Our sample consists of 60 upper and 59 lower modern humans deciduous second molars (Table 1). Of these, 55 teeth, including 23 udm2s and 32 ldm2s, are from 39 Baka individuals. The Baka deciduous second molars were collected by Ramirez-Rozzi after the Baka children naturally shed and donated them with the consent of their families. The rest of our sample was composed of individuals from different geographical regions. The comparative sample includes 22 upper and 11 lower teeth from Europe, six upper and seven lower teeth from Africa, five upper and three lower from Asia, two upper and four lower from the Near East, and two upper and two lower from South America, which is quite a large sample considering the paucity of infant and juvenile specimens in osteological collections. More so, the meta-data for the comparative sample is often incomplete since the individuals come from archaeological collections (Table 1). Another limitation of the sample size are the stage of wear and state of preservation. To successfully obtain a complete EDJ surface the wear cannot exceed stage 452. Dental specimens showing a moderate abrasion of the horn tips (stages 3 and 4) were virtually reconstructed (see µCT acquisition and data segmentation below) before data collection. Specimens with wear stage greater than 4 or with decay in the occlusal area were either excluded from the sample or used only for the analyses of the crown and cervical outlines. The heavy wear, the poor state of preservation and occasional dental treatment made a large number of the Baka dental collection unusable for our study. The teeth were naturally shed and thus they were used approximately until the age of 11 which explain the advanced degree of wear. In some cases, the molars were kept by the Baka children or their families until the next visit of Ramirez-Rozzi, which likely led to the poor state of preservation of the teeth at the moment of image data acquisition (i.e. for the presence of numerous and deep cracks). Additionally, the Baka’s traditional dental treatments are quite invasive, and entail the drilling of large cavities later filled with natural substrates9.
µCT acquisition and data segmentation
The dental datasets were imaged at three different facilities. The Baka teeth were scanned mainly at the Plateforme Imageries du vivant, Université de Paris, using a Micro CT-scanner PerkinElmer, Quantum FX (voxel size 20 - 40/30 - 59/10 - 20 µm, 90 kV, 16 mA) and at the Hard Tissue Research Unit, College of Dentistry (NYU), with a SCANCO Scantron 40 Micro-CT scanner (voxel size 12 µm, 70 kV, 275 mA, 200 ms). The comparative sample was scanned at the Vienna Micro-CT Lab, Austria, with a VISCOM X8060 NDT scanner (voxel size 21 – 60 μm, 110 – 140 kV, 280 – 410 mA, 1400 – 2000 ms, 0.75 mm copper filter). X-ray images were taken from 1440 different angles. Using filtered back-projection in VISCOM XVR-CT 1.07 software, these data were reconstructed as 3D volumes with a color depth of 16,384 grey values.
The µCT data were then imported into Amira 2020 (www.fei.com) and virtually segmented to separate the enamel from the dentine, pulp and from the surrounding material (i.e., air, alveolar bone). In case of a slight abrasion of the dentinal horn tips, the specimens were virtually reconstructed by using the “brush” tool and extending the contours of the dentine into the empty area.
In case both left and right dm2s of one individual were available, we preferred the left ones among our whole sample. However, if better preserved, we used the right tooth after virtual mirroring to the left side. Since there is no scientific evidence indicating the existence of directional asymmetry in human dentition, we assume that the choice of left teeth should not affect the results of this study53–55.
Reorientation and outline collection
After segmentation, the surface models of the crowns were reoriented in Geomagic Design X 64 (www.3dsystems.com) following an established protocol 56–58. The crown and cervical outlines were collected from the reoriented surface models and projected onto the cervical plane. Afterwards, the outlines were further processed using Rhinoceros 6 (www.rhino3d.com) and were split into 24 segments by 24 equiangular radial vectors originating from the centroid of the outline area. At the intersections of the outline and the radii, twenty-four pseudo-landmarks were placed.
