The stratigraphic complex of the layers A5–A6 covers 58 m² at the cave entrance and has been excavated in many different sectors since 1988 and more extensively from 2006 to 2008. It is sandwiched between the Uluzzian layer A4 above and the sterile layer A7 below. This finely layered sedimentary succession made of frost-shattered breccia, Aeolian silt and sands is markedly different in the density and number of anthropogenic signatures (bones, with signs of anthropic modification, hearths, flint flakes) as a consequence of changes in settlement dynamics. A thin, flat charcoal layer (A5) overlies a loose stony layer with loamy fine fraction (A5+A6). Below, a dark layer (A6) is recognizable over the whole excavated zone, with constant dense indications of anthropogenic activity. Combustion structures are plentiful in A6 and there are a smaller number in A5 and A5+A6 (Peresani et al., 2011). The chronometric refinement offered by the ¹⁴C data puts A5 at 43.6–43.2 ky BP and the boundary with the Uluzzian layer A4 above at 43.6–43.0 ky BP (Higham et al., 2009). Red-deer, ungulate and some carnivore bones all bear signs of cultural modification (Peresani et al., 2011).

The lithic industries yielded groups of flint flaked artefacts, with highest frequency in A6 and lowest in A5, with a total of around 6,000 with a module (length + breadth) ≥ 4 cm. Several flint types were exploited, albeit with varying frequency. At a first glance at their texture, structure, colour and the morphology of the exterior surface, the flints were obtained from the Late Jurassic to Middle Eocene carbonatic formations in the western Lessini Mountains. Maiolica and Scaglia Rossa (Cretaceous) have the highest frequencies, but high frequencies are also shown by the Tertiary carbonatic sandstones, Scaglia Variegata (Cretaceous) and its varieties, while there are lower frequencies from oolitic limestone (Jurassic) and other formations of Tertiary age. These lithic assemblages show the variety of raw materials exploited in this region, where provisioning may have occurred at a range of 5–10 km from the site. Flint is also contained in loose coarse stream or fluvial gravels, slope-waste deposits, and soils: for this reason, they became a major resource to be exploited close to the cave. There was also very occasional exploitation of old patinated artefacts collected as potential cores from elsewhere. Indeed, flaking was performed on every type of raw material, regardless of its mechanical properties or the manner in which the flint was introduced into the production-manufacture system.

The artefacts analyzed in the present study amount to a total of 1629 with a module ≥ 4 cm (Table 1). Given to the quantity of lithics recovered during recent fieldwork, this number can be considered a reliable representative sample of such population (Centi, 2008–2009 and Di Taranto, 2009–2010). Studies in progress on the rest of the assemblage confirm the technological variability identified at this first step. They show fresh surfaces over all faces and retouched edges, although natural modifications ranging from the marginal to the invasive of the unmodified and retouched edges affect 6–8% of the total assemblage. The spatial distribution of these alterations, based on the degree of edge modification, becomes denser progressively towards the inner cave.

To reconstruct the reduction sequences, morpho-technical and morpho-metric analyses were carried out on the cores and the completed blanks, which are the most significant by-products (in terms of their technological role), as well as some refitted pieces. As concerns the Levallois method, the conceptual and analytical approaches of the present study have been inspired by the work of E. Boëda (1994) and have also taken in consideration broader criteria for defining Levallois predetermined products (Grimaldi, 1996 and Guette, 2002), previously used in the context of this region (Peresani, 2001 and Peresani, 2012). Technological descriptions of both major and minor lithic production sequences and the key typological features of retouched blanks are given below.

Almost all stages of the Levallois reduction sequence are represented here by the flakes which are mostly the results of the initial shaping of the cores, and the maintenance and re-preparation of Levallois cores (predetermining flakes, core-edge removal flakes, platform renewal, core repairing), designed to eliminate imperfections and accidents (Table 2). Cortical flakes are fairly flat and elongated, resulting from the exploitation of cobbles using the unipolar modality, following parallel planes. Over 200 Levallois end-products reflect the technical aims that were envisaged at initial reduction, and pursued throughout the whole process of exploitation, and also provide good evidence for recognizing options and ways of predetermining shape and size.

