The state of preservation of the four retouchers from level VII at Labeko Koba is generally good. Although nearly all the objects are affected by different taphonomic agents, some of which affect all the faunal remains at the site (Villaluenga, 2013), the surface alterations do not prevent their technical analysis. In only one retoucher (LK.15G.247.230), the use area is partially covered with calcite. The surface alterations affecting the retouchers consist of erosion and rounding of the external bone cortical tissue, due to particularly important hydraulic processes in certain levels in the deposit, which experienced intense leaching (Arrizabalaga, 2000a: 66). Two of the objects also display dehydration fissures. These types of alterations are hydrothermal (climatic or weathering, depending on the terminology of different scholars), the consequence of repeated desiccation and rehydration processes (Behrensmeyer, 1978 and Brugal, 1994). They may be the result either of the exposure of the bones to climatic agents (sun, wind, cold, rain…) before entering the deposit or their removal from the sedimentary environment and later exposure to changing temperature and humidity conditions.

Of the four retouchers in level VII, one has been classified taxonomically as medium size ungulate tibia shaft (LK.11D.242.142). A second tool was anatomically described as a femur shaft from the medial side. However, taxonomical identification was not totally accurate, and probably this fragment belongs to a red deer (C. elaphus) (LK.11 C.297.4). Third, retoucher (LK.15G.230.47) is characterised by its size (62 × 32 × 7 mm), feting with a large ungulate limb bone shaft morphology. Even though, the relevant thickness (7 mm) of the internal side does not reach the medullar cavity or porous surface. Finally, the fourth implement (LK. 9D. 225.401) does not display any particular anatomical characteristics and has been assigned to the small-sized group (Table 1).

The cross-section of the objects is convex-concave except for one that is semi-rectangular. They are between 62 and 100 mm long, while their width is very regular, between 23 and 34 mm, which gives the impression that the bone fragments were selected from limb bone shaft fragments produced during butchery activities (see Section 6.1) (Table 1. Fig. 2). They are between 7 and 13 mm thick. Some of the objects exhibit the traces of percussion to break the bone. This is the case of the tibia fragment of Capra pyrenaica (LK.11D.242.142). It was fractured by direct percussion while the bone was still fresh as shown by the helical fracture plane and the point of impact at one end of the object (Lyman, 1994). Later, the shape of the tool was modified slightly by retouching, also by direct percussion, especially at the end of the ventral and lateral face. The use areas are always on the cortical (external) face of the bone. They consist on scores nearly always with V-shaped cross-sections that may be symmetrical or more often asymmetrical and of variable depth, from very fine notches to others that cut deeply into the bone (Fig. 3). The lack of symmetry in the incision planes indicates the direction in which the tool penetrated in the bone matter (abrupt plane) and the extraction of the matter that this caused (extraction plane) as it came out (Malerba and Giacobini, 2002 and Mallye et al., 2012). Microscopic observation systematically appreciates a series of fine striations that start on the edge of the extraction plane and correspond, as has been shown experimentally, to the slippage of the lithic tool when it comes out of the bone matter, in this way indicating the use of kinematics. The same traces have been described on Aurignacian retouchers at sites like Isturitz, La Ferrassie and Castanet (France) (Castel et al., 2003, Schwab, 2002 and Tartar, 2009). Inside of the scores, flint particles still conserved, visible with a microscope, have been described at other sites, proving that these bone tools were used with lithic materials (e.g., Abrams et al., 2014 and Mallye et al., 2012).

The use-wear marks are arranged in more or less oval areas. Three retouchers have two areas on their surface while the fourth has only one. After end butchery process, including bone breakage with marrow extraction purposes, no preparation before the use of the objects can be observed. The size of the use areas is very homogeneous, both among the different retouchers and on those retouchers with two use areas. They are between 24 and 13 mm long and 14 and 9 mm wide (Table 1). The double areas are always at opposite ends of the objects and the distances between them range from 42 to 32 mm. The distance from the use areas to the ends of the retouchers varies from 2 to 23 mm. The latter parameter shows that there is no generalised laterality of the use areas as recognised at other Aurignacian sites like La Ferrassie (Castel et al., 2003). At Labeko Koba, in contrast, some use areas are found near to one of the edges of the object whereas others are more or less central. Except for one of the two use areas on one object, whose scores are vertical (parallel to the long axis of the object) (LK.15G.230.47), the other six use areas display horizontal scores (perpendicular or slightly oblique to the long axis). The direction of the scores is directly related to the direction of the object when in use, as its long axis may be held perpendicular to the lithic edge (horizontal scores) or parallel to it (vertical scores) (Schwab, 2002, Schwab, 2014 and Vincent, 1993).

