Introduction

Studies of projectile impact marks (PIMs; after O’Driscoll and Thompson 2014) on bones still play a minor role in zooarcheological analysis. Recent research has identified punctures and the occasional presence of embedded lithic fragments as relevant features for recognizing archeological hunting injuries (Bachechi et al. 1997; Boëda et al. 1999; Bratlund 1991, 1996; Churchill et al. 2009; Fernández-Crespo 2016, 2017; Fischer 1985; Gaudzinski-Windheuser 2016; Leduc 2012; Milo 1998; Morel 1998; Münzel and Conard 2004; Nikolskiy and Pitulko 2013; Noe-Nygaard 1974; O’Driscoll and Thompson 2018; Péquart et al. 1937; Pitulko et al. 2016; Pöllath et al. 2017; Rust 1943), but the design of a systematic protocol for the study of this rare archeological material is just beginning. Besides, the commonly accepted assumption that PIMs would be barely recognized (Castel 2008; Letourneux and Pétillon 2008; Noe-Nygaard 1974; Smith et al. 2007), along with the limitation of experimental tests and the lack of uniformity in terminology and methodology, has hindered the identification of PIMs from becoming a standard in zooarcheological analysis. New impulse in this direction has recently come from several experimental studies focused on projectile traumas (Castel 2008; Letourneux and Pétillon 2008; O’Driscoll and Thompson 2014; Pétillon and Letourneux 2003, 2008; Pétillon et al. 2011; Smith et al. 2007; Wood and Fitzhugh 2018; Yeshurun and Yaroshevich 2014) and from the application of quantitative methods in bone surface modification analyses (Arriaza et al. 2017; Bello and Soligo 2008; Bello et al. 2009; Boschin and Crezzini 2012; Courtenay et al. 2017; Maté-González et al., 2015, 2017; Moretti et al. 2015; Pante et al. 2017; Val et al. 2017; Wallduck and Bello 2018; Yravedra et al. 2017). Nevertheless, only few studies have been specifically designed to determine differences between marks produced by projectile impacts and those generated by taphonomic effects (O’Driscoll and Thompson 2014, 2018), while most of them focused on cut mark characterization through the application of high-resolution 3D quantitative methods (Bello and Soligo 2008; Boschin and Crezzini 2012; Courtenay et al. 2017; Maté-González et al. 2015, 2017; Oxilia et al. 2017; Pante et al. 2017; Wallduck and Bello 2018; Yravedra et al. 2017). These methods were also recently applied to different bone lesions as percussion damages related to the extraction of bone marrow and knapping marks due to shaping and retouching of lithic artifacts (Abrams et al. 2014; Bello et al. 2013; Van Kolfschoten et al. 2015).

In a previous study, we explored the potential of 3D microscopy in distinguishing PIMs from other taphonomic marks, developing a diagnostic framework based on experimental data (Duches et al. 2016). This widely applicable method is based on the collection of reproducible morphometrical parameters that can be processed through statistical analysis, allowing for a reliable distinction of PIMs from cut marks and carnivore tooth marks. This work provided not only specific evidence for the identification of hunting traumas caused by Late Epigravettian projectile implements but also comparative data for taphonomic analyses of bones from other Late Glacial contexts (Duches et al. 2016). The present article aims to continue this research applying 3D morphometric analysis to zooarcheological bone surfaces, in order to test the following: (a) confirm the validity and feasibility of the method; (b) evaluate the reliability of the experimental record; (c) verify the bearing of post-depositional alterations in PIMs identification; and (d) estimate the preservation of PIMs qualitative features on archeological remains.

Material and methods

Late Pleistocene adaptations in the Alps

The peopling of northeastern Italy after the Last Glacial Maximum was a gradual process marked by the progressive colonization of new territories that had been previously abandoned, following changes of vegetation and animal distribution (Naudinot et al. 2014; Ravazzi et al. 2007). Late Epigravettian penetration, firstly limited to valley floors and high plateaus around 500 m asl (Aimar et al. 1992; Peresani et al. 2011), reaches mid-altitude territories during the second part of the Late Glacial interstadial with the full development of a logistical occupation network at the limit between coniferous woods and alpine prairies (Bertola et al. 2007). This organization corresponds to a seasonal mobility strategy that sometimes included sites that are functionally complementary to each other and located at different altitudes (Bertola et al. 2007). Technological analysis of the lithic industries from the most relevant sites has highlighted a progressive simplification of the lithic production system throughout the Late Glacial interstadial (Montoya 2008; Montoya and Peresani 2005) associated with the persistence of standardized lithic backed tools used as projectile implements (Duches et al. 2017). The existence of specific retouched forms, characterized by distinct metric and morphological parameters, reveals as many mental templates developed by the Epigravettian craftsmen in response to specific functional requirements and constraints (Duches et al. 2017). These constraints fit within the conception of the bow-arrow delivery system, whose generalized use in the Epipaleolithic complexes of Europe has been hypothesized by many scholars (Caspar and De Bie 1996; Cattelain 1997; Pelegrin 2000; Plisson 2005; Valentin 2008). Functional analyses carried out on Late Epigravettian industries of northeastern Italy (Lemorini and Rossetti 2004; Lemorini et al. 2006; Peresani et al. 2011; Ziggiotti 2006, 2008) confirmed the exclusive use of lithic backed tools as elements in hunting projectiles.

Cornafessa rock shelter

The Cornafessa rock shelter, located at about 1240 m asl on the Monti Lessini Plateau (Trentino), is a Late Epigravettian site of recent discovery (Fig. 1). It opens in a wide overhanging Late Jurassic limestone wall (Oolite di San Vigilio; Avanzini and Masetti 2007), about 28 m long and projecting up to 7 m. The Paleolithic settlements at Cornafessa rock shelter were discovered in 2014 after a test-pit; an archeological excavation was carried out in 2015 by MUSE-Science Museum of Trento. This first excavation, involving a total area of 6 m2, has established the existence of a layered filling deposit characterized by at least one anthropic unit, placed at a depth of about 2 m from ground level. The layer, dated to the Younger Dryas (charcoal-LTL15976A: 10,080 ± 70 BP, 10,050–9350 cal BC at 95.4%; charcoal-LTL15977A: 10,171 ± 70 BP, 10,200–9450 cal BC at 95.4%; dates modeled in OxCal v.3.10 using IntCal13 calibration curve; Reimer et al. 2013), has yielded a Late Epigravettian lithic industry and 827 faunal remains in a good state of preservation. The lithic assemblage consists of 130 artifacts ≥ 1 cm, belonging to different phases of the reduction sequence. At present, there are no retouched tools, except for few undetermined fragments of backed points and bladelets. The archeological investigation are still in progress.