Levallois production concerns a main reduction sequence managed through the unipolar recurrent method, but also involves branching out and other levels of variability: one example is the modification to exploit cortical flake-cores and a second case is the procedure to change from unipolar to centripetal towards the final phase of core exploitation. In addition, technological variants and various predetermining devices have been identified within a given modality.

The assemblage also contains evidence of bladelet production, through the exploitation of cortical ridges, as demonstrated by two refitted pieces (Fig. 4), and of four bladelet cores-on-flake.

A core (fig. 4 n. 6) was exploited on the dorsal face of a large, possibly laminar, flake. At least ten bladelet scars originate from a single striking platform made on a truncation at the middle of the blank. Some of these are overshot, some are hinged and parallel: all testify the frontal reduction of the original volume. The distal parts of several other laminar detachments are visible along the left side and testify to an earlier exploitation of the flake. A few final elements are related to the maintenance of the striking platform.

A second core (fig. 4 n. 7) was exploited on two faces (A and B). The first phase saw the preparation of a striking platform and the initial exploitation (face A) of a partially prepared ridge formed by the intersection of the ventral and dorsal faces of the original flake. The scars of three bladelets are visible: the first runs along the crest and the others hinge at the same height. The next phase saw the rejuvenation of the striking platform by means of two detachments and the production of one further bladelet (also hinged) from a brute ridge adjacent to the earlier one (face B). This ridge has an obtuse angle (about 120°) and is formed by the intersection of the dorsal face and the negative scars of a crest shaped on the back.

A third core (fig. 4 n. 8) has one platform, the creation of which involved the removal of the butt from the original flake. A ridge formed by the intersection of the dorsal and ventral faces was used to remove a bladelet, followed by a series of hinged detachments. The sequence ends with an attempt to widen the flaking face. Tiny flake removals are related to the maintenance of the distal convexity of the core face.

The last core is on a cortical flake that has been shaped into a transverse scraper. The butt was removed to create a striking platform. As in the preceding cases there was also exploitation of an unprepared edge made from the intersection between the ventral and dorsal faces. Here, two bladelets were produced. The core is of similar size to the cases described above.

The bladelets have flat (five), punctiform (four), and faceted (two) butts, with marked bulbs ascribable to direct percussion with a stone hammer. The sagittal profile is straight or slightly convex and its section ranges from triangular to polygonal. Edges show straight and regular outlines and remain unretouched. The elongation index as calculated on eight pieces is 2.8.

Evidence of discoid production is provided by cores and flakes relating to various stages of the reduction sequence: cortex removal/initial shaping (four), centripetal (seven) or tangential flake removal (core-edge removal flakes and pseudo-Levallois points (six)), change in core orientation (one axial crest flake). Some refitting provides evidence of in situ reduction. These flaking products are, for the most part, whole with fresh edges, considerable thickness (average 1.3 cm), quadrangular form and slight elongation. The butts are almost always flat and tilted. The percussion technique was direct with a hard hammer. Two flakes, one cortical and one centripetal, were respectively transformed into a denticulate and a double scraper. The cores are centripetal and heavily reduced: four out of five are below 3.5 cm in length.

The assemblage also contains cores which have been modified to provide a series of edges, used for recurrent production of small multidirectional and bidirectional flakes, alternating between two faces or with a short series from a single face. A hard hammer was used for percussion. The identification of the products from this procedure was difficult due to their rarity and the small size of the flakes.

Retouched tools are mostly scrapers, while the others are points (four), notches (one), denticulates (seven) and a few retouched flakes. Simple scrapers (49) prevail over double (six), convergent (six) and transverse types (five). Generally, tools have been mainly shaped on Levallois blanks and of these, more on unipolar than centripetal flakes (Table 3). Other blanks are cortical and, incidentally, various by-products arising from the Levallois core exploitation: predetermining, core repairing, platform renewal or flakes from accidents. Finally, the blank type of a very few scrapers and denticulates has not been determined due to the invasiveness of the retouching.

