Between the flake and the blade: Associated systems of production at Riparo Tagliente (Veneto, northern Italy)

The Riparo Tagliente site (Verona, Italy) shows three macro phases in which high technological variability can be observed. The aim of this study is to evaluate the specific role of the Middle Paleolithic blade production within this variability. Preliminary results show a complex scenario in which the role of the blade is strictly linked with flake production through mixed reduction systems. Two different approaches were used for analysing the lithic assemblages from the site. The first analysis focused on the identification of the reduction systems by determining the techniques, methods and concepts underlying the entire chaîne opératoire. The second approach concentrated on analysing blade production in order to identify its variability. Evidence of blade technology from the Middle Pleistocene (MIS 8-6) has been found in northern Europe (France, Belgium). Later, during MIS 5 blades can be found over a larger area, this time also including north-western Germany and the central-southern part of France. A third period (MIS 4-3) marks the appearance of laminar production in southern Europe, including in the Italian peninsula. Based on the present state of research these three phases appear to be on-and-off events without clear evolutionary continuity. By repositioning the sequence of Riparo Tagliente within the Italian context we can observe that at the end of the Mousterian period the technological patterns differ greatly, with laminar production being one of its most evident expressions. The origin of this fragmentation is questionable.


Introduction
By simplifying what is produced through lithic production, we can identify three possible categories of products: flakes and blades, both produced by knapping operations (débitage), and shaped tools (hand axes, choppers), the result of shaping operations (façonnage). If shaping operations involve a conceptual modelling structure of a block of raw material, the dichotomy flake-blade is, at the macroscopic scale, a double variant of the same theme, which entails the separation of a piece from its original volume. The Middle Paleolithic marks the emergence and development of a variety of knapping methods aimed at producing predetermined blanks within which the blade occupies a not-yet defined role.
This paper addresses the issue of the technological complexity that characterizes Middle Paleolithic reduction systems and investigates the role of elongated products within the Neanderthal techno-cultural baggage. In addition to Levallois production, the sequence of Riparo Tagliente shows the use of various reduction systems aimed at obtaining a mixture of flake and blade blanks. Because of this a comparison of the morpho-technical characteristics of Levallois and non-Levallois elongated products was carried out.

The blade phenomenon in the Middle Paleolithic
From a global point of view, blade production dates back to the Middle Pleistocene. The first evidence of blade production was found in Africa at two sites, Kathu Pan (Wilkins & Chazan 2012) and Kapturin (Johnson & McBrearty 2010), both approximately 500,000 years old ( Figure 1).
The Amudian complex in the Middle East is the second oldest evidence of blade production and dates back to MIS 9 and MIS 8 (Mercier & Valladas 2003;Barkai et al. 2005).
Subsequently, in a second phase (MIS 7-6), the expansion and differentiation of blade production over a larger area took place, which included the internal part of Syria and the southern area of the Caucasus. This second phase gave rise to several other lithic industries known by various names: the Hummalian (Le Tensorer 2005; Richter et al. 2011), Pre-Aurignacian (Bordes 1977), Hayonim (Meignen 2011), and Djruchula-Koudaro industries (Meignen & Tushabramishvili 2006;2010).
The third and final phase is that of the well-known case of the northern European blade production observed at several sites dating back to MIS 8 and MIS 7 (Révillion 1995).
By contrast, there is no evidence of blade tool production in Asia, at least during the Middle Pleistocene (Boëda et al. 2013;Li & Bodin 2013;Peng et al. 2014). The easternmost assemblages containing volumetric blade technology have been documented at Khonako in Tadjikistan and date back to around 170 ka .
All of these industries have in common the presence of blades, but differ strongly in the rest of their productions (Meignen 1994;. In short, during the Middle Pleistocene at least three blade production epicentres differentiated in space and time can be observed. As far as we know these spatial, chronological and technological differences suggest a convergence phenomenon (Figure 1). (2017)   In short, we can observe how the oldest expressions of the laminar phenomenon occurred within the northern borders for a long time (MIS 8-6) while the southern regions were still dominated by the production of flakes ( Figure 1).

