The specimens, originally described by Tillier (1980) for their external size and morphology, represent the crown of a left and a right upper first molar (ULm1 and URm1), and of a left upper second molar (ULm2). It is likely that, in all three cases, the roots were in advanced resorption. In 2009, the specimens have been imaged by high-resolution microtomography (μCT) at the Centre de Microtomographie of the Université de Poitiers (equipment X8050-16 Viscom AG; camera 1004 × 1004), according to the following parameters: 120 kV, 0.45 mA current, 32 integrations/projection, and a projection each 0.24°. The final volumes have been reconstructed using DigiCT v.2.3.3 (DIGISENS) with an isotropic voxel size of 21.57 μm³.

Using Avizo 6.1 (Visualization Sciences Group Inc.) and MPSAK 2.9 (in Dean and Wood, 2003), a semi-automatic threshold-based segmentation has been carried out following the half-maximum height (HMH; Spoor et al., 1993) and the region of interest thresholding methods (ROI-Tb protocol; Fajardo et al., 2002). From that step, the fossil crowns have been virtually uncapped (Olejniczak et al., 2008) and each tissue visualized separately. In order to assess tissue proportions, the blank volume between the cervical reference plane and the dentine was then considered as dentine. Despite their degree of mineralization and the local extent of occlusal dental wear (notably, on the ULm1), in all cases the endostructural signal was quite distinct, with high contrasts between enamel and dentine (Fig. 1a).

The following seven linear, surface, and volumetric variables describing tooth tissue proportions were digitally measured or calculated: Ve, the volume of the enamel cap (mm³); Vcdp, the volume of the coronal dentine, including the coronal aspect of the pulp chamber (mm³); Vc, the total crown volume, including enamel, dentine, and pulp (mm³); SEDJ, the enamel-dentine junction (EDJ) surface (mm²); Vcdp/Vc (= (Vcdp/Vc) × 100), the percent of coronal volume that is dentine and pulp (%); 3D AET (=Ve/SEDJ), the three-dimensional average enamel thickness (mm); 3D RET (=3D AET/(Vcdp)^1/3), the scale-free three-dimensional relative enamel thickness (for methodological details, see Kono, 2004, Martin, 1985 and Olejniczak et al., 2008).

The results from the virtual analysis of the three specimens from Tighenif have been directly compared to the microtomographic-based quantitative evidence from the poorly worn maxillary deciduous molars of the Neanderthal child of Roc de Marsal (Bayle et al., 2009a) and to a modern human dental sample (MH) from a medieval cemetery at Usseau, France, including 4 virtually unworn upper m1s and 3 upper m2s. All comparative specimens have been scanned with the same equipment as the Tighenif sample (for methodological considerations, see Olejniczak et al., 2007).

Limitedly to the m2s of Tighenif and Roc de Marsal, four additional measures of radial enamel thickness have been taken on the buccol-lingual (BL) virtual sections through the mesial cusps-dentine horns and compared to the estimates obtained on histological ground by Grine, 2005 (Fig. 1). On the Algerian m2, dentine is exposed on the tip of the cusps, but lateral wall enamel is intact. Accordingly, apical enamel thickness has not been measured, while specific estimates have concerned the buccal and the lingual sides, respectively. The variables are as follows: LTB, the maximum linear enamel thickness assessed on the buccal side of the mesio-buccal cusp, perpendicular to the EDJ at a point approximately 1 mm cervical to the dentine horn apex; LTL, the maximum linear enamel thickness assessed on the lingual side of the mesio-lingual cusp, perpendicular to the EDJ at a point approximately 1 mm cervical to the dentine horn apex; h, the maximum linear thickness of occlusal enamel assessed on the mesio-buccal cusp, perpendicular to the EDJ; i, the maximum linear thickness of occlusal enamel assessed on the mesio-lingual cusp, perpendicular to the EDJ (for methodological details, see Kono, 2004 and Martin, 1985).

For all 2D and 3D measures, intra- and inter-observer tests for accuracy run by two observers provided differences of ≤ 4%.

The ULm1 from Tighenif is extensively worn (stage 4; Smith, 1984), with evident occlusal dentine patches. The URm1, which still preserves a small root fragment, is slightly less worn (stage 3; Smith, 1984), with at least three flattened cusps and spread dentine spots. Compared to the m1s, the ULm2 is only poorly worn (stage 1-2; Smith, 1984), with small dentine exposure on lingual cusps (Fig. 2). This latter crown bears three accessory traits: one small cuspule-like structure inserted between paracone and metacone, originally described as a double buccal groove (Tillier, 1980: 415); a small cuspule set on the mesial aspect of the paracone; and a Carabelli's trait on the mesio-lingual aspect (grade 5 of the ASUDA scoring system; Turner et al., 1991).

The virtual reconstruction of the outer enamel surface (OES) and of the EDJ of the three specimens is shown in Fig. 3. The EDJ is an interface playing a crucial role during tooth development, as most of the morphological features expressed at the OES are mapped on the EDJ at the embryonic stage, while others are the results of overgrowth only affecting the enamel surface (Skinner et al., 2008). Moreover, the virtual exploration of the EDJ morphology allows the identification of subtle features of potential taxonomic value (Macchiarelli et al., 2006 and Skinner et al., 2008).

