1Prix de la SAP 2023 / SAP prize 2023

General introduction

2When dealing with human remains recovered from either archaeological or forensic contexts, the main task of the anthropologist is to establish the biological profile of the individuals concerned, of which their biological sex is a fundamental component. This is a crucial step as it can affect the assessment of age and stature (Olivier et al., 1978; Brooks and Suchey, 1990), and above all contributes to interpretations about the identity of the deceased. In the case of a funerary ensemble, sex estimations help to identify palaeodemographic patterns and to characterise the health and living conditions of past populations, as well as contributing to discussions about the funerary practices and spatial organization of a necropolis in the light of this biological data (Duday et al., 1990; Blaizot and Castex, 2005; Duday, 2009; Boucherie, 2020). In forensic cases, the purpose is quite different: sex estimations facilitate police investigations by narrowing down the victim’s identity possibilities (Christensen et al., 2014). These implications mean that employing a method that yields an erroneous sex estimation can produce a cascade of effects on the anthropological conclusions (Bruzek and Murail, 2006; Klales, 2020).

3Anthropologists can routinely apply easy-to-use, reproducible and reliable sex estimation methods that have been developed from the os coxae (Bruzek and Murail, 2006). Based either on morphoscopic criteria (Bruzek, 2002; Klales et al., 2012; Santos et al., 2019) or on measurements (Murail et al., 2005; Bruzek et al., 2017), these methods can achieve a classification accuracy above 95%. However, their reliability can be compromised if the os coxae is absent, highly fragmentary or incomplete (Waldron, 1987; Bello et al., 2002). For example, the probabilistic DSP method (Diagnose Sexuelle Probabiliste) requires at least four measurements on the hip bone to obtain a result (Murail et al., 2005; Bruzek et al., 2017), and the final accuracy can decrease depending on the available variables (Quatrehomme et al., 2017; Chapman et al., 2020). Such situations occur in archaeology, but also in forensic contexts, where human remains are often affected by drastic taphonomic alterations in the aftermath of natural disasters, homicides or terrorist attacks (Christensen et al., 2014).

4In cases when the os coxae is not present or not optimally preserved, anthropologists can either proceed to a secondary sexual diagnosis (Murail et al., 1999) or perform a primary sexual diagnosis using other skeletal elements such as the skull or long bones (Jantz and Ousley, 2005; Spradley and Jantz, 2011; Curate et al., 2017). However, neither of these approaches are exempt from limitations: the first is only suited to funerary assemblages and cannot be used when isolated human remains are discovered, which is frequently the case in forensic contexts; the second necessarily implies having complete bones. With cremated remains, the process of identifying their biological sex is even more challenging due to the recurrence of fragmentation, shrinkage, warping and modifications to the texture and relief of the bones (Thompson, 2002; Depierre, 2014). Their altered aspect, combined with the small number of bones recovered or the absence of diagnostic features, limit the application of standard sex estimation methods (Rodrigues et al., 2021). Moreover, methods based on an ossified os coxae cannot be applied to immature remains. Despite numerous efforts to explore the sexual dimorphism of various immature skeletal elements (Schutkowski, 1993; Molleson et al., 1998; Cardoso, 2008; Stull et al., 2017), the accuracies obtained are very often below the recommended 75% (Moore, 2013). To resolve such situations, genetic analyses are usually seen as the only option to identify the sex of immature individuals, but even so, they cannot be implemented systematically, especially on the scale of a funerary ensemble, due to budget and technical constraints linked to contamination and degradation phenomena (Raffone et al., 2020).

5Consequently, in such frequently encountered situations, i.e., fragmentary adult skeletons, burnt remains and immature remains, found either isolated or within an assemblage, no sexual diagnosis can be made and the biological sex remains unknown, thus limiting subsequent archaeo-anthropological or forensic interpretations.

Objectives of this doctoral project

6These observations point to a need to expand the range of available tools by designing an alternative sex estimation method which is suited to fragmentary remains, reproducible, reliable and based on a skeletal element that is both highly dimorphic and has a high rate of taphonomic survival.