Journal of Lithic Studies
As far as the Italian peninsula is concerned, current studies report the first evidence of blade production in the final phases of the Middle Paleolithic, more specifically in MIS 4 and in the first part of MIS 3 (Figure 1). The geographic distribution of both non-Levallois and Levallois blade production does not appear to be linked to a specific area or environment. In fact these productions can be found all throughout the Italian peninsula.
The only exception to this late appearance in the Italian peninsula is the site of Cave dell'Olio (Fontana et al. 2009;Fontana et al. 2013). This site is, at the present, the only one dating back to MIS 9, representing the only proof of blade production in the Italian Peninsula during the Middle Pleistocene.
While it is now certain that blades were produced during the Middle Paleolithic, the production of bladelets, obtained by means of an independent reduction system, is less evident and occurred just in the final phases of the Mousterian period. Some bladelets production has been noted at the sites of El Castillo and Cueva Morin in northern Spain (Maíllo Fernández 2001;Maíllo-Fernández et al. 2004), at Champ Grand (Slimak & Lucas 2005) and Combe Grenal in France (Faivre 2012), Fumane (Peresani et al. 2013) and Grotta del Cavallo in Italy (Carmignani 2010) and Balver Höhle in Germany (Pastoors & Tafelmaier 2010).
Some geographic areas, such as the Balkans and Greece, and the Iberian Peninsula, do not seem to be influenced by this phenomenon, both during its earliest and more recent phases, completing the fragmentary and irregular overview that emerges from the data in our possession.
Although this absence can be attributed to a lack of research, especially for the Balkan region and Greece, this is certainly not the case for the Iberian Peninsula for which there is a much larger amount of available data.
The Riparo Tagliente site, which is presented in this paper, is part of the last phase of the Middle Paleolithic blade phenomenon and shows an articulated techno-cultural repertoire consisting of mixed flake and blade reduction systems.

The site of Riparo Tagliente
Riparo Tagliente is a rock shelter located in the Veneto region in northern Italy ( Figure  2). It was first excavated in the 1960s by the Museo Civico di Storia Naturale di Verona (Pasa & Mezzena 1964;Zorzi 1962;Zorzi & Mezzena 1963) and subsequently in collaboration with the University of Ferrara (Bartolomei et al. 1982;1984). The Mousterian collection under examination here comes from these excavations. Research at the site is still ongoing currently under the direction of Federica Fontana from the University of Ferrara. Sediment, macrofaunal, microfaunal and pollen analyses date the Mousterian sequence between MIS 4 and the beginning of MIS 3 (Arzarello et al. 2007;Cattani & Renault-Miskovsky 1989;Thun-Hohenstein & Peretto 2005). The stratigraphy, excavated by artificial layer, is composed of a Mousterian sequence and an Epigravettian sequence separated by erosion. The 1960s excavation procedures, which paid much attention to sedimentary details, have enabled us to determine light patterns of internal evolution of the lithic industry. The Mousterian sequences have been found in two different locations known as 'Internal shelter' and 'External shelter' (Figure 3). The Internal shelter comprises 18 layers (52 to 34) and extends over 8 m 2 while the External shelter comprises 13 layers (46 to 34) and a larger surface area (16m 2 ).

Sorting procedure and methodology
In Medieval times the shelter has been used as a refuge. These occupations caused a partial destruction and reshuffle of the deposits on a quite large area both for the Epigravettian layers as well as for the Mousterian's ones.
For these reasons a preliminary check of the material and stratigraphy has been focused on eliminating the squares and the layers considered not reliable. After the check we have considered as being reliable just four squares coming from the Internal shelter (Q 614, 615, 634, 635) and four squares coming from the External shelter (Q 5, 6, 8, 9) (Figure 3). In the same way the layers 34 and 35, have been as well excluded from our analysis because of the presence of contamination coming from the Epigravettian layers. After the sampling, our analysis has been concentrated on the layers going from 52 to 36 on an area of 9 m 2 . We have selected all flakes (complete or broken) bigger than 15 mm. All cores, core fragments, tools, tool fragments and all blades and blade fragments are selected regardless of their size. The distribution of the material across the sequence show different concentration of the material that has been possible to group in three macro phases called Lower layers, Intermediate layers and Upper layers ( Figure 3). Five layers show a high density of stone artefacts (more than 200 pieces). Three layers contain less than 5 pieces and can therefore be considered as sterile ( Figure 3). The lithic products of Riparo Tagliente were analysed using a technological approach. The knapping system analysis follows the same principles as those of the chaîne opératoire analysis, which is supported by the quantitative presentation of technological categories (Inizan et al. 1995). The definition suggested by Boëda (1994) was adopted for the Levallois concept. Given the absence of the refitting reconstruction of the reduction sequences we used the mental refitting method (Pelegrin 1995). The techniques were identified according to experimental studies carried out by Pelegrin (1991;2000). Volumetric and Levallois blade productions were distinguished by means of volumetric structure analyses (Boëda 1990). In terms of the Discoid production, we used the definition put forward by Boëda (1993; as well as also taking in consideration broader criteria (Peresani 1998;Slimak 2003). Diacritical analyses were applied to cores and blanks as a means to reconstruct the chronological order of the scars (Dauvois 1976).

