CDV-Tour 1 has been imaged in 2008 by X-ray microtomography (μCT) at the Centre de Microtomographie of the University of Poitiers. Two acquisitions have been run to scan the entire shaft and the records have been merged after volume reconstruction. Scans were realized by a X8050-16 Viscom AG (image intensifier with a 1004 × 1004 12 bit camera and a pixel size of 147 μm) according to the following parameters: 120 kV voltage, 560 μA current, 32 integrations/projection, and 1200 projections (each 0.30°). The final volume was reconstructed using DigiCT v.2.4.2 (Digisens) in 8-bit format, with an isotropic voxel size of 107.8 μm.

Because of globally good outer and inner bone preservation quality and the lack of any sedimentary infill of the medullary cavity, a semi-automatic threshold-based segmentation with manual corrections has been successfully carried out following the Half-Maximum Height method (HMH; Spoor et al., 1993) and the region of interest thresholding protocol (ROI-Tb; Fajardo et al., 2002) and by taking repeated measurements on different slices of the virtual stack (Coleman and Colbert, 2007) by Avizo v.6.2. (Visualization Sciences Group Inc.) and ImageJ v.1.46a (Rasband, 2010).

To orient the specimen anatomically, we used the position of the linea aspera. The approximate position of the midshaft was confidently defined on comparative ground through the morphology of the linea aspera and the shaft outline in medial and lateral views. The 50% cross-section precise location is usually less clearly indicated in fossil human femora displaying subcircular diaphyses and lacking a true pilaster, such as CDV-Tour 1; however, this region shows little longitudinal changes in morphology along several centimetres, making modest errors in the location of these sections not critical for the structural characterization of the specimen (Sládek et al., 2010 and Trinkaus and Ruff, 2012).

Cross-sectional geometric (CSG) properties of CDV-Tour 1 were measured on four selected virtual sections (A to D) respectively corresponding to: the uppermost (A, 62.4 mm above the midshaft) and the lowermost (D, 33.4 mm below the midshaft) complete sections found along the shaft in proximo-distal direction; the midshaft (C); the mid-distance between the uppermost and the midshaft sections (B) (Fig. 3). As CDV-Tour 1 misses both articular ends, we could not directly measure its biomechanical length. However, if the average value of ∼407 mm obtained for a sample of eight European Neanderthals (La Chapelle-aux-Saints 1, Feldhofer 1, La Ferrassie 1 and 2, Fond-de-Forêt 1, Palomas 92 and 96, Spy 2; in Trinkaus and Ruff, 2012, table A1) is used as a proxy for estimating the biomechanical length of CDV-Tour 1, the segment A-D (95.8 mm) should approximately correspond to the shaft portion comprised between 42% (distal) and 65% (proximal) of its biomechanical length, the section B corresponding to ∼57.5%.

In addition to the anteroposterior (a-m) and mediolateral (m-l) outer diameters (in millimeter, both measured directly on the original shaft by a dial-equipped vernier caliper to the nearest 0.1 mm, and also on the virtually reconstructed specimen), the following CSG parameters have been considered by using a custom routine developed in R v.2.11.1 (R Development Core Team, 2011): total area (TA, in mm²); cortical area (CA, in mm²); percent of cortical area (%CA); second moments of area about the m-l and a-p axes (Ix, Iy, in mm⁴); polar second moment of area (J, in mm⁴); section modulus about the m-l and a-p axes (Zx, Zy, in mm³); polar section modulus (Zp, in mm³); maximal and minimal second moments of area (Imax, Imin, in mm⁴) (Ruff, 2008).

In order to place CDV-Tour 1 within the pattern of variation reported for the European Late Pleistocene fossil record, comparative data for %CA and the Ix/Iy ratio measured at 50% and 65% cross-sectional level were obtained for a number of Neanderthal (Neand.) and Upper Paleolithic (EUPH) human femora from the database recently published by Trinkaus and Ruff (2012), integrated by fresh evidence on a Neanderthal femoral shaft from the MIS 4 French site of Les Pradelles (Mussini et al., 2012) (Table 1). For the comparative assessment of the pilastric index (diameter a-p/m-l × 100), we used the same samples detailed in Table 1 integrated by data from the following specimens: Santa Croce 1 and Stadelhöhle 1, among the Neanderthals, and Barma Grande 6, Caviglione 1, and Předmostí 3, 4, 9, 10 and 14, for the Upper Paleolithic sample (Trinkaus and Ruff, 2012, table A13).