7The cranial base, defined as the occipital and temporal bones including the bony labyrinth (the osseous moulding of the inner ear), presents encouraging features. Firstly, visual differences between males and females have been previously noted (Ascádi and Nemeskéri, 1970; Walker, 2008), although not yet thoroughly quantified. Secondly, contrary to the os coxae, this cranial region is composed of very resistant anatomical elements (i.e., basilar part, occipital condyles, petrous part of the temporal bone) that are likely to be less altered by taphonomic processes and more likely to be preserved even on fragmentary skeletons (Bello et al., 2002). Thus, the first objective of this doctoral research was to explore and quantify the sexual dimorphism of the cranial base and the bony labyrinth in an adult assemblage, applying a morphometric approach, in order to propose a new alternative tool for sexual diagnosis that can be rapidly and easily used for fragmentary skeletons when the os coxae is not usable, but the cranial base is entirely or partly preserved. This study focused on a Western-European assemblage, as previous studies have examined South American or Indian populations more thoroughly (De Paiva and Segre, 2003; Saini et al., 2012).

8Another feature of interest in the cranial base is that it is made up of skeletal elements that undergo pre-pubertal maturation (i.e., foramen magnum and bony labyrinth). Therefore, the second objective of this research was to investigate the sexual dimorphism of the cranial base and the bony labyrinth on immatures to assess their potential value for the sexual diagnosis of juveniles. This type of research has very rarely been done on the immature occipital and temporal region (Veroni et al., 2010) and is unprecedented for the immature bony labyrinth, even though this specific anatomical structure raises an important issue for biological identification. The bony labyrinth is assumed to reach its mature size and shape in utero between the 18th and 31st foetal week (Jeffery and Spoor, 2004; Richard et al., 2010), which implies that it could be used to develop an age-independent sex estimation method applicable to both immatures and adults. This would avoid the need for anthropologists to fall back on destructive and costly genetic analyses to assess the sex of immature remains.

9Furthermore, considering the anatomical location of the bony labyrinth, which is naturally hidden within the petrous part of the temporal bone, its added value for sexing burnt human remains should not be underestimated (Norén et al., 2005; Masotti et al., 2013; Uhl, 2016). Our research explores the effect of the burning process on the morphology of this anatomical structure in order to assess its usability for the sexual diagnosis of cremated bones.

10Finally, we consider that virtual acquisition of the identified crania under study with a surface scanner is appropriate for two reasons:

11- to assess the consistency between measurements made on dry bones and virtual measurements made on 3D cranial surfaces as well as the applicability of established predictive models in a virtual environment;
- to quantify the sexual dimorphism of the occipital and temporal regions in adult individuals, in terms of both size and shape, through a geometric morphometric approach.

Study section 1: Investigating the sexual dimorphism of the cranial base in immatures and adults using a traditional osteometric approach

Materials and methods

12With a view to designing a new sexual diagnosis tool suited to fragmented remains, a large skeletal assemblage composed of individuals of known sex was compiled. Seven osteological collections curated in four different Western European countries (Belgium, France, Switzerland and Portugal) were used. Altogether, 611 skulls were examined, including 537 adults (>20 years) (261 females, 276 males) and 74 immatures (<20 years) (39 females, 35 males) (Table S1). Pathological individuals were excluded. These identified osteological assemblages were established either by museums or universities after the partial or complete dismantling of 19th-century cemeteries (the Simon, Schoten, Châtelet, Lisbon and Coimbra collections) or by medical institutions following prior autopsies, or under current body donation programs (the Olivier and Nice collections). Most of the individuals studied lived during the 19th and 20th centuries (n=519) and 92 were alive in the 21st century. Sex is known for each individual whereas the exact age is known for 86% of them.