Lithic technology
Our database contains a total of 2315 débitage removals and 75 cores. The raw material used is good quality flint from local sources (<5 km). The flint was collected mainly in secondary position in the form of pebbles and to a smaller extent in primary position as roundish nodules (Arzarello et al. 2007). Production mainly comprised flakes and to a lesser extent blades (Table 1). Hard hammer direct percussion was the only technique used in all the reduction systems. The abundance of cortical flakes proves that the initial stage of knapping activities was carried out at the site (Table 1).
In terms of the knapping products, all the layers show a high degree of homogeneity as shown by the large number of Levallois flakes derived from centripetal and unidirectional methods (Table 1). Generic unidirectional and centripetal flakes are numerous. Unidirectional flakes, the number of which falls in the upper layers, represent the only element of discontinuity across the sequence. Blade production is distributed in similar percentages throughout the sequence and is composed of both Levallois and non-Levallois blades (Table  1). Production also includes convergent, orthogonal, bidirectional and Kombewa flakes that are present in small numbers throughout the sequence. The apparent homogeneity observed when analysing the knapping products will be partially invalidated when we turn our attention to the analysis of the cores.

Lower layer reduction systems
The Lower layers contain 468 lithic pieces of which 25 are cores. As is the case with the end product, the cores indicate that the Levallois is the main reduction system, which is predominantly expressed in the centripetal method and secondarily in the unidirectional method ( Figure 4). The purpose of using the Levallois unidirectional system was to produce mostly flakes. Few blades are associated with this system.
The second most adopted system is based on the exploitation of cortical thick flakes by means of the Kombewa system ( Table 2). The exploitation can be limited to a singular detachment or to a short sequence of detachments ( Figure 5). The preparation of the cores is limited to a partial correction of the lateral convexities of the flaking surface.

Lower layers
Intermediate layers Upper layers t52 t50 t49 t48 total t46 t44 t42 t40 total t36 (total) Levallois centripetal - Two cores show a unidirectional reduction system composed of two different exploitation yet interconnected phases, which we termed Unidirectional Type 1 (Figure 6). The first phase exploits the larger surface of the volume through a short unidirectional sequence and has two complementary functions: to produce quadrangular, slightly elongated flakes and to reduce the thickness of the adjacent surface, which will be exploited by a second unidirectional sequence (second phase). The exploitation of the thinner side of the volume, already reduced in thickness during the first sequence, allows for the production of small  (Figure 6). The configuration of cores is limited to a partial preparation of the lateral convexities carried out by means of a series of orthogonally-oriented detachments with regard to the main flaking direction. An isolated core shows a bidirectional exploitation starting from two opposite striking platforms. The variability of the production systems in this unit is also composed of two Discoid cores and two sub-pyramidal cores. The sub-pyramidal cores are aimed at producing thick convergent flakes ( Figure 5). Four cores follow a reduction system based on the exploitation of orthogonal alternated surfaces that can be associated with a SSDA system (Forestier 1993) or with an opportunistic method, sensu Arzarello (2003).