Despite some size differences between the smaller left and the larger right m1s (B-L: 9.9 vs. 10.7 mm, respectively; Tillier, 1980), and the extensive occlusal dental wear partially masking the lingual half of the left crown, a comparative analysis of their EDJ surfaces reveals a close morphological resemblance in terms of subtle structural features. This is true for the general appearance of the basin ridges and, limitedly to paracone and metacone, for the cusp pattern and the relatively high and protrusive marginal ridges. The intact EDJ of the second molar clearly shows the imprint on the dentine of the three accessory traits appearing at the OES (Fig. 3). Notably, at the EDJ, the external feature between paracone and metacone corresponds to a small but distinct cuspule. Conversely, the single mesial cuspule hardly appreciable occlusally because of wear, corresponds to two lesser cuspule-like traits lying on the mesial marginal ridge (Fig. 4).

Comparative dental tissue proportions in Tighenif, the Neanderthal child from Roc de Marsal, and a modern human sample (MH) are shown in Table 1. Limitedly to the m1s from Tighenif, notably the left one, it should be pointed out that the values concerning the volume of the enamel cap (Ve), the volume of the coronal dentine (Vcdp), the total crown volume (Vc), the Vcdp/Vc percent ratio, and, of course, the enamel thickness estimates (3D AET and RET), are affected by occlusal dental wear and have to be considered as minimum estimates only.

Despite the absolute and relative amount of enamel loss respectively testified by their low enamel cap volumes (Ve) and high Vcdp/Vc ratios, both m1s from Tighenif display a significantly larger total crown volume (Vc) compared to the Neanderthal and the modern specimens, as well as a larger volume of coronal dentine, including the coronal aspect of the pulp chamber (Vcdp). For Vcdp, Neanderthals are typically characterized by a coronal dentine significantly larger than observed in extant humans (Bayle, 2008, Bayle et al., 2009a, Macchiarelli et al., 2006 and Olejniczak et al., 2008). In Tighenif m1s, a difference of 17.2% and 15.2% should be noted between the crowns for Vc and Vcdp, respectively, the right crown being the larger one, as already noticed for the bucco-lingual and mesio-distal diameters by Tillier (1980). In these two crowns, the same pattern is also shown by the portion of the EDJ surface (SEDJ) which is unaffected by wear. With this respect, the Algerian fossil specimens show the absolutely largest values within the comparative sample considered in this study, associated to a lesser difference of 7.6% between the left (smaller) and the right surfaces.

In terms of inner structural morphology, a clearer pattern for the Tighenif dental sample emerges from the analysis of the Um2 (Table 1). In this case, all volumes of the African specimen systematically and significantly exceed the comparative estimates, as does the EDJ surface. However, in relative proportions, the percent of coronal volume that is dentine and pulp (Vcdp/Vc: 62%) closely fits the modern figure (range: 59–62%), not the estimate for the immature from Roc de Marsal (66%).

As noted above, both 3D AET and RET values for the Um1s from Tighenif are biased, while those assessed for the second molar are reliable. For the 3D average enamel thickness (AET), the value of the Algerian specimen falls again within the modern variation range represented in our study, in any case well above the figures of Roc de Marsal. Conversely, Tighenif occupies an intermediate position for the 3D relative enamel thickness (RET).

For all linear variables describing cuspal enamel thickness through the protocone-paracone virtual section (LTB, LTL, h, i), the values of the maxillary m2 from Tighenif are systematically closer to the modern human figures (data from Grine, 2005) than to the estimates obtained for both Roc de Marsal's m2s (Table 2). In all cases, fossil and modern, the enamel is thicker on the lingual aspect of the protocone than on the buccal side of the paracone (LTL > LTB). However, while in Tighenif and the reference modern sample the same topographic pattern is distinctly shown also by the maximum linear thickness of occlusal enamel (i > h), in the available Neanderthal representative the difference between protocone and paracone is slight, with no directional pattern.

For each fossil (Tighenif and Roc de Marsal) and modern (three specimens) m2 crown, a 3D virtual perspective in five views of the enamel thickness topographic variation is rendered in Fig. 5 by means of a thickness-related colour scale (in mm). This visualization technique maps the local enamel thickness on the outer enamel surface and permits to comparatively appreciate the structural contrasts (Kono, 2004, Macchiarelli et al., 2008 and Macchiarelli et al., 2009).

Comparative cartographies clearly evidence the whole similarity between the Algerian fossil and the modern human pattern in enamel topographic distribution, while the absolutely and relatively thin-enameled Neanderthal specimen clearly sets apart. Nonetheless, while in the three modern m2s measured in the present study (range total crown volume: 273–327 mm³) the thickest enamel is found mesially, at the base of the protocone (1.25 mm), in Tighenif (418 mm³) it corresponds to the distal marginal ridge (1.28 mm).