Measurements

13To reduce the effects of subjectivity induced by morphological scoring, we used an exclusively morphometric approach. After reviewing the literature, 73 measurements were selected to characterize the morphometry of the occipital and temporal bones and their structural elements (i.e., the basilar part, the foramen magnum, the occipital condyles and the mastoid process; see figure 1; Table S2). Due to the poor standardisation of the measurement terminology, especially for the mastoid process (Petaros et al., 2015), the measurements corresponding to all available definitions of length and width of this element were all included in the study. Measurements were recorded in millimetres using a sliding calliper or a metric tape. To assess repeatability, variables were collected a second or third time by the author, at intervals of two weeks to a month, from 238 individuals. Additionally, all the measurements were collected by three other observers from 62 individuals belonging to the Châtelet, Schoten and Nice collections. To assess the repeatability and reproducibility of the protocol, an intraclass correlation coefficient (ICC) was calculated for each variable. Measurements were excluded from the set when the ICC was below 0.70, the threshold above which moderate reliability is indicated (Koo and Li, 2016).

Statistical treatment

14This study is not the first to examine the sexual dimorphism of the occipital and temporal bones. However, some shortcomings were noted among these previous studies:

15- sexual dimorphism of the occipital and the temporal bones was investigated separately: very few predictive models combine both measurements (Holland, 1986; Gapert, 2009; Macaluso, 2011, Say-Liang-Fat, 2014);
- statistical processing of the metric data used to construct the predictive models was not systematically clear and rigorous (e.g., absence of intra- and inter-observer error tests; no consistency in univariate or multivariate analysis, or in the use of cross-validation and independent validation samples; infrequent use of a decision threshold other than 0.50).

16Consequently, the statistical analyses for this study were carried out with caution and following a strict protocol to ensure the highest reliability of results.

17They were performed separately for the immature and adult assemblages. For the immatures, the sexual dimorphism of the occipital and temporal regions was examined for two age groups: pre-pubertal individuals (between 6 and 13 years old, n=12) and pubertal individuals (between 14 and 20 years old, n=40).

19Predictive sexual diagnosis models were then developed in R (version 3.4) using logistics regression and a stepwise approach. To create equations for both immature and adult assemblages, we applied the following procedure: firstly, models were developed and cross-validated with a training sample; secondly, the models were validated with a totally independent sample (namely the validation/test sample) made up of individuals from the Coimbra collection. Multivariate models were created combining occipital and temporal measurements, but also considering only occipital variables or temporal variables, or various combinations of these (e.g., measurements of the right part of the occipital bone with measurements of the left temporal bone). The aim was to devise models that can be adapted to different states of preservation of the cranial base (either complete or fragmentary). The number of measurements included in each model had to remain reasonable (<15) to ensure ease of use.

20For each model, accuracy percentages, error rates and sex bias were evaluated for the training sample after leave-one-out cross-validation (LOOCV), and for the test sample by applying four different thresholds to the posterior probabilities: 0.50 – as commonly used in anthropology; 0.70; 0.80 and 0.95 as used for DSP (Murail et al., 2005; Bruzek et al., 2017). The three latter thresholds allow the sex estimation method to include the possibility of indetermination (Santos et al., 2019; Galeta and Bruzek, 2020).

21The applicability and reliability of the models were assessed in fine with an application sample (n=29) made up of complete or fragmented crania from two Belgian archaeological collections (Coxyde site - 12th-15th century; Koekelberg site - 19th century) and one forensic context (Bois du Cazier identification mission, see Polet et al., 2024) (n=29) for which sex was initially known either by DSP or genetic analyses.

Results and discussion

22For the immature assemblage, the results were unfortunately inconclusive as the samples by age groups was too small and many data were missing. This underlines the fact that identified immatures are very rare in osteological collections and that, if present, they can be highly fragmented. However, we observed that:

23- only two measurements were significantly different between girls and boys among pre-pubertal individuals (p<0.05), males having larger measurements than females;
- 14 measurements, especially around the mastoid process, were significantly different between girls and boys among pubertal individuals (p<0.05), males having larger measurements than females.

24An attempt was made to create predictive models, but they proved unusable due to over-fitting.

25In the adult sample, each measurement, except the foramen magnum index, was found to be significantly different between the sexes, males having larger variables than females (p<0.05). Similar observations had previously been made in other European populations (Kemkes and Göbel, 2006; Gapert et al., 2009a; 2009b; Macaluso, 2011; Say-Liang-Fat, 2014; Amores-Ampuero, 2017; Chovalopoulou and Bertsatos, 2017).

