Fumane Cave (Lessini Mountains, Veneto) records the MP–UP transition with a finely layered sedimentary succession preserving several Late Mousterian levels, covered by Uluzzian and Proto-Aurignacian levels (Broglio et al., 2006, Higham et al., 2009, Peresani, 2012 and Peresani et al., 2011a). The Latest Mousterian is recorded in a stratigraphic complex named A5–A6 composed of layers A5, A5 + A6, and A6. The bone scraper was found in layer A5 + A6 in the rear of the cave (square 120c), associated with dispersed charcoal fragments, flaked stones, bones, and few hearths. The chronometric refinement of the ¹⁴C data sets layers A5 and A5 + A6 to 42.3–39.8 ky ¹⁴C BP. The bone scraper has been subjected to taphonomic, techno and use-wear analyses, the results of which have been compared to the experimental data. Taxonomical and anatomical determinations were made after comparison with the faunal collection stored at the University of Ferrara (Laboratory of Archaeozoology and Taphonomy). Taphonomic alterations have been analysed at low and high magnification. Measurements of the specimen included length, width and thickness. For the interpretation of the traces we have referred to the available scientific literature (Blasco et al., 2008, Fischer, 1995, Olsen, 1989, Shipman and Rose, 1988 and Villa and d’Errico, 2001). The functional analysis has directly been carried out on the specimen, without the use of resin replicas. The appearance, development, location, and distribution of manufacturing and use-wear traces have been observed with a stereoscope Nikon-SMZ-10 (magnification 0.75–70 ×) and a metallographic microscope Nikon Eclipse ME 600 (magnification 10–200 ×). The interpretation of technological and functional modifications has been based on both criteria defined in the scientific literature (Christidou, 1999, Legrand, 2007 and Maigrot, 1997) as well as on the experimental comparison.

A total of 3096 remains of mammal species were identified among the 171,257 bone fragments recovered in layer A5 + A6. The number of identified bones is low due to the high rate of bone fragmentation (more than 88.8% are smaller than 2 cm) and combustion (47.8%). These conditions are due to a set of depositional and post-depositional processes also noted in other fully anthropic contexts (Morin, 2010). The number of bones smoothed by different natural mechanisms, like trampling, rolling, and microscraping is very low (5.4%). Root marks are the most common alterations identified on bones (26.8%), followed by surface microfissures from weathering and/or weathering cracks (20.4%), manganese stains (17.6%) and exfoliations (6.4%). Some concretionary fragments (17.6%) were concentrated near the external area at the limit of the drip-line, while the few with trampling abrasions (0.5%–total remains 919) in the east part of the cave are associated with the highest stoniness of the layer. There is high incidence of burned remains that could suggest the use of butchering refuse, rich in fats, as fuel for fire. Amongst the burned fragments, 88.0% were subject to moderate combustion (200–500 °C, black/brown), while 12% were calcined (> 700 °C, grey/white). At a general level, these patterns and phenomena are similar to evidence from the Mousterian unit A9 with Discoid technology, and to other anthropogenic deposits in the region (Romandini et al., 2014).

The macromammal assemblage is composed mainly of red deer and roe deer, elk, ibex and chamois, and giant deer have also been recognized. Fox, wolf and bears are the most numerous carnivores. Avifaunal bones from different species also bear traces of human modifications (Peresani et al., 2011b). With the exception of the wolf, carnivore remains show skinning and defleshing butchering marks. This data, together with the scarcity (No. 71) of carnivore gnawing on bones as well as a consistent sample of anthropic marks, suggest the anthropic nature of bone accumulation in this layer (Peresani et al., 2011a). Intense intentional breakage of undetermined bones is also testified by many percussion cones (No. 1951) and notches as well as by impact scars (n^o 466). Most of the ungulates yielded anthropic traces, especially red and roe deer. When considering the whole ensemble of faunal remains, butchering traces have been identified on 1442 remains, determined at the level of the species (butchering rate is the 47% of the total NISP) and 1954 determined bones (1.2% of the total undetermined). Tibias, femura and metapodials, radii and humeri from Cervids (red deer, roe deer, and very large specimens of giant deer or elk), and at lesser extent from other ungulates fragmented from extracting marrow were used as retouchers (total number A5 + A6 layer: 48; total number from the A5–A6 complex: 131). While some of the surfaces were scraped before being used (probably to remove residues of meat and tendon or the periosteum), no further preliminary preparation is documented (Jéquier et al., 2012). Punctiform and linear impressions, striations and pits usually characterize functional areas on retouchers. Up to three different functional zones can be identified on each retoucher, although the majority of the pieces show a single functional zone. Interestingly, 28 out of the 516 fragmented shafts > 5 cm of length sampled in that study (Jéquier et al., 2012) were recycled as retouchers, whereas only one was turned into the scraper described in this paper.