The use intensity of the retouchers is one of the most often discussed aspects of these objects. It can be assessed by measuring two parameters: the size of the use area and the penetration in the matter caused by their use, according to the experimental data offered by some researchers (Mallye et al., 2012). It has been seen that the use areas on these retouchers are relatively homogeneous in size, and this does not depend on the size of the objects, as the largest retoucher (100 × 34 × 10 mm) has a smaller use area than some smaller retouchers from this level (e.g., LK.9D.225.401). At the same time, the formation of only one of the six use areas has involved the modification of the natural bone surface, causing a relatively large pitted area. However, the length of the implement would have allowed its use to continue, by starting a new use area at the opposite end, but this did not happen. The sizes indicate that the retouchers were used with a similar intensity, and this was, in any case, exhaustive. A use fracture affected the active zone in only one retoucher (LK.15G.230.47) but did not stop the object being used as shown by the fracture plane that cuts the use-wear marks. In the other retouchers, no use fracture affected the development of the use areas.

Like the objects from the Proto-Aurignacian level, most of the retouchers in the Early Aurignacian levels show different degrees of alteration of the bone. These levels were also affected by water action to a greater or lesser extent (Arrizabalaga, 2000a: 66). This has caused the erosion of the outer cortical tissue of the bones, but very few of the use areas of the retouchers have suffered damage. Some of the objects also display solution pits (Villaluenga, 2013). This type of erosion is related to biochemical processes (the solution of elements in the bone tissue by the circulation of acidic solutions), related to low intensity water action (the objects were exposed to water or conditions of high humidity at the time of their deposit or after becoming buried) (Laafar, 1994). However, the alterations have only affected the interpretation of four of the retouchers, where calcite and/or erosion on the external surface of the bones have affected the use areas (the objects LK.15H.164.88; LK.13 J.188.174; LK.13I.143.37; and LK.13F.197.243). Taphonomic alterations have not prevented the measurements of the use areas. Most of the objects are complete, just in three cases postdepositional fractures have been identified (LK.9D.146.170; LK.13G.203.129 and LK.9D.180.293), and the last one also revealed recent post-depositional fracture (excavation or storage accident).

None of the four retouchers found in level VI could be taxonomically identified. However, two of them have been classified as a medium-sized ungulate, and in each case limb bone shaft fragments were used as support. Fourth tool (LK. 9D.180.293; 62 × 32 × 13 mm) would have been made on a large animal bone fragment. Twelve retouchers have been identified in level V, it has been possible to ascribe two of them to a species and anatomical element, LK.15H.173.122 and LK. 9D.146.170 correspond to C. elaphus metatarsal and humerus shafts. Exceptionally, a third retoucher/pressure-flaker (LK.15F. 204. 98) was made on a R. pyrenaica horn fragment. The other seven objects could not be identified, although according to their size and shape, three of them should produce on large ungulates limb bone shafts, including two tibia fragments. Finally, the last tool was made on a medium-sized ungulate large shaft fragment (LK.13F.197.243; 96 × 29 × 9 mm). Out of 16 retouchers in level IV, a single tool (LK. 13 J.140.27) could be taxonomically classified as a medial tibia of a large bovid (B. primigenius or B. priscus). The other fifteen pieces were classified according to their size. Eight of them were made on medium-size ungulate bone fragments, including three humerus or femur fragments. A single piece was made on a large animal shaft, and it was not possible to determinate the support of the remaining eight implements. In two cases due to the small size and other two cases, bone surface modification caused by dissolution did not allow an accurate identification (Villaluenga, 2013).

