuous process that starts with Levallois-like unidirectional exploitation and ends with radial detachments of flakes (Baumler, 1987). At other sites small assemblages do not allow for such examination; however, some diachronic patterns can be observed. At Crvena stijena, notable is the shift from the predominant use of Levallois in XXIV (likely MIS 5a) to the habitual production of naturally backed blanks in subsequent Levels XXII–XVIII. In Krapina’s typologically homogeneous assemblage dominated by sidescrapers, a diachronic pattern during the short time span of MIS 5e is seen where patterns of blank production varied from a high frequency of Levallois blanks to the wider use of cortical backed blanks. If one bears in mind the curation and stratigraphic difficulties of this assemblage it may be difficult to discern whether these are two different flaking trajectories (Simek & Smith, 1997) or part of the same sequence that parallels Zobište as suggested by some (Baumler, 1988). Within 10 stratigraphic units, not chronologically distinguishable, two peaks in lithic densities led to the hypothesis there were at least two occupations of this site (Simek & Smith, 1997).

In sum, MIS 5 industries show the use of two main blank production methods, Levallois on one side, at times accompanied by volumetric-like blade production methods, and on the other side a tendency to produce thick cortical blanks. Both technological contexts share a common emphasis on the production of scrapers, often increasingly resharpened. In a broad sense, this technological diversification in a typologically similar context corresponds to the Ferrassie vs. Quina Charentian dichotomy in SW France (Geneste et al., 1997). Unlike in SW France, the two technological methods co-occur and are not exclusive; rather, assemblages show fluctuations in the importance of producing one blank morphology or another – large and thin Levallois blanks vs. thick and cortical blanks.

Scraper-Rich Assemblages Common in MIS 5

Regardless of the blank production method, typologically all assemblages during this period are rich in scrapers; however, they differ in tool production intensity and the intensity of tool modification. In some assemblages, tools comprise a smaller percentage and/or do not exhibit particularly extensive resharpening (e.g., Zobište, Asprochaliko, Temnata Dupka), while others have a higher frequency of tool production, more elaborate production of scrapers, and heavier edge modifications (Crvena stijena, Theopetra). Between assemblages with predominant Levallois use, there is a difference in the frequency of retouched tools and the intensity of retouch that may stem from more economical concerns, namely different degrees of reduction that in turn may fluctuate according to functional aspects of the site or settlement patterns. For instance, assemblages that share the production of often elongated blanks through the Levallois method—Asprochaliko, Zobište, and Theopetra—have different patterns of tool use intensity. The rich assemblages of Asprochaliko and Zobište have fewer tools and less edge resharpening on scrapers, while Theopetra’s smaller assemblage has less stone tool production intensity, in terms of blank production, but is abundant in tools, many of which exhibit heavier edge modification. This said, a perspective that takes into account the use-life of a tool may explain the variation between assemblages that predominantly use the Levallois flaking method, analogous to the variation in other MP contexts, namely in SW France where the difference between the Ferrassie and Typical Mousterian facies is understandable as a consequence of economic factors that would more or less stress frequent tool use and reuse (Dibble, 1995; Faivre et al., 2014).

Earlier studies of the Balkan MP emphasized the particular typological signature of Crvena stijena levels attributed to MIS 6 and MIS 5 (XXVIII–XXIV). It featured the use of Levallois, various side-scrapers, among them double and convergent forms, bifacially and unifacially retouched, frequent ventral thinning on scrapers, and the analogous truncated-facetted pieces (potentially evidencing small flake production). These features are shared with the contemporaneous Theopetra and Karain assemblages, and further to the east related to similar techno-typological phenomena in the Zagros Mousterian (Kozłowski, 1975, 1992, 2002; Panagopoulou, 1999). Its western counterpart would be Ferrassie Mousterian (Dibble, 1991; Geneste et al., 1997). This similarity across extensive territory supported an idea of a unified culture among Neanderthals in Europe and further to the east, potentially suggesting migrations (Kozłowski, 1992, 2002). Further reinforcement of this idea came from a paleoanthropological perspective that supported the eastward ‘expansion’ of Neanderthals during MIS 6/5 (Hublin, 2002; Kozłowski, 1992, 2002).

In sum, MIS 5 assemblages show greater technological variability with various blank production methods being used simultaneously at times. Flaking methods introduced during the pre-Eemian MP (at first production of thick blanks as in Velika Balanica and later the Levallois method) appeared as distinct techno-economic concepts that were practised independently. Once they became incorporated into the MP repertoire, these concepts were integrated into the MP technological package (Kuhn, 2013; Pettitt, 2003). These blank production methods appear to be concomitantly used at any one time (period/occupation?) at some sites, for instance at Crvena stijena XVIII or Krapina (potentially Petrovaradin as well, pending future age estimations: D. Mihailović, 2009a). The regional differentiation in the distribution of these patterns suggests a connection with central Europe (D. Mihailović, Milošević, et al., 2022),while in the eastern Balkans (Bulgaria) thus far only the Levallois component is present, with some early presence of bifacial flaking (see below).

Earlier descriptions and interpretations of the Balkan MP (Ivanova, 1979; Kozłowski, 1975) labeled these industries according to technology (Levallois Mousterian) or sometimes typology (Charentian), and at times the same assemblage would be placed in the two different facies (for instance, Crvena stijena XXIV). We would like to stress the tendency for scraper production in most of these assemblages. The principal difference, however, is in the blank morphology and technological process used to obtain blanks for scraper production, namely Levallois with a tendency to produce elongated blanks, rarely accompanied by blade production, and a less elaborate way of producing thicker and often cortical blanks. Both blank morphologies (wide and thin blanks and relatively thicker blanks) have a potential for further edge exploitation (Kuhn, 1992; Lin et al., 2013; Meignen et al., 2009; Pettitt, 1992; Turq, 1989) but differ in which aspect of blank shape, area or thickness, is used for extending edge use. This said, the major theme of these earlier assemblages, one may suggest, is an emphasis on the resharpening potential of a blank, where patterns of techno-typological production are directed towards tool use and extending tool use life. Depending on the prevalent flaking method, Levallois blanks/blades and/or naturally backed knives were selected for retouch. Notwithstanding the obvious functional background of these technological solutions, one may wonder if other factors would affect selection of one option over another, especially if both offer blanks with long-term use potential. Whether these differences indicate distinct cultural groups and can be explained from a culture-historical perspective (Boëda et al., 1990; Delagnes & Meignen, 2006; Geneste et al., 1997), or whether, as some have suggested, these differences reflect responses to changing environmental circumstances and the concomitant changes in mobility or hunting practices remains to be determined.

Late Middle Paleolithic Variability and Regional Differences

Variability in Flaking Methods

As the presence of Neanderthals in the region during the MIS 4 is not well presented, our knowledge of the late variability of their industries relies on MIS 3 assemblages and stratified sites where the features of the late MP are discerned when viewed relatively to earlier deposits. The overall technological theme in this period is the use of centripetal methods of flaking but with less frequent and less standardized Levallois production and instead more often discoidal and less formal radial flaking methods. However, whether there is any directionality, at a chronological scale, in regional variation, or in how flaking practices relate to toolkit organization, may be more difficult to grasp.

The southern coastal zone offers numerous sites, many of them stratified, for temporal changes to be examined. The facies frequently assigned to the majority of MP industries from these sites (particularly in the late MP) is the Micromousterian (Asprochaliko, Crvena stijena, Mujina cave, Bioče, Klissoura) (Fig. 7). The small size of the artifacts has been recognized as a particularity of industries at Crvena stijena (Benac & Brodar, 1958; Brodar, 1962) and in Asprochaliko’s upper level (Higgs & Vita-Finzi, 1966). Later, when assemblages with artifacts of smaller sizes in sites along the Mediterranean coast of the Balkans (Greece, Montenegro, Croatia) were discovered and described, analogies were made with the former two sites based on artifact size alone. Consequently, the Micromousterian designation remained restricted to the southern region. However, some assemblages outside of the southern Balkans are also composed of artifacts of comparably small sizes, for example, Petrovaradin with an average size of 25 mm (D. Mihailović, 2009a), Velika Pecina (Karavanić, 2007), and even Temnata Dupka (Drobniewicz, Ginter, Kozłowski, et al., 2011), which shows a small, ~ 3 cm average blank size. However, these have rarely been allocated to the Micromousterian and related to the Mediterranean MP. The small size of flakes in an assemblage may stem from several different but not mutually exclusive factors—such as raw material size, flaking reduction intensity, prominent and intentional small flake production—and so a more detailed examination of these factors is required prior to (or instead of) simply assigning them all to a single Micromousterian unit. For instance, at many sites in the Mediterranean region where the Micromousterian designation has been frequently used, one factor that is constant is the small locally available raw material that largely configures the lithic assemblage and appears as a major explanatory factor of the small artifact size. However, within this context, intersite differences in artifact size and assemblage structure, as observed at Crvena stijena and Bioče, are due to raw material conditions, namely the distance and the source type (primary versus secondary), as well as to reduction intensity (Dogandžić & Đuričić, 2017). Moreover, at stratified sites, if one assumes that raw materials remain roughly constant, there is still a certain degree of variation in flaking methods, tool production, and use that is more relevant for understanding the variability of MP industries than the artifact size alone. This said, we should look for techno-typological patterns beyond artifact size as a common denominator for these industries. For instance, at sites that show evidence of raw material constraints, primary driving factors of techno-typological variability are shifts in the predominance of flaking methods (Dogandžić & Đuričić, 2017; Gowlett & Carter, 1997; D. Mihailović & Whallon, 2017; V. Papaconstantinou & Vassilopoulou, 1997; Papagianni, 2009; Sitlivy et al., 2008). Changing flaking methods are furthermore partially responsible for the microlithic appearance of some industries (Asprochaliko) (E. Papaconstantinou, 1989).