Intermediate layers reduction systems
The Intermediate layers of Riparo Tagliente show some elements of continuity with the Lower layers such as the persistence of the Discoid and SSDA systems. The centripetal Levallois continues to be the predominant reduction system, however, the plasticity of the Levallois concept finds greater variability here than it does in the Lower layers. The centripetal and unidirectional methods are supported by a bidirectional exploitation while the use of the preferential method, totally absent in the Lower layers, is well represented here (Figure 7).
No Kombewa cores were noted in the Intermediate layers. The presence of Kombewa flakes in these layers could indicate the export of the cores outside the site or they could derive from other flaking operations such as the configuration of a Levallois surface based on the exploitation of the ventral face of a flake. The absence of pyramidal and unidirectional system type 1 methods is a further element of divergence compared to the Lower layers.
In the Intermediate layers the most common production system consists of a unidirectional system which tends to develop around the edge of the core following a semirotating rhythm (Unidirectional core type 2) (Figure 7). There is no or minimal flaking surface preparation. The maintenance of the core convexities is evident in some debordant blades and plunging laminar blanks. The end products consist of elongated thick blanks.

Upper layer reduction systems
The lack of cores roughly sums up the reduction systems in the Upper layers. However, based on the end products, we can see a certain continuity with the Intermediate and Lower layers represented by a large number of Levallois flakes. As for the Intermediate layers, the Levallois concept shows great variability expressed in the convergent, unidirectional and bidirectional methods (Figure 8). The unidirectional semi-rotating system (Unidirectional Type 2) is only observed in one core.

Retouched pieces
Three different categories were established in order to study the retouched pieces. Each of these categories corresponds to the number of transformation degrees undergone by the blanks: low, medium and high degrees (Figure 9).
The low degree describes a marginal retouch of the perimeter of the piece, which does not modify the cutting edge nor the morphology of the pieces in any way. The medium degree consists of a retouch that modifies the morphology of the cutting edges, but not the structure of the piece.
The high degree refers to the structural modification of the blanks, which completely or partially transforms their original morphology.
Transformation through retouching can be noted in all layers. The Lower layers show the highest percentage of transformation while the lowest percentages are observed in the Intermediate and Upper layers ( Table 3).
The retouching phase shows different degrees of transformation in terms of the débitage classes. Besides a few rare exceptions, high levels of transformation are mainly observed in the cortical and generic flakes found in all three layer groups (Table 4). On the other hand, Levallois flakes only show slight modifications just like in the blade production ( Figure 9). (2017)

Blade tools across the sequence
Blade production is similar throughout the sequence with a slight increase in percentages in the Upper layers (Table 5). The blades can be described as being well preserved. Proximal fragments are the most numerous (Table 6).
Within the sequence different production systems can produce elongated blanks, both deriving from the unidirectional and bidirectional Levallois systems as well as from unidirectional non-Levallois systems (Unidirectional Type 1 and 2). Therefore the main aim of the study was to verify whether this variability was due to a predetermined intention to produce differentiated tools by using different reduction systems or whether this was only the result of opportunistic behaviour.
By observing the morphological characteristics of the blades and those of experimental representatives it was possible to distinguish two main blades categories: Levallois blades and non-Levallois blades. The blade fragments which could not be attributed to a specific category and blades with mixed characteristics that could have pertained to any category were placed in a third category termed 'undefined blades'.
The parameters taken into consideration when defining these categories were: types of platform, knapping surface angles, cutting edge angles, transversal cross-section, longitudinal profile, length-width ratio and width-thickness ratio.
Most of the blades fell in the 'undefined blade' category (Table 6). Levallois blades and non-Levallois blades are found in all layers in similar frequencies. From a morphometric point of view there is certain overlapping between the non-Levallois and Levallois productions in as far as the length-width ratio is concerned (Figure 10). (2017)   Conversely a significant difference is evident in their width-thickness ratios ( Figure 11). This difference is also noticeable when we compare the angle of the cutting edges. In the Levallois blades the opening of the angles are concentrated between 10° and 35°, while the non-Levallois blades show wider angles of the cutting edges, ranging between 35° and 55° ( Figure 12). Figure 11. Riparo Tagliente. Levallois and Non-Levallois blade thickness-width ratios.

