Introduction

Late Antiquity is traditionally viewed as a transitional period between Classical Antiquity and the Middle Ages, marked by significant shifts in economic, social and cultural structures. Spanning from the 3rd to the 6th century AD, a period also associated with large population movements that spurred intense cultural exchanges and significantly impacted the everyday lifeways of Mediterranean populations1,2.

Within this broader context, Sicily was a prosperous region during the 4th and early 5th centuries AD, primarily due to its grain trade3,4. Later, particularly after the occupation of Carthage by the Vandals in 439 AD, trade networks were disrupted, and Sicily fell into economic decline. The island was the target of frequent military incursions between 455 and 470 AD; the Vandals raided the Sicilian coast almost every year, and the Vandal kingdom likely controlled parts of the island. Shortly after the political collapse of the Western Roman Empire and Odoacer’s subsequent accession to power in 476–477 AD, Sicily was formally ruled again from Italy in exchange for a yearly tribute to the Vandal king settled in Northern Africa. The island was anchored to Theodoric’s Italy from the very beginning of his reign (491 AD) until 536 AD, when it was incorporated into the Eastern Roman Empire3,5,6. The resulting instability and military activity increased mobility across the region, likely introducing new cultural and demographic influences, including interaction with Germanic groups6.

Relatively little is known about local-scale dynamics, cultural interactions and everyday life during this turbulent period. Current knowledge of human lifeways in Late Antique Sicilian communities remains rather limited (with the exception of an isotopic study on Syracuse7), likely due to the absence of isotopic datasets comparable to those recorded in other Italian territories8,9. To address this research gap, we employed stable isotope analyses to reconstruct the diets and spatial mobility of individuals buried at the Late Antiquity rural rock-cut cemetery of Scorrione. This recently excavated site yielded well-documented anthropological material. A smaller group of individuals from the burial site of Cisternazzi (Ragusa), in the ancient city’s suburb, was also included in our study to broaden the regional perspective.

In large areas of Sicily (such as the Iblean Plateau), the most widely attested vestiges of rural settlement are the remains of rock-cut cemeteries, formed by funerary chambers carved into natural rock walls and by open-air graves carved on the top of rock outcrops. Generally speaking, these sites have remained visible and accessible throughout history, which made them a preferential target for all sorts of anthropic alterations, from grave looting to reuse or complete destruction10. For these types of sites, skeletal remains are often absent or heavily altered, fragmented and dislocated from their original position.

The Iblean territory was characterised by a dense network of settlements extending from the plateau to the sea and is marked by the presence of rock-cut burial areas; it is likely that all sites fell within the scope of large, landed estates inhabited by small farming communities specialising in grain and olive oil production (Scerra forthcoming). The cemetery sites of Scorrione and Cisternazzi are located in the Ragusa region, within the eastern Iblei region in southeastern Sicily (Fig. 1). Both represent recently excavated sites with burials in their original deposits. The Scorrione site is unique for the presence of non-local ‘Germanic’ grave goods (more in Supplementary Material). Given the geographic proximity of the sites, we assume that they share similar ecological conditions11,12. The burial sites consist of rock-cut chambers (hypogea or catacombs) of different forms and sizes (more on the archaeological context in Supplementary Material). Such types of burial types were traditionally attributed to local families and communities13,14,15,16 and can be regarded as a rural counterpart of larger underground catacombs used by urban populations (e.g. Syracuse or Palermo7,17).

Fig. 1: Location of studied sites.
figure 1

a Map of Sicily showing its location relative to the Italian peninsula. b Detailed map of Sicily showing the location of the Scorrione (N 36° 47′ 13.793″ E 014° 49′ 13.784″) and Cisternazzi (N 36° 54′ 12.11″ E 014° 41′ 27.922″) burial sites.

Full size image

Anthropological evidence for this period and region remains limited, thus restricting our understanding of population demography, health and wellbeing. Only a few studies have reported osteological results, including Scorrione, for Late Antique Sicily18. Approximately half of the examined individuals in hypogeum D were younger than 25 years, and none lived beyond 40 years of age. Skeletal indicators, including osteoarthritis, robust bone morphology, and Schmorl’s nodes, suggest hard physical labour for some individuals, and the relatively young age at death suggests potential poor living or health conditions. Nonetheless, some individuals appear to have maintained comparatively good health. A similar pattern has been documented at the Licata site, suggesting broader variability in health status within these Late Antique Sicilian communities19.

Stable carbon (13C/12C) and nitrogen (15N/14N) isotope ratios, commonly expressed using delta notation relative to a standard (δ, ‰), when measured in collagen extracted from bone and tooth dentine samples, primarily provide information on the sources of dietary protein, although the carbon signal also includes a comparatively small contribution from carbohydrates and lipids20,21. Stable carbon isotope measurements on bone carbonate or tooth enamel reflect the carbon signal from the macronutrient dietary mix21,22.