Fig. 1
figure 1

Location and picture of Cornafessa rock shelter

Full size image

Analytical methods

Taxonomic and skeletal identifications were based on the complete alpine fauna reference collection of MUSE-Science Museum of Trento. Taphonomic analyses were carried out using a stereomicroscope (Leica M 165C with magnification from 0.75× to 125×), a scanning electron microscope (ZEISS EVO 40 XVP) and a coupled energy dispersive X-ray detector (EDX) at MUSE-Science Museum of Trento. The SEM was operated in variable pressure mode (chamber pressure ~ 14 Pa), enabling backscattered electron images to be obtained at magnifications between 70× and 500× without the application of a conducting layer on the specimen. EDX microanalysis was carried out to determine the elemental composition of surface inclusions, using an Oxford Instrument x-act PentaFET precision detector and Aztec 3.1 SP1 software.

In order to identify the nature of surface alterations and to distinguish different taphonomic agents, we referred to well-established literature (Binford 1981; Blumenschine and Selvaggio 1988; Capaldo and Blumenschine 1994; Domínguez-Rodrigo et al. 2009; Domínguez-Rodrigo and Piqueras 2003; Fernandez-Jalvo and Andrews 2016; Fisher 1995; Outram 2002; Pickering and Egeland 2006; Potts and Shipman 1981; Selvaggio and Wilder 2001; Shipman 1981; Shipman and Rose 1984; Villa and Mahieu 1991). The descriptive criteria and the terminology used for PIMs, follows O’Driscoll and Thompson (2014) as already discussed in Duches et al. (2016).

The 3D analysis of bone surfaces was carried out using a Hirox Digital Microscope KH-7700, with MXG-10C body, OL-140II lens and AD-10S Directional Lighting Adapter (Arrighi et al. 2016; Boschin and Crezzini 2012; Crezzini et al. 2014; Moretti et al. 2015) at the University of Siena. Metrical parameters of drags were recorded according to criteria already published (Duches et al. 2016, Fig. 4), using one cross-section per mark in its median part (Fig. 2). In addition, the ratio between breadth at the top (BT) and depth of cut (DC) was calculated (RTD). Measures and profiles taken on archeological specimens were compared to drags produced experimentally by Late Epigravettian projectiles on mouflon carcasses (Duches et al. 2016). Of the total 35 drags available, only 7 could be analyzed through 3D microscopy due to the statistical need of recording all dimensional parameters. The marks were indeed frequently incomplete because of either the total removal of a bone edge or the absence of a continuous floor in the case of spongy or trabecular bones.

Fig. 2
figure 2

Measurements taken on drag marks and cut marks profiles. DC, depth of cut; BT, breadth at the top of the cut; BF, breadth at the floor of the cut; GD/SD, greatest and smallest distances from the middle of the floor to the edges; OA, opening angle

Full size image

Besides, the values were compared to experimental cut marks (n = 93), of which 27 were inflicted on cattle autopodials (22 using an unretouched flake, 5 using a retouched implement) (Boschin and Crezzini 2012) and 66 were produced with unretouched flakes during skinning experiments on cat carcasses (Crezzini et al. 2014). All bone modifications considered on this paper (including the samples used for comparisons) have been directly scanned by two of the authors (FB and JC), who also collected all the measurements processed in the various statistical tests (Table S1). Measurements were collected on the cross-sections using the software directly installed in the digital microscope and then processed using the PAST software (Hammer et al. 2001).

A rib of Ursus arctos from Cornafessa rock shelter and four modern U. arctos ribs belonging to different age classes were analyzed by microfocus X-ray computed tomography (μCT) at the Multidisciplinary Laboratory of the “Abdus Salam” International Centre of Theoretical Physics (Trieste, Italy), using a system specifically designed for the study of paleontological and archeological materials (Tuniz et al. 2013). The μCT acquisitions of the specimens were carried out by using a sealed X-ray source (Hamamatsu L8121-03) at a voltage of 110 kV, a current of 90 μA, and with a focal spot size of 5 μm. The X-ray beam was filtered by a 1-mm-thick aluminum absorber. A set of 1440 projections of the sample were recorded over a total scan angle of 360° by a flat panel detector (Hamamatsu C7942SK-25) with an exposure time/projection of 1.5 s. The resulting μCT slices were reconstructed using the commercial software DigiXCT (DIGISENS) in 32-bit format and an isotropic voxel size of 18 μm. Using the Avizo v.8 software (Visualization Sciences Group Inc.), a semi-automatic threshold-based segmentation was carried out in order to separate the bone component from the interstitial air between the trabeculae (Bouxsein et al. 2010; Coleman and Colbert 2007; Fajardo et al. 2002).

Experimental data and 3D analysis in PIM identification

As already mentioned, a previous ballistic experimentation provided comparative data for hunting lesions identification, with specific clues for the distinction of trauma caused by Late Epigravettian projectiles on medium-sized ungulates (Duches et al. 2016). In particular, 160 arrows armed with lithic backed points and backed (bi)truncated bladelets have been shot against five complete carcasses in a realistic simulation of game hunting, generating a total of 70 hunting injuries on bones. Each lesion has been classified by following O’Driscoll and Thompson (2014) and analyzed, when possible, through 3D digital microscopy with the acquisition of multiple morphometrical parameters useful for statistical processing and comparison (Duches et al. 2016).