As for the 280 deer fragments recovered in layer A5 + A6 (Fig. 1A), coordinates have been recorded for ungulates remains, which are almost totally concentrated on the east side of the investigated area, close to the NE stone wall. In the remaining part of the investigated area, red deer bones are present and become sporadic in the external area to the SE (Fig. 1B). In A5 + A6, bigger faunal remains (1823 remains ≥ 3 cm of length) are concentrated with a frequency of 10–28 remains per square (Fig. 1C). Most of the carnivore remains with anthropic traces are located at the entrance to the tunnels B and C. The spatial distribution of bones with impact traces (notches, percussion marks and cones) shows concentrations of intentionally modified remains and percussion cones on the eastern side, along the inner part until tunnel C (Fig. 1D–E).

Experimental retouched edges have been shaped on eight flakes produced for recovering of marrow as a primary goal, using direct percussion on 6 fresh radii extracted from 3 red deer carcasses previously defleshed by vultures. The bones were handled and placed on a base during percussion (Fig. 2a1–a6). The location of the impacts affected the morphology of the resulting flakes. Namely, spiral fractures longitudinal to the long axis of the diaphysis can be considered the best way to open the bone and extract the marrow. These fractures were produced by striking on the medial-lateral part of the diaphysis, close to the epiphyses. Diversely, transversal fractures created by impacting the postero-dorsal side of the bone are not functional for marrow extraction. Normally, only one blank was morphologically suitable for producing a notched edge through marginal direct percussion. In very few cases, up to three or four notches have been obtained.

Fat and periosteum were scraped off the bones in order to make technological work easier. This activity produced typical striations, visible on the negatives even after direct percussion. Stone hammers and flint blocks of different size, weight (1.5–0.5 kg) and texture were used for retouching (Fig. 2a1–a6). On the thickest portions of the edge, generally in their medial or lateral sides, retouch was carried out easily with an edge close to 60°. This task was facilitated by holding the bone against a base (Fig. 2a2–a6). In four cases, while retouching the bone blank through alternate retouch, the diaphysis broke after 4–5 retouches were produced.

Experimental scrapers were very effective on different materials (fresh and dry hide, fresh and dry wood, fresh bone with meat), and use-wear traces started to develop after 5 minutes. They were held with one or both hands, and both the dorsal and the ventral edges of the scrapers were successfully used in unidirectional, convergent-divergent, or alternate movements. Wood and bone scraping produced intense rounding much quicker than hide or meat processing (Table 1; Fig. 2 and Fig. 4a–b). While diagnostic differences can be identified in the invasiveness and the development of rounding on the edge according to the different materials processed and the specific motions performed with the experimental tools (e.g., scraping hide vs. scraping wood), all the activities carried out have developed modifications oriented and distributed according to the movement performed with the scraper (Table 1; Fig. 3a–f). Differences in the invasiveness of the use-wear traces have been observed according to the high or low working angle, while very flat rounding developed while working wood. Different use-wear traces have always developed along the used edge and never on the medullar ridges of the bones, as those were not in contact with the worked material during the activity.

On the basis of the experimental results, we suggest that the portion of diaphysis exploited for the manufacture of RF1395 was deliberately shaped by direct percussion on the anterior and posterior surfaces, successively. The breakage planes suggest that the bone was retouched when still fresh. On the dorsal face of the bone, it is possible to observe a sequence of five detachments obtained using the 60° natural angle of the radius, and its central mesial edge is characterised by a slightly narrower angle of 55° due to the absence of the trabecular frame along the medullar channel. Notches and detachments on the anterior face are thin, moderately invasive, and disposed in sequence (Fig. 4A). On the exploitable edge, a “pseudo” percussion cone is still visible (Fig. 4A/#), whereas on the medullar face (Fig. 4B) four invasive notches overlap the previous opposed ones, removing their impact points. In one case, the notches on both faces share a common impact point, which is still visible along the bone edge (Fig. 4C). The suite of gestures reveals the exploitation of the two opposed faces, hierarchically related from the medial diaphysis angle. While retouches on bone shafts can be produced by carnivore activity, the latter are generally smaller and isolated and always associated with punctures pits and score-pits (Blumenshine, 1995, Domínguez-Rodrigo and Piqueras, 2003 and Fischer, 1995).

At high magnification, the surfaces of the bone scraper RF1395 are characterized by various post-depositional alterations, such as traces of dissolution caused by roots, tiny carbonate concretions, dehydration cracks, and soil striations. On the ventral surface, detachments can be observed along the right edge and along dehydration cracks. Post-depositional edge rounding is also developed along the natural and retouched edges of the tool as well as on the ventral surface and along the edges of the marrow cavity (Fig. 3g). These post-depositional modifications also characterize unmodified bone fragments coming from the layer A5 + A6 of Fumane Cave. On the experimental tools, rounding is always associated with striated areas; their distribution, orientation and characteristics are diagnostic of the different movements performed and materials worked with the bone scraper during the activity (e.g., fresh and dry wood, fresh and dry hide, meat and bone). While, on RF1395, developed rounding is never associated with densely striated areas at a macroscopic level rounded, striated spots with tiny depressions oriented according to the striations have been identified at high magnification (Fig. 3m–n). These micro use-wear traces are compatible with modifications produced on experimental tools after processing dry wood for a short time. On this basis, we cannot exclude that post-depositional modifications affecting the surfaces of RF1395 might have altered the eventual and more developed traces produced after using the bone scraper on wood.

