Dietary inequality marker reveals 10,000 years of gender and cultural disparity in Europe

Abstract

Diet is a key to evaluating social and health inequalities over time, as it reflects disparities in access to resources often linked to socioeconomic and gender factors. Stable carbon and nitrogen isotopes, while semi-quantitative, typically limit intersite comparisons, as the results are tied to local baseline isotope values. In this study, we overcome this limitation by applying the interdecile ratio—a metric from economics—to isotope data from 12,281 individuals across 393 European sites over millennia. Our isotope-based dietary inequality index reveals the nonlinear evolution of dietary disparities over time and across different geographical areas. Sex-based disparities are evident throughout all time periods. Male individuals are consistently overrepresented in the upper deciles, indicating greater access to animal proteins, while females dominate the lower deciles, reflecting more restricted access. Neolithic societies exhibit homogeneous diets at the population level, but animal protein consumption tends to differ between men and women. As expected, Bronze Age carbon interdecile indexes mark increasing dietary inequality, likely linked to agricultural advances and social hierarchies. Dietary disparities peak during Antiquity, although the gap between the sexes narrows slightly. This diachronic analysis highlights the complex interactions between diet, social structures, and gender and provides a robust framework for comparative studies of health inequalities in archaeology.

Significance Statement.

Understanding dietary inequalities in past societies has long challenged archaeologists due to methodological limitations. We introduce an innovative approach using the interdecile ratio (IDC) applied to isotopic data from 12,281 European skeletons spanning 10,000 years. This method allows diachronical and intersite comparisons revealing recurring sex-based patterns: women are more frequently in lower deciles for animal protein intake, while men dominate higher nitrogen isotope deciles—except during the Copper Age, where the pattern is less clear due to fewer sites. By normalizing local variation, IDC overcomes traditional limitations and provides a robust framework to investigate social stratification in past societies, offering crucial historical context for understanding long-term gendered and broader food inequalities.

Introduction

Dietary choices served as powerful indicators of societal inequalities, reflecting complex interactions between gender, socioeconomic status, and cultural norms (1). In contemporary societies, among the 3 billion people suffering from chronic malnutrition (2), access to highly sought-after foods such as meat and fish strongly correlates with modernization indicators, particularly per capita income and urbanization rates (3–6). Unlike wealth, which can be accumulated and transmitted, access to sufficient food is an immediate and fundamental need, and its insecurity disproportionately affects vulnerable populations (7). These people face heightened challenges, including issues of both quantity and quality of available food, uncertainties surrounding food supply, and experiences of hunger. Such limited access to fresh, nutritious food leads to health disparities in modern and preindustrial times, including malnutrition, infectious and parasitic diseases, chronic illnesses, and reduced life expectancy (2, 8). However, the historical depth and persistence of dietary inequalities, especially those related to gender, remain poorly understood across different time periods and cultures (see details in the Supplementary Information, SI2: Heath and diet).

Archaeological research offers unique insights into past dietary practices through the analysis of human remains, providing a window into long-term patterns of resource distribution and social stratification. The study of ancient populations typically relies on the analysis of animal and plant remains from archaeological sites, as well as isotope proxies retrieved from human remains. Stable isotope analysis of bone collagen, particularly carbon (δ¹³C) and nitrogen (δ¹⁵N) ratios, has emerged as a powerful tool for reconstructing individual dietary patterns. While these analyses do not reveal the nutritional quality or the symbolics attached to food products, they effectively indicate the proportion of different food categories in past diets. Nitrogen isotope ratios positively correlate with animal protein intake, reflecting higher values in consumers of marine or freshwater foods due to extended aquatic food chains (9). Meat consumption itself is often linked to notions of power and control, historically and ethnographically documented as more frequent among men than women when differences are observed, and associated with higher social status, whether demonstrated or inferred for prehistoric periods (1, 10). Carbon isotope ratios, meanwhile, provide insights into the types of plants consumed. C4 plants like millet, sugarcane, and corn show elevated δ¹³C values, while C3 plants and freshwater resources yield lower values (11). The symbolic value of these foods varies: millet is often considered an “inferior” grain in historical and contemporary contexts (12–14), while maize generally holds a more favorable status during the periods relevant to our assemblage. Marine resources, on the other hand, display a status that fluctuates over time, depending on trophic level and cultural perceptions (15). This variability, associated with higher δ¹³C values, makes the interpretation of carbon isotopes more context dependent, whereas nitrogen isotopes offer a more straightforward reflection of animal protein intake.