Following the original description and interpretation of the isolated dental sample from Tighenif, because of their proportions, external morphology, occlusal and interproximal wear patterns, it is likely that the three deciduous upper molars are from the same individual (Tillier, 1980). With this respect, the qualitative and quantitative results derived from the present microtomographic-based 3D virtual analysis of their endostructure do not reject such interpretation. In particular, while the left and right first molars show some linear, surface, and volumetric differences (Tillier, 1980 and present Table 1), their enamel-dentine junction surfaces fit in terms of morphological details. Nonetheless, the preservation conditions of the two specimens do not allow a conclusive statement on this matter, and additional finer structural analyses, likely via phase contrast synchrotron radiation microtomography (see Smith and Tafforeau, 2008), should test this point in the future.

Our attempt to characterize the pattern of tooth tissue proportions and enamel thickness distribution of the morph represented by this early Middle Pleistocene North African sample faced the unavoidable limits represented by the preservation conditions of the m1s and, of course, small sample size. Nonetheless, it is noteworthy that, besides the evidence from a few Neanderthals (Bayle, 2008, Bayle et al., 2009a, Bayle et al., 2010, Bondioli et al., 2010, Macchiarelli et al., 2006, Macchiarelli et al., 2007 and Toussaint et al., 2010) and even fewer anatomically modern fossil specimens (Bayle et al., 2009b and Bayle et al., 2010), the extent of deciduous tooth endostructural variation is simply unreported for any Early and Middle Pleistocene human extinct taxon. At the best of our knowledge, the only 3D estimates of tissue proportions available so far concern two H. erectus s.s. permanent molars from Trinil, whose relative enamel volumes better fit the modern human rather than the Neanderthal condition (Smith et al., 2009).

Currently available 3D imaging evidence shows that, compared to the extant human condition reported so far, Neanderthal deciduous teeth possess relative thinner enamel, absolutely and relatively larger dentine volumes, larger and more complex enamel-dentine junction, larger pulp chambers, thicker roots (Bayle, 2008, Bayle et al., 2009a, Bayle et al., 2010, Bondioli et al., 2010, Macchiarelli et al., 2006, Macchiarelli et al., 2007 and Toussaint et al., 2010). In this perspective, certainly for the upper m2, our results allow a preliminary assessment of the primitive inner structural condition characterizing the deciduous molars of the early Middle Pleistocene North African groups likely related to H. heidelbergensis.

The comparative analysis of the percent of the crown volume that is dentine and pulp (Vcdp/Vc) shows that, relative to the enamel, the proportion of dentine and pulp is higher in the Neanderthal specimen considered in the present study, but more balanced in both Tighenif and the modern reference sample. Limitedly to the structural organization of the lower deciduous molars, but not of the incisors (Bayle et al., 2010), the same pattern distinguishes Neanderthals from both anatomically modern fossil and extant humans (Bayle, 2008, Bayle et al., 2009a, Bayle et al., 2009b, Bayle et al., 2010, Macchiarelli et al., 2006, Macchiarelli et al., 2007 and Toussaint et al., 2010). Accordingly, Tighenif is closer to the latter groups.

A growing body of evidence also shows that, while the Neanderthal and modern human dentitions are comparable in terms of total volume of enamel (e.g., Olejniczak et al., 2008), Neanderthals possess a more complex EDJ surface (Macchiarelli et al., 2006 and Toussaint et al., 2010; for the permanent molars, see Skinner et al., 2008) and greater dentine proportions (Bayle, 2008 and Bayle et al., 2009a), finally resulting in lower average (AET) and relative (RET) enamel thickness values (on historical ground, cf. Smith and Zilberman, 1994). Here again, the evidence from Tighenif fits the modern human condition. In fact, the differences in enamel thickness topography recorded on the Tighenif's m2 between the thicker protocone and the thinner paracone are in accordance with the functional pattern expected for the modern deciduous upper molars, likely related to the development of a helicoidal occlusal wear plane (Grine, 2005).

In sum up, present comparative morphological observations and 2-3D measures concerning the topographic repartition pattern and global distribution of the enamel (site-specific thickness and total volume), and the tooth organization as reflected by crown tissue proportions (notably, those between dentine-pulp and enamel), all indicate that, in relative terms, the structural signature virtually extracted from the ca. 700 kyr deciduous teeth from Tighenif closely resembles the modern human condition. While, at this preliminary stage, these results do not legitimate any conclusive statement on the taxonomic status of the Algerian sample because of the current lack of similar information on other Early-Middle Pleistocene samples (except for the developmental study of the dentition of the sapiens-like Middle Pleistocene child from Jebel Irhoud, Morocco; Smith et al., 2007), they, nonetheless, identify the likely primitive deciduous tooth pattern for both Neanderthals and anatomically modern humans.

Accordingly, based also on recent suggestions about the role of genetic drift in shaping Neanderthal morphology (e.g., Weaver, 2009), we predict that the inner structural morphology of the deciduous molars from Middle Pleistocene western European series, such as Tautavel, in France, or Atapuerca Sima de los Huesos (SH), in Spain, better fits the primitive, not the derived Neanderthal condition.