The objects are always formed by between a half and a quarter of the total circumference of the bone shaft. They are, therefore, convex-concave in cross-section, while some of them tend towards rectangular (Fig. 4). The lengths of the retouchers vary from 41 to 177 mm; the widths from 16 to 63 mm, and the thicknesses from 6 to 26 mm. The use areas are on the external bone surface of all the retouchers. The marks are grouped in areas with an oval shape; 23 of the objects possess a single use area, whereas the other six display two areas. In the latter cases, the active areas are at opposite ends of the objects (proximal and distal areas). The size of the use areas ranges from 8 to 25 mm in length and from 4 to 28 mm in width (Fig. 5). The maximum distance between the areas and the nearest end of the object is 18 mm, while the minimum distance is 2 mm. The retouchers with double use areas display a separation between the areas varying from 12 to 42 mm. Except for the area on one object with vertical impressions (parallel to the long axis of the bone), in the other 34 use areas, the traces associated with the use of the bone are horizontal (perpendicular or slightly oblique to the long axis). A total of 15 of the 35 active zones were prepared by scraping before the objects were used. In these cases, the scraping was parallel to the long axis of the bone. The time when the scraping was carried out, which was always before the retouchers were used, is shown by the superimposition of the use marks, as the functional scores are always above the scrape marks (Fig. 6 and Fig. 7). The initial observation of the size and shape of the scraped areas, which are always associated with the active zones and only seen beneath the use marks, leaves no doubts about their technical nature. The measurement of the size of the scraped surfaces, about the areas that they contain, supports this observation. All the use areas occupy between 20 and 40% of the area previously scraped. This result means the idea that the scraping was related exclusively to processing the meat on the bones, as proposed for retouchers from some Mousterian sites (Daujeard et al., 2014) can be discarded. The scraping did not occur before the preliminary fracturing of the bone, which is logical as the point of impact is not seen in this part of the bone shaft. In contrast, the effect of the scraping on the cross-section of the bones is inappreciable. Therefore, its purpose was not to regularise the surface or prepare a working plane but was probably to eliminate remains of the periosteum and/or fat or meat, which might affect the functionality of the retoucher. The size of the areas, which is always small in comparison with the available surface areas of the bone fragments, and the faintness of the use marks show that these retouchers were not used until they became exhausted. Five of them present a moderate impact on the active surface (with a slight modification of the cross-section of the bone) and only one retoucher exhibits pits in the active areas indicating an intense use involving a change to the curvature of the external surface (Mallye et al., 2012). This object was also used as an intermediary piece. In the other bones with traces, the functional, technical action did not change the external shape of the bone. The four retouchers with fractures on their edges affecting the development of the use areas were abandoned without using the opposite end of the bone fragment, which was still functional.

Three variables have been analysed to discriminate possible operational patterns in procuring bone fragments exclusively intended to be used as retouchers, from those related to the fracturing of bones for nutritional purposes. First is the correspondence between bone shafts used as retouchers, in comparison to most frequent anatomical element and taxa abundance per level (Table 6 and Table 7). Second considers bone retouchers supports fracture pattern (Fig. 8) against limb bone shafts. Finally, third assesses the size and shape of the retouchers in relation to the rest of the bone fragments. Nearly all supports were ungulate limb bone shafts, specific areas with a concave profile in their external face were selected (where the use areas are located). This morphology is common to all studied levels (VII, VI, V and IV) and should be in connection with the functionality of the objects. On this point, some authors have suggested that they were used to re-sharpen lithic tools (Mallye et al., 2012 and Patou-Mathis et al., 2002). In the case of Labeko Koba in the following point are considered the possible ways in which these objects were used. A comparative analysis of retouchers taxonomical attribution and taxa abundance in each level at Labeko Koba (see Table 6) shows that bone supports belong to most frequently hunted animals (bovids, C. elaphus and E. ferus). In consequence, in VI, V and IV levels, bovid bones were chosen. However, for level VII, there is no correspondence between the most common taxa (E. ferus), and the fragments chosen as blanks for the retouchers (bovid bones). Nevertheless, the large amount of remains brought by carnivores (Crocuta crocuta spelaea) and the remains of Ursus spelaeus, (Villaluenga, 2013) have completely altered the taxonomical composition of this level, which hinders its direct comparison with the other three assemblages (levels VI, V and IV) of exclusively anthropogenic origin. At the same time, in the upper levels (VI, V and IV) a clear correlation is seen between the animals hunted and the source of the retouchers, as bovids became increasingly important in the faunal assemblage (42.99% NISP in level IV and 71.25% NISP in level IV). In general terms, bones from large and medium size ungulates are used more than those from small ungulates (see Table 2, Table 3, Table 4 and Table 5). Most of the objects display marks caused by breaking the bones by direct percussion on the fresh bone, with helical fracture planes and impact marks (Lyman, 1994). These kinds of fractures are identical to those seen on bone fragments not selected to be used as retouchers (remains of consumed animals) (Villaluenga, 2013). Therefore, at Labeko Koba, as at most sites with retouchers, both Aurignacian and Mousterian (e.g., Giacobini and Malerba, 1996, Mallye et al., 2012, Schwab, 2002 and Vincent, 1993), there does not appear to be a specific fracturing of bones to make retouchers. Citations to a particular debitage of long bones to make objects of a certain size that were later used as retouchers have only been made for the Mousterian levels at the site of Axlor (Spain) and Scladina Cave (Belgium) (Abrams et al., 2014 and Mozota, 2009).