Isotope-based diet reconstruction is made possible due to differences in isotopic abundances among potential food groups. A broad classification separates C3 and C4 protein sources according to their δ13C values. These arise from differences in isotopic discrimination between plants following C3 or C4 photosynthetic pathways. The food products from terrestrial animals consuming these plants may also be classified as exhibiting C3, C4, or mixed δ13C signals23,24,25. Human consumers typically exhibit δ13C values of c. -20‰ measured in collagen when fed exclusively C3 protein sources, and of c. -6‰ when fed exclusively C4 protein sources. However, these values can vary geographically and depend on local food baselines25,26,27, and in our research, we rely on our limited dataset and previously published data to define a reference baseline. Similarly, human carbonate δ13Ccarb values are typically c. -14‰ or c. 0‰ for humans eating C3 or C4 food sources, respectively24,28. Exclusive dietary contributions from marine food sources result in human collagen or carbonate δ13C values similar to those observed under exclusive C4 consumption29.

Stable nitrogen isotope ratios (δ15N) measured on human collagen provide information on the source of dietary protein. In ecological contexts, there is a relatively well-established positive relationship between δ15N values and the trophic level of a species. This is often the basis for distinguishing human terrestrial consumption of plant vs. herbivore protein, although human crop and animal management practices (e.g. manuring, irrigation and foddering) may result in food values that break ecological relationships between the trophic level and δ15N values30,31,32. An offset of c. 3.5‰ is commonly employed in ecological studies between dietary protein and animal tissues; however, this may vary according to species33,34. Previous studies on human diet-to-collagen δ15N offsets proposed values ranging from c. 5 to 6‰33,35. This is in agreement with an offset value of 5.5 ± 0.5‰ derived from an analysis of population-level isotopic data, which we employ in the present study21.

Adjusting for isotopic fractionation, stable oxygen isotope ratios (δ18Ocarb) measured in bone and enamel carbonate mainly reflect ingested water consumed directly as a liquid or via foods, although there are also some signal contributions from structural oxygen in foods and respired oxygen36,37. Spatial differences in the δ18O values of water sources resulting from temperature variations, precipitation levels, altitude, and distance from the coast are explored for studies of human mobility36,38. However, the processing of liquids and foods during preparation and storage (e.g. cooking or fermentation) may complicate the interpretation of human δ18Ocarb by fractionation processes that alter the δ18O signal of local water39,40,41. When it comes to human spatial mobility studies employing δ18O, tooth enamel is often the preferred material for analysis because it is more resistant to diagenetic alterations than bone carbonate, although the latter is also often employed36,42. Enamel δ18Ocarb values represent the time of crown mineralisation (with the first molar mineralising between birth and c. 3 years of age43,44) and can suggest an individual’s place of origin.

In ancient Sicily, human consumers had access to C3, C4, and marine food sources. However, their consumption may have varied by region and period. During Late Antiquity, archaeobotanical studies show that the majority of consumed cereals were C3 (barley and wheat)45. By the end of Late Antiquity in the 6th century AD, barley appears to have surpassed wheat in importance46. Consumable C4 plants would have included cereals like millet and sorghum, as sugarcane did not reach Sicily until the Arab period. Archaeobotanical and isotopic research has not yet detected the presence or consumption of C4 plants (Panicum milliaceum, Setaria italica, or Sorghum) in western Sicily during prehistory47,48, since earlier research focused only on the Bronze Age. In contrast, in eastern Sicily, macrobotanical remains of domesticated Setaria Italica were identified at the early medieval settlement of Punta Secca49. The δ13Ccoll values from the Late Antique site of Akrai in southeastern Sicily indicate that some herbivores were likely fed C4 plants, although these represented only a minor share of the studied faunal samples50. The direct consumption of wild species of foxtail millet (Setaria verticillata) by humans was confirmed by dental calculus analysis at the Late Palaeolithic site of San Teodoro, which is located in eastern Sicily51. The δ13Ccarb measurements on bone carbonate from archaic Syracuse also suggest the consumption of C4 carbohydrates by some individuals52. Based on current evidence, foxtail millet appears to have been present since prehistoric times, particularly in the eastern and southern regions of the island.

While the most identified taxa in archaeozoological assemblages are cattle, pigs and sheep/goats, the representation of species appears to vary by site type and location. A moderate prevalence of sheep/goats can be observed during Late Antiquity50,53,54. Other archaeozoologically identified species in early medieval contexts included domestic poultry, cattle, equids, turtles, and fish and molluscs (limpets and cuttlefish)49.

Marine food sources were accessible and likely consumed, but probably not on a large scale, especially among the poorest segments of local populations55. At two sites dated to the Archaic period, Himera and Syracuse, marine sources were not isotopically detected in human diets48,52. At the site of Punta Secca, located at a distance of 20–30 km from Cisternazzi and Scorrione, a skeleton of a female dated to the mid-7th century AD showed isotopic values consistent with an overall C3 plant-based diet, although potentially with minor contributions from terrestrial and marine animal proteins45. The absence of marine products in the diet of the Sicilian population was also attested at medieval Segesta (12th–13th century AD)56.

The closest chronological and geographical dietary parallel for Scorrione and Cisternazzi is reserach on individuals from the Late Antique catacombs of Santa Lucia in Syracuse. Isotopic results indicated that the Christian community dating to the period between the 5th and 7th century AD consumed freshwater fish in small amounts, and despite the close proximity to the sea, the consumption of marine products was not readily apparent7.