However, comparing isotope data across different sites presents significant challenges. Nondietary factors such as manuring (16, 17) or other husbandry and agricultural practices, climate (18), and metabolic stress (19, 20) can influence isotope compositions and therefore local baselines for a given time period. While these factors typically affect entire populations uniformly, except in cases of individual mobility or species-based agricultural practices, they complicate intersite comparisons (21, 22).

This is where the application of the interdecile ratio (IDC or D90/D10) becomes particularly relevant. The IDC is a statistical indicator commonly used in economics to measure inequality (23). It compares the ninth decile (D90), the threshold above which the top 10% of values lie, to the first decile (D10), the threshold below which the bottom 10% fall. Importantly, D90 and D10 refer to the distribution thresholds, not the means of these groups, which limits the influence of outliers. The IDC is thus calculated as:

$$\mathbf{I}\mathbf{D}\mathbf{C}=\mathbf{D}\mathbf{90}/\mathbf{D}\mathbf{10}$$

As a relative measure, IDC captures the spread between distributional extremes, independently of absolute values or local baselines. Unlike the Gini coefficient, which considers the entire distribution and has been used in archaeology to quantify dispersion (e.g. house sizes (24, 25)), the IDC focuses specifically on the extremes. By comparing the highest and lowest deciles within each archaeological population, this standardized measure provides a robust assessment of dietary inequality that can be compared across sites and time periods. This form of normalization reduces the influence of local baselines, whether due to agricultural practices (manuring, fertilization, and irrigation) or environmental factors. This makes it more suitable for assessing dietary inequalities (24), as it reduces the influence of local baseline variation and allows direct comparison of the highest and lowest deciles across sites and periods. Although IDC does not explain the biogeochemical causes of isotopic variation or measure dietary inequalities or socioeconomic structures, it does highlight internal heterogeneity within groups. In all the studies gathered for the database used here, the authors of these publications have interpreted disparities in carbon and nitrogen isotopes within a single time period, culture, and population as variations in animal protein and cereal consumption. Assuming individuals from the same site have comparable isotopic baselines, disparities in access to foods, such as animal proteins (nitrogen) or marine resources and cereals (carbon), can be interpreted by examining the distribution of isotopic values within each population. These variations may therefore serve as indirect indicators of dietary inequalities, which can in turn reflect broader socioeconomic structures. Each step in this interpretive process is subject to limitations and requires cautious assumptions. The IDC operates under the working assumption that individuals from the same site share comparable isotopic baselines. This assumption can be affected by intra-site chronological differences, the presence of mobile individuals, or the introduction of imported foods.

This study examined the proportion of male and female individuals in different deciles to investigate potential variations in women's and men's access to animal proteins. This differential access is a common topic of discussion in many past and present cultures worldwide and is often interpreted as being gender based, including in prehistoric Europe (1, 10, 26, 27). However, this approach to prehistoric societies is based on the assumption that their sex/gender systems were similar to those of contemporary Western societies (28, 29), which normalizes a binary division of bodies (30). As the sex of the individuals in our database was determined by methods that categorized them as either male or female, this is a limitation of the present work. Therefore, when interpreting observed differences between male and female individuals as evidence of gender inequalities within a population, the term “gender” will be used.

Here, we analyze isotope values from 12,281 adult individuals of known sex from 673 European sites, grouped into 393 distinct chronological locations across 40 countries in Western Eurasia and the Mediterranean basin. Our dataset, compiled from published sources (Materials and methods, Datasets S1 and S2), meets both qualitative criteria (stable isotopes from bone collagen, reliable C/N ratio, adult population of known sex) and quantitative requirements (minimum series of 10 individuals with known sex). The IDC ratio effectively measures dietary inequalities by assessing the gap between the deciles with the highest and lowest access to animal proteins (for nitrogen), or to marine foods or millet, depending on the context (for carbon; Fig. S2). Our objectives are (i) to employ the IDC as an analytical tool for comparing food-related inequalities between sites while mitigating baseline effects related to environmental changes or cultural practices; (ii) to establish a European reference framework for each major chronological period and to assess whether the principal trends observed align with expected patterns; and (iii) to investigate potential differences in access to animal protein between men and women, based on the working assumption of a binary sex/gender system. This comprehensive analysis allows us to explore the magnitude of variation in isotope values within societies over a 10,000-year time period in Europe (Dataset S3), revealing persistent patterns of dietary inequality that often correlate with biological sex disparities that we interpret as gender based.