26Thirteen predictive sexual diagnosis models were established (table 1). AD1 is a model suited to a complete cranial base (where both occipital and temporal bones are preserved), while AD2 to AD4 are exclusively composed of occipital variables, AD5 to AD7 of temporal variables, and models AD8 to AD13 are suited to different states of preservation of the cranial base. The measurements included in each predictive model are shown in Table S3. The 4 best performing models can reach 85.1% to 86.8% of correct classification after cross-validation in the training sample, and 85.1% to 92.5% accuracy in the validation sample, with a decision threshold of 0.70 (table 1). This threshold was found to be the best compromise to reach the highest percentage of accuracy, the lowest error rate (<15%) and the smallest proportion of indetermination (<33%). With 0.80 and 0.95 thresholds, accuracies were high, but the number of indeterminate individuals was too large (>65%). The accuracies obtained in this study, even with a 0.50 threshold, outperformed those obtained in previous research that combined occipital and temporal measurements (Say-Liang-Fat, 2014) or considered the occipital (Gapert et al., 2009a; 2009b; Macaluso, 2011; Chovalopoulou and Bertsatos, 2017) and temporal bones separately (Kemkes and Göbel, 2006; Bernard and Moore-Jansen, 2009; Jaja et al., 2013). The models developed with temporal measurements only (AD5, AD6 and AD7) achieved higher accuracies than those established with the occipital bone (AD2, AD3 and AD4). The lower degree of sexual dimorphism found on the occipital bone has been previously underlined (Gapert, 2009; Macaluso, 2011), and explained by its early maturation required to complete the neural network, and by its functional role as an architectural interface between the neurocranium and the face (Humphrey, 1998; Liebermann et al., 2000). When applied to complete or very fragmented archaeological and forensic specimens, the best performing predictive models gave good correct classification results (>80%), with the 0.70 threshold.

Study section 2: Quantification of sexual dimorphism of the cranial base in a virtual environment

Materials and methods

27In parallel with the conventional osteometric approach, a subsample of 100 crania from the overall corpus, belonging to 9 subadults and 91 adults, was digitized with a surface scanner NextEngine 3D Scanner© (Table S4). Among these, 50 were females and 50 males. Most of them lived during the 19th and 20th centuries although 16 of them were alive in the 21st century. To obtain a complete 3D model of each skull, two 360° acquisition positions were used: one in norma facialis, one in norma basilaris. To be sure of having a complete image of the occipital bone, two additional single scans were acquired in norma basilaris. The scans were then processed for trimming, alignment and fusion with Scanstudio©.

Results and discussion

Dry versus virtual measurements of the cranial base

31When the predictive models (AD1 to AD13) were applied to the 91 adults in our set, using virtual measurements of their cranial base, the accuracy percentages ranged from 82% to 95%, with a 0.70 threshold. Final sex attributions for both approaches showed, for each model used, substantial agreement with a Cohen’s kappa above 0.70 (Landis and Koch, 1977).

Geometric morphometric analysis of sexual dimorphism of the adult cranial base

32Centroid size was significantly different between males and females for each module considered, males having a systematically larger size than females (p<0.01). This result echoes previous studies that found cranial morphology to be largely influenced by size (Rosas and Bastir, 2002; Franklin et al., 2006; Kimmerle et al., 2008), and made similar observations on both occipital and temporal bones (Guyomarc’h and Bruzek, 2010; Chovalopoulou et al., 2013; Milella et al., 2021). Procrustes distances were only significantly different between sexes when looking at the right temporal module (p=0.03 versus p=0.15 on the left temporal module). For this module, polarization between groups occurred significantly along PC3 (13.66% of variation) and PC4 (9.08% of variation), even though a substantial overlap was still visible (figure 3). Surface warps illustrating variations along these two PCs showed that males, with negative values for both PCs, had a longer, wider (medio-laterally), and more anteriorly and posteriorly projected mastoid process than females. With positive values for both PCs, the process in females is smaller and more obliquely orientated relative to the Frankfurt plane (figure 4). Even though these computed shape consensuses are forced extrapolations, they allow the subtle differences existing between sexes in the cranial base to be discerned in a replicable manner and across all spectra of variations.