As a main methodological outcome, drag and puncture marks have been recognized as the only diagnostic categories in PIMs identification. Crucial for their interpretation are the following elements: (a) the location of the mark; (b) the presence of flaking and/or cracking as secondary features of both categories; (c) the occurrence of specific morphometrical values, different for drags and punctures (detailed below); and (d) the presence of stone embedded characterized by diagnostic impact fractures.

Drag marks were the most frequent experimental trauma (50%) and, consequently, the most likely preserved on zooarcheological material. This category was recorded mainly on ribs (40%), vertebrae (34.3%), and on long bones such as the radius and ulna (14.3%). On the most resistant bones the drag profile is extremely sharp and clean, with frequent unilateral flaking (48.6%) and cracking (34.3%) (Fig. 3). In the case of a total removal of the bone edge, drags are characterized by a polygonal shape (Duches et al. 2016, Fig. 7). The 3D analysis compared drags to carnivore tooth marks and cut marks (produced both by unmodified flakes and retouched implements) using the parameters listed in the analytic method section (see also Fig. 2). Drags turned out to be significantly different from other taphonomic marks, especially considering the breadth at the cut floor (considerably wider than cut marks while comparable to that of tooth marks), the depth of cut (considerably higher than other taphonomic marks), the RTD index (significantly different suggesting a more U-shaped profile than other taphonomic marks), and the opening angle (narrower and more standardized in drags, ranging from 65.8° to 94.5°) (raw data and statistical analysis in Duches et al. 2016).

Fig. 3
figure 3

Experimental drag marks on mouflon bones made by Late Epigravettian lithic projectiles. a and b: stereomicroscope image, 3D rendering and cross-section of drags on two specimen of radius; c: stereomicroscope image, 3D rendering and cross-section of a drag on rib

Full size image

Puncture marks were the second most frequent experimental bone injury (24.3%) and were recorded on the entire skeleton. In the case of the complete perforation of the bone, the projectile left an imprint characterized by a more or less regular polygonal outline. Cracking occurs in the 48.1% of the punctures. The 3D analysis compared punctures to carnivore pits and corrosion cavities considering the following parameters: opening area (A), volume (V), depth of cavity (D), and the ratio between the opening area and volume of pits (RAV) (Duches et al. 2016). Puncture marks resulted significantly deeper than carnivore tooth pits and characterized by a statistically lower RAV (Duches et al. 2016, Fig. 14). Morphological features such the contour of the opening area and the shapes of slopes and floor led to easily distinguish impact punctures from corrosion cavities.

Besides, it is noteworthy that some peculiar features of both drags (drag with sharp edges and polygonal shape) and punctures (polygonal outline) turned out to be strictly related to Late Epigravettian projectiles, confirming the hypothesis of a significant relationship between projectile design and PIMs morphometrical features (Duches et al. 2016).

Thus, the outcomes of 3D digital analysis offer new clues in PIM identification and distinction from other taphonomic marks, qualifying drags and punctures as reliable archeological indicators of prey capture modalities.

Results

Cornafessa rock shelter bone assemblage

The faunal assemblage analyzed includes 827 bone fragments, with a total weight of 108.7 g (Table 1). Only 18 fragments could be taxonomically classified (Table 2). The low NISP value is due to the high rates of bone breakage (more than 96.5% are smaller than 2 cm) and thermal alteration (25%). These conditions are consistent with a set of syn- and post-depositional processes observed in bone assemblages from other Late Epigravettian contexts of northeastern Italy (for a synthesis, see Bertola et al. 2007; Fiore and Tagliacozzo 2005). The remains altered by mechanisms of mobilizations, like trampling, are few (0.2%). The most common cause of degradation is root etchings (2.9%), with a variety of patterns and dimensions. Some fragments show manganese staining (0.7%) and concretions (0.4%). No traces of carnivore and rodent tooth marks are recorded on bone surfaces. At present, the small sample examined does not allow for relevant paleo-ecological considerations, except for the coexistence in the same site of faunas typical of rocky habitats as ibex and species featuring lower elevations (grasslands, wetlands, and forests) as red deer and bear. The scarcity of the sample and the fair preservation of bone surfaces, corresponding to weathering stage 1 or 2 according to Behrensmeyer (1978), allowed us to recognize few stages of carcass exploitation. Defleshing and periosteum removal are well represented by scraping marks along the surface of unidentified medium and large mammals’ diaphysis altered by fire (n = 6; 0.7%; Table 2); conversely, no butchering cut marks or percussion marks have been identified on this small zooarcheological assemblage.

Table 1 General anatomy and taphonomic details of bone remains from Cornafessa rock shelter
Full size table
Table 2 Overview of the Cornafessa rock shelter bone assemblage
Full size table

Analysis of the projectile impact mark

Except for the scrapings, the only anthropic mark identified is a notch located on a rib of U. arctos, which exhibits some morphological traits consistent with a PIM (Fig. 4). The bone is fragmented and corresponds to a rib probably in the range between the 4th–8th: the body (corpus costæ) with the external costal groove (sulcus costæ), the cranial and the caudal edges have been mostly preserved, as has the neck (collum costæ) and the tuberosity of the muscoli longissimi; conversely, the tubercle (tuberculum costæ) and the head (caput costæ) are very deteriorated.

Fig. 4
figure 4

Ursus arctos rib from Cornafessa rock shelter with taphonomic features highlighted. Green lines: roots marks; red lines: scraping marks; blue lines: trampling marks; black arrow: PIM. a: PIM and root etchings; b: scraping marks

Full size image

Morphometric features are consistent with both a juvenile bear of regular dimensions (1–5 years old; 80–100 kg) and an unusually small adult bear (about 80 kg), as suggested by a reference bear skeleton from Trentino stored in the historical zoological collections at the MUSE-Science Museum of Trento (record code: MSN 105533). The μCT analysis of the Cornafessa U. arctos rib has allowed to detail its age-class. As already reported by Boschin et al. (2015), this method can indeed provide bone micro-structural information useful to estimate the age at death of animal remains. The microstructure of juvenile and young U. arctos ribs, compared to that of adult samples, shows a more porous cortical portion and thicker trabeculae (Fig. 5). Such condition is probably related to the increase in bone volume fraction in growing bones, before the architectural adaptation to external loading during life (Tanck et al. 2001). According to its bone microstructure (Fig. 5e), the Cornafessa specimen is probably a juvenile individual.