Results

A nonlinear trend over time

Table 1 synthesizes the key dietary patterns and biological sex distributions across all chronological periods, highlighting the persistent male bias in the highest consumption decile (D90) and the varying degrees of dietary inequality as measured by IDC values. Both carbon (IDC_carbon) and nitrogen IDCs (IDC_nitrogen) strongly vary within time periods (Fig. 1). Preagricultural isotopic values, although limited in number (n = 13), exhibit considerable diversity in dietary habits. Temporal variations are evident in the IDC_nitrogen curve, with higher peaks of dietary inequality during the Iron Age, Antiquity, and Late Middle Ages (mean IDC = 1.24 to 1.27) compared with other time periods (mean IDC = 1.23; SD = 0.03). While the magnitude of IDC_carbon (1.01 to 1.46) is smaller than that of nitrogen (1.06 to 2.45), it exhibits significant variation (KW = 55.71; P = 8.9e-09, Fig. 1).

Table 1.

Synthesis of dietary patterns and gender distribution across time periods in Europe.

Chronology	Sample size (n)	Sex ratio (M:F)	IDCnitrogen	IDCcarbon	Key dietary features
Preagricultural	203	1.26	1.23	1.09	High variability in dietary habits; significant sex imbalance in N D10
Neolithic	1,242	0.95	1.18	1.04	Most egalitarian distribution; terrestrial resources dominant. Female-biased sex ratio; significant sex imbalance in N D90, D10, and C D90
Copper Age	316	1.39	1.19	1.04	Similar to Neolithic patterns, strong male bias in the sex ratio
Bronze Age	603	1.25	1.21	1.10	Marked shift in carbon isotope values; introduction of millet; significant sex imbalance in N D90 and C D10
Iron Age	931	1.11	1.26	1.08	Peak in dietary inequality; significant sex imbalance in N D90, D10, and C 90
Antiquity	1,939	1.47	1.24	1.06	High dietary inequality; decline in millet consumption; strong male bias in the sex ratio; significant sex imbalance in upper and lower C and N isotope deciles
Early Middle Ages	3,591	1.17	1.21	1.07	Reintroduction of fish consumption; significant sex imbalance in upper and lower C and N isotope deciles
High Middle Ages	1,733	1.5	1.23	1.08	Continued dietary stratification, strong male bias in the sex ratio; significant sex imbalance at the extremes
Late Middle Ages	907	1.37	1.27	1.09	Peak in Nitrogen IDC variation; significant sex imbalance in N D90, D10, and C 90
Modern	816	1.36	1.23	1.07	Persistent inequalities; significant sex imbalance in C D90
All periods	12,281	1.26	1.23	1.07

Sex ratio expressed as males/females; IDC, average interdecile ratio (D90/D10); D90, the highest consumption decile.

Fig. 1.

D90/D10 interdecile for carbon (IDC_carbon) and nitrogen (IDC_nitrogen) since the Mesolithic in Europe (12,281; 393 sites). EMA, Early Middle Age; HMA, High Middle Age; LMA, Late Middle Age. The black dotted line shows the average values per time period. The numbers in parentheses denote the number of sites (only sites with >10 sexed adults are considered). The statistical results correspond to the comparisons of the extreme deciles two by two without adjustment: the significant probability P-values are indicated by asterisks (Dunn's test). The number of asterisks represents the number of 0s after the decimal point: *P < 0.05; **P < 0.005; ***P < 0.0005. The Kruskal–Wallis test was conducted with all groups taken together (ddl = 8).