Scrapers have been made from different types of flakes: simple scrapers using cortical flakes, mainly from the initial decortication, and from unipolar recurrent flakes; a small number are recorded on naturally backed predetermining flakes and, occasionally, on various blanks detached in core maintenance, Kombewa-type flakes, but also flakes from accidents. Similar to simple scrapers, cortical and Levallois flakes were mainly used as blanks for convergent and transverse tools. The retouch is simple, moderately invasive, with less marginal or strongly invasive interventions. Four scrapers were thinned on the dorsal or ventral face. Concerning the other tools, it has been shown that Levallois semi-cortical blades and flakes were used for making pointed items; notches and denticulates were made on quite thick, even cortical flakes, but also on recycled broken flakes, discoidal flakes and various by-products; two pieces are thinned on the ventral face; the thinning of unretouched blanks involved cortical, Levallois, Kombewa-type and platform trimming flakes.

Contrary to these trends, the few tools with marginal, partial or discontinuous retouching have been shaped on second choice blanks such as predetermining flakes, occasionally cortical flakes and two recycled pieces.

The lithic industry found in the A5–A6 complex provides a large data set to allow inferences to be made about the final Levallois assemblage and other types of products. The aim of extracting elongated Levallois convergent blanks with symmetrical and highly functional lateral edges is also highlighted by the incidence of retouching, which is higher among these flakes – and even on those from the initial decortication/production phase – than for others. The use of the unipolar method towards the end of the reduction sequence shows that these technical aims were constantly pursued, in accordance with a procedure that had been well-integrated into the system of production. This might explain why in the course of some sequences there was technical variation, which involved reduction through a series of blows in oblique and, at times, perpendicular directions relative to the former series. The aim of this method was clearly to reduce the impact caused by expensive re-preparations of core convexities. The turning to the centripetal method when the reduction sequence ended, but before core discard, implies minor reshaping of the surfaces to control the volume of the raw material at the same time as extracting more predetermined blanks, despite being less morphologically normalized than unipolar blanks. This production decision was influenced by two factors: the first being a change in the criterion of the main aim of yielding blades, not shorter than a given length. The second relates to optimizing resources by possibly changing production technique because of the qualitative deficiencies of some flint types, thereby reducing the risk of accidents and facilitating re-preparations.

Making a comparison of the Levallois technology between A5–A6 and the chronologically closest layers at Fumane, there is a greater similarity with A11, A10V and A10I: unipolar reduction focuses on blades with more constrained morpho-metric calibres, although differences have been observed. The modality in A11 combines bidirectional and orthogonal patterns until core deactivation, whereas in A5–A6, there is lateral expansion of the trimmed striking platforms for the striking of convergent blade series. Again, in the final step of the reduction sequence this “widened” unipolar pattern is replaced by the centripetal procedure, up to final deactivation (Peresani, 2012).

This Late Mousterian sequence at Fumane broadly fits other evidence in the North Adriatic region where Final Mousterian practice is based on the Levallois recurrent unipolar modality (Peresani, 1996). At the final steps of the reduction sequence, the unipolar modality may have been replaced by the centripetal to extract the last, small flakes. Tools are scrapers and points largely made from the Levallois blades. Similar situations can be seen across the Italian peninsula, where blades and elongated flakes, flat and with thin converging margins, have been noted at various sites, such as Castelvicita in Campania (Gambassini, 1997) and Riparo l’Oscurusciuto layer 1 (43.8–42.2 ky BP), where the laminar Levallois production seems to be employed in the making of scrapers (Boscato et al., 2011).

