Prehistory and Protohistory in Sicily. A Geometric Morphometrics Approach to Study the Biological History of Early Human Peopling of the Island

Gabriele Lauria

https://orcid.org/0000-0002-0122-7093
Department of Biological, Chemical and Pharmaceutical Sciences and Technologies (STEBICEF), University of Palermo, Via Archirafi 18, 90123 Palermo, Italy

Luca Sineo

https://orcid.org/0000-0001-8634-2295
Department of Biological, Chemical and Pharmaceutical Sciences and Technologies (STEBICEF), University of Palermo, Via Archirafi 18, 90123 Palermo, Italy

Introduction

The biological history of the human peopling of Sicily presents many interesting aspects. Its proximity to both the Italian Peninsula and North Africa and its unique physical geography have made the island a crossing point for several living species (e.g., Leithia melitensis, Aquila fasciata, Chamaerops humilis, Pinus halepensis etc.) (Bosch 2010; Massa 2011; Surdi 2011, Pignatti 2011). Sicily is the largest island in the Mediterranean Sea and is itself divided into a mosaic of habitats (due to coastlines, sea valleys, small cliffs, hills, and mountain ranges), each with its own microclimate (Ruggieri 1973; Massa et al. 2011). At the same time, it is isolated biogeographically and is characterized by numerous endemic species of flora and fauna (Pignatti 2011). Environmental adaptation and/or natural selection (climatic factors, bottleneck effect, genetic drift, gene flow, or random evolutionary processes caused by neutrality) have produced profound coevolutionary changes in all species in the region (Sineo et al. 2015). Homo sapiens was no exception and even before the intense cultural and biological contributions left by the Greek, Carthaginian, Roman, Byzantine, Islamic, and Norman/Swabian colonizers (Lauria et al. 2023), early Sicilians were subjected to and strongly influenced by the environmental conditions. These conditions quickly and significantly influenced the island’s gene pool and the human phenotype as early as Prehistory (D’Amore et al. 2009, 2010; Galland et al. 2019).

In recent years, the debate about the early human peopling of Sicily (and the related dispersal patterns) has almost exclusively focused on the archaeological evidence. The possibility for and degree of admixture caused by ancient migratory flows have been only investigated in limited anthropological studies conducted on a short time spans (D’Amore et al. 2009, 2010; Galland et al. 2019). The archaeological based settlement hypotheses assume an early colonization of Sicily’s northern coast during the Palaeolithic, initially driven by economic needs and the related subsistence strategies (Beretta et al. 1986; Beloch 1989; Di Salvo et al. 2008; Barucco et al. 2021). This was followed by a new colonization and partial substitution during the Mesolithic (Galland et al., 2019). The island was then significantly impacted by Neolithic and post-Neolithic demic movements (Fernandes et al. 2020) that determined the colonization of the hinterland and of the southeast coast during protohistoric periods. These movements were driven by cultural and social factors (Serratti 2000; Di Salvo et al. 2008) such as the foreign invasion of the island which was carried out by several populations (Siculi, Ausoni, Morgeti and Elimi – Holloway 2002).

Recent craniofacial morphometrics analyses (Lauria and Sineo 2023) that considered migratory flows and population influx have provided a more comprehensive approach to understanding settlement patterns. These analyses go beyond archaeologically based settlement hypotheses by merging previous archaeological evaluations and paleoclimatic studies with an anthropological approach.

This study expands the sample upon earlier morphometric work (D’Amore et al. 2009, 2010; Galland et al. 2019; Lauria and Sineo 2023) and provides an overview focusing the period from the Upper Palaeolithic to the Iron Age (14.500 B.C.E. – 900-800 B.C.E.). To reconstruct the biological history of the early human settlement in Sicily we analysed the shape variation (polarity and magnitude) of facial features in order to associate the changes in cranial morphology with population influx. Hypotheses were formulated by performing a Procrustes coordinates PCA to understand the separation between the specimens and the groups reasonably. Instead, PCA is an exploratory analysis we use its similarity with the discriminant analyses to draw our conclusions.

Materials and methods

For this study, human skulls from Sicilian Prehistory and Protohistory were considered. Approval to examine the skulls was issued by Department of Sicilian Cultural Heritage and the Gemmellaro Archaeological Museum. These skulls were divided into six periods (Table 1) based on the 14C cal B.P for SanTeodoro 1 (Sineo et al. 2002), Molara (Leighton 1999) and Oriente (Modi et al. 2022) and based on archaeological artifacts (industry and pottery) for SanTeodoro 2, Uzzo, Marcita and Polizzello (Leighton 1999) (Table 2). Sample selection was based on a detailed bibliographic review of previously published archaeological and historical works as well as previous anthropological and palaeoecological studies (Table 3). Additionally, the A. Salinas, Baglio Anselmi, and L. Bernabo’ Brea’ museum catalogues were consulted. A preliminary visual examination of the skeletal remains was carried out to assess their morphological completeness and to determine if they were suitable for inclusion in this study. To avoid any error related to approximation, broken, incomplete (skulls not in anatomical connection and/or lacking landmarks necessary for the GM study discussed below) and restored skulls were a priori excluded and not considered for the study.