Among inland locations, the largest difference in carbon isotope values is recorded during the Bronze Age, Iron Age, and Early Middle Ages at North Italian sites in the Po Valley (Bovolone, Ostiglia La Vallona, Olmo di Nogara, Arano di Cellore, and Ballabio) and next to Verona (Fig. S1, S2A). In these sites, the top decile shows about 1.35 times more range in carbon values than the low decile (D10). This difference likely reflects the dietary upheavals of this period linked to intensified soil fertilization and an increase in millet consumption (12).

While carbon isotope variations highlight dietary changes linked to agricultural practices, nitrogen isotope data reveal other patterns of dietary inequality. The most striking example comes from the Late Medieval sites in Greenland (Anavik, Sandnes), where the top decile shows 2.4 times more variation in nitrogen isotope values than the bottom decile (Fig. S1, S2B). These differences reflect both distinct diets and varying proportions between the two locations, with δ¹⁵N values suggesting the presence of individuals with high marine food consumption and others with a more terrestrial tendency. When considering only inland sites located at least 40 km away from the sea, nitrogen isotope variations are significantly smaller, as expected given the fewer trophic levels in terrestrial food webs (Fig. S7). However, two Italian sites (Opi Val Fondillo and Sulmona Saint Lucia (31)) appear as outliers with ratio values around 2, due to the presence of two distinct groups within the same location (Fig. S2).

Neolithic individuals are characterized by the lowest variability at the population level and the smallest difference in food intake between deciles across both isotope systems. Despite spanning several millennia and diverse cultures and geographic regions, subsistence processes appear most egalitarian during this period: the top 10% have access to enriched carbon food sources only 1.04 times more than the bottom 10% for carbon and 1.18 times for nitrogen on average. As previously observed, this observation implies significant uniformity in Neolithic diets throughout Europe, predominantly relying on terrestrial resources, except for sites along the Danube River where subsistence strategies still involve freshwater resources (Fig. 1).

Copper Age IDC values, for both carbon and nitrogen, strongly resemble those of the Neolithic, but this dietary homogeneity ends with the Bronze Age, which marks a notable shift toward unequal distribution of carbon and nitrogen inputs (Fig. S2). Beyond the establishment of agriculture and animal domestication, the growing complexity of economic and political systems (32), including the emergence of states and princely tombs, appears to have intensified wealth inequalities from the Bronze Age onwards (24, 33). This increase primarily stems from three geographic areas: Northern Italian sites, Greece, and East European sites, where the extreme decile shows between 1.46 and 1.10 times higher δ¹³C values compared with the lowest decile. This ¹³C enrichment may reflect new diets (introduction of millet), the development of nonlocal economies with imported products, and/or emergence of new agricultural practices. These changes prove more disruptive than the later introduction of New World C4 plants like corn and sugar in the Modern era (Fig. 1). The subsequent reduction in the IDC_carbon during Antiquity reflects the widespread Roman rejection of millet cultivation (34, 35) (Figs. 1 and S2).

A sex bias within the database

An average of nearly 31.3 individuals per site over 10 different chronological periods is observed. A significant majority of males characterize these data (masculinity rate of 126%, χ² = 166; P = 2.2e−16; Table 2, Fig. 2). While various factors such as climate, resource availability, competition, attractiveness, population density, and genetic parameters can potentially bias sex ratios, demographers generally consider sex ratios at birth to be stable over time within a population (36). The observed variations here thus likely result from funerary or sampling biases. This bias does not appear to be restricted to a specific geographical area (Fig. S5). Six geographical areas only involve males. They are related to medieval Christian contexts (Varnhem, Dunes Abbey, Gloucester, Finglesham, and St James Abbey (33, 36–38)) and the accidental La Tène site of Cornaux-Les Sauges (38) (Fig. 2), which shows that in medieval contexts, this bias is influenced by the sampling of religious sites (10 monastic or convent sites out of 179 early, classic, and late medieval sites) (39). Additionally or for other time periods, there could be (i) a preference for sampling male graves, perhaps due to a more frequent or abundant presence of grave goods in male burials, although we did not find literature to support this hypothesis or (ii) a bias related to a lesser preservation of the postmenopausal skeleton in relation to osteoporosis (40). Conversely, the Neolithic period is characterized by an overrepresentation of females (Table 2). This could be due to the rarity of these graves, leading to more systematic sampling, or to the existence of differential funerary practices favoring female individuals, as recently documented in Portugal (37). Therefore, apart from the oldest prehistoric periods (Mesolithic and Neolithic) that are significantly different from the others, it appears that males are predominantly analyzed. This is especially true for the Bronze Age, Copper Age, Antiquity, and High and Late Middle Ages. This distribution constitutes an important bias that must be taken into account in the results.