Fumane represents a Neanderthal referential site during MIS3, due to the bulk of evidence like lithic and bone remains related to multiple occupational events. At the site, the use of bone splinters for retouching lithic artefacts is well known, and bones from hunted animals represented an alternative raw material, which Neanderthals expediently exploited.

Our experimental activity as well as the results of the technological and taphonomic analyses, supports the interpretation of the retouched bone RF1395 from Fumane as a scraper. In particular, the technological analysis has pointed out that the medial portion of the radius was selected and retouched through an array of gestures carried out on the two opposite faces of the bone. This part of the diaphysis was preferred for its thickness and for the narrow exterior curvature of the bone. The interpretation of the deliberated modified edge has been based on morphological and technological indicators, such as shape and location of the retouches, their uniform distribution, the presence of percussion bulbs, and the regularity of the edge profile. The appearance of the bifacial, planar, continuous, and overlapping retouches along the edge of RF1395 differs from most typical taphonomic traces (e.g, carnivore gnawing, rodents activity, trampling, and sediment pressure) and from the anthropic modifications produced after green bone breakage for marrow recovery (Backwell and d’Errico, 2001, Blasco et al., 2008, Brain, 1981, Domínguez-Rodrigo et al., 2009, Fischer, 1995, Galán et al., 2009, Olsen, 1989, Shipman and Rose, 1988 and Villa and d’Errico, 2001). Rather, modifications on RF1395 can be compared with the continuous unifacial or bifacial retouches made on thin dihedral edges of the midshafts from Gran Dolina (Rosell et al., 2011), with the invasive and bifacial detachments regularly positioned on the side of some Acheulean bone artefacts (Biddittu and Segre, 1982 and Segre Naldini et al., 2009) and clustered in a distinct pattern (Anzidei et al., 1989), on the bone fragment from Vaufrey Cave (Vincent, 1988), but also with the continuous scaled removals observed on the bone from Blombos Cave (d’Errico and Henshilwood, 2007). Nevertheless, Fumane scaper is different from the denticulated artefacts and the pointed tool from Abric Romaní (Carbonell et al., 1996), as well as and the trihedral pick made on rhinoceros tibia from Gruta Nova de Columbeira (Barandiarán et al., 1971), which marks out a variability in the way these flakes were modified. Our experimental results have also demonstrated that the angle and length of the edge of the bone scraper was potentially functional for cutting or scraping various materials (e.g., hide and wood). Although post-depositional agents partially affected the way use-wear traces preserved on the tool, a brief utilization of the scraper for woodworking cannot be excluded on the basis of the microscopic analysis and experimental comparison.

In order to better characterize the technological aspects related to the production of the bone scraper, the possibility that the retouch represented an outcome of the production of bone retouchers has also been considered, as this was documented at the site Axlor, level D (Spain) (Mozota Holgueras, 2009). While the technological analysis of 130 bone retouchers carried out at this site show that in six cases tool blanks were prepared by percussion on the lateral sides and/or on the proximal parts, none of the 48 bone retouchers coming from levels A5 and A5 + A6 and 83 from level A6 of Fumane were extracted by percussion (Jéquier et al., 2012). Additionally, the detachments along the sides as well as on the proximal extremity of bone retoucher blanks are always more isolated than the ones characterizing the retouches on RF1395.

As a bone retouched tool, RF1395 is unique in the A5-A6 stratigraphic complex as well as across the whole Late Mousterian sequence at Fumane. The technological characteristics of this tool clearly indicate that during the Middle Palaeolithic at Fumane, the exploitation of bone as raw material involved the use of direct percussion through a transfer of technical knowledge from flint knapping. While gestures are comparable to the operational sequences involved in flint knapping, this does not necessarily imply that this technique was used in the absence of a more effective one. Also, while side-scrapers are the most abundant category of retouched lithic tools, no morpho-types similar to the bone scraper are documented in the lithic industry at the site. In various Lower Palaeolithic contexts the use of bone as raw material for tool manufacturing has been related to the lack of appropriate lithic raw material in the surrounding area (Anzidei, 2001, Dobosi, 2001 and Gaudzinski et al., 2005). Yet, archaeological evidence for bone modification by retouch is absent during the whole Middle Palaeolithic at Fumane. Consequently, we cannot exclude that the use of bone scrapers might have represented the result of sporadic events although the direct percussion represents the primary technique used in the manufacture of stone artefacts at this site. In conclusion, our results support the idea that the technological performances of Neanderthals involving the use of bone as raw material were variable but still at nascent level.