By using an original (μ)CT record, we also comparatively detailed cortical bone thickness variation at eight specific sites (anticlockwise: posterior, postero-lateral, lateral, antero-lateral, anterior, antero-medial, medial, postero-medial) of the 50% and 65% cross-sections in CDV-Tour 1, La Ferrassie 1 (left), Rochers-de-Villeneuve 1 (50%), Spy 8 (Volpato et al., 2012b), Cro-Magnon 1 (left; Puymerail, 2011 and Puymerail and Macchiarelli, 2009), and in the Magdalenian skeleton from Chancelade (right femur; Bondioli et al., 2010, Puymerail, 2011 and Puymerail and Macchiarelli, 2009), as well as in an extant human reference sample.

In our study, the extant human condition is illustrated by 20 femora measured from medical CT records performed at spatial resolutions ranging from 350 to 791 μm (Puymerail, 2011 and Puymerail et al., 2012). This sample (EH) pools adults of both sexes from 19th century France and the Imperial Roman graveyard of Velia, Italy, stored at the Musée de l’Homme of Paris, the University of Poitiers, and the Museo Nazionale Preistorico Etnografico “L. Pigorini” of Rome.

Three-dimensional cortical thickness topographic distribution in CDV-Tour 1 for the whole portion A-D has been virtually rendered for each anatomical projection by a chromatic scale increasing from dark blue (thin) to red (thick), where bone thickness is defined for each point on the periosteal surface as the distance to the closest point on the endosteal surface (Bayle et al., 2011, Bondioli et al., 2010, Mazurier et al., 2010, Puymerail et al., 2012, Volpato et al., 2011, Volpato et al., 2012a and Volpato et al., 2012b).

Technical advances in the qualitative and quantitative structural characterization through functional imaging of the femoral diaphyseal organization have made possible the realization of morphometric maps suitable for synthetic visual appreciation and comparison (Bondioli et al., 2010, Puymerail, 2011 and Puymerail et al., 2012). Accordingly, the A-D periosteal surface portion of CDV-Tour 1 has been mapped onto a cylinder, whose diameter corresponds to the maximum width of the original surface of the shaft, virtually unzipped along a predefined vertical line along the anterior aspect, and then unrolled into a plane. Thickness values have been standardized between 0 and 1 (methodological details in Bondioli et al., 2010). For the specific purposes of the present study, we directly compared the structural signature characterizing CDV-Tour 1 to the morphometric maps obtained for the 42–65% shaft portion of their respective biomechanical lengths for the Neanderthal femora La Ferrassie 1 (left) and Spy 8 and the Gravettian Cro-Magnon 1 (left). With this respect, we also provided comments about the μCT-based morphometric map already generated for the Chancelade 1 femur (Bayle et al., 2011, fig. 6; Bondioli et al., 2010, fig. 3; Puymerail and Macchiarelli, 2009). Finally, since it is possible to perform generalized additive models (GAM) of interpolation to obtain consensus maps by merging individual morphometric maps into a single dataset (Wood, 2006), we also included in the comparative analysis the consensus map summarizing the whole variation expressed by our recent comparative sample (EH; Puymerail, 2011 and Puymerail et al., 2012). To generate the maps, we used the pattern of relative thickness repartition (local morphometric properties) obtained by a custom routine developed in Scilab v.4.1.2 (The Scilab Consortium) with the support of the R v.2.11.1 (R Development Core Team, 2011) packages mcgv (Wood, 2004, Wood, 2006 and Wood, 2008), spatstat (Baddeley, 2008), and gstat (Pebesma, 2004).

Although the femoral shaft portion CDV-Tour 1 represents only approximately 50% of its estimated original biomechanical length (204.7 mm of ∼407 mm), it preserves enough morphological and structural information to allow its characterization and reliable attribution to an adult Neanderthal individual, more likely a male. Besides its general outer morphology (notably, the absence of a pilaster and its general robusticity), this is suggested by the comparison of the antero-posterior and medio-lateral diameters at midshaft (Table 2, Fig. 4a) and the pilastric index (Fig. 4b) of CDV-Tour 1 to figures of the Late Pleistocene European Neanderthals (Mussini et al., 2012 and Trinkaus and Ruff, 2012), and European Upper Paleolithic (Trinkaus and Ruff, 2012) and recent humans (Puymerail, 2011 and Puymerail et al., 2012). However, in this respect, the range of variation displayed by the European Upper Paleolithic femoral sample is noteworthy: it mainly reflects a positive allometric effect in which a-p rigidity increases faster than m-l rigidity (slopes greater than 1.0; Trinkaus and Ruff, 2012). For these parameters, CDV-Tour 1's closest fits are the femora known as La Ferrassie 1 and 2, La Quina 5, and Saint-Césaire 1 (in Trinkaus and Ruff, 2012). Conversely, it shows no overlap for the pilastric index exists with the reference sample used in the present study, nor for the outer diameters, which were measured at 65% of the estimated biomechanical length, where CDV-Tour 1 largely exceeds our extant figures (Table 2).