Fig. 5
figure 5

Longitudinal and transversal cross-sections of five bear ribs. a: 2 months old individual; b: young individual not fully grown; c: adult individual; d: adult individual of small body size; E: Cornafessa specimen

Full size image

The surface of the specimen was altered by root etching, which produced easily recognizable grooves (Fig. 4a) (Lyman 1994). Rare manganese staining and trampling marks with the characteristic U-shape are also present (Domínguez-Rodrigo et al. 2009). Longitudinal scraping marks in lateral and medial portions of the rib (Fig. 4b) suggest that the bone was partially worked on and then abandoned. The scrapings are in fact fully consistent with the traces found on other Late Epigravettian bone tools (Cristiani 2008; Gurioli 2008). Moreover, the only bone point yielded from Cornafessa rock shelter shows very similar scrapings in all the worked surfaces (Fig. 14).

The PIM, located in the middle of the rib’s external surface, consists of a drag with two small flint fragments embedded. The mark is about 6 mm long and presents a wider breadth in its initial part. Although unilateral flaking is visible around the projectile’s entrance point, the drag profile is extremely sharp and clean for its entire length (Fig. 6a, b). Cracking is only observable around the stone embedded at the top of the mark (Fig. 6c). Though this one was visible at low magnification, the existence of a second fragment fixed within the drag’s floor and covered by sediment was revealed only by μCT (Fig. 7). These two fragments are perfectly aligned to the drag’s orientation and parallel to each other, whereas they show different depths. Unfortunately, it was not possible to classify any of them either in terms of tip portion or fracture typology.

Fig. 6
figure 6

Stereomicroscope images of Cornafessa projectile impact mark. a Evidence of straight trajectory of the lesion. b Wider breadth of the drag in its initial part and flaking of one side of the mark. c Evidence of directional microstriations on the drag’s floor

Full size image
Fig. 7
figure 7

μCT virtual renderings of Cornafessa rib showing the stone fragments still embedded in the bone (a, b) and metallographic microscope images of the lithic fragments using bright field illumination (c, d)

Full size image

SEM analysis showed several other small lithic fragments embedded in the drag’s floor and let us evaluate the conservation of PIM qualitative features (Fig. 8). As clearly visible in Fig. 8, the projectile’s impact dug a trapezoidal groove, causing flaking on one side and leaving directional microstriations on the opposite wall. EDX spectra from four selected areas allow to characterize the lithic inclusions and distinguish them clearly from the embedding bone tissues (Fig. 9). The mark has been documented from different orientations because of the uneven topography of the drag and the consequent problem in the collection of elemental data from shielded areas. The large silicon and oxygen peaks in Fig. 9, spectra 1 to 3 (S1–S3), indicate the presence of silicon dioxide (SiO2), confirming the visual attribution of the fragments to the Lower Cretaceous Maiolica flint lithotype (Bertola 2016; Bosellini et al. 1978; Weissert 1981). Analyses of the surrounding matrix (Fig. 9f) showed the presence of calcium and phosphorus peaks typical of bone. Smaller peaks representing carbon and aluminum are probably associated with environmental contamination from soil and handling. Furthermore, EDX maps of elemental distribution distinctly localize stone embedded in areas marked by high silicon levels and isolate them from the surrounding calcium-rich matrix (bone) (Fig. 9c, d, g, h).

Fig. 8
figure 8

SEM secondary electron images of Cornafessa projectile impact mark

Full size image
Fig. 9
figure 9

a, e SEM secondary electron images of Cornafessa projectile impact mark with the location of the four areas analyzed by energy dispersive X-ray spectroscopy (EDX; S1–4). The mark was documented from different orientations because of the uneven topography of the drag and the consequent problem in the collection of elemental data from shielded areas. b, f EDX spectra corresponding to the four areas analyzed; the horizontal axis of the spectrum is the X-ray energy and the vertical axis is the total number of X-rays counted. The peak at zero keV is created by the instrument electronics as a reference. c, g BSE images and combined elemental maps shown as false color: phosphorus in light blue, silicon in blue, oxygen in green and calcium in orange. d, h Discrete EDX maps of elemental distribution

Full size image

The measurements taken on the median cross-section of the drag are: depth of cut (DC) = 478 μm, breadth at the floor (BF) = 112 μm, breadth of the top (BT) = 556 μm, opening angle (OA) = 34.6°, greatest and smallest distances from the middle of the floor to the edges (respectively GD and SD) = 309 and 247 μm. The morphology of this drag resembles that of experimental PIMs (Duches et al. 2016): it is deeper than all observed cut marks and has a narrower opening angle (Figs. 10 and 11). Moreover, the floor of the drag is wide. The value of the RTD index (ratio between BT and DC) lies within the variability of experimental drags and outside that of cut marks. A principal components analysis was performed on measurements from the whole sample (raw data in Electronic Supplementary Material) considering the specimen from Cornafessa separately from experimental PIMs (Fig. 12). All the measurements are considered in this analysis: DC, BT, BF, GD, SD, and OA. The resulting outcome counts for 89.1% of sample variability. PC1, which counts for the 71.2% of variability, is mainly related to the size of marks: the greater the values are, the wider and deeper the marks, but the narrower the opening angle. Regarding PC1, drags are significantly different from cut marks (Mann-Whitney U test: U = 2; p = 1.35E−05). PC2, which counts for the 17.9% of variability, is related to the shape of marks: the greater the values are, the more V-shaped and the more shallower the cross-sections and the wider the opening angle. The Cornafessa drag is clearly far from the variability of cut marks while it is close to (although still outside) the convex hull of experimental PIMs. The position of Cornafessa specimen in the PCA scatter plot is mainly influenced by its depth and narrow opening angle, confirming data presented in the boxplots (Fig. 12).