Journal of Lithic Studies
Both for the Levallois and non-Levallois productions, six techno-functional categories were observed, all based on the morphological structure and the organization of the cutting and non-cutting edges ( Figure 13). Only completely intact blades were analysed; minimally fractured pieces were also excluded.
In general, we can observe how blade production at Riparo Tagliente focused on the production of objects with differentiated techno-functional characteristics rather than the making of a mono-tool ( Figure 14).
By comparing the classes of blade we can see how blades with a peripheral cutting edge (S1 Type) are attributed mainly to Levallois blades. On the contrary, debordant blades (S3, S4 type) are more frequent among the Non-Levallois blades. Convergent blades (P1 type, P2 type) are rare in both categories ( Table 7). The undefined blade category does not show any specific tendency except for the scarce presence of convergent blades, as was the case in the Levallois and non-Levallois blades. (2017)

Discussion and conclusion
Despite the apparent substantial homogeneity of the Riparo Tagliente sequence, some differences can be observed in the reduction systems used. The main characteristics, common to the whole sequence, are the use of the Levallois concept and the production of elongated blanks. Other common features such as the presence, even though sporadic, of the Discoid and SSDA systems are shared by the Lower and Intermediate layers. This homogeneity, which is evident in the end-products, masks the presence of some differences, these mainly visible in the cores.
The greatest variability in the reduction systems used can be observed in the Lower layers ( Figure 15). In the Intermediate and Upper layers, the fall in the number of reduction systems is replaced by an increase of the variability of the Levallois concept, which is expressed by means of the centripetal method as well as the convergent, bidirectional and preferential methods (Figure 15). débitage phases were seldom retouched. This can be linked to the anticipation of the variability of the end products for flakes as well as blades already preconceived in the production systems. This aspect emphasizes the substantial difference with the more standardized blade productions of the Upper Paleolithic where differentiation of tools is usually mostly achieved during the retouching phase. Based on our data, at Riparo Tagliente, Levallois and non-Levallois reduction systems coexisted producing elongated blanks, different in their morphological and technological characteristics as a direct result of the different reduction systems used to obtain them. Both reduction systems are aimed at producing blades and flakes rather than blades in a systematic way. This differentiation in production can be observed in the Levallois unidirectional and bidirectional end products as well as in the unidirectional Type 1 reduction system. By observing the Riparo Tagliente sequence within the context of the Italian peninsula it is therefore possible to make a number of general observations (Figure 16).
The first observation is that the blade phenomenon in the Italian peninsula appeared at some point between MIS 4 and the beginning of MIS 3 and therefore later than in the south of France where blade production is first recorded as early as MIS 5 ( Figure 16). The data from Riparo Tagliente fit will within this framework.
The second observation is that, as far as we know, there is no trace of local nor internal evolution. In fact, blade production seems to appear 'simultaneously' from north to south in the Italian peninsula and is always associated to other types of reduction systems of which the Levallois is the most common. (Figure 16).
As already noted for the rest of Europe, the production of blades did not entail a particular raw material preference. Blades were made from all types of raw materials (flint, chert, limestone, quartzite) and their different forms (pebbles, nodules, slabs, core flakes). Various reduction systems were used in the production of blades. Blades can be produced exclusively by means of a Levallois concept, as in the cases of Grotta di Castelcivita (Gambassini 1997), Riparo del Poggio (Caramia & Gambassini 2006), Barma Grande (Yamada 1997), and Riparo Mochi (Grimaldi & Santaniello 2014;Yamada 2004), or by 'volumetric' reduction systems, as is the case at the sites of Santa Croce (Arrighi et al. 2009), Grotta Reali Peretto 2012), and Grotta del Cavallo (Carmignani 2010). Occasionally the two systems were used together as has been noted at Riparo dell'Oscurusciuto (Villa et al. 2009) and Riparo Tagliente .
In short we can observe how during the MIS 4 and MIS 3 there is widespread production of blades produced by means of original knapping systems or as in the case of the Levallois by a readjustment of this concept oriented towards the production of elongated products.
Given the current state of knowledge there is still much to be learnt concerning the causes of this technological change.
Middle Paleolithic blade productions cannot be considered as monolithic entities. This 'non universal' phenomenon contrasts with other types of production systems such as the Levallois or the Discoid system, with which it coexisted and which contrastingly show a greater geographic diffusion and chronological continuity.
Understanding the role of blade production during Middle Paleolithic requires a systematic approach, which takes into account both the techno-functional aims and the evolution of the reduction systems.