Blade and bladelet volumetric production is little represented in respect to the Levallois at Oscurusciuto, but at Grotta del Cavallo, a lamellar industry is noted in the sectors FIII–II, developed through the production of local flint plaquettes (Carmignani, 2010). Other lamellar production evidence has been, at times, identified to date at Riparo Tagliente, where a specific flint type was used in blade production from level 37 upwards (Arzarello and Peretto, 2005); at San Francesco on the Ligurian coast (Bietti and Negrino, 2007); at Grotta Breuil on the Latium Coast; and at Grotta Reali, again in the South of Italy (Peretto, 2012). There is no chronometric evidence for the lithic assemblage found in a single layer at San Francesco, which contains many Upper Palaeolithic tools like truncated and retouched blades made on blades produced by the exploitation of three different types of cores: prismatic, recurrent Levallois and prismatic with lateral crest (Tavoso, 1988). Although the assemblage from the most recent layers at Grotta Breuil shares a group of features with other local Mousterian assemblages, it shows that, there, the laminar flaking method used unipolar cores made from small pebbles with no relation to the appearance of Upper Palaeolithic tools (Rossetti and Zanzi, 1990). At Grotta Reali, the Mousterian sequence yields evidence of blade production in layer 5, based on the recurrent Levallois method (mostly unipolar) and on the detachment of blades/bladelets from prismatic unipolar cores. In the light of this variability, the features and the variety of the retouched tools do not differ from a typical Mousterian profile comprised of scrapers and denticulates. This assemblage dates between 44.5 and 39.4 ky BP (Peretto, 2012), falling in a wide temporal interval which sees the Final Mousterian replaced by the Uluzzian and the Proto-Aurignacian in the South of the Peninsula (Moroni et al., in press). More data are also required to set the chronological position of the assemblage found in the second stratigraphic complex (layers 2abc and 2β/2γ), where blade production comparable to layer 5 is also recorded (Peretto, 2012). The hypothesis claiming cultural persistence at Grotta Reali needs further chronometric confirmation for two reasons: the radiocarbon age of layer 2γ (40.3–37.2 ky BP), which brackets a time range containing the start of the Aurignacian in the South of the Peninsula (at least from 40.4 ky BP) and the lack of Campanian Ignimbrite across the sedimentary succession (Giaccio et al., 2008).

Returning to Fumane, a break with this apparently deep-rooted use of the unipolar Levallois method is recorded in the oldest Uluzzian layer A4 where, rather, flakes and cores were made using the centripetal instead of the unipolar recurrent modality. Levallois blades and laminar flakes are therefore sporadic and polygonal or fan-shaped flakes of variable sizes feature in the lithic assemblage, in addition to other flakes issuing from other poorly curated methods, the incidence of which increases at the top of the Uluzzian sequence. As well as side-scrapers and points, the retouched tools include backed knives, splintered pieces and one end-scraper. There are very few denticulates and marginally retouched points on Levallois flake. The backed knives are on thin flakes, which can also be cortical, with the back either straight or convex. The frequency of splintered pieces in A3 is double that in A4. The shapes vary, as do the thicknesses (Peresani, 2012).

During the Final Mousterian, in the North of Italy, the Levallois method continues to enjoy a role of primary importance, although based on the standardization of laminar production with careful control of the technical parameters. Blades spread as a general phenomenon in the Levallois techno-complexes in the Old World and in many cases are associated with Upper Paleolithic formal tools. In fact, the effort devoted to producing elongated flakes/blades appears as a feature that is shared across many European regions in the interval between 50 and 40 ky BP. In southern Caucasus, in Georgia, unipolar Levallois reductions mark levels 10 to 5 dated between 50 and 39 ky BP at the site of Ortvadle Klde (Adler et al., 2006) and produce suitable blanks for making simple or convergent scrapers. In the Crimea the Micoquian was replaced by the western Crimean Mousterian, whose industries are concentrated on laminar products and elongated points, obtained through Levallois and volumetric methods and transformed into simple scrapers or into unifacial foliated points (Chabai, 2000). In the Middle Danube area, a similar discontinuity marks the replacement of Micoquian with the techno-complexes with blades and points like the Bohunician (Škrdla, 2003) and the Bachokirian (Kozłowski, 1979 and Teyssandier, 2006), founded on the use of uni- and bipolar Levallois and blade volumetric methods. In the South-East of France, the production of elongated flakes and normed points using the same system was noted in the Neronian, where scholars have identified a pattern of blade/point and another of bladelet/micropoint (Slimak, 2007) that do not find a point of comparison in the context of the classical Middle Paleolithic industries.

Nevertheless, the lack at Fumane of the co-presence of Levallois technology and Upper Paleolithic tools (as seen in this Europe-wide spectrum of industries), clearly marks out the complex A5–A6 and highlights the replacement with Uluzzian technology. This replacement might be considered abrupt, similar to that at the other aforementioned Italian sites, but the complementary use of other methods like the bladelet making seen at Fumane as well as at Grotta del Cavallo and Grotta Reali, seems to record a deviation from the homogeneity typically expressed in the Mousterian Levallois industries of this period.