Methods

Material

In this study, skeletal remains of 17 adult individuals from 10 graves at the burial sites of Scorrione and Cisternazzi were sampled for stable carbon, nitrogen, and oxygen isotope analyses of bone collagen and tooth and bone carbonate to reconstruct the human diet and spatial mobility (Table 1). For Scorrione, all available identified adult individuals from the first two excavation seasons (Hypogeum D and F) were included in our study (n = 13). An osteological report for the studied individuals was previously published18 and its summary is given in Supplementary Material. For Cisternazzi, the highly commingled remains limited the number of sampled individuals to four adults. These samples were selected from available materials held in the Superintendence of Cultural Heritage of Ragusa archives and were limited to individuals with known age and sex (brief osteological data57). A summary of the employed methodology and osteological results is provided in the Supplementary Material.

Table 1 List of samples
Full size table

The Scorrione site was of particular interest for the research presented here due to the availability of osteological material preserved in its original deposit. However, this material was nonetheless impacted by natural processes, especially regular flooding of the valley, which led to the mixing of bone assemblages within the graves. Due to the commingled nature of multiple burials and the poor preservation of complete skeletons, the sampling strategy for Scorrione focused on long bones (see Supplementary Material). Femurs were primarily sampled for stable carbon and nitrogen isotope analyses of bone collagen; when femurs were unavailable, other long bones or mandibles were used (Table 4). All skeletal elements were macroscopically screened to avoid sampling near any visible pathological lesions.

For Cisternazzi, the sampling focused on the long bones of the lower limbs. The exceptional preservation of two complete skeletons in Grave 33 allowed for the additional sampling of ribs and a molar alongside femur samples (Table 4).

Finally, four faunal samples (Table 4) from chronologically associated archaeological layers (to avoid possible contamination from more recent upper layers) were also isotopically analysed. We applied ZooMS analysis to two samples with successfully extracted collagen (for more information, see Supplementary Material).

Pre-treatment and analytical measurements of collagen samples

Collagen extraction was made in the Laboratory for Stable Isotope Analysis at Masaryk University (CZ). We employed a modified Longin protocol58 for collagen extraction (Supplementary Material). Extracted collagen was transferred to a tin capsule for combustion and elemental and isotopic analyses at the Isotopic Research Laboratory at the Centre for Physical Sciences and Technology in Vilnius (LT).

Between 0.5 to 1 g of cortical bone samples was taken for collagen extraction. These were mechanically cleaned of any macroscopic contaminants and soil residues. The samples were then placed in acetone for 1 h and washed three times in an ultrasonic bath for 15 min using ultrapure water. Samples that were visually classified as well-preserved (all samples except SCOD_7_D1, SCOF_7_A1 and SCOD_6_B1) were covered in a 35% HCl (Penta, p.a.) solution for 1 h at room temperature. For poorly preserved samples (SCOD_7_D1, SCOF_7_A1 and SCOD_6_B1), this step was skipped. All samples were then demineralised at room temperature in 0.2 M HCl (hydrochloric acid) with solution changes on alternate days. Once the samples were fully demineralised, they were washed five times in ultrapure water. This organic remnant was gelatinised in an HCl solution set at pH = 3 for 24–48 h at 70 °C. The gelatinised organic mix was decanted into clean tubes, filtered using Ezee filters, and finally freeze-dried for 48 h.

Extracted collagen was transferred into a tin capsule for combustion and elemental and isotopic analyses. Collagen stable carbon and nitrogen isotopic composition were measured using a Thermo Delta V Advantage mass spectrometer in a Continuous Flow III system (Thermo Scientific) connected to an EA Flash EA1112 elemental analyzer (Thermo Scientific). Isotopic ratios are reported using delta (δ) notation relative to VPDB (The Vienna Peedee Belemnite) and AIR (atmospheric nitrogen) standards for δ13Ccoll and δ15Ncoll, respectively. Reported δ13Ccoll and δ15Ncoll values were calibrated using a multipoint linear regression based on certified international reference materials (USGS24 (graphite, δ¹³C: –16.05‰ ± 0.07‰ (VPDB)); IAEA 600 (caffeine, δ¹³C: –27.77‰ ± 0.04‰ (VPDB), δ¹⁵N: +1.0‰ ± 0.2‰ (AIR)). International standards were measured with collagen samples in the same sequence. These were loaded after every 12 samples. Each sample was measured once. Analytical uncertainty was 0.2‰ for δ15N and 0.15‰ for δ13C.

Pre-treatment and analytical measurements of carbonate samples

Carbonate samples were pre-treated in the Laboratory for Stable Isotope Analysis at Masaryk University according to a modified Miller et al. protocol59 and isotopically analysed at the Laboratory of the Czech Geological Survey in Prague (CZ).

Samples with weights of 10–20 mg were taken from compact bone or tooth enamel. The samples were powdered and left in a 0.1 M H3COOH (acetic acid) solution for 4 h. Every 30 min, samples were shaken. The samples were then cleaned five times in ultrapure water and dried.