The shape of selected bone fragments (see above) is elongated, however, elongation index reveals a high variation (since level IV, L/W = 0.45 to level V, L/W = 5.2). Considering the mean ratio of bovid bones (the largest group, 103 shaft fragments) elongation index decreases slightly, from a ratio of 2.75 in level VII to 2.35 in level IV. This change is thought to indicate a more intense use of the bone remains. A comparison between shaft and retouchers length (L/W index) in VII, VI and V levels, shows how shaft fragments are longer than the retouchers. Nevertheless, opposite pattern appears in level IV. This level yielded the largest number of retouchers (NR = 17) and limb bone diaphysis fragments (shafts) with anthropogenic modifications (NISP = 111), in this set of remains, implements (retouchers) are slightly more elongated than shafts (mean L/W = 2.69 and 2.56, respectively) (Fig. 8). This inversion in L/W index the length of the fragments and tools can be related to the techno-functional change in use of bone raw material at Labeko Koba (Fig. 8). In early Aurignacian chrono-cultural contexts (VI, V and IV), bones were used as fuel. As an example, only in the later stage of early Aurignacian (level IV) have been identified 1671 burnt fragments. Furthermore, the length of 1560 fragments is below 3 cm. This behaviour consisted of the selection of diaphysis and epiphysis fragments and systematic fragmentation to expose the highest quantity of bone tissue to fire (Villaluenga, 2013, Villaluenga et al., 2012 and Yravedra et al., 2005) (Table 6 and Table 7). This practice, also observed at other sites, both Mousterian and Upper Palaeolithic in age, and characterised experimentally (e.g., Costamagno et al., 2005, Théry-Parisot et al., 2004 and Villa et al., 2002) at Labeko Koba might reflect technical behaviour related to an increasingly cold climate (Iriarte-Chiapuso, 2000). Otherwise, preservation of certain number of shafts to be, some of them, used as bone retouchers indicate a particular behaviour among EUP humans.

The site of Labeko Koba was excavated in the course of a salvage operation, but the most careful methodological procedures were observed, and all the sediment was sieved with water, which means the results obtained in the study of the lithic assemblage can be regarded as representative. The deposit contained a large number of remains of lithic reduction (over 22,000 pieces, of which nearly 1500 are tools) in its different levels. The levels with the largest number of objects are levels V and IV (Early Aurignacian) and VII (Proto-Aurignacian), in that order. However, the massive presence of Dufour backed bladelets in the latter level means that it is the one with the most tools. Flint is by far the most common raw material in all the levels, and its percentage only reduces slightly (85%) in the Chatelperronian layer (lower level IX). The main sources for the flint were in the Ebro valley (the Urbasa and Treviño Outcrops, both of them about 30 km far away from Labeko Koba), on the other side of the Atlantic/Mediterranean watershed from the site of Labeko Koba (Tarriño, 2000).

From its start, the Labeko Koba lithic assemblage is dominated by a high proportion of blades. Indeed, the percentage of blades decreases from the oldest levels (over 72% of the whole objects) to the most recent one (over 41%). Regarding fragments of blades and bladelets, the evolution is similar. Contradictory with this inverted leptolithisation process, linear and punctiform butts, which are associated with microblade lithic reduction, display an evolution in accordance with what might be expected and steadily become more common (Fig. 9). The levels in Labeko Koba contained very different numbers of lithic tools, and only three of them (VII, V and IV) have yielded enough for a statistically significant analysis. However, the evolution in modes of retouching is very consistent, at the site itself and on a more regional scale. Thus, the abrupt mode, basically on backed bladelets, is very common in Labeko Koba Proto-Aurignacian level (VII) at the expense of the simple mode. From level VI to the top of the sequence (all of them associated with Early Aurignacian), the simple mode increases whereas the abrupt mode is gradually replaced (by the burin mode in level VI, as well as by the simple mode) (Fig. 9). This evolution runs parallel to the classic characterisation of the Early Aurignacian, the abundance of endscrapers and blades or bladelets with simple retouching. In our opinion, the link between the most common elements in the osseous assemblage (retouchers on long bone shaft fragments) and lithic retouch is directly related to the typological characteristics of lithic assemblage.