In total, 16 adult (Ubeleker 1989; Scheuer and Black 2000; Buikstra and Ubelaker 1994) human skulls (Table 4) from eight different settlements were selected (Figure 1, Table 2). The spatial bias in the map (Figure 1) is mainly due to a lack of specimens from the southeastern part of the island, which saw scarce colonization (lack of settlements) of Homo before the historical periods. Despite the small number of specimens and settlements, the sample effectively represents the inhabited regions of Sicily before the historical conization.

With the aim of associating the diachronic changes in cranial morphology (Relethford 2002, 2004a, b; Roseman 2004; Taubadel and Lycett 2008; von Cramon-Taubadel 2008; 2009a; 2009b; 2009c; 2014; Betti et al. 2009; Smith 2009; 2011; von Cramon-Taubadel and Weaver 2009; Matsumura et al. 2018; Manthey and Ousley 2020; Klingenberg 2022) with population migrations, we performed a 3D geometric morphometrics (GM) comparative analysis. This study followed the approaches proposed by several authors (Bruner and Manzi 2004; Bruner 2007; D’Amore et al. 2009, 2010; Baab et al. 2010; Smith 2011; Fredline et al. 2012; Matusmura et al. 2018, 2022; Galland et al. 2019; Hubbe et al. 2020; Lauria and Sineo 2023; von Cramon-Taubadel and Lycett, von Cramon-Taubadel 2008, 2009a, 2011, 2014, 2017; Grine 2023a, 2023b; Gunz and Freidline 2023, Ribot et al 2023). Sex was estimated by evaluating the skulls’ morphological characters according to Acsàdi-Nemeskèri (1970) as reported in Walrath (2004) and Minozzi and Canci (2015). The 3D models from Grotta di SanTeodoro, Grotta della Molara, and Grotta D’Oriente were acquired using computed tomography (CT) while the specimens from Grotta dell’Uzzo, Ragusa, Partanna, Marcita, and Polizzello were obtained through photogrammetric reconstruction following the protocol proposed by Lauria et al. (2022). The method consists of a Structure-from-Motion (SfM) photogrammetry that uses a single camera to capture chromatic details and reconstruct shape. A series of photos were taken forming circles around the target that will completely be covered and reconstructed by allying the photos and building a cloud of point before the polygonization of the model in the end.

Photogrammetric and CT models were scaled with the software used to build the models (Metashape 1.5.1 for photogrammetry and Slicer 5.2.2 for CT) and after exported in PLY format. The GM analyses were based on a configuration of 26 Landmarks (Buikstra and Ubelaker 1994) (Figure 2) positioned on the suture boundaries (Landmarks Type 1) and on the anthropometric points (Landmarks Type 2) (Bookstein 1991). Using the software “Landmark3.6” (Wiley et al. 2007), 78 Raw Coordinates (RCs) were acquired for each specimen and were then analysed using the software programs “MorphoJ 2.0” (Klingenberg 2011) and “PAST 2.0” (Hammer and Harper 2001).

MorphoJ was employed to subject the RCs (exported from “Landmark”) to a Generalized Procrustes Analysis (GPA) (Dryden and Mardia 2016) to remove the effects of translation and rotation and to standardize each specimen to unit centroid size (Gower and Payne 1975, Rohlf and Slice 1990, Goodall 1991). The resulting Procrustes Fitted Coordinates (PFCs) were visualized using shape change graphs, including Lollipop and Wireframe Graphs, which allow three-dimensional forms to be visualized in two dimensions (Hammer and Harper 2008; Klingenberg 2013). These graphs illustrated the shape changes from a starting shape (the mean shape in the sample) to the target shape (the most extreme of the specimens) (Klingenberg 2013). These graphs were used to capture shape variation (direction and magnitude) across time (Harvati et. 2007, 2010, Bruner and Ripani 2008, Baab et al. 2010, Galland et al. 2016, 2019, Lauria and Sineo 2023). To highlight the positions of the specimens and groups within the sample (Hammer and Harper 2008), the RCs (exported from MorphoJ) were procrustized again in PAST to perform a Principal Component Analysis (PCA) based on a Covariance Matrix (represents the change of each variable relative to the others, including itself). The PCA separately analysed the PFCs of the single specimens, the average PFCs of each site, and the average PFCs of period.