Aliquots of samples and standards were weighed into 12 mL borosilicate Labco Exetainer vials and then flushed using a helium flow (100 mL/min for 5 min). H3PO4 (phosphoric acid; Sigma-Aldrich puriss p.a., ≥99%) was added manually to each vial, and the sample remained exposed to the acid for 1 h at 70 °C. The stable carbon and oxygen isotope composition of released CO2 was measured using a Delta V Advantage IRMS (Thermo Scientific) coupled to a GasBench II. The measurement sequence for 16 samples began and ended with international standards NBS-18 (δ¹³C: –5.014‰ ± 0.050‰ (VPDB), δ¹⁸O: –23.2‰ ± 0.1‰ (VPDB)), IAEA 603 (δ¹³C: +2.46‰ ± 0.01‰ (VPDB), δ¹⁸O: –2.37‰ ± 0.04‰ (VPDB)) and two in-house standards (STD Delta, Calcium carbonate chelometric standard, Sigma-Aldrich, 398101 (δ¹³C: –10.03‰ ± 0.05‰ (VPDB), δ¹⁸O: –20.55‰ ± 0.1‰ (VPDB); JLs-1 Limestone, Geochemical Reference Material, (δ¹³C: 2,0‰ ± 0.05‰ (VPDB), δ¹⁸O: –4.5‰ ± 0.1‰ (VPDB)). When the standards showed no drift during the measurement, a single normalisation of the whole set was performed, and for every tenth sample, a sequence of measurements was repeated. The measurement uncertainty was ±0.1‰ and ±0.15‰ for δ13C and δ18O, respectively. The values for δ13Ccarb and δ18Ocarb are both reported relative to the VPDB standard, with the latter calculated from values relative to the VSMOW (Vienna Standard Mean Ocean Water)60 by equation (δ18Ocarb(VSMOW) = δ18Ocarb(VPDB) × 1.03091 + 30.91). To estimate the δ18Odw value of ingested water, we employed the same approach employed by Cocozza and colleagues61. Carbonate δ18Ocarb values were converted into phosphate values62 (δ18Ophos =  δ18Ocarb(VSMOW × 1.0322 - 9.6849). From this, we obtained the δ18Odw value for drinking water following (δ18Odw = δ18Ophos × 1.55 - 33.49)63.

Because bone carbonate is exposed to post-depositional alteration to a greater extent than enamel36,37, we excluded from further analysis bone samples that did not pass collagen preservation criteria. In the discussion chapter, greater confidence is placed in the results obtained from enamel samples. When we discuss bone carbonate values, isotope-based interpretations are supported by other archaeological records. In our study, a conservative error margin of ±2‰ was applied to the calculated values of ingested δ18Odw to account for potential diagenetic impacts37. Breastfeeding complicates the interpretation of estimated δ18Odw values from tissue formed during this period, as it introduces an isotopic offset that is difficult to quantify accurately. However, based on evidence from Late Antique societies64, it is generally assumed that weaning was completed by ~3–3.5 years of age, which correlates with the mineralisation period of the crown in first molars. In modelling spatial mobility for M1 samples from individuals SCOD_6_A1 and CIST_33_A3, we followed previous approaches to correct for breastfeeding effects on δ18Odw values by adding and comparing two different offset values (−2‰ and −3‰)65,66,67.

Radiocarbon dating

Two individuals (GR 7A and GR 40E) were selected for radiocarbon dating to confirm and sharpen archaeological dating. Samples were pre-treated at the Radiocarbon Laboratory of the Max Planck Institute of Geoanthropology (DE). Bone demineralisation followed the same protocol as that employed for stable isotope analysis. Following demineralisation, the samples were placed in 1% NaOH (1 h at room temperature), rinsed (nine times) in ultrapure water, and placed in 1% HCl (1 h at room temperature). Samples were filtered using pre-combusted 2.7-μm pore size glass filters. Radiocarbon measurements on collagen samples were done at the 14C-Analysis Laboratory of the Max Planck Institute of Biogeochemistry (DE) using a MICADAS AMS system from Ionplus (CH) following established procedures68,69,70.

Radiocarbon measurements on collagen samples were produced at the 14C-Analysis Laboratory of the Max Planck Institute of Biogeochemistry (DE) using a MICADAS AMS system from Ionplus (Switzerland) following established procedures68,69,70.

Statistical methods and Bayesian modelling

R programming language software (version 4.3.2) was employed for data analysis71, while ArcGIS Pro (version 3.3) was employed to map the studied sites. We employed R-based Pandora & IsoMemo software tools (https://isomemoapp.com/) to reconstruct past human diets and spatial mobility. The Bayesian spatial smoothing model AverageR was employed to map the spatial distribution of δ18Odw values using modern water precipitation estimates from the Cluster-based Water Isotope Precipitation Model— RCWIP61,72,73. Human estimates of δ18Odw were compared to the AverageR baseline using the LocateR mapping tool to assign spatial residence and origin probabilities. We associated an uncertainty of 2‰ to human δ18Odw values, which we believe to be conservative.