Fig. 10
figure 10

3D rendering and median cross-section of the Cornafessa drag

Full size image
Fig. 11
figure 11

Distribution of values of three parameters measured on the analyzed sample. RTD, ratio between the breadth at the top and the depth; DC, depth of cut; OA, opening angle; PIMs: n = 7; experimental cut marks made by unretouched flakes: n = 88; experimental cut marks made by retouched implements: n = 5; CF, Cornafessa; PIMs, projectile impact marks; UF, unretouched flakes; RI, retouched implements

Full size image
Fig. 12
figure 12

PCA performed on the sample. Components 1 and 2 count respectively for 86.1 and 9.8% of the sample variability. Blues squares: cut marks made by unretouched flakes; red plus: cut marks made by retouched implements; green crosses: experimental PIMs; blue asterisk: Cornafessa drag

Full size image

Discussion

Methodological outcomes and ballistic considerations

The morphometric features of the archeological drag are consistent—even if not fully correspondent—to the experimental ones (Duches et al. 2016), confirming its interpretation as a PIM. This result also proves the usefulness of 3D in distinguishing PIMs from other taphonomic marks and the reliability of the experimental data assumed as reference in this work. With regard to cut marks, a clear distinction is defined by statistic: drags depth is significantly higher than that of cut marks (Mann-Whitney U test: U = 9; p = 1.963E−05); RTD index of drags is lower than that of cut marks (Mann-Whitney U test: U = 128; p = 0.007); lastly, the opening angle is narrower in drags (Mann-Whitney U test: U = 112; p = 0.004). In comparison to indentations found on bone or antler hammers used as knapping tools (Bello et al. 2013; Van Kolfschoten et al. 2015), drags generally look different in shape, being deeper (and sometimes wider) than scores measured by Bello et al. (2013) and Van Kolfschoten et al. (2015, supplementary data). The Cornafessa drag, in particular, is deeper (0.58 mm) than any other score described by these authors. Since only length, depth, and width were measured on knapping marks, a statistical comparison with PIMs is difficult. In particular, it would be necessary to know the RTF/RTD indexes of knapping scores.

Post-depositional alterations of bone surfaces—in the form of roots etching, corrosion, manganese staining, and trampling marks—do not compromise the feasibility of this method on the Cornafessa specimen, while the effects of weathering and fluvial abrasion in PIMs identification remain difficult to evaluate. Nevertheless, even considering the possibility of an alteration of the original profile in archeological drags, some morphometric parameters, such as depth, opening angle, and RTD index, remain crucial in PIMs identification. With regard to the conservation of qualitative features, SEM analyses show that some microstriations on the walls of the drag are preserved.

Concerning the small lithic fragments embedded, this new archeological evidence contributes to the debate around the diagnostic significance of this feature in drags identification. Although some authors suggested that small lithic fragments could be trapped in bones either during butchering actions or projectile impacts (Milo 1998; Parsons and Badenhorst 2004), there are hitherto no evidence of experimental or archeological stone embedded in cut marks. We can thus support the hypothesis that only a considerable amount of force could plunge lithic fragments into bones (O’Driscoll and Thompson 2014), making this feature a relevant indicator for drags identification. It is however necessary to consider that stone embedded were also documented in knapping marks left on bones during shaping and retouching of lithic artifacts, as the result of high-energy impacts (Abrams et al. 2014; Bello et al. 2013; Bordes 1974; Mallye et al. 2012; Olsen 1989; Van Kolfschoten et al. 2015). In these cases though, the association of lithic fragments with a series of characteristic features of knapping damages (i.e., gouges, pits, scores, scratches, and internal striations; Van Kolfschoten et al. 2015) allows for an easy recognition of taphonomic causality both on experimental and archeological material (Abrams et al. 2014; Bello et al. 2013; Bordes 1974; Mallye et al. 2012; Olsen 1989; Van Kolfschoten et al. 2015). Even though the fragments embedded in Cornafessa drag are small and difficult to classify, they offer some elements of interpretation. We can assume, in fact, that the deeper one results from the initial impact of the lithic point that dug a wider and deeper groove before the breakage, while it is not clear if the second, more superficial, belongs to the point itself or to a different implement assembled on the same shaft. Given the belonging of the fragments to the same lithotype (i.e., Maiolica; Weissert 1981), EDX spectra are inconclusive in distinguishing different implements. Based on their alignment and different depth, however, it seems difficult to ascribe the second piece to the same backed point, already fragmented. It is more likely that another implement arranged behind the point hit in turn against the bone surface, leaving a slice of the cutting edge exposed from the hafting. Such evidence thus supports the hypothesis of a Late Epigravettian composite projectile, with a backed point hafted as piercing element and backed (bi)truncated bladelets functioning as lateral cutting elements. This arrangement has already been suggested by many authors for the Late Epigravettian of northeastern Italy, on the basis of backed tool morphological features and use-wear analysis (Agogué and Dalmeri 2005; Duches et al. 2017; Ziggiotti 2008). Considering drags produced experimentally by Late Epigravettian backed implements, we verified that composite projectiles could create drags with variable width and depth along the profile, due to the actions of multiple cutting elements against the bone (Fig. 13).