At these sites, and particularly at sites that afford assessments of temporal tendencies, a rather general trend is observed. In earlier phases (MIS 5), blank production is more frequently organized towards the production of blanks that are suitable for scrapers, usually large and long Levallois blanks, and/or thicker blanks convenient for resharpening. In contrast, subsequent phases of the MP feature centripetal flaking with a rather low incidence of Levallois production (Papagianni, 2009). The reduction pattern where Levallois flaking produced less standardized radial forms and remnant discoidal cores in the later stages of reduction is apparent in earlier periods (MIS 5, e.g., Zobište, Asprochaliko), but what differentiates later, MIS 3 assemblages in general, is their less pronounced Levallois character, less of a tendency towards the production of larger implements, and overall production that is not aimed at large blank manufacture for transformation into scrapers. This broad two-part division of MP industries, where non-laminar and non-Levallois assemblages are present in the upper/later stages, has been proposed for Greek assemblages firstly, following the sequence at Asprochaliko, and later at other southern sites (Papagianni, 2009; Runnels, 1995). At some stratified sites changes are likewise marked by the diminution of flaking methods directed towards the production of thick blanks for scraper production (e.g., Crvena stijena) (Dogandžić & Đuričić, 2017; D. Mihailović, 2017a; D. Mihailović & Whallon, 2017) and the decline of elongated Levallois products for scraper production (e.g., Asprochaliko) (E. Papaconstantinou, 1989; V. Papaconstantinou & Vassilopoulou, 1997). Instead, a production pattern with less formal centripetal flaking, exploitation of cores until their exhaustion, and alternative flake production methods, for example, secondary blank production via cores on flakes, are rather common. Crvena stijena’s late MP deposits (Levels XIV–XII) feature industries with increasing use of discoidal and radial flaking and fewer Levallois products. Most cores are small and reduced, there are cores on flakes, and the particularity is the presence of cores that resulted in the production of small, often elongated products (Dogandžić & Đuričić, 2017; D. Mihailović & Whallon, 2017) (Fig. 7: 4–7).

In the upper Mousterian at Asprochaliko, the so-called ‘Asprochaliko flaking method’ (E. Papaconstantinou, 1989) is similar to discoidal (Bailey et al., 1992; Boëda, 1997) as it produces short and thick blanks, using a ramified production where blanks are used as cores. This production of small pointed pseudo-Levallois points, from Kombewa-like cores or the truncated-faceted method, is considered an indicator of the purposeful production of small blanks (E. Papaconstantinou, 1989; V. Papaconstantinou & Vassilopoulou, 1997). In any case, pseudo-Levallois are very common and almost 30% of all retouched blanks are of this morphology (Fig. 7: 8–11). This is in contrast to somewhat more elongated Levallois element production in the earlier (MIS 5) phases at this site, suggesting a more distinct diachronic change in blank production (Papagianni, 2000, 2009). The small MIS 3 assemblage at Mujina cave contains very few pieces diagnostic of any particular technology of blank production, though the site is interpreted as a workshop which probably reduces the technological variability (Karavanić et al., 2008). Informal cores for small flake production are present as well. Broadly, at Kalamakia, the main temporal shift in the flaking method is the decreasing use of Levallois in upper levels (Darlas, 2007).

These data on temporal variation suggest a change to less ‘standardized’ production methods, at least compared to Levallois production of large blanks/blades, and indicate situational and less formal blank production through centripetal and often discoidal methods. How data from stratified sites at Klissoura and Bioče compare to this is not entirely clear. At Klissoura, the middle and upper parts of the sequence (most likely MIS 3) have evidence of centripetal and discoidal flaking compared to earlier assemblages with notable blade production. Levels at the top of the MP sequence feature many blade and bladelet cores. These are similar to the upper levels at Crvena stijena. At Bioče, levels of the upper sequence (MIS 3?) are characterized overall by centripetal flaking methods. The predominance of discoidal flaking methods in the latest phases is replaced by a slightly higher presence of Levallois products, with the occurrence of blade and elongated products, and flake production through truncated-faceted pieces (Dogandžić & Đuričić, 2017). The chronology of these levels, however, is not yet determined. Additionally, the lithic assemblage of the still-undated (but likely belonging to MIS 3) open-air site of Campanož in Dalmatia primarily consists of discoid and centripetal flaking methods, with very few tools (Banda et al., in press).

The temporal patterns in lithic variation described here are principally based on stratified sites in the southern Mediterranean (Dogandžić & Đuričić, 2017; D. Mihailović & Whallon, 2017; Papagianni, 2009; Sitlivy et al., 2008). This brings us to the question of whether a parallel trajectory is observed in other regions within the Balkans, or whether industrial variations are inherent to geographical regions. It may not be completely unexpected that different regions see contrasting developments, given that in the Balkans the mountainous regions running across the peninsula may have posed barriers to population movements and therefore have structured variation and development of lithic industries, particularly in a north–south direction. Besides population movements, we can suppose that these different environmental conditions may have triggered different technological responses. The central and northern Balkans, however, do not provide many stratified sequences or reliably dated MIS 3 sites that would enable worthwhile comparisons of temporal trends. Some parallels can be observed: for instance, at Pešturina, the late MP does not exhibit a sharp change in technology but contains many denticulates, unlike MIS 5 deposits. At Šalitrena cave, the Levallois method is present in the late MP, but not dominant compared to discoidal and centripetal methods. Additionally, at some other stratified sites in the northern regions potentially the same trend is observed; for instance, at Vindija (Croatia), the most striking temporal trend is the use of Levallois flaking methods in the lower level K, which U-Th dating places in MIS 5e, but Levallois flaking is almost absent in the upper, MIS 3 assemblages, which, furthermore, contain many denticulate and notched tools (Blaser, 2002; Karavanić, 2007; Karavanić & Smith, 1998). At Temnata Dupka, levels III (MIS 4?) and II (MIS 3) show abundant discoidal products, simple flake production, cores on flakes, small residual cores (globular), many retouched flakes, and a relative increase in denticulate tools, but also contain volumetric and Levallois blade production. This pattern is rather similar to the one observed in the south and therefore suggests that this pattern may not be regionally specific.