The Bayesian software ReSources was used for isotope-based diet reconstruction. This software is an updated version of the FRUITS software61,74. We employed a ReSources mixing model to generate two estimates. One for overall caloric contributions from each food source (cereals, pulses, and animal food sources) for individuals from Scorrione 'mean' and another for protein only contributions to see the animal:plant protein ratio in the diet. The second was modeled for Scorrione 'mean', 'outliers' and Cisternazzi groups. The implemented model followed approaches similar to previous studies61,75,76. Full model descriptions are given in the supplementary file (Supplementary Material).

ZooMS and osteological analysis

Basic osteological determination followed the archaeofaunal manuals (Table 2)77,78,79.

Table 2 Archaeofaunal samples
Full size table

ZooMS analyses were conducted on two bone samples (SCOD_AN1 and SCOD_AN2), selected based on collagen preservation criteria appropriate for stable isotope analysis. The sample SCOD_39_AN1 did not yield a sufficient quantity of collagen and was therefore excluded from ZooMS analysis.

All laboratory procedures were done at the CEITEC facilities in Brno, Czech Republic. Chemical pre-treatment of the samples followed the established protocol80 based on previously published protocols81,82. Enzymatically digested collagen extracts were purified using Pierce™ C18 Tips, subsequently spotted onto MALDI-TOF plates, and analyzed using a Bruker UltrafleXtreme MALDI-TOF mass spectrometer. Mass spectra were processed using FlexAnalysis software (Bruker Daltonics) and interpreted with mMass software83. Taxonomic identification was performed by comparing the identified peptide mass fingerprints with published ZooMS reference libraries81,82.

Results

Radiocarbon measurements

Radiocarbon measurements were conducted on two bone collagen samples from Scorrione (Table 3). Both samples provided a sufficient amount of collagen for analysis84.

Table 3 Radiocarbon results
Full size table

For the young adult individual 7A, the calibrated radiocarbon date ranges from 382 to 534 AD (95% credible interval). Archaeological data based on the typology of grave goods dated the grave context to the early and mid-5th century AD85, which is in agreement with the radiocarbon date range. Individual 40E is dated to 428–542 AD (95% credible interval). The use of the grave is dated by stratigraphy to the period between 425 and 500 AD, and the typology of the grave goods is consistent with the radiocarbon result. These results match the chronology for the Scorrione site, dating from the early 5th to early 6th century AD (Supplementary Material)86,87.

Reconstruction of human diets

Five out of 24 bone samples did not produce sufficient collagen for analysis (Table 4). The elemental carbon content of the samples varied from 39.5 to 45.6%. The elemental nitrogen content of bone collagen in archaeological samples varied from 14 to 16.5%. All collagen samples fell within the commonly accepted C:N atomic elemental ratio range of 2.9–3.684,88.

Table 4 Elemental (wt% of C and N, and atomic C:N ratio) and isotopic data for Scorrione and Cisternazi
Full size table

Five bone carbonate samples that did not meet the collagen quality criteria were removed from further analysis. Summary statistics of the stable isotope analysis for Scorrione and Cisternazzi are available in Supplementary Material.

To establish a local food isotopic baseline, we combined previously published isotopic faunal data for Sicily dating from the Bronze Age48 to the Medieval period56 with those from our Late Antique study. A comparison of isotopic results for the two faunal samples from Scorrione (SCOD_AN1 and SCOD_AN2) with previously published datasets indicates that they both fall within the modelled 95% distribution range of herbivores and omnivores in Sicily (Fig. 2). Mean isotopic values and associated standard deviations (SD) for three animal food groups—marine fish, terrestrial herbivores and terrestrial omnivores were subsequently used in the quantitative reconstruction of human dietary patterns (Figs. 3a, 4).

Fig. 2: Faunal isotope data.
figure 2

Faunal isotopic data from Scorrione (Late Antiquity) compared with available published data from Himera (Bronze Age)48 and multiple sites (Medieval period)56. Coloured ellipses indicate the 95% distribution range for each animal group. Source data are available in the Supplementary Material.

Full size image
Fig. 3: Distribution of isotopic results.
figure 3

a Plot with δ13Ccoll values versus δ15N values in collagen. Individuals 40D and 6 are ‘outliers’ from the main Scorrione cluster. The mean and standard errors for each food category were calculated from faunal samples (Fig. 2); b Plot with δ13Ccoll values and δ13Ccarb values for Scorrione (red), Cisternazzi (blue) sites, and animal samples from Scorrione (yellow).

Full size image
Fig. 4: Bayesian dietary reconstruction.
figure 4

a Bayesian estimates of dietary sources contributions for Scorrione individuals (outliers 40D and 6 excluded). Boxes represent 68% and whiskers 95% credible ranges. Horizontal continuous lines represent the mean and dashed horizontal lines the median. b Bayesian estimates of protein contributions to diet for Scorrione (outliers 40D and 6); Scorrione individuals (outliers excluded) and Cisternazzi.