The comparison of the data obtained by this study will allow any differences and similarities between the Proto-Aurignacian and Early Aurignacian retouchers at Labeko Koba to be identified (Table 8). The first detail is the number of objects found in each of the periods represented at the site. As described above, the study of the lithic and osseous assemblages reveals changes in the intensity of occupation through different phases, including a more intensive occupation in the Proto-Aurignacian than in the Early Aurignacian. This assumption is proportionally inverse to the number of retouchers in each level. Indeed, the number of retouchers in level VII (Proto-Aurignacian) is identical to its number in level VI, the horizon with lowest occupational intensity (level VI, NR = 4). This cannot be explained regarding a mere preference for the use of this kind of tool in the Early Aurignacian, as numerically large assemblages are known since Middle Palaeolithic. It seems more likely that the unequal proportions of retouchers at Labeko Koba in the different levels reflect certain variability in the modes of lithic technology. The varying retouch mode (the abrupt mode in the Proto-Aurignacian and the burin mode in Early Aurignacian) may, therefore, contribute to explain the different numbers of retouchers in each level. Taxonomical and anatomical choice of bones to be used as retouchers also varied significantly between the Proto-Aurignacian and Early Aurignacian levels. In the latter (levels VI, V and IV) a strict correlation exists between the species hunted and consumed (bovids, C. elaphus and E. ferus), middle size ungulates and selected anatomical elements (limb bones with a preference for the humerus and femur). In contrast, Proto-Aurignacian groups preferentially selected large size ungulate limb bone fragments, the horse is most abundant species in the faunal assemblage (Table 6). However, it should be borne in mind that the small number of objects and the possible taphonomic bias caused by the impact of carnivores in level VII might influence this view (see Section 6.1).

Regarding the size and shape of the retouchers, variation is seen in their size, although in this case it is not diachronically linear. Thus, the size of shaft fragments is larger than the fragments selected for use as retouchers in levels VII, VI and V. However, in level IV, the retouchers are larger than the general shaft fragments. The inversion in the trend does not take place in the change from the Proto-Aurignacian to the Early Aurignacian occupations, but instead is identified in the upper part of the sequence. This apparent change in the pattern of choice of bones may be related to paleoenvironmental conditions. Palynological data confirm that level VII formed in more temperate conditions than the upper part of the sequence, gradual worsening in the climate (stadial taxa and the low proportion of arboreal species) (Iriarte-Chiapuso, 2000). As a consequence during late occupations (levels V and IV) bone remains were smashed to be used as fuel, with a single exception, shafts used as bone retouchers support (Villaluenga et al., 2012 and Yravedra et al., 2005). The last difference between Proto-Aurignacian and Early Aurignacian retouchers in Labeko Koba is seen to be in the work surfaces conditioning. None of the use areas on the Proto-Aurignacian retouchers displays any traces of the preparation of the active area, may be explained by the size of the sample (NR: 4). In contrast, out of the 35 use areas on the Early Aurignacian retouchers, 15 were prepared by scraping (use marks are superimposed to scraping marks) (Fig. 7). It should be noted that these differences between the Proto-Aurignacian and Early Aurignacian retouchers at Labeko Koba do not correspond to increasing complexity in the form of this kind of object. From a strictly technological point of view, retouchers in each techno-complex display no conceptual differences, as neither do the Mousterian retouchers (Daujeard et al., 2014). In all cases, they are tools whose volume is not shaped specifically, except Mousterian levels at Axlor (Mozota, 2009) and Scladina Cave (Abrams et al., 2014). Tasks in which retouchers were used required minimal technical requirements (size, shape, weight…), moreover abundance of shafts produced during butchery process produced a great availability of appropriate supports, made an exclusively technical fragmentation unnecessary. Finally, morpho-technical characteristics manifest a correspondence between Proto-Aurignacian and Early Aurignacian objects. Traces characteristics and orientation, perpendicular or slightly oblique, to the long axis of the bone, show no significant differences in kinematics, intensity and duration of the activities (Fig. 9).