Always considering that PCA is an exploratory analysis that allows for the formulation of a hypothesis based on the visualization (Le Maître and Mitteroecker 2019) and that discriminant analyses were not applicable (MANOVA and CVA are possible only when the variables are less than the specimens, Bronstain et al. 2006, Hammer and Harper 2008), we used the similarity between PCA and other discriminant analyses (Lauria and Sineo 2023) to reasonably understand the separation between the specimens and the groups. PAST was also employed to create a cluster analysis (CA) (UPGMA cluster procedure and Euclidean distance matrix) (Saitou and Nei 1987), leaving the software to recognize the outgroup.

Finally, on the PFCs, were conducted two “one-way ANOVA tests”, the Levene’s test for homogeneity of variance (Levene 1960; Derrick et al. 2018) and the more robust Shapiro-Wilk test (Shapiro and Wilk 1965) to evaluate the null hypothesis (H0) of equal multivariate means between the groups.

Results

Eigenvalue and percentage of variance

The plots of the PFCs (Single Specimens: Figure 3a; Averages of Each Site: Figure 3b; Averages of Each Period: Figure 3c) show that only the first two Principal Components (PCs) are significant. In detail, for the single specimens, the % of variance is 38,317% and 26,252% and the eigenvalue stands at 37,4197 and 26,252%, respectively. For the average of each site the % of variance is 38,188% and 26,717% while the eigenvalue are 33,4632 and 23,4111. Finally considering the averages of each period the values are 43,531% and 24,462% for the % of variance and 28,2558 and 15,8780 for the eigenvalue. In general, both the eigenvalues and the percent variance decrease gradually, converging toward similar values (Figure 3d-f), with a slightly more marked trend between the first two PCs (Figure 3). This suggests greater variations between the Paleolithic and the Mesolithic components (hunter-gatherers of the Stone Age) compared to the farmers-shepherds of the Metal Ages.

Shape variation

Considering that sample size and sample composition (number of specimens and sexual dimorphism) can influence the analysis, the shape variations show that the most notable changes are located on the neurocranium and the lower face, while minor changes affected the upper face (landmarks 4 and 19–22), the nose, and the orbits (Figure 3g). The frontal bone generally increased in size, becoming more elongated and lower (Figure 3h-i), while the width remained almost unchanged (Figure 3l). The parietal bones slightly reduced in size, showing decreases in length, the height of the posterior portion (Figure 3i), and in the width of the superior part (Figure 3l). The nose became slightly elongated (Figure 3g), however its size and position remained almost unchanged (Figure 3i-l). The orbits showed minimal changes in size and position (Figure 3l), only moving slightly downward.

Principal component analyses and cluster analyses

The PCA performed on PFCs of the single specimens (Figure 4a-b) shows the two Upper Paleolithic specimens from San Teodoro (Incarbona et al. 2010a) lying exactly on the negative axes of PC1, separated from the other specimens. In contrast, all the Mesolithic hunter-gatherer specimens from Molara, Uzzo, and Oriente are located on the positive axes of PC1. The Uzzo specimens are positioned on the negative side of the PC2 axes and those from Molara and Oriente are on the positive side of the PC2 axes. Notably, the Oriente specimens appear separated from both the Molara ones and all other individuals. In general, the hunter-gatherer specimens from the Sicilian Stone Age occupy an inhomogeneous morphospace surrounding the specimens representing the farmers-shepherds of the Sicilian Metal Ages. The farmers-shepherds of Ragusa (Copper Age/Eneolithic), Marcita (Bronze Age), and Polizzello (Iron Age) are all grouped in a homogenous morphospace near the centre of PCs axes (Figure 4a-b).

Further PCA analyses, based on the averages of the PFCs (average of each site: Figure 4c-d; average of each period: Figure 5a), reinforce the previous findings; the Paleo-Mesolithic sites of San Teodoro, Molara, Uzzo, and Oriente each occupy different quadrants in the PCA, surrounding the Copper, Bronze, and Iron Age sites of Ragusa, Marcita, and Polizzello, respectively (Figure 4c-d). Among these last three sites (all positioned on the positive PC1 axis), it should be noted that the Iron Age site of Polizzello appears slightly distant from the sites of Ragusa and Marcita, which are close to each other but separated by the PC2 axis (Figure 4c-d). These results also show a significant distance between the Paleolithic and Mesolithic periods (both hunter-gatherers, separated by the PC1 axis; Figure 5a), which in turn are far from the Copper, Bronze, and Iron Ages (all farmers-shepherds), grouped together in the upper right quadrant (positive PC1 and PC2 axes) close to each other but each in their own position (Figure 5a). The CA results (Figure 5b), based on the averages of the PFCs for each period, confirm the PCA findings. San Teodoro is automatically recognized as the outgroup, while the Paleolithic group retains some affinities with the Mesolithic specimens, which cluster separately from both the Paleolithic and the more recent farmers-shepherds from the Metal Ages. Additionally, the Prehistoric Eneolithic/Copper Age and the Protohistoric Bronze Age are still clustered together and, in turn, separated from the Iron Age group, which is chronologically closest to the most recent historical periods.