Full size image

The isotopic results from Scorrione and Cisternazzi suggest a predominantly C₃ terrestrial diet with a potential small contribution from C4 plants. Overall, the δ13Ccoll values are lower at Cisternazzi than at Scorrione (Fig. 3). A statistically significant difference between the sites (Welch Two Sample t-test; p value = 0.01136) can possibly be explained by a slightly larger consumption of C4 plants at Scorrione, but may also be the result of differences in food isotopic values arising from different farming practices89. At both sites, human collagen samples show relatively low δ15Ncoll values in comparison to other sites in Italy from the same period (Fig. 7). For most individuals at Scorrione and Cisternazzi, their δ15Ncoll values approach those observed for local herbivores (Fig. 3a), indicating limited consumption of animal protein and a mostly plant-based diet. However, two male individuals from Scorrione (6 and 40D, mean of δ15Ncoll = 10.1‰ ± 0.02) stand out as outliers (identified using a univariate robust statistical approach employing the median and median absolute deviation), with clearly higher δ15Ncoll values (Fig. 3a). No statistically significant difference in δ13Ccoll or δ15Ncoll was observed between males and females at Scorrione, while the small sample size at Cisternazzi did not allow for statistical testing.

Human and animal δ13Ccoll and δ13Ccarb values for Scorrione and Cisternazzi were compared with the linear reference model proposed by Kellner et al. to visually estimate dietary contributions of carbohydrates/lipids and protein from C3 versus C4/marine sources90 (Fig. 3b). Human isotopic values are generally aligned along the C3 protein line, with two values (individuals 29 and 7A, mean of δ13Ccarb = −6.6‰ ± 0.1) that show relatively elevated δ¹³Ccarb values. These could arise from higher contributions from C4 carbohydrate sources (possibly millet).

Quantitative caloric estimates of source contributions were obtained using a Bayesian mixing model. Estimates were made for the average Scorrione population, excluding two outliers (Fig. 4a). The results for the overall population model indicate that 15 ± 17% of consumed calories were obtained from terrestrial animal products, 48 ± 26% from C3 and C4 plants and cereals, and 34 ± 21% from pulses. Modelling results suggest that pulses, protein-rich plants, might have provided an important share of dietary protein (Fig. 4a).

Estimates of the ratio of plant to animal protein intake were also made for the Scorrione population mean (excluding the two outliers), as well as separately for the outliers and for the Cisternazzi population for comparison. The model indicates that the proportion of animal protein (both terrestrial and marine) relative to total plants protein was similar at both sites, averaging ~1:2. For the two Scorrione outliers, however, the ratio approaches 1:1, suggesting a relatively greater reliance on animal-derived foods (Fig. 4b).

Spatial mobility

The modelled local δ18Odw value for the site of Scorrione is −5.7‰, while the value for Cisternazzi is −5.8‰. To identify non-local individuals, we considered a ±2‰ uncertainty range to define the local baseline (−3.7 to −7.7‰ for Scorrione and −3.8 to −7.8‰ for Cisternazzi)37. Most individuals at both sites have bone carbonate δ¹⁸Odw values that fall within the local range, and only three humans (Scorrione 7A, 40D and 29) and one animal show δ18Odw values outside the local baseline range (Fig. 5a). Estimates of drinking water δ18Odw values were made for the first molar’s enamel δ18Ocarb values from two individuals (Cisternazzi 33A and Scorrione 6). These are plotted with and without breastfeeding offset corrections (we considered as correction −2 and −3‰; for more details, see the Methods section). Corrected values for individual 33A’s first molar fall within the local range, while corrected values for the first molar of individual 6 fall outside the local range and suggest a non-local origin for this individual (Fig. 5b).

Fig. 5: Distribution of δ18Odw results.
figure 5

a Plot with distribution of δ18Odw values in bone carbonate. b Plot with distribution of δ18Odw values in enamel for Scorrione (red), Cisternazzi (blue) and faunal samples (yellow). Dashed lines represent 2‰ range relative to site mean value (full line).

Full size image

Among the Scorrione samples, we chose two likely non-local individuals (7A, 40D) showing the lowest δ18Odw values and one individual (39B) with the highest δ18Odw value to estimate their residence areas during the period of bone formation (Fig. 6). Residence probability maps based on individual δ18Odw values indicate that all three individuals likely spent time during their life outside southeastern Sicily.

Fig. 6: Probability density maps for selected individuals.
figure 6

Maps display the the probability of place of origin/dwelling for selected individuals (7A, 40D, 39B and 6 without any offset correction in a first molar, with an offset −2‰, and offset −3‰ to δ18Odw). Areas shown in green represent the highest probability of sharing the same δ18Odw, with red areas indicating the lowest probability.

Full size image

Probability residence mapping for individual 6 was done without any offset correction to the estimated δ18Odw value and with added −2‰ and −3‰ offsets (Fig. 6). Offset-corrected δ18Odw values place the origin of individual 6 in Central or Eastern Europe.

Discussion

Archaeobotanical studies provide evidence for the grain cultivation of main staples such as barley and wheat in prehistoric and early medieval Sicily46,47,49. Isotope data for prehistoric Sicily also suggest that human diets were mainly based on C3 plant protein48,52. The isotope results for Late Antique individuals from eastern parts of Sicily (Scorrione and Cisternazzi (this study), Santa Lucia7) would also suggest a mainly C3 terrestrial diet, although with small contributions from C4 plant sources for some individuals.