33Significant allometry results were found for both individual modules of the right and left temporal bones (p<0.001), suggesting that the shape differences observed are largely determined by size. When the allometric component was removed, no significant shape differences were found between males and females (p=0.98 for the right temporal module; p=0.93 for the left temporal module).

34The lack of shape differences found between sexes in the occipital bone agrees with other studies (Guyomarc’h and Bruzek, 2010; Milella et al., 2021). It can be linked to stronger genetic developmental control of this specific area, where phenotypic variations need to be somewhat reduced to ensure that the functionalities of the neural system are maintained (Humphrey, 1998). However, we cannot exclude that sexual dimorphic differences are smaller in this study due to the older age of the females in our sample, whose skulls, according to some authors (LeBoff and Glowacki, 1999), tend to become masculinized with the onset of menopause.

35The demonstration of significant allometry on the temporal bones, associated with the absence of shape differences when this allometric component is removed, indicates that the variations found on the right temporal module are largely driven by size dimorphism, males having larger crania than females. As illustrated by regression graphs (Boucherie et al., 2022a, figure 8), this size-driven sexual dimorphism may be accentuated in our sample by the presence of a few particularly large males.

36As seen in figure 4, size entails variations on the right temporal bone specifically around the mastoid area. This finding is consistent with the observations of Lahr and Wright (1996) and Rosas and Bastir (2002), who revealed a significant association between the development of bony superstructures and cranial size.

37This size dimorphism found on the cranial base needs to be further understood in the light of functional, biomechanical and ontogenetic considerations. Distinct body composition and energy requirements between females and males (Rosas and Bastir, 2002; Franklin et al., 2006) can explain why biomechanical forces impact differently cranial superstructures like the mastoid process, known to be a dense and resistant bony element to which numerous neck muscles are attached. The allometric component of sexual variations found on the temporal bones can also be linked with the different ontogenetic trajectories between sexes after the onset of puberty, when the growth period is considerably longer in males (Viðarsdóttir et al., 2002; Bulygina et al., 2006).

38Nonetheless, it is still important to consider that the patterns of shape variation found in our sample, which are largely influenced by size, are sample-specific and might turn out to be very different in other populations where size is less important, for example (Viðarsdóttir et al., 2002). Similarly, if the tendency towards masculinization of female skulls with the onset of menopause is confirmed (LeBoff and Glowacki, 1999), the mean older age of the females in our sample could explain the lack of sexual differences found in both the overall cranial base module and the occipital module.

Study section 3: Exploring sexual dimorphism of the bony labyrinth in immatures and adults

Materials and methods

39Sexual dimorphism of the bony labyrinth was examined in 47 immatures and 74 adults (Table S5). Due to its naturally hidden location within the petrous part of the temporal bone, the morphology of this structure can only be studied by tomodensitometry (figure 5). As it was not possible to CT-scan every skull of the overall sample, a study sample made up of images of different types (dry crania and crania with soft tissues) was compiled from:

41Only the right side of the structure was observed as its asymmetry has not been clearly demonstrated (Spoor and Zonneveld, 1995; Marcus et al., 2013; Osipov et al., 2013).

42The CT-scans at the Erasme hospital were done with a Siemens Force (384 RAM, spatial resolution of 0.24 mm) using a slice thickness of 0.4 mm every 0.1 mm. These parameters were used to obtain windowing specifically on the petrous part of the temporal bone, as routinely done in clinics for inner ear reconstructions on living patients. At KULeuven, the images were acquired with full body scanning, without specific windowing on the inner ear, with a Siemens Definition Flash and a slice thickness of 1 mm every 0.9 mm. To ensure better visualization of the structure and to comply with confidentiality and anonymity requirements, the CT-scans were framed directly on the bony labyrinth. For more details on the acquisition parameters see Boucherie et al., 2021.