Geographical Variation

Other stratified sites in the eastern part of the peninsula (Bulgaria), however, show a slightly different picture. Bacho Kiro and Kozarnika have rather homogeneous industries with a varied use of Levallois and discoidal methods and no sharp techno-typological changes, except maybe at Temnata Dupka. Eastern regions, nonetheless, are relevant for another prominent phenomenon of regional variation in lithic industries in the Balkans, and that is the occurrence of bifacial elements. The so-called ‘East Balkan Mousterian’ with bifacial leaf points has been defined as a regional group covering only areas east of the Carpathian mountains, due to its overrepresentation in Bulgarian sites. It is also known as the Mouselievo-Samuilitza complex (Ivanova, 1979; Ivanova & Sirakova, 1995; Sirakova & Ivanova, 1988). Several sites in the northern areas of the Balkans have single examples of bifacial implements—Šalitrena terrace (B. B. Mihailović et al., 2014), Risovača cave in Serbia (Gavela, 1988; D. Mihailović, 2017b), Vindija in Croatia (Karavanić & Smith, 1998; Zilhão, 2009)—and there are several specimens of leaf-points in Kamen (Bosnia), though these latter are somewhat less typical and rather resemble bifacial scrapers (Basler, 1963, 1979; Ivanova, 1979). Petrovaradin likewise has scrapers with bifacial retouch (D. Mihailović, 2009a). This phenomenon is generally associated with MIS 3 and the late Mousterian. Several sites, however, have some evidence of bifacial production in the form of roughout pieces and somewhat less typical bifacial points even earlier than MIS 3, potentially MIS 5 (Temnata Dupka, Kozarnika), which suggests a longer duration of the use of bifacially flaked pieces (Drobniewicz et al., 2011a, 2011b; Guadelli et al., 2005). Notwithstanding regional variation within bifacial production in central and northwestern Europe (Bosinski, 1967; Jöris, 2004; J. Richter, 2000; Ruebens, 2013; Soressi, 2002), it appears that there is a broader geographic delineation of bifacial production in Europe traced along topographic features: mountain regions of the Balkans, the Apennines, and the Pyrenees demarcate Europe’s MP bifacial production, leaving most of the central Balkans outside of this area. There are, however, exceptions: Shiroka Polyana, a high-altitude open-air site in the Rhodopes, has numerous leaf-shaped points (Ivanova, 1994; Ivanova & Sirakova, 1995), and surface assemblages in Peneios river in Thessaly (Runnels, 1988) and a couple in Albania (Runnels et al., 2009) have dozens of Szeletian-like foliates.

This said, an additional aspect to the regional differentiation of the production of bifacially flaked implements is whether their use and production is a characteristic of a particular site type, namely open-air versus in-cave sites, where bifaces appear in smaller numbers. Even at the eponymous site of Samuilizta, the number of foliates is small (n = 10) compared to hundreds of leaf-shaped elements at the open-air site of Mouselievo (Kozłowski, 2003). What remains to be further explored is the question of how bifacial industries, with a likely long history in the eastern Balkans (e.g., Kozarnika), are related to similar phenomena in other parts of Europe (e.g., Micoquian), and particularly their relationship with transitional industries like the Szeletian or the Lincombian-Ranisian-Jerzmanowician.

Notable at several sites is the production of blades and bladelets in the latest phases of the MP. Blade production in Temnata Dupka, Samuilitza in the late MP phases, Bacho Kiro, take the form of a volumetric production and Temnata additionally includes Levallois bidirectional method. These can potentially be examined in the context of transitional industries and their potential role in the Initial Upper Paleolithic (Kozłowski, 2004; Tsanova, 2012). It is still an open question whether these industries were produced by Neanderthals independently or represent an outcome of the interaction of the local population with the newcomers who were entering the continent; there is, however, growing support for the idea that Initial Upper Paleolithic is made by the newcomers themselves (e.g., Hublin, 2015).

Increased elongated blank and blade production is noticed in the south, in Bioče and Klissoura, and at Klissoura and Crvena stijena there are cores for producing small elongated blanks (bladelet) (Dogandžić & Đuričić, 2017; D. Mihailović, 2017a; D. Mihailović & Whallon, 2017; Sitlivy et al., 2008). This occurrence should be examined in the context of the transitional Uluzzian industry, which is best known in southern Italy but is also present in northern Italy (Famane) and Greece at the site of Klissoura. Once more, there is a question of whether this technological behavior is part of the MP technological repertoire, and there is increasing evidence that it was (D. Mihailović, 2017a; D. Mihailović et al., 2017; D. Mihailović & Whallon, 2017; Peresani, 2012; Peresani et al., 2016). It is difficult to envision the influence of modern humans in the production of these elements, given that modern humans were not present in the southern Balkan region until late (see 5.2.3).

Expediency as a Typological Signature in the Late Middle Paleolithic

The principal typological theme in the late MP is the decreasing frequency of scraper types, sometimes associated with reduced tool production in general. Since scrapers, on the one hand, and notches and denticulates on the other, represent the two major categories in the MP tool repertoire (Dibble, 1988), increases in one will necessarily result in a relative decrease in the other group. In contrast to earlier stages, which saw a higher level of scraper production, including at times more regular resharpening, many assemblages of more recent phases feature less emphasis on frequent and repetitive tool edge resharpening (Crvena stijena, Klissoura, Asprochaliko, Mujina cave, Pešturina, Samuilitza). The concomitant increase in the number of denticulates, notches, and simple retouched tools in the uppermost levels is noticed in many sites (Crvena stijena, Mujina cave, Temnata Dupka, Pešturina, Klissoura). For assemblages in the Adriatic region this typological character prompted the proposition of Denticulate facies, represented in caves (Crvena stijena, Mujina cave) and open-air sites in Dalmatia (Basler, 1975b, 1983; Batović, 1988; Karavanić et al., 2008a, 2008b; Rink et al., 2002). It remains to be discerned how dominant these tool types are compared to known Denticulate Mousterian facies, which seem to be more restricted in time and space (Theodoropoulou, 2008; Thiébaut, 2005, 2010). Are they similarly confined in geographical distribution and chronological framework? As for the latter matter, it appears, however, that the incidence of denticulate tools is not restricted to the Mediterranean region (they appear in greater numbers in Temnata Dupka, Pešturina, Vindija) and one cannot speak of a regionally constricted phenomenon as previously thought. Regarding the temporal confinement, with the potential exception of Temnata Dupka (MIS 4?) (Kozłowski, 2002), an increase in denticulate tools is a phenomenon related to the later stages of the MP.

An increase in denticulate tool types, however, is often coupled with an increased occurrence of damaged pieces, thus suggesting that denticulate edges are likely the result of post-depositional processes that produced unintentional edge modifications (Kolobova et al., 2012; Thiébaut, 2005, 2010). An increase in edge damage and pseudo-tools is expected in gravelly sediments, in anthropic sediments with contact between artifacts, or as a result of trampling (Flenniken & Haggerty, 1979; McBrearty et al., 1998; S. P. McPherron et al., 2014; Nielsen, 2011), with the gravel size being one of the major factors affecting the frequencies of edge damage (S. P. McPherron et al., 2014).

At Crvena stijena and Bioče a higher proportion of damaged pieces is observed in levels with a higher incidence of denticulate tools (Dogandžić & Đuričić, 2017). An increase in denticulate tools in Mujina cave parallels an increase in pseudo-tools and damaged edges (Šprem et al., 2020) and particularly in levels with large and dense gravel (Šprem et al., 2020) (Fig. 7: 12–14). Open-air sites on the coast of Dalmatia are also to be questioned as their deposition is not clear and potential reworking that could cause damage is not testable. Moreover, it has been suggested that the dominant type at these open-air assemblages is scrapers (Vujević, 2011; Vujević et al., 2017). One should not, however, forget that this pattern may result from biased selection at open-air sites. Likewise, at Pešturina, the edges of denticulate tools appear to be attributable to taphonomic processes (D. Mihailović & Milošević, 2012). This said, there is a high possibility that taphonomic factors play a role in the frequencies of denticulate pieces in all these assemblages. Before interpreting their significance in terms of industrial variability, edge damage should be examined to account for any taphonomical processes for each assemblage.

Still, denticulate tools are common in contexts of greater reliance on more expedient tool production and use (though there are cases of the opposite: Picin et al., 2011). In Level XII of Crvena stijena and Level 2 of Bioče the percentage of denticulates is high, though scrapers remain more frequent and therefore these assemblages are not defined as a Denticulate Mousterian. At Crvena stijena more denticulate tool production in the most recent Middle Paleolithic level XII is associated with less retouching and less resharpening and with more frequent lightly retouched flakes. Likewise, Level XII, and particularly XIV, have many finely retouched thin flakes (raclette) along with cores that testify to small flake production. By contrast, numerous denticulates co-occur with a high frequency of scrapers and heavier retouch in Bioče upper levels (Dogandžić & Đuričić, 2017). An increased presence of raclettes, backed pieces and retouched flakes is noted in upper MP at Šalitrena cave (B. Mihailović, 2017a, 2017b, 2017c). In the latest, uppermost levels at Klissoura, the low production of tools is linked with fewer scrapers and an increase in denticulate tools, though still in very low amounts (Sitlivy et al., 2008). Mujina cave likewise has very few retouched tools in general. As for the neighboring regions, a trend where scrapers are 'replaced' by denticulates is noted in Vindija (Janković et al., 2006; Karavanić & Smith, 1998), Riparo Mochi (Grimaldi & Santaniello, 2014; Stiner & Kuhn, 1992), and to some extent at sites in coastal Latium (Mussi, 2001).