Given the absence of archaeozoological evidence for marine foods, elevated δ13Ccoll values are more plausibly attributed to a minor consumption of C4 plants or of animals feeding on C4 plants. C4 plants from the study region (Sicily) and period include foxtail millet (Setaria italica), whose domesticated variant was identified using archaeobotanical analysis at early medieval Kaukana in southeastern Sicily49. Previous stable isotope studies on human remains have hinted at the possible minor consumption of C4 plants during the Archaic period in Syracuse52 and in Late Antiquity in Agrigento91. Common millet (Panicum miliaceum) and foxtail millet (Setaria italica), both dry-adapted species, can thrive in less favourable conditions with a higher ambient temperature and aridity compared to other crops92. C4 plant cultivation in the southeastern coastal region could be attributed to these environmental factors, since our study area (as is the case with Kaukanea and Agrigento) is located within an arid humidity belt11,12 in which the mean annual temperature does not drop below 15 °C92,93.

The comparatively low δ15N values observed for most individuals from Scorrione and Cisternazzi (Fig. 7; ‘Scorrione-mean’), together with Bayesian dietary estimates, indicate the low consumption of terrestrial animal protein with a considerable share of plant protein derived from pulses (Fig. 4). Current research supports the importance of legumes in the human diet especially at rural sites and states that it was important addition to cereals94. Overall, we conclude that most of the examined individuals followed a predominantly plant-based diet with only minimal input from animal protein (aquatic or terrestrial). Diachronic isotopic studies from Rome spanning the Imperial to early medieval periods reveal a temporal shift indicating a decline in the consumption of animal protein during the early medieval period95,96. This trend has also been noted for sites such as the cemeteries of Ambra Aradam and Piazzale Ostiense, as well as the rural cemetery of Castro dei Volsci8,95,96.

Fig. 7: Isotopic data for Late Antique Italy.
figure 7

Distribution of isotopic data for selected sites dated to the 4th and 6th centuries AD in Italy7,8,9,95,115,116. Regional faunal baselines (Sicily, Rome and Veneto) are displayed for two faunal categories (herbivores and omnivores). Source data are available in Supplementary Material.

Full size image

A direct comparison between mean human and herbivore δ¹⁵N values from Scorrione shows similar results and rather minor amounts of terrestrial animal products in the diet (Fig. 7). However, two Scorrione outliers and several selected individuals from the urban cemetery of Santa Lucia in Syracuse (the geographically and chronologically closest reference site) display elevated δ¹⁵N values. These values indicate a greater intake of animal-derived foods7.

This variability in animal protein consumption across Sicilian populations raises the question of whether there was a difference in access to animal products between urban/rural or rich/poor communities. By maintaining adequate meat production relative to grain yields, a diet primarily based on cereals may have been made more affordable for the majority of the population97,98,99,100.

The reduced consumption of animal products observed at Scorrione and Cisternazzi may reflect the more modest lifestyle of a poorer rural population as well as the impact of the crises following the collapse of the Western Roman Empire, which impacted agricultural practices and the food supply in rural and peripheral areas (for the Megarian area101; southeast coast3,102). The poor health status for some of the individuals buried at Scorrione (5th–6th century AD), as also indicated by their young ages at death, and pathological and activity markers (Supplementary Material), suggests that the population was probably comprised of individuals undertaking hard physical work, and supports the hypothesis of a deprived rural community18. From an economic point of view, meat was an expensive commodity97, and the population in rural areas likely had to exploit local sources. Nutritionally, animal products could have been supplemented or replaced by legumes (i.e., Vicia sp. is common in archaeobotanical contexts in Southern Italy), as they have a high protein content than other plants94,103,104,105. Additional isotopic data from other rural sites must be collected to ensure a more in-depth analysis, allowing comparisons across different strata of society or temporal evaluations among distinct regions of the Roman Empire.

Male and female individuals from grave 7 (individuals 7B and 7A) were buried together with various grave goods that could be grouped into two funerary assemblages85 (two belt sets with other finds as rings, glass beads and ceramics with a varied state of preservation; for more information and references, see Supplementary Material). Radiocarbon and typological dating place individual 7A in the second quarter of the 5th century AD (Table 3). The grave goods are compatible with the typological inventory recorded in early Vandal-period Africa, where East Germanic and steppe peoples (Vandals and Alani) settled in the 5th–6th century AD85. Both individuals share the same non-metric dental trait (shovel-shaped incisors), which might suggest that they were genetically related (see Supplementary Material)18,43,106.