Measurements

43Each right bony labyrinth was reconstructed in 3D by manual segmentation with Avizo 7.0©, using a 550–650 Hounsfield unit to separate the bone structures from air and any soft tissue present. In accordance with the Spoor and Zonneveld (1995) protocol, these reconstructions were used to reformat the 2D slices in a transverse and sagittal plane where measurements could be accurately recorded using ImageJ v.1.8.

44To characterize each structural element of the bony labyrinth (i.e., the cochlea, the vestibule and the three semicircular canals - SCC), 43 measurements were selected from the literature, 28 linear and 15 angular (see Table S1 of Boucherie et al., 2021 for the full list). Thirty-seven of these measurements were previously defined by Spoor and Zonneveld (1995), the relative radius of curvature of the 3 SCC by Spoor et al. (2003) and the two variables of the vestibule by Marcus et al. (2013), whereas measurement of the vestibular index is unprecedented. Linear measurements are variables of height, width and radius of curvature whereas angular measurements concern the torsion angles of each canal and the angles between the labyrinthine elements (figure 6).

Statistical treatment

46The normality and homoscedasticity of the data distribution were checked, and non-parametric tests were then used. Firstly, differences between bony labyrinth morphology in immatures and adults were investigated with Mann-Whitney U-tests to construct predictive models of sexual diagnosis for a structure expected to reach maturity in utero.

47As significant differences were found for six variables (p<0.05), sexual dimorphism of the bony labyrinth was subsequently examined separately in immatures and adults. Mann-Whitney U-tests were used to compare female and male measurements, with a p-value threshold of 0.05. Two types of predictive analyses were then carried out with R:

48- univariate and multivariate linear discriminant functions were established based on the set of significantly dimorphic variables in immatures and adults;
- univariate and multivariate logistic regression, with a direct and stepwise approach, were used for immatures and adults.

Results and discussion

51Discriminant functions established for the immature and adult samples have already been published (Boucherie et al., 2021). The results obtained through logistic regression are detailed below.

52In the immature sample, nine measurements were found to differ significantly between boys and girls (p<0.05), including six linear measurements (TLIa, Vestibl, ASCw, PSCh, LSCh, LSCw), two ratios (PSC-R and LSC-R) and one angular measurement (VSC<LSCt; see Table S4 in Boucherie et al., 2021). The values for girls were smaller than for boys. Eight predictive sexual diagnosis models were then constructed (table 2). Comprising two to seven variables, the four best performing models (LO-SUB4, 5, 7 and 8) achieved 72.7-84.2% accuracy in the training sample and 72.7-80% in the validation sample. Again, the 0.70 threshold was the best compromise to attain the highest rate of accuracy (>75%), the lowest error rate (<27%) and the smallest proportion of indeterminates (<30%). This study is the first to design predictive models using the immature bony labyrinth, and the results are promising. Uhl (2016) had previously tested discriminant functions on 23 immatures from the identified Dart collection, where she found 52.2% of accuracy, but the equations used had initially been established from adult individuals.

53In the adult sample, significant differences between the sexes were demonstrated for five measurements (p < 0.05): two linear (TLIp and PSCh), two ratios (Vestibl/w and LSCh/w) and one angular measurement (ASCs<PSCs; see Table S5 in Boucherie et al., 2021). The values for females were lower than for males. Seven predictive models, made up from two to five measurements, were established (table 3). The three best models (LO-AD 4, 6 and 7) achieved 80.5 to 82.9% of correct classification, using a 0.70 threshold. Even though comparisons are somewhat restricted because of differences in measurement sets, acquisition protocols and statistical parameters, these results are fairly similar to those found by Osipov et al. (2013) for an identified Cretan population.

54However, accuracies for the validation sample were significantly lower, at 50-55%, which is too close to equal probability to be reliable. There are two possible reasons for these results: 1) a time-related difference between individuals in the training sample who lived during the 19th century and those in the validation sample who were alive in the 21st century; 2) the different nature of the images, i.e., framed CT-scans made on dry crania versus full body CT-scans made on crania with soft tissues and their associated lower resolution, which could increase measurement discrepancies. Therefore, we do not recommend using these models at present, because the influence of means of acquisition on such a small anatomical structure as the bony labyrinth needs to be further explored.