The strong similarity of CDV-Tour 1 to the Neanderthal estimates and its difference with respect to both Upper Paleolithic and recent samples are revealed by the cross-sectional geometric (CSG) parameters summarized in Table 2.

In particular, for the anteroposterior (Ix) and mediolateral (Iy) second moments of area (Fig. 5a), CDV-Tour 1 approximates the lowest Neanderthal values at both 50% and 65% cross-sections (greater resistance to medio-lateral bending charges), and clearly is set apart from the fossil and extant modern human estimates, systematically showing higher ratios (Trinkaus and Ruff, 1989, Trinkaus and Ruff, 1999, Trinkaus and Ruff, 2012, Trinkaus et al., 1998 and Volpato et al., 2012b). At 50%, the difference for the Ix/Iy ratio between CDV-Tour 1 and the Neanderthal median equals 0.11, whereas those from the Upper Paleolithic and the extant samples correspond to 0.65 and 0.35, respectively. A similar pattern is found also at 65%, the differences in this case corresponding to 0.01, 0.24, and 0.17, respectively.

Following a recent reassessment of femoral diaphyseal properties through Pleistocene Homo, it is likely that changes in cross-sectional shape recorded in the middle portion of the diaphysis during the Late Pleistocene, expressed by a relative antero-posterior vs. medio-lateral reinforcement, reinforcement, more closely reflect shifts in pelvic shape and proximal femoral proportions rather than differential ranging (Trinkaus and Ruff, 2012; contra Beauval et al., 2005 and Trinkaus et al., 1998).

Compared to other CSG parameters, the same time-related trend analysis (Trinkaus and Ruff, 2012) suggests a modest informative value in terms of biomechanics provided by the relative cortical area (%CA), an indicator that shows significant intra- and inter-group variation and no evident evolutionary pattern from Middle to Late Pleistocene humans (Trinkaus and Ruff, 2012; for H. erectus, see also Puymerail et al., 2012). The cortical area provides a measure of resistance to axial loads (Ruff et al., 1993), and the combination of cortical and total area reflects the differential subperiosteal deposition and endosteal resorption of bone, principally during development (Ruff and Hayes, 1983 and Ruff et al., 1994). Accordingly, as also suggested by the %CA comparative plots at 50% and 65% cross-sectional level shown in Fig. 5b (Table 2), percent cortical area should mostly represent a measure of differential developmental and aging processes, not a subtle mechanical indicator (Trinkaus and Ruff, 2012).

For the shaft portion 50–65%, we compared the amount of cortical bone volume in CDV-Tour 1 (34.8 cm³, i.e., 76% of the total volume) to the estimates obtained for the Neanderthal femora known as La Ferrassie 1 (left: 38.5 cm³ = 81%) and Spy 8 (right: 27.4 cm³ = 79%; Volpato et al., 2012b), to the Upper Paleolithic specimens Cro-Magnon 1 (left: 43.3 cm³ = 85%) and Chancelade (right: 30.7 cm³ = 84%), and to the figures from our reference sample (average: 28.0 cm³ ± 5.5 = 77% ± 5.7). However, this approach did not reveal any possible Neanderthal vs. modern (fossil and/or extant) pattern.

In order to tentatively identify at specific cross-sectional levels a recurrent model in bone topographic distribution characterizing the Neanderthals and/or the morphologically modern humans considered in this study, we also used our (μ)CT record to measure linear cortical thickness at eight sites along the 50% and 65% sections in CDV-Tour 1, La Ferrassie 1, Rochers-de-Villeneuve 1 (50% only), Spy 8, Cro-Magnon 1, and Chancelade (Bondioli et al., 2010, Puymerail, 2011 and Puymerail and Macchiarelli, 2009), and in the recent human sample. The results are summarized in Table 3. At midshaft, cortical bone is thicker posteriorly in virtually all investigated cases, including among the recent humans, but the sequences of relative bone distribution at the remaining sites are quite variable among the fossil specimens as well as within the comparative reference sample (model based on the site-specific average values), with no apparent recurrent model. Interestingly, while measured at only six sites (i.e., excluding the posterior and the anterior cortices), a decreasing pattern in cortical bone thickness at midshaft close to that displayed by CDV-Tour 1 (Table 3) has been established for the left femur of the 2.5–3 years old Neanderthal child from Roc de Marsal (lateral > postero-lateral > medial = postero-medial > antero-lateral > antero-medial; Macchiarelli et al., 2007, fig. 4). However, it has been noted that in this case also, the structural model shown by this Neanderthal immature does not clearly differ from the condition found in samples of recent human femora from individuals < 6 years (Cowgill, 2010, Macchiarelli et al., 2007 and Volpato et al., 2007).