The δ18Odw value obtained from the femur of individual 29 is close to the lower threshold value for the local baseline (Table 4 and Fig. 5a). Together with individual 7 A, both individuals showed comparatively high δ13Ccarb values (-6.5‰), potentially linked to C4 carbohydrate consumption (Fig. 3b). Similar to the two individuals from grave 7, the incisors of individual 29 were also shovel-shaped. Our results suggest the possibility that at least these three individuals (7A, 7B and 29) were a part of the same community and might have been genetically and/or culturally related, as indicated by the combination of historical, archaeological, anthropological, and isotopic data. Based on historical and archaeological records, it could be hypothesised that this community may have been related to the eastern populations present within the Vandal kingdom or Roman army; however, this assumption needs to be proven by further research, such as aDNA analysis. Given the available data, it is possible to formulate two hypotheses. The first, based on radiocarbon dating and the militaria recorded in Scorrione grave 7, may regard the ‘easterners’ as barbarian recruits of the Roman army and/or their relatives, probably relocated in Sicily around the time of the military conflicts in 427–432 AD, in which Germanic and Hun generals took part107. The second hypothesis is based on a comparison between the grave goods at Scorrione grave 7 and the early Vandal-period finds in present-day Algeria, and it would be related above all to the population of the African Vandal kingdom, which held a strong influence, and probably also some garrisons, on Sicilian soil from 439 AD onward.

Grave 40 from Scorrione, containing at least five individuals, was likely continuously used from 425 to 500 AD. The δ15N ( = 10.1‰) value of individual 40D is significantly higher than that of the rest of the individuals from grave 40 (n = 4; mean = 6.9 ± 0.4‰). We hypothesise that this male individual may have belonged to one of the different chronological phases of grave use, and/or he was part of a different social strata than others, following different food strategies, namely a higher reliance on animal protein. It is possible that burial 40D belongs to a relatively late period, corresponding to the final period of the cemetery, similar to grave 6 with comparable δ15N values. A ceramic lamp from this grave was certainly deposited after 480 AD. Male individual 40D also falls outside the local baseline for δ18Odw values and possibly spent his adulthood outside Scorrione (Fig. 6).

Skeletal remains in grave 6 were poorly preserved, but the available material suggests a single individual. The proposed origin of this individual, based on modelling δ18Odw enamel value, is attributed to the central and eastern half of Europe, which is consistent with the presence of grave goods associated with central ‘barbaric’ European contexts, specifically Lower Austria and Moravia (context information in Supplementary Material)85. Individual 6 is one of two outliers with higher δ15N values than the rest of our dataset, suggesting a greater consumption of animal proteins. A comparison with 5th–6th-century European sites, including cemeteries in Moravia, Austria, Germany, and Hungary, revealed similar diets with a reliance on terrestrial animal protein108,109,110,111,112. The combination of grave goods suggests a dating to the second half of the 5th century85. During this time span, two events may have favoured the arrival of Central European individuals to southeastern Sicily. Roman operations against the Vandals (457–468 AD), where military forces allegedly included Huns or Goths, or Odoacer’s and later Theodoric’s resumption of Italy’s control over Sicily (from 476/77 AD and from 491 AD, respectively).

Scorrione individual 39B had the highest δ18Odw value (−3.8‰) and is on the threshold of the local baseline (Fig. 5a). The mapping of respective residence probability map suggests that the individual may have lived in Malta or North Africa (Fig. 6). It is important to note that the value was measured in bone carbonate, and that such an outlier could potentially result from post-mortem bone alteration. However, such a result would not contradict the contemporary political situation and/or the existence of trade contacts with Northern Africa during Late Antiquity4,113.

It is plausible to consider that the movement of individuals, craftsmen, and traders occurred in the vicinity of Scorrione. The mixing of different ethnicities and peoples due to the military actions of various Germanic tribes in Late Antiquity can also be presumed3,6. Likewise, the role of the later Roman army as a mobility driver for ‘barbarian’ populations is well known114.

In conclusion, stable oxygen isotope analysis of human skeletal remains from the site of Scorrione, together with osteological, archaeological, and historical evidence, offered insights into the spatial mobility patterns of its burial population. Based on a combination of stable isotope analysis and archaeological results, we identified at least one non-local individual (grave 6), suggesting connections to Central Europe. Additionally, a group of individuals at Scorrione, linked through shared dietary patterns, grave goods, and isotope-based mobility modelling, suggests ties to regions outside Sicily, including mainland Europe and North Africa. In contrast, all individuals from the Cisternazzi site appear to be of local origin. Overall, isotope-based reconstruction of dietary habits at both sites revealed a heavy reliance on plant foods, possibly including legumes. The limited consumption of animal proteins at Scorrione and Cisternazzi could likely be attributed to the persistent effects of the crises following the collapse of the Western Roman Empire on human lifeways. The pronounced differences in δ15N human values between our dataset (from the Scorrione and Cisternazzi sites) and other Late Antique Italian sites revealed the presence of a special dietary case, although limitations in current research do not allow a full exploration of the potential variations according to settlement type (urban vs. rural) or region.

Limited and fragmentary as it may be, our dataset from the rock-cut cemeteries in the Ragusa province already challenges established paradigms. Evidence from Scorrione reveals a complex and diverse community, where individuals of different cultural and social backgrounds, origins, and dietary habits may have coexisted. However, our small-scale study also highlights the need for further research on spatial mobility and dietary patterns at rural sites in order to gain a more comprehensive understanding of social dynamics in Sicily between Late Antiquity and the early Medieval period.

Statement and declarations

This paper presents partial results of the first author’s doctoral research project.