Table 2.

Synthetic distribution of isotopic series (probability compared with a natural theoretical population with a balanced sex ratio).

Chronology	Number of sites	Number of individuals	Male (n)	Female (n)	Sex ratio	χ2	P-value
Preagricultural	13	203	113	90	1.26	2.61	0.106
Neolithic	47	1,242	604	638	0.95	0.93	0.334
Copper Age	15	316	184	132	1.39	8.56	0.003
Bronze Age	24	603	335	268	1.25	7.44	0.006
Iron Age	35	931	490	441	1.11	2.58	0.108
Antiquity	54	1,939	1,155	784	1.47	71	3.60E−17
Early Middle Ages	104	3,591	1,938	1,653	1.17	22.6	1.98E−06
High Middle Ages	46	1,733	1,041	692	1.5	70.283	2.20E−16
Late Middle Ages	29	907	524	383	1.37	21.92	2.84E−06
Modern era	26	816	470	346	1.36	18.84	1.42E−05
All periods	393	12,281	6,854	5,427	1.26	165.81	2.20E−16

P-values lower than 0.05 (marked in bold) are considered significant. This means that the sex ratio during this period differs significantly from the expected ratio of an equal number of men and women.

Fig. 2.

Sex ratio of the populations grouped by time periods. The solid line indicates a sex ratio of 1:1; below: females are more numerous; above: males are more numerous; dotted line: average sex ratios. Horizontal bars show statistically significant differences between time periods, probabilities are symbolized by asterisks (Dunn's test; the number of asterisks indicates the number of 0s after the decimal point, *P < 0.05; **P < 0.005; ***P < 0.0005).

Biological sex inequalities in extreme deciles

The analysis of extreme deciles provides compelling evidence for dietary inequalities between individuals identified as males or females that persist across time periods. The populations belonging to the highest decile (D90) consistently exhibit a higher proportion of males compared with females for all time periods, with significant nitrogen isotope variations from the Neolithic to the Late Middle Ages (Fig. 3). We tested if the sex ratio within the upper deciles (D90) was significantly different from the global sex ratio of the population, whether for carbon and/or nitrogen. The results show consistently over time an overrepresentation of males in the deciles corresponding to a preferential access to animal proteins (for nitrogen isotopes) as well as marine food and/or millet (for carbon isotope ratios), particularly from the Neolithic onwards (Fig. S5). The average nitrogen isotope ratios of the whole population also differ between men and women for all time periods, especially during the preagricultural time (Fig. S5). For carbon isotopes, this difference is partly explained by the consumption of aquatic foods by friars, but the differences remain significant when religious sites are excluded. The overrepresentation of males in this first decile compared with the average sex ratio across the entire population is significant for carbon isotope ratios in the Neolithic, Bronze and Iron Ages, Antiquity, Early, and High and Late Middle Ages (Figs. 3, S5, and S6, Dataset S4). Overall, there are 2.16 men for every woman in the top decile of nitrogen (492 females and 1,062 males, χ² = 88.47; P < 2.2e−16), especially during the Middle Ages. In contrast, the lower decile exhibits an overrepresentation of females (862 for 678 males, χ² = 76.62; P < 2.2e−16), especially during the Neolithic period and the Early Middle Ages, for nitrogen isotope values. In addition to the overrepresentation of males on one side (D90) and females on the other (D10), the overall trends are not continuous and show variations over time for both markers (Fig. 3). These variations suggest that these differences are mostly not only biological (see details in the Supplementary Information) but also result from different cultural or social practices over time (1, 10, 41).

Fig. 3.