55Despite these limitations, an attempt was made to use these predictive models for cremated adult remains of known sex (Boucherie et al., 2022b). When applied to a sample of six petrous bones before calcination (with a threshold of 0.70), most of the models classified males and females accurately. However, on calcined bones, they produced a systematic misclassification of male individuals. This echoes previous observations made by researchers on other burnt skeletal elements such as the humerus, femur, talus and calcaneus (Gonçalves et al., 2013; Gonçalves et al., 2020; Rodrigues et al., 2021). Despite its naturally protected position within the petrous part of the temporal bone, the bony labyrinth structure can be affected by fire, at least when calcination is reached, producing internal cracking (figure 7). This aspect was also underlined in this study when directly comparing unburnt and burnt bony labyrinths: a high percentage of shrinkage for some measurements was found, and a significant difference in centroid size was seen only in the male group (Boucherie et al., 2022b). This is consistent with observations made by earlier studies that highlighted heat-induced changes in burnt human bones (Gonçalves et al., 2013; de Oliveira Coelho, 2015).

General discussion and conclusions

56This project offers significant advances in understanding human sexual dimorphism of the cranial base within a Western European metapopulation. By combining occipital and temporal measurements for the first time, it provides a new tool of sexual diagnosis characterized by advantages (i.e., ease of use, low cost, reproducibility, reliability) that make them suitable for analyses of fragmentary adult remains in archaeo-anthropological and forensic contexts. The models developed on the adult cranial base (AD1-AD13) can be used as from now by anthropologists wanting to assess the biological sex of a skeleton in which the cranial base is preserved but the os coxae is unusable. Their classification accuracies (85.1-86.8%, with a 0.70 decision threshold) certainly cannot compete with those obtained by DSP on the os coxae (95% of reliability with a threshold of 0.95, an error rate and sex bias below 5% and 0.8%). However, this classification performance appears to be an appropriate alternative tool for primary sexual diagnosis. This study also contributes to give a detailed list of reproducible measurements of the cranial base that can be used with confidence when applying a secondary sexual diagnosis to a funerary ensemble. A free R package is to be released in the next few months to facilitate the use of predictive models within the scientific community. The package will include a tab for definitions of the variables and a tab for analysing a target individual where measurements of the cranial base can be entered and the decision threshold chosen. We recommend using it only for Western European adults and with a threshold of at least 0.70. For the immature sample, given the inconclusive results, further investigations with a larger sample of identified immatures are still needed to explore the sexual dimorphism of the occipital and temporal bones in more depth, and its ontogeny throughout the pubertal peak. The difficulties encountered here (i.e., small sample, numerous missing data) nevertheless had the advantage of again underlining the necessity, for such projects, of continuously including infants in identified osteological collections.

57Furthermore, the digital cranial acquisition performed during this project provided an opportunity to improve our knowledge of the modes of expression, in size and shape, of sexual dimorphism of the adult cranial base. Through geometric morphometrics, the major influence of size on cranial morphology was confirmed and the functional and ontogenetic particularities of both occipital and temporal bones were disentangled (Boucherie et al., 2022a). In parallel, this study showed that virtual acquisition of cranial base measurements, using a NextEngine surface scanner and lhpFusionbox software, entails slight discrepancies compared to the conventional osteometric approach, suggesting that each 3D cranial surface model is consistent with the original dry cranium. The predictive models established for sexing the adult cranial base were also found to be effective on digitized crania. These results illustrate the advantages of a more frequent use of virtual tools, such as surface scanners or photogrammetry, in both archaeo-anthropology and forensic anthropology: to avoid over-manipulation of extremely fragile specimens, to assess the biological profile of a specimen remotely and collegially and to build up digital archives of numerous identified individuals. On this latter point, the digital cranial models acquired in this project on individuals of known sex and age are available on request for any other investigations (Table S2).