One-way ANOVA

The Levene’s test for homogeneity of variance to evaluate the H0 returned in all the case p (same) value < 0.001 (in a significance level of α=0.05), in detail PFCs specimens p (same)=3,569E-31, PFCs sites p (same)=1,915E-10 and PFCs periods p (same)=9,223E-12 that rejects the H0. In the same way the Shapiro-Wilk test returns PFCs specimens p (same)=6,067E-19, PFCs sites p (same)=1,6E-12 and PFCs periods p (same)=1,016E-7 that also in this case rejects the H0.

Discussion

Overall, the gradual decrease in eigenvalues and the percent variance suggests a slow progressive increase in variability (Figure 3) caused by primitive low-density migrations. These migratory events of the Paleolithic and Mesolithic (Stone Age hunter-gatherers) exhibit spatial clustering distinct from the later Metal Age farmer-shepherds.

The shape variations across prehistory and protohistory (Figure 3g) show a general trend of dolicefalization of the neurocranium (becomes more elongated and narrower). In contrast, the lower face became more elongated but wider, while the nose and orbits largely maintained their size and position. This pattern suggests phenotypic and genotypic changes due to neutral forces such as environmental factors and the stochastic forces that generally evolve in a neutral manner (Smith 2011). These acted in parallel with the population influx caused by migratory flows from the content (Betti et al. 2009). The scatterplots generated by the PCAs (Figures 4–5) and by the CA (Figure 5) show a decrease in the biological distance from the Paleolithic to the Iron Age, with short distances between the Sicilian Metal Age groups. Specifically, the hunter-gatherer migratory flows of the Stone Ages (Paleolithic and Mesolithic) were always characterized by sporadic, low-density migrations that produced cyclical and discontinuous occupations of the island. From the Eneolithic/Copper Age onward, there was a slow increase in both frequency and density of migratory flows, continuing through the Bronze and Iron Ages. Although these demic migrations produced some population discontinuity, the concentric arrangement indicates a limited but constant degree of admixture among the mentioned groups. This implies no significant morphometric variation between Prehistory and Protohistory populations. Finally, these dynamics produced a not negligible allometry supported by the one-way ANOVA tests, that both returned p (same) values < 0.001 that rejects the H0.

Although some of our landmarks (such as glabella or inion) involves sexually dimorphic cranial areas males and females have not influenced the scatterplots. Indeed, no differences were observed between groups from all sites. Male and female individuals from the same site are often very close to each other (Galland et al. 2019). When evaluating the variation between human groups arriving to a localized geographical region (like an island), it is important to consider that the genetic pool is often stressed by genetic drift phenomena such as the bottleneck and founder effect (Manica et a. 2007). In addition to these stochastic forces, adaptive changes (such as the masticatory-inducted phenotype) are in parallel impacted by cultural variations with the same plasticity but with a slow degree of diversification (Harvati and Weaver 2006). In particular, patterns of the cranial vault and the upper face are evolving largely neutrally (Smith 2011). Nevertheless, the large differentiation of facial shapes during the centuries could not only be explained by adaptive changes but also by the arrival of new genetic components (Betti et al. 2009).

Conclusion

According to the paleoclimatic data (during the last glacial peak, Sicily was characterized by a steppe or semi-steppe environment and extremely low rainfall values) (Incarbona et al. 2010a, b; Sadori et al. 2008) a stable occupation by Homo sapiens of the island was possible not before the Upper–Paleolithic. Right in that period early migratory flows arrived from the continent on the northwest coast and continued exclusively along the northern coastline, moving east to west in a counterclockwise direction. PCA analyses show that, during the Sicilian Stone Age, the hunter-gatherer’s colonizers cyclically occupied the land establishing settlements in the proximity of caves, close to the coastlines. Moreover, the scatterplot displays that only during the transition to a mobile-forager/semi-sedentary ecology during protohistoric period allowed a gradual increase in the frequency and density of migratory flows. Although the degree of admixture was limited, semi-migratory farmers-shepherds began abandoning caves for small villages. This shift marked the start of the colonization of the southeast cost and hinterlands, a process that was only completed in the first historical periods with Greek and Punic colonization (700 B.C.E.).