Sex ratio within a time period for all individuals (in black), as well as in the upper and lower deciles k, for carbon (in orange) and nitrogen (in green) indexes. When estimating the sex ratio within deciles for each time period, individuals were first classified into the lowest and highest deciles based on their relative position within each site. This site-specific approach accounts for local isotopic baselines before aggregating data at the broader temporal scale. As a result, intersite baseline differences are effectively normalized, allowing for meaningful comparisons across regions and periods. The statistical results correspond to the comparisons of the extreme deciles two by two without adjustment: the significant probability P-values are indicated by asterisks (χ² tests, Dataset S4), and the number of asterisks represents the number of 0s after the decimal point (*P < 0.05; **P < 0.005; ***P < 0.0005.).

A key advantage of IDC_nitrogen is its ability to normalize local baseline variations while maintaining sensitivity to social disparities. By comparing the highest and lowest deciles within each population, this ratio effectively controls for regional and temporal fluctuations in baseline isotope values, allowing for robust cross-cultural and diachronic comparisons. This methodological strength is particularly apparent when analyzing sites with distinct environmental conditions or agricultural practices, where direct comparisons of raw isotope values would be problematic. The ability of IDC_nitrogen to reveal persistent dietary inequalities across diverse contexts highlights its potential as a powerful tool for investigating social stratification in archaeological populations.

Discussion

Food-related social determinant

Our results reveal persistent inequality in access to animal protein in Europe over the last 10,000 years. The prevalence of women in the bottom decile of nitrogen isotope ratios highlights a systematic restriction of animal proteins, confirming previous anthropological observations of growth deficiency and stress markers affecting women more (10, 42). Although this trend has often been suggested (10, 43), our study provides the first large-scale quantitative evidence. Gender-based food inequalities may be rooted in a variety of cultural practices: food taboos, cosmological beliefs, misperceptions of protein needs, and social norms that favor female deprivation in favor of men (1, 10, 44). Meal segregation, both spatial and temporal, with males given precedence, reinforces these disparities (45), a phenomenon observed in both Europe and Africa (46). However, our data indicate that these inequalities are not ubiquitous. In 5% of the studied sites, females had preferential access to animal proteins.

It is important to note that the isotopic differences observed between men and women cannot be attributed to inherent metabolic differences between the sexes. If that were the case, we would expect to see a consistent gap between male and female isotopic values across all chronological periods and geographical contexts. However, our data show that these differences vary significantly depending on the period and cultural context (Fig. 2), with particularly pronounced disparities during the Neolithic and the Middle Ages, but less so during Antiquity. This pattern, which is evident in our extensive dataset, suggests that, despite the male-dominated social structures described in historical sources, access to food resources was not consistently determined by gender in all socioeconomic and cultural contexts. During certain time periods or in certain cultures, class and age have been more structuring factors of societies and therefore, of dietary inequalities. In ancient Roman societies (47), for example, meals were shared between males and females of the same social class, but food procurement was closely linked to social status. This was less true for Hellenistic societies, which tend to show more dietary inequalities between male and female in our database (47). These results highlight the value of integrating large isotopic datasets to complement and refine interpretations derived from textual evidence and traditional archaeology, showing that dietary inequality could be modulated by site-specific and intersecting cultural, economic, or environmental factors. This temporal and contextual variability confirms that the observed disparities are primarily the result of cultural and social factors rather than physiological differences (see details in Supplementary Information 6).

During the Bronze and Iron Ages, males are overrepresented in the top carbon and nitrogen deciles. This indicates higher millet and meat consumption—an observation that initially seems contradictory. However, a detailed analysis reveals that men consuming millet exhibit lower nitrogen isotope ratios than others. This confirms the existence of a dual dietary discrimination during this time: one probably based on social status (12), where lower-status men consumed less meat and more millet, and another gender based, where women—regardless of their social status—consumed less animal protein and millet than men. Regional variability in millet consumption further suggests the influence of cultural preferences, exchange networks, and social interactions, in addition to environmental constraints.

The lack of sufficient data for preagricultural societies (n = 13) does not allow us to discuss the onset of these inequalities and which cultural event (if any) it could be related to, such as the rise of agriculture (26, 48) and/or herding (49). Yet, rather than assuming a linear onset of inequality, it is important to consider that social hierarchies may have fluctuated over time, as seen in seasonal variations among some Paleolithic, historical, and modern societies (50). However, our results indicate that hunter-gatherers already exhibited a marked disparity, as shown by differences in δ¹⁵N between males and females (Figs. 3 and S5). This trend echoes observations in various hunter-gatherer societies worldwide (1, 10, 51), such as the Hazda (East Africa), where males consume ∼40% of their diet as animal protein, compared with only 1% for females (41).