Landmark collection on the EDJ
To collection of the landmarks from the enamel-dentine junction (EDJ), followed established protocols 56–58. For the ldm2s, we placed eight landmarks (LM) on the five main horn tips and three at the deepest points between Metaconid and Entoconid, Protoconid and Hypoconid, and between Hypoconid and Hypoconulid. Afterwards, 23 curve semilandmarks (sLM) were placed to represent the EDJ marginal edge. To ensure homology, we traced the EDJ occlusal edge by creating a spline curve, ignoring all the accessory cusps. For the udm2s, seven LMs were placed on the four horn tips, and on the deepest points of the central fovea and distal fossa, and the deepest point of the disto-lingual marginal ridge. The EDJ marginal edge was resampled by 47 sLMs. The LM collection was carried out in the EVAN Toolbox 1.75 (www.evan-society.org), which uses the bending energy technique for the sliding of sLM 59–61.
The geometric morphometric analyses were performed separately for each set of landmarks, resulting in four different analyses per tooth type, namely: 1) cervical outlines; 2) crown outlines; 3) EDJ and 4) the combined dataset of EDJ and cervical outline. First, landmark configurations had to be normalized via General Procrustes Analysis (GPA)62,59,63. The translation and rotation of the landmark sets were not necessary for the outline configurations. We run the Principal component analysis (PCA) on the Procrustes shape coordinates and visualized the shape changes along the principal components by means of warpings, using the Thin-Plate Spline technique64,65. The analyses of the size were carried out using the natural logarithm of Centroid Size (lnCS). In the case of the combined EDJ and cervical outline, this is a measure of the 3D dentinal crown, including crown height, and therefore more representative than linear measurements. Furthermore, we performed a multivariate regression to analyze the influence of size on the shape of the dentinal crown. We explored shape covariation between udm2s and ldm2s as well as between different features of the same dental types by means of the 2-block partial least squares analysis (2B-PLS). Further statistical analyses, were carried out using R Studio (www.r-project.org), PAST 4.03 (www.softpedia.com) and SPSS (www.ibm.com). We performed the analysis of variance on Procrustes shape coordinates of larger subsamples, i.e. the Baka and the Europeans, to assess their degree of variation in the morphological expression of udm2s and ldm2s. Mann-Whitney test was used to test significance in size differences between the Baka and the Europeans, and between the Baka and the rest of the sample. To assess the size differences on populational level across the whole sample, Kruskal-Wallis test was used. Prevalence of various non-metric traits in our sample was analyzed by Chí2 test.
Non-metric traits were evaluated based on the Arizona State University Dental Anthropology System (ASUDAS)22,66. This system was originally developed to identify traits on the outer enamel surface, but many of the non-metric traits are visible on the EDJ as well. However, because of the high degree of wear of the Baka sample and root resorption, we focused on four among the most informative dental traits reflecting human neutral genetic variation21,67:
The fourth cusp of the upper molars that forms a separate region of the occlusal aspect, the trigon. Its manifestation varies from grade 0 to 622. In this study we used dichotomous classes to represent the hypocone degree of expression: none/light (0-3) and moderate/heavy (4-6).
- Carabelli cusp
The Carabelli cusp is an accessory cusp occurring on the protocone (mesio-lingual cusp) of the upper molars. This trait has been used as a diagnostic trait for European populations40, however according to more recent studies, there is no manifestation difference between populations22. According to ASUDAS, the Carabelli cusp’s expression varies between grades 0 to 7, but for this study we used two categories: none/light (0-4) and moderate/heavy (5-7).
The entoconulid, or cusp 6, can be found in the distal area of the occlusal aspect of lower molars between the hypoconulid and the entoconid. The manifestation of the entoconulid can be expressed between grades from 0 to 5 that are divided in our study into none/light (0-2) and heavy (3-5).
The metaconulid, or cusp 7, can be found on the lingual aspect of the lower molars between the entoconid and metaconid. Its manifestation can be expressed with 6 grades of 0, 1, 1A, 2, 3 and 4. We dichotomized the scoring values to none/light (0 to 1A) and moderate/heavy (2 to 4).