In our series, the average cortex measured at ∼65% of the biomechanical length is universally thicker than at midshaft (Table 3), the highest value being displayed by the left femur Cro-Magnon 1 (10.0 mm), the same specimen that at midshaft shows the thinnest average cortex (6.7 mm) among the six fossil femora considered in the present study. In terms of relative bone distribution and decreasing patterns, also at this cross-sectional level no distinction is possible between Neanderthal and modern figures.

As a whole, when combined, these specific volumetric and punctual linear measurements support the suggestion that variation in the relative distribution of cortical bone within the femoral shaft is primarily affected by age- (and likely sex-) related differential endosteal resorption and subperiosteal deposition dynamics (Gosman et al., 2011, Macchiarelli, 1988, Ruff and Hayes, 1983, Ruff et al., 2006b, Ruff et al., 2006a, Ruff et al., 1994, Trinkaus et al., 1994 and Trinkaus and Ruff, 2012). However, the extremely limited number of fossil specimens examined so far does not allow any conclusive statement on this matter, and only extensive future research can test the present findings.

Preliminary functional imaging of the inner organization of the femoral diaphysis through the planar representation of its structural pattern has revealed some differences in local morphometric properties between Neanderthals and fossil and extant humans (Bondioli et al., 2010, Puymerail, 2011 and Puymerail and Macchiarelli, 2009). In order to ascertain its structural signature, in Fig. 6, we compared the standardized morphometric map of cortical bone thickness variation in CDV-Tour 1 (portion 42–65%) to the maps obtained for La Ferrassie 1, Spy 8 (Volpato et al., 2012b) and Cro-Magnon 1 (Puymerail, 2011 and Puymerail and Macchiarelli, 2009), which were generated through the virtually unzipped and then unrolled projection of the same portion. The variation expressed by our recent reference sample is also summarized into a single consensus map (EH; Puymerail et al., 2012). In all cases, the plots represent the relative thickness of the cortex at different levels of the shaft.

The structural signature of CDV-Tour 1, displaying a slightly oblique distinct reinforcement concentrated along the medial aspect of the mid-proximal shaft portion associated with a narrow bony thickening along the linea aspera, closely resembles those of the two Neanderthal specimens considered in the present analysis (Volpato et al., 2012b). This pattern, which expresses the Neanderthal condition of medio-lateral strengthening of the proximal femoral shaft (Trinkaus and Ruff, 2012), contrasts with that found in Cro-Magnon 1, where the thickest values are concentrated along the posterior aspect in relation to a fully developed pilaster, and both medial and lateral aspects show only moderately thick bone. It is noteworthy that a similar signature to that of Cro-Magnon 1, also expressing a greater resistance to axial bending and torsional loads, characterizes the femoral diaphysis of the Magdalenian adult skeleton from Chancelade (Bondioli et al., 2010, fig. 3; Puymerail and Macchiarelli, 2009). In contrast, the structural condition provided by the consensus map of the recent sample features three extended major reinforcements, respectively corresponding to the posterior pilaster (where, as in Cro-Magnon 1 and Chancelade, the thickest bone is found around the midshaft), to the proximal medial and the proximal lateral aspects of the shaft (Puymerail et al., 2012). However, whereas the first two features are systematically found in recent human femora, the degree of expression of the lateral thickening is quite variable. Within this comparative context, the Gravettian Cro-Magnon 1 displays a distinct signature from both Neanderthals and recent humans.

The accidental discovery of the CDV-Tour 1 partial human femoral shaft and associated faunal remains during speleological exploration of “grotte de la Tour”, one of the numerous cavities forming the site complex of La Chaise-de-Vouthon, in Charente (Debénath, 2006), confirms the fossiliferous richness of the Middle to Late Pleistocene sedimentary deposits partially filling the karstic network opening along the Tardoire river, and points to their value for future prehistoric and paleoanthropological investigation. Similar to other contexts reported from the same region (e.g., Beauval et al., 2005, El Albani et al., 2011 and Tournepiche et al., 1996; for a review, see Discamps, 2011), the currently available record supports the preliminary interpretation of a carnivore den for this cavity, and the mammal assemblage recovered so far is compatible with a chronological attribution to the MIS 3. The multiple tooth-marks left on the femoral shaft confirm a close interaction in the region between human groups and large carnivores, especially in the use of natural shelters (Beauval et al., 2005).

Despite the incompleteness of the specimen, our analytical and comparative analysis of CDV-Tour 1, integrating information from classical morphology, cross-sectional geometric properties, and the microtomographic-based 2-3D quantification of its inner structural organization, including the functional signature revealed by its morphometric map of cortical bone topographic variation, convincingly shows that this partial femoral shaft is from an adult Neanderthal individual. Of course, only future research at the “grotte de la Tour” under controlled conditions will confirm the reliability of our assessment concerning its chronological and taphonomic context and allow the identification of additional human remains possibly still preserved in this promising cavity.