Social stratification and changing dietary practices

Carbon and nitrogen isotopic differences also reflect social and environmental differences in access to food resources. While carbon IDC remain similar between coastal and continental sites, a significant change appears in the Bronze Age (Fig. S7), coinciding with the introduction of new agricultural practices such as millet cultivation (52). These innovations seem to play a major social role, particularly in Italy (Po Plain, Germany (Saxony-Anhalt)) and Greece (Sparta), where food inequalities are most marked. This hierarchy related to plant type consumption could exist elsewhere but remain invisible due to the absence of C4 plant crops.

In postagricultural societies, fish consumption becomes marginal (53), before reappearing from the Early Middle Ages, probably under the influence of religious practices (54). Regarding nitrogen, interdecile differences are more related to fish consumption, which does not hold the same symbolic value as meat, than to strict social inequalities (Figs. 1, S8, and S9). However, when religious sites are removed from our analysis, biological sex dietary differences remain significant throughout these periods, indicating that these disparities transcend religious contexts and reflect broader social patterns. Thus, a high IDC_nitrogen in coastal sites does not necessarily mean increased social stratification. Similar patterns are observed in communities exploiting freshwater resources, such as the Danube sites of the Mesolithic and Neolithic or Sakhtysh on the banks of the Volga (Fig. 1).

Hierarchy and food luxury

The Neolithic is distinguished by relative dietary equality, despite an overall reduced consumption of animal proteins. Nevertheless, even in this context, men benefit from privileged access to these resources (Fig. 3), which is consistent with archaeological markers of male privilege (26, 55, 56). This hierarchy became more pronounced over time: from the Bronze Age onwards, meat became a luxury product reserved for the upper classes, dominated by men. However, if the Copper Age seems to show some dietary continuity with the Neolithic, it is during this period that the sex ratio shows an overrepresentation of men (Fig. 3). Dietary inequality reached its peak during the Iron Age, Antiquity, and the Late Middle Ages, reflecting culinary practices differentiated according to social status and a possible increased use of fish (57, 58). Again, while Antiquity is characterized by strong social stratification, it also marks a minimal level of dietary inequalities between sexes. The global variation in dietary inequalities observed between different periods may be distorted by the overrepresentation of certain cultures. The Iron Age in Western and Northern Europe was much longer than in Southern Europe, where sites dating back to 800 BC are classified as Antiquities.

Our results thus highlight the intersection between social status and gender in access to animal proteins throughout history. The accumulation of new data will allow us to refine these trends and better understand the evolution of food practices and inequalities on a European scale.

IDC_nitrogen: a key indicator of dietary inequalities

The study of inequalities has long fascinated archaeologists (59, 60), particularly with the rise of the Gini coefficient, which allows the analysis of economic disparities across periods and cultures (24, 61). However, these indices are mainly based on material criteria (size of dwellings, funerary furniture) and offer only an indirect view of food and health inequalities (50). Our study proposes a new approach by relying on isotopic data from human bones, a direct and dynamic source of information on ancient lifestyles: nutrition is a key factor in socioeconomic status and a powerful marker of social inequalities (62).

By applying the IDC to these data, we obtain a measure comparable to contemporary indicators of inequalities, such as those used by the OECD for income (7, 63). This ratio has several advantages: it allows examining the extremes of a population without being biased by average values, facilitates diachronic comparisons, and avoids the standardization problems that make intersite comparisons difficult. Our results show that this indicator varies significantly across periods, with some being more egalitarian than others.

The robustness of IDC_nitrogen as a dietary inequality marker is particularly evident in inland sites located more than 40 km from the coast (Fig. S7). In these areas, where marine resource consumption is minimal, IDC_nitrogen effectively captures socioeconomic disparities in access to terrestrial animal proteins. In contrast, IDC_carbon values in these contexts are more susceptible to confounding factors such as agricultural practices and C4 plant consumption, as illustrated by the fluctuating patterns observed in Po Valley sites during the Bronze Age (52).