58This unprecedented study investigating the sexual dimorphism of the bony labyrinth in both immatures and adults tends to indicate that a sexing method based on this element cannot be age-independent (Boucherie et al., 2021). Differences found between immature and adult labyrinthine morphology suggest that ontogeny is not strictly complete in utero, awaiting further evidence from a larger sample. Although the predictive sexual diagnosis models that were developed for immatures and adults gave promising results, the differences in accuracy found between the training and test samples, especially for the adults, suggest that additional validations using larger and more homogenous CT-scan assemblages, especially in terms of voxel size, are needed before they can be applied with confidence. The potential added value of the bony labyrinth in sexing cremated remains was also rejected because of the observed systematic misclassification of adult males linked to shrinkage phenomena (Boucherie et al., 2022b). Given these results, other options seem more appropriate for cremated remains: establishing predictive models adapted to burnt bony labyrinths; exploring metric bone reconstruction based on chemometric data obtained from infrared spectra (Gonzalez et al., 2020); and testing proteomics, ideally on carbonized remains (Buonasera et al., 2021).

59Finally, thanks to the implementation of rigorous statistical processing, this study contributes to the discussion on the fundamental requirements that need to be met in anthropology when developing a new sexing method. It underlines the necessity of setting up a standardized and rigid analytical framework that would be expected for any future development of a sexual diagnosis method. At the European scale, where efforts are under way in forensics to implement accreditation and standards, this point is becoming crucial. By providing an alternative reliable sex estimation method, this doctoral study contributes to the improvement of identification tools used to analyse past populations and contemporary missing persons. Anthropologists will now be able to perform a sexual diagnosis on more individuals: fragmentary and damaged skeletons will no longer be left aside. This means that more biological data will be collected, allowing a better and more precise understanding of living conditions, gender and funerary practices in past populations (Boucherie, 2020). In the current forensic context where the number of missing persons is rising, especially due to armed conflicts and migration crises, the implementation of a sexing method that is easy, rapid, and cost-free to use may also prove to be of considerable value.

Supplementary information

60https://journals.openedition.org/​bmsap/​14372?file=1

Acknowledgements: I would first like to express my gratitude to the Société d’Anthropologie de Paris that did me the honour of awarding me its 2023 Ph.D. prize. My warmest thanks to my thesis directors, Dr. M. Vercauteren and Dr. C. Polet, and also to Dr. Ph. Lefèvre, for their steadfast support. This anthropological research on methodological aspects would not have been possible without examining identified osteological collections. I thank the different curators for granting me access to the material under their care: Dr. P. Semal (Royal Belgian Institute of Natural Sciences - Châtelet & Schoten collections); Pr. M. Besse and Dr. J. Desideri (University of Geneva - Simon collection); Dr. S. Wasterlain (University of Coimbra - Coimbra collection); Dr. S. Garcia (MUNHAC - Lisbon collection); Dr. M. Friess (Muséum National d’Histoire Naturelle, Paris - Olivier collection); Pr. G. Quatrehomme (Côte d’Azur University - Nice collection). For CT-scan acquisitions and images, I am thankful to Pr. P. Scillia, Pr. N. Sadeghi and Ch. Van Humbeeck (Department of Radiology, Erasme Hospital, Brussels) and to Pr. W. Van de Voorde, Dr. J. Wuestenbergs and Dr. W. Coudyzer (Department of Forensic Medicine of the Katholieke Universiteit Leuven). This doctoral research was supported by a teaching and research assistant contract within the CReA-Patrimoine of the Université libre de Bruxelles and by short mission grants provided by the Fonds de la Recherche Scientifique (FNRS) and the Fédération Wallonie-Bruxelles (FWB). I am also grateful to Dr. Ph. Collart (BIOPS, ULB), Ir. Rudy Ercek (LISA, ULB), Dr. T. Chapman (RBINS & ULB), Dr. D. García Martínez (Computense University), A. Thibeault, Dr. L. Nogueira (Côte d’Azur University), Dr. J. Dumoncel (University of Poitiers) and Dr. K. Salesse (Masaryk University) for their valuable help with this research. I thank the anonymous reviewers for their comments that contributed to the improvement of the paper.