Fig. 13
figure 13

Experimental drag marks on bones made by Late Epigravettian composite projectiles. a: stereomicroscope image of a drag on radius; b: stereomicroscope image of a drag on rib

Full size image

Regarding the weapon system used to launch this projectile, further discussion is required. The hypothesis of a general widespread in Europe of the bow-arrow delivery system during the Late Upper Paleolithic has been assumed by many scholars for different chrono-cultural contexts (Caspar and De Bie 1996; Cattelain 1997; Duches et al. 2017; Pelegrin 2000; Plisson 2005; Serwatka 2018; Valentin 2008). Although Stelmoor findings (12,600–11,500 cal BP; Beckhoff 1968; Litt and Stebich 1999; Rust 1943) represent the oldest archeological evidence, considerations supporting a previous introduction of this technology in Europe deal with the progressive miniaturization of lithic projectile elements, the disappearance of spear-throwers from archeological contexts, the transformation of hunting strategies, and the evolution of climatic and environmental conditions (Ammann et al. 2013; Bignon 2008; Caspar and De Bie 1996; Cattelain 1997; Duches et al. 2017; Naudinot et al. 2017; Nielsen 2013; Pelegrin 2000; Plisson 2005; Valentin 2008). With regard to the Late Epigravettian of northeastern Italy, technological analyses testify a general simplification of the lithic production systems between 17,000 and 11,600 years cal BP, with a shift in technical investment from shaping on the core to a subsequent shaping on the derived flake blank (Duches et al. 2017; Montoya 2008; Montoya and Peresani 2005). In this framework, the persistence of standardized lithic backed tools used as projectile implements suggest the need for easily repairable projectiles through the substitution of functional elements (Duches et al. 2017). Since the interface between shaft and point represents the fundamental fulcrum of the projectile (Bergman et al. 1988; Carrère 1990; Friis-Hansen 1990; Kooi and Bergman 1997; Pétillon et al. 2011), thus a correct fit of the armature will profoundly condition its capability to perforate and wound: a discrepancy between shaft diameter and implements’ width could cause indeed the rebound of the projectile (hilt effect sensu Guthrie 1983; Pétillon et al. 2011). Shape and dimensions of Late Epigravettian backed pieces (Duches et al. 2017) suggest a shaft diameter not wider than 0.8 cm, excluding it from the range documented in archeology and ethnography for javelins or spear-thrower darts (Cattelain 1997; Churchill 1993; Cundy 1989; Davidson 1934; Ellis 1997; Hughes 1998; Shott 1997; Thomas 1978). Consistent data are also provided by Late Epigravettian antler and bone projectile points: complete items from Dalmeri and Tagliente rock shelter show a maximum width rarely beyond 1 cm in front of a high variable length (Bertola et al. 2007; Cristiani 2008; Gurioli 2008). Bone point from Cornafessa rock shelter (Fig. 14) confirms this tendency, showing a width corresponding to the expected range for arrow tips on the basis of ethnographic (Cattelain 1997, 2004; Churchill 1993; Ellis 1997; Shott 1997; Thomas 1978; Witthoft 1968), experimental (Caspar and De Bie 1996; Cattelain and Perpére 1993; Geneste and Plisson 1986, 1990; O’Farrell 2004; Soriano 1998), and paleo-ballistic data (Bergman et al. 1988; Carrère 1990; Friis-Hansen 1990; Kooi and Bergman 1997). Using as reference 150 complete backed points from Dalmeri rock shelter (raw data and statistical values in Duches et al. 2017), TCSA (tip cross-sectional area, after Hughes 1998) values are largely lower than the expected ones for arrowheads on the base of ethnographic data (Fig. 15); considering the weights, ranging between 0.05 and 1.50 g (Duches et al. 2017), Late Epigravettian projectiles fall into low-power/high-velocity projectiles which are consistent with the bow and arrow delivery system (Hughes 1998; Pétillon et al. 2011; Serwatka 2018). Lastly, spear-throwers are completely absent from Late Epigravettian archeological record.

Fig. 14
figure 14

Bone point from Cornafessa rock shelter. a: intentional breakage of the base and hafting traces on the inner surface; b: longitudinal striations and hafting traces on the mesial part; c: ochre residues and root etchings on the base; d: close-up of scraping marks related to fabrication; e: faceting of the profile of the distal part; f: longitudinal striations and rounding of the tip

Full size image
Fig. 15
figure 15

Boxplot showing the comparison between TCSA values of Late Epigravettian backed points, ethnographic arrowheads, and ethnographic dart tips. BP: Late Epigravettian backed points from Dalmeri rock shelter (n = 150); A: ethnographic arrowheads (n = 118 after Thomas 1978); D: ethnographic dart tips (n = 40 after Thomas 1978 and Shott 1997)

Full size image

As suggested by the location of the Cornafessa injury and the projectile’s entrance point, the arrow probably hit the bear from behind with an oblique trajectory headed toward its vital organs. Just like butchery marks, bone injuries might be unintentional and undesired. Hitting osseous tissues may increasingly damage the arrow and compromise its penetration ability. As such, we can suppose that this shot erroneously hit the rib but could anyway cause the bleeding to death of the animal. Some authors even suggest that a more “dorsal” or “lateral” angle of the shot could help the projectile’s penetration of the ribcage in the case of medium or large animals (Frison 1998; Pöllath et al. 2017).

Archeological evidence of bear hunting

It is nowadays generally recognized that the frequent association of bear and cultural remains in Paleolithic sites does not necessarily demonstrate an interaction between the two species (Auguste 2003; Bona et al. 2007; Cvetković and Dimitrijević 2014; Fourvel et al. 2014; Gargett 1996; Stiner 1998; Torres et al. 2007). The sharing of ecological and ethological characteristics led in fact bears and human to live in the same habitats and shelters, cohabiting in a good part of Europe all through the Pleistocene (Arribas and Palmqvist 1999; Brugal and Fosse 2004; Pacher 2003; Quilès 2004; Stiner 1999; Wolverton 2001). The taphonomic history of bear faunal assemblages is therefore the most significant aspect in investigating the interaction between ursids and Paleolithic human groups. If the presence of cut marks and percussion marks confirms the anthropic exploitation of this animal in multiple periods and cultural contexts (Abrams et al. 2014; Arilla et al. 2014; Auguste 2003; Binford 2002; Brugal and Fosse 2004; Enloe 2012; Fosse et al. 2002; Germonpré and Hämäläinen 2007; McLaren et al. 2005; Miracle 2007; Pitulko and Kasparov 1996; Romandini and Nannini 2012; Romandini et al. 2018; Stiner 1994; Tillet and Binford 2002; Turk 2014; Wojtal et al. 2015), the information concerning the modality of bear procurement is limited.