When this typological signature is integrated with the previously described technological tendency in the later MP phases, a techno-economic model emerges (Dogandžić & Đuričić, 2017). Stone tool production in most cases did not target scraper production and resharpening as a major techno-economic strategy (unlike in the earlier phases). A similar trend is observed in sequences in southwest France, where during earlier phases lithic production is oriented towards scraper production and use (Mellars, 1965; Monnier & Missal, 2014), and we see that a similar trend potentially exists across the Balkans, at least in its southern regions. However, with a spatially and temporally patchy record, we are not entirely confident that the industries in regions beyond the Mediterranean zones (particularly in the east) follow the same pattern. For instance, at Temnata Dupka, throughout the sequence, tools are rare and the majority of the toolkit is comprised of retouched flakes, with scrapers and denticulate often in similar proportions, and at Bacho Kiro scrapers are the main tool type, whose proportions decrease in the latest phases. The previous importance of scraper production is the recognizable quality of earlier periods that is rarely seen in the later MP developments.

In sum, late MP in the Balkans features flaking methods that are already practised throughout the earlier MP periods, centripetal flaking, exploitation of flakes as cores, and in some instances in the later phases, the production of blades through Levallois or volumetric methods. The major theme is the general and varied use of centripetal flaking methods with fluctuating, though usually low, presence of Levallois that is rarely if ever directed towards large flake production. The increased reduction intensity of cores, which then proportionally reduces the number of Levallois blanks at the expense of radial and core edge flakes, especially when the reduction continues to a radial/discoidal manner, is one of the major factors that may affect this pattern; however, reduction intensity as usually measured does not always increase in these instances (Dogandžić & Đuričić, 2017; Papagianni, 2000, 2009). On the other hand, the flaking technology trend in the late MP is accompanied by the lesser importance of retouched elements and the renewal of their working edges.

Many late MP stratified sites in other regions testify to the frequent use of Levallois and discoidal methods of production, though in changing proportions (Faivre et al., 2014; Meignen et al., 2009; Peresani, 2012; Picin et al., 2014). The order of the predominance of blank production methods has become an important factor in recent debates on the existence of chronological trends of Mousterian industries and their relationship with climate and subsistence (Discamps et al., 2011; Peresani, 2012; Picin & Carbonell, 2016; Stiner & Kuhn, 1992). It is worth examining the reasons behind changes in the relative importance of various debitage systems. If population crashes can be envisioned during MIS 4, material culture is expected to reflect these demographic changes with a lack of innovation and a loss of technological traditions in low population densities (Boyd & Richerson, 1985; Premo & Hublin, 2009; Premo & Kuhn, 2010; J. Richter, 2000). Likewise, climatic deterioration may have induced changes in subsistence and mobility strategies that would require changing techno-economic strategies. While certain trends do appear (e.g., low Levallois components in many later MP sites), directionality in technological behavior in Balkan assemblages might be more difficult to demonstrate, as variation in stone tool production is rather low (this is particularly challenging when chronological information is lacking). Changes in strategies rarely involve the exclusive use of one flaking method over another. If this were the case, one might link these patterns to cultural groups delimited in time and space (e.g., Jaubert et al., 2011). This is less likely to explain the record. The shift seen in the late MP might be associated with situational, techno-economic behavior that would favor increasing the number of blanks and new edges rather than tool portability as expressed in repeated edge modification (Delagnes & Meignen, 2006; Delagnes & Rendu, 2011).

Biogeography of the Middle and Early Upper Paleolithic Hominins in the Balkans

Parts of Pleistocene Europe underwent frequent episodes of human depopulation and recolonization resulting in only a partial occupation of certain areas (Dalén et al., 2012; Dennell et al., 2011; Hublin & Roebroeks, 2009; Roebroeks et al., 2011). The position of the Balkans for understanding this population history is somewhat contradictory; its refugial character and the presence of migration corridors would suggest a continuous and rich record of human occupation and yet the available record implies an almost sporadic human presence. This situation calls for an evaluation of the causes of the apparently sparse record, in particular, whether it simply reflects a research bias or whether indeed it is a true pattern and, if so, what can account for it. We will make an attempt, based on the evidence at hand which is admittedly rather patchy, to illustrate the occupational history of the Balkans during the Middle and early Upper Paleolithic, both chronologically and spatially. Occupations in the early Upper Paleolithic are relevant for contextualizing the Neanderthal demise and population replacement.

Genetic studies on past human demography primarily centre around the question of effective (breeding) population size, which varies widely (Briggs et al., 2009; Fabre et al., 2009) and does not relate to census population size in a straightforward way (Bocquet-Appel & Degioanni, 2013). It is generally considered, however, that the Neanderthal metapopulation was rather small, that it experienced several bottlenecks (Dalén et al., 2012; Reich et al., 2010), and that low population density characterized Neanderthal occupation of Europe (Churchill, 2014; Dalén et al., 2012; Lahr & Foley, 2003; Roebroeks et al., 2011; Stiner et al., 1999). From the archaeological perspective, evaluating past demography relies on parameters related to the intensity of occupations at a regional and site level—a number of sites per time period, various proxies of occupational intensity, patterns of radiocarbon dates, and so forth (Bocquet-Appel, 2000; Bocquet-Appel & Degioanni, 2013; Conard et al., 2012; Dogandžić & McPherron, 2013; French, 2015; French & Collins, 2015; Mellars & French, 2011, 2013; Surovell & Brantingham, 2007). Linking these parameters to population estimates is not a straightforward task, largely due to issues of site sedimentation rates and erosional episodes, the character of site use, varied collection curation and report/publication practices coupled with different excavation methods (Conard et al., 2012; Dogandžić & McPherron, 2013); most sites from the Balkans do not provide fully reliable data on find densities, both lithics and faunal, and consequently do not allow for data comparisons between sites across this region. This said, we will focus on the presence and absence of human occupation in a given period assuming that factors affecting the preservation of sites are acting equally across the regions and do not represent a significant bias. The archaeological record in general increases through time; any departure from this pattern is relevant for population biogeography. Of course, further research may invalidate any pattern inferred from the current archaeological record.

Staying Warm—No Clear Evidence of a Refugium in MIS 4

The earliest phases of the MP (before and during MIS 6) are represented by a small number of sites. Overall, the Balkan data shows a peak in site abundance in MIS 5 and then again in MIS 3 with an apparent interruption during MIS 4 (Fig. 6). Therefore, Neanderthal higher occupation intensities in MIS 5 and later repopulation of the Balkans in the MIS 3 conforms to the wider European pattern where the most intense Neanderthal habitation is related not only to warm climatic conditions but also to periods of low climatic fluctuations (Lahr & Foley, 2003; J. Richter, 2016; Stringer et al., 2003). In the Balkans, this may be expected even for the southern regions: for example, at Theopetra, human occupation is testified to during warm climatic periods, such as the Last Interglacial, but there is almost none during cold events that left traces in the form of frost actions (Karkanas, 2001; Karkanas et al., 2015). Additionally, at a site level, anthropogenic features can be used as indications of site-use intensity. Combustion features are common at several sites, in particular during earlier periods (MIS 5), for example, Crvena stijena’s (Morley, 2007) and Theopetra’s (Karkanas, 2001; Karkanas et al., 2015; Tsartsidou et al., 2014) Last Interglacial levels with superimposed colored sublayers—whitish, reddish, blackish rich in charcoal elements, representing accumulations resulting from several burning episodes; Krapina, where Gorjanovic-Kramberger observed sub-layers with fire-places and burnt bones within the 8–9 m of Krapina’s stratigraphic complex dated to MIS 5e; Lakonis I sequence, dated to the post-MIS 5e period, exhibits intensive use of hearths. Intense fire traces at Temnata Dupka, in contrast, are present in levels with one TL date of a later, MIS 4 stage (given the overall pattern of lack of MIS 4 occupations and the lack of more reliable chronology, this one should be treated with caution). This pattern of more regular use of fire in warm climatic phases is noted in southwest France as well; the explanation of this pattern, though still controversial, is Neanderthals’ lack of the technology to create fire and their reliance on natural fires, which are more frequent in warm and humid conditions (Sandgathe et al., 2011).

The basic premise of the refugia hypothesis is that if a region were a refugium for humans, then their presence should be continuous regardless of the climatic conditions; more specifically, even during the cold climate, human populations would occupy the region. The observed paucity in the record of Neanderthal occupation during MIS 4 fits with the MP record more generally of Europe (Davies, van Andel, et al., 2003; Stringer et al., 2003), where Neanderthal populations occupied only limited regions during MIS 4 (e.g., SW France: Turq, 1999) and are greatly reduced even in other southern refugia, Spain and Portugal (d’Errico & Sánchez Goñi, 2003).