Evidence from different Mousterian sites suggests intensive exploitation of both cave bear and brown bear through targeted visits at watersides or in caves used for lair. The finding of consistent bear bone accumulations associated with a high rate of anthropic marks and a prime adult dominated spectrum supports in fact the hypothesis of active and directed hunting at Biache-Saint-Vaast in France (Auguste 1992, 1995, 2003), Taubach in Germany (Bratlund 1999; Gaudzinski-Windheuser and Roebroeks 2011), Krapina in Croatia (Miracle 2007), and Rio Secco Cave in Italy (Romandini et al. 2018).

Regarding the Upper Paleolithic, there is scarce information from Aurignacian contexts with the notable exception of the Slovenian site of Potočka zijalka. This cave, frequented by humans during the Early Upper Paleolithic, has yielded several osseous projectile points mostly fabricated from ursid bones (Hofreiter and Pacher 2004; Jéquier 2016), associated to a small Protoaurignacian lithic assemblage (Moreau et al. 2015). The zooarcheological record was mainly represented by ursids (in particular Ursus spelaeus, sensu lato), but its accumulation was in most cases natural (Pacher 2001; Pacher and Marinelli 2004; Rabeder et al. 2004). Nevertheless, the identification of a cave bear shoulder blade with a pathology indicating an artificial traumatic origin is noteworthy (Withalm 2004). Wound size and shape led the researcher to infer that the injury was caused by a bone peak, very similar to those found in the Aurignacian layers. Moreover, the position and direction of the channel suggests a perforation from behind with a down-to-upside inclination. Finally, the clear evidence of healing attests to hunting failure (Withalm 2004).

Human exploitation of both cave bear (U. spelaeus, sensu lato) and brown bear is well documented in Gravettian sites of Central Europe (Münzel and Conard 2004; Wojtal et al. 2015). Even though in the Gravettian layers of Hohle Fels Cave (Swabian Jura) and Deszczowa Cave (Krakowsko-Częstochowska Upland), the majority of the accumulated bear remains is the result of natural deaths during hibernation, butchering marks are present on all skeletal elements and document all the phases of the carcasses’ dismemberment (Münzel et al. 2001; Münzel and Conard 2004). Thanks to the presence of fetal horse bones and cut marks on juvenile (about 1 year old) cave bear remains, Münzel (2004) could demonstrate winter occupation of Hohle Fels by Paleolithic hunters and thus confirmed vulnerability of the bears during hibernation. Some exceptional hunting evidence is represented by a projectile impact mark recorded on the lateral process of a cave bear thoracic vertebra dated to 27,830 ± 150–140 uncal BP (Münzel 2004; Münzel and Conard 2004; Münzel et al. 2001). The puncture, characterized by an embedded flint, confirms the active predation of cave bears and suggests a winter hunting during hibernation on the basis of the lesion’s location and shooting angle (Münzel and Conard 2004). The authors also proposed a spear or lance with flint tip as hunting weapon. Finally, the success of the hunt is demonstrated by several cut marks recorded on the spinal process of the same vertebra (Münzel and Conard 2004).

Conspicuous bear bone assemblages also come from Early Epigravettian sites located on the Berici Hills in northeastern Italy, which testify to both bear hibernation and human occupations. Cave bears frequented this area during the Last Glacial Maximum and Epigravettian hunters repeatedly visited the Covolo Fortificato di Trene and the Buso Doppio del Broion Cave for the exploitation of this hunting resource. Taphonomic analysis revealed in fact several anthropic traces on cave bear remains (Romandini and Nannini 2012) stratigraphically associated with shouldered and backed points, some of which characterized by diagnostic impact traces (Broglio et al. 2009).

The Late Epigravettian evidence from the southern slope of the Alps are on the contrary very scanty, testifying the occasional exploitation of U. arctos in faunal assemblages generally dominated by medium-sized ungulates (Bertola et al. 2007; Fiore and Tagliacozzo 2005). In Tagliente (Rocci Ris et al. 2005), Biarzo (Bertolini et al. 2016) and Dalmeri (Fiore e Tagliacozzo 2008) rock shelters, all located in northeastern Italy, taphonomic evidence is mostly related to skinning actions and, rarely, to defleshing and disarticulation. From Dalmeri rock shelter, multiple cut marks on third phalanges also attest to the deliberate removal of bear claws (Fiore and Tagliacozzo 2008).

The only direct information about Late Glacial bear hunting comes from the alpine site of Grotte du Bichon (La Chaux-de-Fonds, Swisse). In this cave, a unique evidence of hunting accident is preserved: a female brown bear skeleton injured by a projectile impact mark was found covering the almost complete skeleton of a 20-year-old man. The archeological interpretation supposes that a female bear was driven out from her den while hibernating with two cubs, wounded, and successively chased in the deepest part of the cave by a hunter who found in turn his death, being injured by the animal (Morel 1998). The abundance of charcoals suggests that the bear was smoked out of the cave and then attacked at the entrance by one or more hunters. The animal displays a flint fragment embedded in a cervical vertebra but it is not clear whether bow and arrow or a spear-thrower were used to inflict the hunting lesion. Since the location of the injury supports the hypothesis of a shot against the animal when it charged standing upon its haunches (Morel 1998), both weapon systems could have been used.

Ethnographical comparison on bear hunting strategies

Ethnographic data show that bear hunting was a pan-boreal practice across the northern hemisphere (McLaren et al. 2005). The synthesis provided by Hallowell (1926), focusing on bear hunting strategies traditionally used from northern Eurasia to North America, offers relevant clues for archeological interpretations. The first noteworthy aspect is the evidence of different annual favorable moments in bear approachability, which considerably affects weapons and hunting techniques employed. Certain predictable aspects of seasonal bear behavior are in fact strategical for the procurement of this prey and hunting success.