It has been difficult to demonstrate evidence for MIS 4 occupations in the Balkans. The best cases to support MIS 4 occupations are Temnata Dupka, Crvena stijena, and Klissoura. The latter two have a continuous record of human occupation across dozens of levels, suggesting habitation even during this stage. At Crvena stijena (XX) and Temnata Dupka (6) TL dates point to MIS 4, both associated with rich evidence of human activities, in terms of the density of remains and the fire features (Ginter et al., 1992; March et al., 2017). We would like, however, to emphasize the need for a more reliable chronology at several sites and the site-specific circumstances such as stratigraphic and/or occupational hiati that would substantiate the MIS 4 occupation. Even if we accept these as reliable indicators of habitations during this cold phase, the pattern certainly appears to show a drop in occupation, whereas a consistent number of occupations, if not an increase, is to be expected if the region was a refugium during this time. Until further reliable chronology can be obtained, our notion of MIS 4 will remain largely devoid of humans.

It is the southern regions of the peninsula that display stratified sites that bear the rich and continuous evidence of human habitation. Examples are long sequences with over a dozen MP layers rich in finds at Crvena stijena, Klissoura, and the stratified site of Bioče, with remarkably high artifact density. This suggests that, as previously suggested, the southern areas are more likely to represent refugia for fauna, flora and humans. But even within the Mediterranean region, there is a climatic and environmental contrast that may be decisive for the likelihood of the continuity of human occupation across different climatic phases; for example, in Greece, it has been observed that the western areas—where water resources are more available and precipitation, and therefore vegetation cover, is higher in comparison to the eastern, more continental regions—have higher Paleolithic site densities (Elefanti & Marshall, 2015; Tourloukis & Harvati, 2017). In the extreme south, the Mani peninsula on Peloponnese, has a relatively high number of MP locations (Tourloukis & Harvati, 2017); however, in this case this may be the result of increased research. Additionally, the site’s attractiveness for habitation depended on its location, topography, and immediate conditions. Avoidance of some cave sites during harsher climatic conditions may be due to their unfavorable position and location, as is suggested for Theopetra, since populations would have favored locations in different topographic settings (Karkanas, 2001). This said, it may be plausible that the Balkan peninsula was not a refugium in its entirety, but that the refugia were confined to the southern regions. These have a relatively more suitable climate and environment than mountainous and northern areas for uninterrupted occupations. Moreover, within these regions, there are smaller localized areas even more likely to be habitation ‘shelters’, as ‘refugia within refugia’ (Gómez & Lunt, 2007), depending on the local environmental factors (Feliner, 2011; Gómez & Lunt, 2007; Tzedakis, 2004).

Neanderthal Demise and Homo sapiens into the Balkans

Recent decades have seen intense research on the intriguing events associated with the demise of Neanderthals and the arrival of Homo sapiens from Africa into Europe. Major debates have focused on the circumstances of the population replacement, the nuances of the chronology of these events, the overlap or geographical separation of the two populations, potential contact, acculturation, competition, and so on.

Support for the idea of temporal overlap between these two groups comes from an analysis of radiocarbon ages across Europe—which suggests that their coexistence, on a continental scale, lasted as long as 5000 years (Higham et al., 2014)—and genetic evidence of their interbreeding (Fu et al., 2014, 2015). Scenarios for the demise of Neanderthals, however, most probably differed across Eurasia, and the extent of their possible encounters should be addressed on a regional/local scale (e.g., Conard et al., 2006; Jöris et al., 2011; Mallol et al., 2012; Pinhasi et al., 2011), with a consideration of the particularities of each region, and with the possibility that the routes, and course of events, followed by the MP–UP transition differed across the continent. The recent genetic evidence suggesting interbreeding, from the Peștera cu Oase (Romania) fossils (Fu et al., 2015; Sankararaman et al., 2012), makes the Balkans particularly interesting for investigating these questions. Here, at least, it seems that Homo sapiens directly encountered Neanderthals.

Still, building a regional model of these events for the Balkans is not a straightforward task. The chronology of the last Neanderthal and the arrival of Homo sapiens primarily relies on radiocarbon ages, and here we are dealing with events at the limit of the method’s range, around 40–45 ka BP, where extra special care has to be taken to address issues of contamination. Thus in assessing the late presence of Neanderthals, one has to expect potential outliers, given that many reported samples have not been treated with the most up-to-date and rigorous pre-treatment methods (i.e., ultrafiltration for bones and ABOx for charcoal). Newer methods almost invariably result in older ages (Devièse et al., 2017; Higham et al., 2006). Caution is likewise recommended when only a few samples are available from particularly critical levels, given the possibility for post-depositional processes to impact the vertical integrity of stratigraphic units. Focusing on broader patterns based on solid data is the first step in building models for this period (Fig. 8).

Fig. 8
figure 8

Location of transitional, Initial, and Early Upper Paleolithic sites. Sites: (1) Šandalja, (2) Bukovac, (3) Vindija, (4) Aurignacian sites in north Bosnia (Lonđa, Danilovića brdo, etc.), (5) Crvena stijena, (6) Bioče, (7) Šalitrena cave, (8) Golema pesht, (9) Crvenka, (10) Orlovača, (11) Peștera cu Oase, (12) Baranica, (13) sites in Romanian Banat (Românești) Dumbrăvița, Coșava, Tincova), 14) Tabula Traiana cave, (15) Kolomnitza, (16) Kozarnika, (17) Klissoura, (18) Franchti, (19) Toplitza, (20) – Temnata Dupka, (21) Bacho Kiro. Color codes: purple: Initial UP; green: Uluzzian, light blue: proto-Aurignacian; blue: Aurignacian; red: CI tephra or cryptotephra; white: Homo sapiens sapiens fossil (Color figure online)

Full size image

No Evidence of Neanderthal Occupation After 44/43 ka BP

There is, however, agreement that the last occurrences of MP, representative of Neanderthals, in most of Europe date to ~ 40 ka BP (Higham et al., 2014). For a long time, the Balkans have figured as a region where some of the last Neanderthals in Europe could be found. This idea comes from a series of radiometric ages from the site of Vindija (Croatia) (Higham et al., 2006; Serre et al., 2004; Smith et al., 1999; Wild et al., 2001) (Table SM1). Most of the dated bones originate from Level G1, a level that has been in focus for its co-occurrence of Aurignacian artifacts and Neanderthal fossils. Bearing in mind significant post-depositional reworking of this level (Bruner, 2009; Karavanić & Smith, 2013; Zilhão, 2009), a more conservative approach of focusing only on the ages of the Neanderthal fossils and not on the other faunal specimens (e.g., bear bones), whose association with the fossil material is unclear, is warranted. One of the Neanderthal bones from G1 was directly dated to 38,580–34860 BP (Higham et al., 2006). More recent data, however, has shown that this date can no longer be taken as an indication of a late Neanderthal presence because the pretreatment did not remove all the contaminants. A new single-amino acid AMS dating method of Neanderthal bones from Vindija places them well before 40 ka BP, and more likely predating even ~ 44ka BP (Devièse et al., 2017).

Archeological evidence of a late persistence of Neanderthals (after 40 ka BP) in the region is difficult to demonstrate. Radiometric dating of MP sites (Table SM1) suggests that, aside from infinite or minimum ages (e.g., Asprochaliko, Smolućka cave, Bacho Kiro), most dates fall earlier than 41–42 ka BP. A late date at Pešturina Level 3 (RTD7231, 33kaBP) is incompatible with the stratigraphy and relative position of the samples (Alex & Boaretto, 2014), and potentially represents an intrusion from Gravettian levels (Alex et al., 2019). Radiocarbon ages from this level do not indicate an MP occupation later than 43 ka BP. An ESR age of the last MP occupation at this site places it at 38.9 ± 2.5 ka BP (Blackwell et al., 2014). With this error range, this age is close to AMS ages of 43–42 ka BP and cannot be considered as a reliable indicator of late Neanderthal persistence. Similarly, a radiocarbon age at Mujina cave made on charcoal is several thousand years younger than dates obtained on bone from any other MP level at that site, as well as several dates on bone, pretreated with ABA method, with unexpectedly young dates (Boschian et al., 2017; Rink et al., 2002) (Table SM1) and all should be viewed with caution. The most reliable chronology still places the latest age at this site at ~ 43 ka BP. Likewise, a late age for the MP in Kličevica likely results from bad collagen preservation and will be further tested (Karavanić et al., 2017), and the one from Vinica at 36 ka cal BP is made on a bone that does not display anthropic modification (Vukosavljević et al., 2022). Another signal of Neanderthals after 40 ka are deposits with Middle Palaeolithic artifacts at Bioče found above the ash layer that presumably corresponds to Campanian Ignimbrite tephra (Vishnevskiy et al., 2020). This evidence, however, is not unambiguous, as more details are needed about the ash layers and the taphonomy of the sequence. U-Th dates on flowstones from Kličevica in Croatia provided an age younger than 40 ka BP (Karavanić et al., 2021). Radiocarbon ages from Dubočka-Kozja cave in Croatia fall at c. 41–37.5 kya cal BP; however, a further dating program is planned to clarify this chronology (Borić et al., 2022).