Hunting during the hibernating season, between winter and early spring, is well documented in ethnography and connected to bear vulnerability. The animal’s habit of using the same lair for several consecutive years (Kurten 1976) and the hunters’ custom of tracking bear during late fall in order to find its den guarantee a potential mid-winter source of food, fur, grease, and bone (Hallowell 1926). The animal is never killed in the den but the hunters generally lure it out using smoke or dogs and then dispatch it through spear or axe (Hallowell 1926). Spears could be either wielded by the hunter or set in a stationary position at which the bear was provoked or tricked into charging (McLaren et al. 2005). The ethnographical documentation of the use of bow and arrow in such hunting situation is on the contrary very rare (Hallowell 1926). This kind of weapon is more frequently employed in the open during warmer seasons, as documented for some Paleo-Siberian peoples (Ainu: Batchelor 1901; Koryak: Jochelson 1905-1908; Gilyak of Sagyhalin: Labbé 1903) and natives of the North Pacific Coast (Algonkian: Denys 1908; Tlingit: Krause 1885). It is also interesting that during the bear festival among the Ainu, the animal is firstly shot at with blunt arrows before being ceremonially slaughtered (Batchelor 1901). In closer hand-to-hand combats, bow and arrow are frequently replaced by spear or lance engaging a very dangerous venture (de Laguna 1972; Hallowell 1926).

The use of traps and snares is frequently documented among ethnographic people and can occur all year round. The construction of traps is strictly related to the bears’ habit of “step-marking,” i.e., their tendency to step over their own footprints while walking their habitual trails (Holzworth 1930; Hummel and Pettigrew 1991). Such behavior facilitates bear trapping and considerably decreases hunting risk. Bear deadfall traps were used among numerous peoples of the Northwest Coast of North America as well as along the Siberian Coast (Emmons 1991; Hallowell 1926; Harrison 1925; Swanton 1905). Butchering activities always occurred in the vicinity of trapping localities that are, in turn, always far away from the residential camps (McLaren et al. 2005).

Regardless of the modality of predation, ethnographic data testify carcass-selective processes and differential transport from the initial kill site: when the frequencies of recovered skeletal remains in sites other than butchering places are compared, indeed, bear trunk elements are always underrepresented. This is interpreted as the habit of leaving the spinal column and the ribcage in the dispatching area (Binford 1983, 2002). Archeological data from two early Holocene residential sites with abundance of bear remains strengthen this assumption: the Kilgii Gwaay site on the American Northwest Coast (McLaren et al. 2005) and the Zhokhov Island site in Siberia (Pitulko and Kasparov 1996) show a similar pattern of bear skeletal element abundance, marked by the strong underrepresentation of vertebral elements, ribs, and hind limbs.

Conclusion

This work represents the first archeological application of a diagnostic method for PIMs identification, based on statistic comparison of morphometrical parameters acquired through 3D digital microscopy. The recognition of the Cornafessa drag mark is particularly relevant because it confirms the method’s validity on the most challenging PIM category. If indeed puncture marks are frequently recognized on zooarcheological material because of the presence of stone embedded (Gaudzinski-Windheuser 2016; O’Driscoll and Thompson 2018; Pöllath et al. 2017), the distinction of drags from cut marks and carnivore tooth scores is much more complicated. Moreover, we proved that different projectiles leave different PIMs and that only a reliable ballistic experimentation could provide the diagnostic elements useful in drag identification (Duches et al. 2016). The presence of small lithic fragments embedded in the Cornafessa specimen enable us to add some methodological considerations concerning the relevance of this feature in drag identification and the potential significance of profile width and depth variations for the reconstruction of projectile design. In this regard, the morphometrical features of Cornafessa specimen compared to the experimental ones suggest that the lesion was caused by a lithic composite projectile, probably characterized by a backed point associated with at least one backed (bi)truncated bladelet.

The identification of this drag on a U. arctos rib enriches the current debate about Pleistocene bear hunting, offering new clues about the strategies employed by Epigravettian groups during the Late Glacial. Several remains from Late Epigravettian sites in northeastern Italy testify to bear processing for the acquisition of fur, meat, and bones (Bertolini et al. 2016; Fiore and Tagliacozzo 2008; Rocci Ris et al. 2005) but none provides clear evidence of bear hunting. If evidence of skin removal suggests a primary access to the carcass for these sites, the reconstruction of procurement season and hunting strategy is more challenging. Seasonal information is not available for the Cornafessa rock shelter yet, but the PIM analysis confirms a direct predation on this bear.

Going beyond the limit of the Late Glacial period, direct information about bear hunting strategies remains sporadic. The drag from Cornafessa rock shelter represents the fourth hunting lesion recognized on bear bones for all the Pleistocene (Table 3). The other evidence from Potočka zijalka, Hohle Fels, and Grotte du Bichon (Morel 1998; Münzel and Conard 2004; Withalm 2004) consist of three impact punctures, interpreted as hunting predation during winter taking advantage of bears vulnerability in hibernation. Given the different location of the lesions, the authors hypothesize a predation toward a bear lying in hibernation for Hohle Fels (Münzel and Conard 2004) and an assault against a bear charging in a standing position for both Potočka zijalka and Grotte du Bichon (Morel 1998; Withalm 2004). Regarding the weapon employed, a spear or lance with osseous or lithic tips are believed for Potočka zijalka and Hohle Fels, while both bow and arrow or spear-thrower are considered for Grotte du Bichon. Nevertheless, ethnographic data suggest that spear or lance are widely preferred by hunters when attacking a bear at the cave entrance after the interruption of its hibernation period.

Table 3 Archeological hunting lesions on bear bones
Full size table

Since Cornafessa rock shelter is not suitable as hibernating den, this new evidence represents a dissimilar attestation of bear hunting strategies, testifying to the predation of a bear in a place different from lair caves. Besides, the PIM location on a rib and the presence of two vertebrae within the identified bear bones possibly suggest the transportation into the site of the whole carcass, an unusual behavior for such a big prey. In this regard, ethnographic and archeological data frequently attest to bear carcass differential transport from the initial kill site to residential site, on the basis of comparative frequencies of recovered skeletal remains marked by a strong underrepresentation of bear trunk elements (Binford 1983, 2002; McLaren et al. 2005; Pitulko and Kasparov 1996).

Finally, since the other few direct evidence of bear hunting suggests a different scenario, i.e., den hunting through spears, this rib represents the first archeological proof of bear predation using bow and arrow.