In conclusion, the evidence of Neanderthal occupation after ~ 43 ka BP is rather scant and for the moment rests on a couple of dates that are of questionable quality. Aside from layers at Bioče that still need to be taphonomically assessed, no MP occupation is found in sequences above the Campanian Ignimbrite tephra at 39 ka BP, which can then be used as a terminus ante quem for the end of Neanderthals in this region.

First Homo sapiens in the Balkans and the Associated Industries

The first fossil evidence of Homo sapiens in the region is represented by the Peștera cu Oase (Romania) fossils, discovered out of archaeological context but with radiometric ages close to 40 ka BP (Rougier et al., 2007; Trinkaus et al., 2003). Its age is compatible with the Aurginacian, and the Aurignacian technocomplex has been identified as an archaeological signature of early Homo sapiens humans in Europe (Fig. 9).

Fig. 9
figure 9

Chronological span of late MP and early UP sites in the Balkans according to currently available ages. The δ18O variations from North Greenland Ice-Core Project (NGRIP) are taken from http://www.gfy.ku.dk/~www-glac/data/gripdelta.dat. Greenland stadials (GS) are marked in blue and Greenland interstadials (GIS) in red. The red line indicates the time of the Campanian Ignimbrite (CI) and the blue bands Heinrich Events 4 (H4) and 5 (H5). Orange circle—age of a modern human fossil from Peștera cu Oase. Uluzzian age estimate based on Klissoura, IUP/Transitional based on Bacho Kiro and Temnata Dupka, and proto-Aurignacian age on Kozarnika and Românești-Dumbrăvița (Color figure online)

Full size image

Initial Upper Palaeolithic (IUP) industries are another proxy for the first H. sapiens attempts at colonizing Eurasia (Bar‐Yosef, 2006; Hoffecker, 2011; Hublin, 2015; Hublin et al., 2020; Kuhn & Zwyns, 2014; Müller et al., 2011; Svoboda, 2005). IUP assemblages incorporate MP features such as Levallois elements, flat core surfaces, and faceted platforms with prismatic blade production and a UP tool repertoire (Kuhn & Zwyns, 2014). Potential IUP industries include the Emirian in the Levant (Boker Tachtit: Marks, 1983); Üçağızlı in Turkey (Kuhn et al., 2009); and Bohunician in central Europe (D. Richter et al., 2008, 2009). In Europe, they are likely related to at least Greenland Stadial 12 (Nigst, 2012, 2019; Nigst et al., 2014) and by HE5 (Greenland Interstadial 13) (Müller et al., 2011). In the Balkans, Temnata Dupka and Bacho Kiro (Kuhn & Zwyns, 2014; Teyssandier, 2008; Tsanova, 2008) have been linked under the name Bachokirian and are considered IUP. At Bacho Kiro, IUP is associated with H. sapiens and an age of 45 ka cal BP, likely even earlier at 47ka cal BP (Fewlass et al., 2020; Hublin et al., 2020). With IUP ages in the Levant around 50ka BP (Boaretto et al., 2021) and their association with H. sapiens fossils (Bosch et al., 2015; Douka et al., 2013), it is evident that these industries are a proxy for east–west dispersal of H. sapiens populations who did not leave descendants to present-day population (Fu et al., 2014). It has been stressed that the industries of Temnata Dupka and Bacho Kiro show a techno-typological continuity from the Middle to the Upper Paleolithic reflected in the transformation of the Levallois method to volumetric blade production and increase in UP tool types (Tsanova, 2008; Tsanova et al., 2012). Additionally, a transitional assemblage from Unit Ia at Lakonis I (Greece) features prismatic blade production, Levallois production of points, and some bifacial elements, similar to the Bohunician, and is chronologically indistinguishable from the latest MP (Elefanti et al., 2006; Panagopoulou et al., 2004).

Of other transitional industries, the Uluzzian is present in the southern Balkans only at Klissoura cave (Greece) (Douka et al., 2014; Koumouzelis, 2001). The Uluzzian is mainly known from Italy and it maintains many MP elements, such as mainly flake production with some bipolar knapping, and exhibits new artifact forms such as lunates (microliths with backed arched retouch), UP tools, bone tools, and pigment (Moroni et al., 2012; Peresani, 2008; Peresani et al., 2016). In Italy, its association with H. sapiens fossils at Grotta del Cavallo has been disputed (Benazzi et al., 2011; Zilhão et al., 2015), while its techno-typological continuity with the MP substrate is reconsidered (D. Mihailović & Whallon, 2017; Peresani, 2008, 2012; Peresani et al., 2016). The CI tephra overlies the Uluzzian at Klissoura as it does in Cavallo (Giaccio et al., 2008) and can serve as a temporal marker with 45 ka cal BP being the likely start of the Uluzzian technocomplex (Douka et al., 2014).

This said, radiocarbon dates from Klissoura (Table SM2) can be considered minimal, and most likely the beginning of Uluzzian can be placed at around 45 ka cal BP (Douka et al., 2014).

Another transitional industry, Szeletian, is not securely confirmed in this region. It is represented by a handful of bifacial leaf-like points found at a few sites (in a disturbed level G1 at Vindija: Malez, 1979; Zilhão, 2009), one at Risovača (Serbia) (D. Mihailović, 2017b; Gavela, 1969), and several at Kamen (Bosnia) (Basler, 1979)). It is also unclear whether industries with elongated leaf-points in Bulgaria (Muselievo, Samuilitza) that also have UP elements (Tsanova, 2008) are related to the Szeletian.

A more conservative view takes Aurignacian as an unquestionable indicator of Homo sapiens. Current knowledge of the technological character of different Aurignacian variants and chronology in the Balkans remains in its incipient stages.

The earliest Aurignacian presence in the Balkans is associated with the so-called Kozarnikian assemblage (Level VII at Kozarnika cave) (Sirakov et al., 2007), dated between 43 and 41 ka BP (Guadelli et al., 2005) and corresponding to Protoaurignacian in western Europe (Bon, 2002; Le Brun-Ricalens, 2005; Le Brun-Ricalens et al., 2009) and Ahmarian in the Levant (Kuhn et al., 2009; Otte et al., 2011; Tsanova et al., 2012). On the left bank of the Danube, in Romanian Banat, several notable sites (Tincova, Românești-Dumbrăvița, Coșava) contain industries that incorporate elements of both Proto- and Early Aurignacian (Anghelinu et al., 2012; Chu, 2018; Sitlivy, 2016; Teyssandier, 2003). Their chronology is indicated by the average TL ages at Românești-Dumbrăvița that place the GH3 level (Protoaurignacian with some Early Aurignacian elements) at ~ 40.6 ± 1.5 ka (Schmidt et al., 2013). Several other Aurignacian sites (Fig. 9, Table SM2) belong either to its earlier phases [Crvenka, At (D. Mihailović, 1992; D. Mihailović et al., 2011; Nett et al., 2021), Baranica (Dimitrijević, 2011), Tabula Traiana cave (Borić et al., 2012)], or late phases of this technocomplex [Šandalja (Karavanić, 2003), Šalitrena (B. Mihailović, 2013; D. Mihailović & Mihailović, 2014)], and several sites in northern Bosnia (Basler, 1979; Montet-White, 1992; Montet-White et al., 1986).

Tephra layers attributed to the CI eruption can serve as a chronostratigraphic marker for these industries. The radiometric chronology of the IUP/EUP at Temanta Dupka and its position under the CI tephra (Bluszcz et al., 1992; Fedele et al., 2003) serves as an indicator of their presence before the eruption. Transitional (Uluzzian) and EUP (Protoaurignacian) industries are found underneath the CI tephra at several sites in Italy (Fedele et al., 2003, 2008; Giaccio et al., 2006, 2008). In the Balkans, this is demonstrated only with the Uluzzian at Klissoura (Koumouzelis et al., 2001; Kuhn et al., 2010).

Population Replacement Scenarios

The gradual disappearance of Neanderthals coincided with the arrival of Homo sapiens in Europe and the climatically unstable MIS 3. Harsh climatic episodes of HEs and the aftermath of the CI eruption at 39,280 ± 110 yr BP might have severely affected Neanderthal habitat (Fedele et al., 2003; Fitzsimmons et al., 2013), and depleted their populations, which contracted their ranges to the southern refugia (G. Bailey et al., 2008; Finlayson et al., 2006; Jennings et al., 2011; Zilhão et al., 2010) and to cryptic refugia in the north, such as Belgium. Ultimately, such conditions might have acted as the leading driver of their extinction. As argued above, it seems that, in the Balkans anyway, the MP ended before this eruption. Neanderthals’ gradual disappearance might have started earlier, namely during the HE5 at 48 ka BP, when their depopulation facilitated repopulation of the region by Homo sapiens (Müller et al., 2011; Staubwasser et al., 2018).

Genetic (Fu et al., 2015; Green et al., 2010; Prüfer et al., 2014) and morphological (Smith et al., 2005, 2016) evidence suggests contact and interaction between the two populations. Archaeological evidence in this region is, however, rare. Understanding population replacement during this period is contingent on chronology, the question of who made transitional industries, and how spatially distributed the populations were across the region.

Regarding the spatial extent of Homo sapiens dispersals, conspicuous is the geographical distribution of UP sites (Fig. 9). The IUP, for the moment, is represented at only two sites in the north of the peninsula. The EUP (Aurignacian) is likewise evidenced in the northern regions, along the Danube corridor, as expected according to models that see the Danube as a migration route from east to west (Conard & Bolus, 2003; Kozłowski, 1992), though this hypothesis requires further support to be corroborated (Chu, 2018). Transitional (Uluzzian) and UP assemblages are located along the southern coastal route, but, with the current state of research, only in Greece and not further to the north along the Adriatic (Karavanić, 2009). The record will be reviewed with respect to the main regions of the peninsula (northern river valleys, central mountainous region, and southern coast).

In the north of the peninsula, along the Danube valley, chronological overlap with Neanderthals is a likely scenario, Neanderthals met their demise by at least ~ 41 ka BP and more likely around 43–44 ka BP, and Homo sapiens with IUP industries entered Europe around 47ka BP.

The overlap is less apparent for late MP and the EUP; the ages of the last Neanderthal occupations in the peninsula correspond to the ages of the EUP (Protoaurignacian at Kozarnika, at 43–42 ka BP). It has been hypothesized, given the distribution of MP and Aurignacian sites in the Balkans, that there was a geographical segregation of the two populations, where Neanderthals lived in the mountainous regions of the Balkans and Homo sapiens settled the river routes, with both populations maintaining their territoriality (Mihailović, 2004; Mihailović & B. Mihailović, 2014a, 2014b). But the question of overlap rests on the nuances of the radiocarbon ages and, for the moment, we lack high-resolution dating across the region. Unless new, more precise, accurate, and reliable dates undoubtedly demonstrate the temporal overlap of Aurignacian, even Protoaurignacian phases, and the last MP, this scenario may not be supported.

The IUP, however, is chronologically synchronous with the MP. In this case, Neanderthal demise might have been associated with competition with the incoming population, eventually suggesting that the combined effect of climate and competition was a decisive factor in their demise. Depending on different population characteristics of both Neanderthals and Homo sapiens, such as size and aggregation, two models of population interaction have been recently proposed (D. Mihailović, 2019): (1) with a strong demographic potential of Homo sapiens, Neanderthals would have retreated from the low-elevation areas inhabited by the newcomers, and into the mountainous areas; (2) alternatively, acculturation of Neanderthals, reflected in transitional industries, would have been possible with low demographic potential of Homo sapiens (D. Mihailović, 2019). The population replacement events in this area include the overlap of the two populations, but understanding the character and dynamics of their interactions certainly requires further data and analyses. The nature of these encounters remains uncertain. While the hybridization scenario does not rule out the possibility of conflictual relations between different populations, especially if there was some level of competition involved, it is important to note that the present archaeological data is not sufficient to allow us to definitively infer the exact nature of these encounters.

In the south of the peninsula, along the Mediterranean coast, the MP ends by 43–41 ka BP, and its last phases are contemporaneous with the Uluzzian at Klissoura. No site with this industry has thus far been found on the Adriatic coast, though it is assumed that it represents a phenomenon that should link the north Italian coast and Greece (Peresani, 2008). The fossil association of this industry with modern humans has been challenged (Zilhão et al., 2015) and the techno-typological assessment rather sees the continuity between the MP and Uluzzian industries (D. Mihailović & Whallon, 2017; Peresani, 2012; Peresani et al., 2016). Again, the population replacement scenario depends on the issue of the Uluzzian makers; if made by modern humans, the temporal overlap would support the idea that the UP elements in Mousterian are a result of acculturation, a scenario possible with a low demographic potential of Homo sapiens (D. Mihailović, 2019). Alternatively, novel elements in Uluzzian represent independent Neanderthal innovations within Mousterian (Mihailović & Whallon, 2017; Peresani et al., 2016). This scenario would also be supported by the current record, which indicates an absence of coexistence between Neanderthals and EUP Homo sapiens in the maritime region and that Neanderthal demise was not associated with new populations in the area. Furthermore, this area was devoid of humans and largely depopulated since the disappearance of Neanderthals. The proximity to the origin of CI eruption and the severe conditions that followed might have had an impact on human survival. Only after several millennia, we see repopulation of the area by EUP populations inhabiting the region, as evidenced by Šandalja (Karavanić, 2003), sites in Albania (Hauck et al., 2016), as well as Franchthi and Kolomnitsa (Darlas & Psathi, 2016).

The spatial distribution of populations provides additional insights into the occupational history during this time. Transitional and EUP industries cluster in two areas: in the south, transitional (Uluzzian) and EUP (Fig. 9) are found in the coastal regions; and, in the north of the peninsula, IUP and EUP sites are distributed in the low hill ranges along the valleys of the Sava, the Danube and their tributaries (Fig. 9) (Karavanić, 2009; D. Mihailović et al., 2011). The central mountainous regions are devoid of sites with EUP, and stratified sites typically show sequences that show a gap from MP to the Gravettian (D. Mihailović et al., 2011). The two only sites with IUP (Bachokirian) present in the Balkans are likewise located in the same area, suggesting similar dispersal routes and settlement areas of Homo sapiens at different periods. Unless site preservation is an issue, IUP sites are rather rare, which raises the question of the intensity of these dispersals and the level of permanency of their settlements, potentially representing the ‘pioneer’ colonization (Davies, 2007; Fu et al., 2014).

The location of Balkan Aurignacian settlements—along the river valleys, in low-elevation locations—follows the observed preference for fluvial environments of EUP sites elsewhere in Europe (Davies, 2007; Hauck et al., 2018; Hussain & Floss, 2015). Their settlements rarely occupied higher elevations (with the exception of the high-altitude locations of Potočka zijalka and Mokriška jama in Slovenia). We may speculate about why they occupied only particular geographical and/or ecological zones (Chu, 2018). We can rule out the possibility that, at least in the Balkans, this results from a territorial segregation from Neanderthals, given that there was no long-term coexistence between the two populations in the region, and the fact that the Aurignacian did not enter the mountainous region even several millennia after Neanderthals disappeared. We may assume that the reason for this geographical, topographical, and environmental preference does not lie in segregation from and competition with Neanderthal populations but rather in landscape choices that are related to settlement systems that would facilitate settling in new environments (Hussain & Floss, 2015). The near absence of EUP sites in the east Adriatic coast may stem from the fact that most of the low elevation areas (the seeming preference of Aurignacian habitation) in the great Adriatic plain were then submerged under water and sites are almost impossible to discover (but see Karavanić et al., 2014, 2016).

Directions for Future Research on Hominin Biogeography in the Balkans

The position of the Balkan peninsula as a region with a disproportionately sparse Pleistocene record relative to its large territory is changing slowly, thanks to the intensification of research in recent decades, with an attempt to discover Pleistocene stratified locations as well as open-air locations with surface finds. This research has been conducted in many countries, including Serbia (Dogandžić et al., 2014; Kuhn et al., 2014; D. Mihailović, 2008; D. Mihailović et al., 2014), Greece (Panagopoulou et al., 2015; Tourloukis et al., 2016), and Croatia (Karavanić et al., 2016).

One of the main questions to be addressed in future work is the intensity of human occupation and the role of geography, topography, and environment in human habitat choice. When considering the topic of refugia during the climatic cycles, it is important to evaluate whether the whole region was equally occupied, and whether some areas experienced discontinuities in habitation or desertion during harsher conditions. Before applying archaeological proxies it becomes necessary to first eliminate research bias and geological gaps; this is feasible only with extended research (e.g., Tourloukis & Harvati, 2017).

The Pleistocene record at hand, however, offers several premises about human occupation in terms of geography and site preference that could be further explored. What differentiates MP occupation patterns LP and UP is the almost regular habitation of mountainous areas of mid- and high-altitude areas, often at sites that lack any LP or early UP occupations (Karavanic, 2000; D. Mihailovic & Mihailovic, 2009; Tourloukis & Harvati, 2017), even though many MP deposits are located at stratified sites in regions of milder topography. Most of these areas remained largely unexplored and they are certainly relevant for two research avenues: the role of refugia, and habitat preference with regard to the mobility and subsistence practices of Neanderthals.

At the two other ends of the chronological spectrum, the LP and UP demonstrate an almost distinct preference for topographically/environmentally different habitats. The lack of settlements in the LP–MP transition (> 200 ka) in the karstic areas is apparent in many regions in Europe, with locations of human habitation at low elevations in river terraces (Hopkinson, 2007). This pattern is further corroborated by recent LP surface finds in Serbia (D. Mihailović et al., 2014; D. Mihailović & Bogićević, 2016), and an open-air site in Greece dates to MIS 14 or MIS 16 (Panagopoulou et al., 2015, 2018). Some gaps, however, should be expected due to geological factors: for instance, long loess sequences in northern areas have accumulated over deposits containing LP occupation (e.g., Romanowska, 2012).

Littoral contexts (northern river valleys and the Mediterranean coast) are also favored in the early UP, as a consequence of populations dispersing along the river corridors and/or the particularities of their settlement and mobility patterns. Aside from the large Danube valley, other river valleys running in a south–north direction (Marica–Nisava and Vardar–Morava corridors) should be considered to be migration routes (D. Mihailović, 2019). Thus, the focus of the new research (survey) endeavors targeting this period should be on geomorphological and topographic aspects, including low elevations, drainage catchments, and lakes. Furthermore, it has been proposed that dense and continuous presence of humans is evident in several regions of Europe with prevalent topographies of river valleys and lowlands with diverse resources (the valleys of Dordogne and Vezere in southwest France, Ardennes in Belgium, and the Middle Danube basin), and positioned in the central areas of the continent, where the northern and southern environmental and climatic regions overlap (Davies, Valdes, Ross, &Van Andel, 2003a, b; Stewart, 2005). In the Balkans, those favorable conditions could be identified in the northern plains with rich water resources, and in southern maritime areas with almost continuous forest presence. Locating ‘attractive’ regions with diverse and/or favorable ecological conditions could be helpful in the process of locating stratified sites. The potential locations are where the northern river valleys meet the mountainous or hilly regions, and, eventually, where most of the stratified sites have been found. For example, the Morava river valley is one potential area (Dogandžić et al., 2014), since, even today, it has a higher average annual temperature than the hilly regions (Ducić & Radovanović, 2005).

Conclusion

Compared to the other regions of Europe, the MP record of the Balkans is inadequate for confidently building large-scale interpretations of Neanderthal adaptations. The obvious impediment to a clearer understanding of Middle Paleolithic variation in the Balkans is the relative scarcity of sites over a large territory. The review presented here provides a provisional model of the variation of MP industries based on occurrences and frequencies of techno-typological elements. Such broad trends can be very informative (e.g., Monnier & Missal, 2014) and appropriate for building preliminary models of industrial variation in the region where the resolution of data and chronology is rather low. This, and the obstacles involved in the study based on a literature review, mean that these models remain rather tentative. Most ideas presented here take the form of hypotheses that should provide a direction for further in-depth research.

We would like to underscore several points stemming from this review. Techno-typological variability of the Balkan MP appears to follow certain temporal trends. The earliest MP phenomena, evident even before MIS 6, are related to the wide use of scrapers and the production of thick blanks, bearing similarities to Quina technological behavior. This is followed by the introduction and extensive use of Levallois technology starting from MIS 6 and the comparable importance of scraper production, tool modification, and various levels of resharpening. Similar trends are observed in SW France, as well as in the east (e.g., Karain). Once these technological methods entered the techno-typological repertoire, they became building blocks of MP variability and were used throughout the early glacial.

MIS 5 industries demonstrate blank production that is more frequently organized towards the production of Levallois elements, the production of elongated blanks, and parallel use of the production of thick blanks, all geared for scraper production. The late MP, however, sees a decline in the use of these strategies. Flaking methods were not necessarily directed towards Levallois or Quina-like blank production, but rather employ informal radial and discoidal methods, blanks were less likely to be retouched into scrapers, and at many sites, denticulates abound. The degree of occurrence of denticulate tools depends on taphonomic factors as well, and it is advisable to evaluate the edge-damage frequencies resulting from taphonomic processes before inferring the typological character of each assemblage. It is notable, however, that many late MP assemblages exhibit higher numbers of these tool types, especially since it appears within a general trend where scrapers are a less frequent tool type compared to earlier assemblages, and where more tools feature lightly invasive retouch.

Regional differences in the variation of MP industries in the Balkans are moderately pronounced. One noticeable contrast is along the east–west axis. In the east of the peninsula, Quina-like methods of thick blank production have not been observed. Furthermore, industries with a high incidence of the production of bifacial products are mainly known from the east of the peninsula (Bulgaria). Additionally, a north–south distinction is observed as well: isolated bifacial pieces are found along the northern valleys of the entire peninsula but not in the inland or the south. This geographical pattern is visible across the continent where industries with bifacial tools appear in the northern regions (Bosinski, 1967; Jöris, 2004; Ruebens, 2013).

The Balkans has a unique potential for examining broader trends in the variation of MP industries at larger, interregional scales (Dibble, 1991), as it provides a valuable link between the rich and well-studied regions of SW Asia and western Europe. The presence of scraper-rich assemblages—with the shifting strategies between Levallois or Quina-like systems that give way to more expedient methods and less reliance on scrapers—is a trend also observed in SW France, for instance. Are there any underlying mechanisms in the observed similarities between these assemblages that are geographically separate? Do these trends transcend local factors responsible for the variations in industries?

Technological fluctuations during the MP did not necessarily involve innovations so much as the recurrence of the primary flaking possibilities, methods that are constituting part of the MP package (White & Pettitt, 1995). A certain chronological dynamism in the shifts in the use of flaking methods is apparent in most parts of Europe (Delagnes & Meignen, 2006; Meignen et al., 2009; Monnier & Missal, 2014; Pettitt, 2003; Rolland, 1977). The task for the Balkans is to correlate lithic variability with faunal, environmental, and climatic studies, and to explore subsistence and mobility to understand the recurrence/reappearance of techno-economic solutions.

Understanding the biogeography of the Balkans is closely related to the question of the refugial character of this region. The available record suggests the peninsula was largely depopulated during MIS 4, rather than being a refugium for human populations during that harsh climatic period. Their late persistence in this region, longer than in other regions in Europe, during the time Homo sapiens entered the continent, cannot be confirmed for the moment. Given the data available thus far, Neanderthals disappeared around 43–42 ka BP, if not even earlier. The influence of climatic conditions can be invoked. Did they make it through the H5 or did this relatively short event cause a demographic vacuum leaving room for the arrival of modern humans? The Paleolithic of the Balkans has an important role to play in these matters. Along the river valleys in the north of the peninsula, there are IUP sites that overlap with the late MP. The same ecological zone was inhabited during the EUP at the time that, or soon after, Neanderthals disappeared from the region. Uluzzian components are present in the late MP in the south of the peninsula and the emergence of the UP seems rather late in this area. The observed patterns could suggest several population replacement scenarios, but a re-examination of the Balkan industries and the new investigations will shed more light on these processes.

This paper aims to offer a comprehensive view of the MP in the Balkans. It hopes to draw attention to the Balkan MP and underline its relevance for the understanding of Neanderthal adaptations in Europe. A region-wide critical disentanglement of major technological and typological trends from spatial and temporal perspectives will hopefully help put the Balkans back on the map of the European MP.