Isotopic ratios are influenced by the foods themselves: for instance, millet, despite isotopic enrichment, was often a low-status staple, whereas marine fish—also contributing to high δ¹³C—was typically linked to higher-status diets in medieval European contexts. In this context, the IDC, by focusing on distributional extremes rather than absolute values, provides a robust measure of relative inequality, reducing the influence of local ecological or agronomic noise. Nonetheless, it cannot entirely eliminate these sources of isotopic variability, and absolute δ¹⁵N values may still reflect environmental or agricultural conditions. Agricultural factors do complicate the interpretation of isotope values in relation to inequality, contributing to isotopic “noise” within populations. Uniform manuring raises overall δ¹⁵N levels without increasing variability, potentially masking dietary differences, whereas uneven manuring may inflate isotopic spread (64), mimicking inequality. Other practices—such as irrigation, crop rotation, foddering strategies, and urban/rural differences in resource access—can further elevate δ¹⁵N and introduce additional within-site variability (16, 21, 22). Accordingly, differences observed between individuals, sexes, or sites should be interpreted with appropriate caution.

Nevertheless, IDC_nitrogen is especially valuable for diachronic comparisons. Unlike carbon isotope ratios, which are strongly influenced by temporal shifts in agricultural practices—such as the introduction of millet in the Bronze Age or New World C4 plants in the Modern era—nitrogen isotope ratios primarily reflect differential access to animal proteins. This makes IDC_nitrogen a consistent marker of social status over time. This trend is particularly evident in our biological sex-based analyses, where IDC_nitrogen consistently reveals a significant overrepresentation of males in the highest decile across all chronological periods (Fig. 3).

Finally, IDC_nitrogen provides a robust relative measure of dietary dispersion, allowing comparison of extremes within populations and enabling cross-site and diachronic analyses while accounting for local environmental variations. Its effectiveness in revealing persistent dietary inequalities makes it a powerful tool for exploring social stratification in archaeological populations. However, the IDC should not be used in isolation: complementary analyses of raw isotopic distributions and intersite comparisons remain essential to fully interpret social and cultural patterns.

Materials and methods

This study encompasses the isotopic analysis of 12,281 individuals from 673 European sites, consolidated into 393 chronological locations (Fig. 4). Carbon and nitrogen isotope data were compiled from multiple database sources, including IsoArch, CIMA, Isotopia, MAIA, and the AIDE project, along with Münster et al.’s compilation for the Neolithic period (43, 65–68) (Dataset S2). The chronological distribution spans from preagricultural to modern times, though postagricultural societies are notably overrepresented in the dataset. The early Middle Ages constitute the largest portion of the corpus, representing 29% of the series.

Fig. 4.

Location of the 393 chronological sites sampled. Circle size represents sample size. Site locations are displayed in jitter to visualize all data. Map created with the Geo Map extension of Orange data mining software (Sources: Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus DS, USDA, USGS, AeroGRID, IGN, and the GIS User Community).

Our sampling protocol was developed to address potential biases in anthropobiological and isotopic data. The effect of sample size on IDC_carbon and IDC_nitrogen values has been estimated using a linear mixed model and a breakpoint analysis (see Supplementary Information, SI4: Biases related to geographical and chronological variations, Figs. S3 and S4). We exclusively utilized bone collagen values, as collagen primarily reflects dietary protein content, whereas bioapatite values correlate with overall diet, including proteins, lipids, and carbohydrates (69, 70). Sample selection required C/N ratios between 2.9 and 3.6 (71), excluding unpublished or unknown ratios. To minimize age-related dietary variations and complications from weaning patterns, immature individuals were excluded from the analysis. Sex determination was crucial for our gendered perspective on dietary inequalities. We included only individuals with confirmed biological sex, determined through skeletal analysis or DNA testing, with confidence levels exceeding 95% in most cases. A small percentage (3%) of probable sex assignments (confidence levels exceeding 80%) were retained, particularly in sites with limited sample sizes. Geographical classification considered coastal proximity, with sites beyond 40 km classified as “inland” and those within 20 km as “coastal.” This distinction reflects historical daily movement patterns of nomadic populations, estimated between 21 and 37 km (72–74), and accounts for the limited influence of regular fish consumption at inland locations.

Statistical analysis employed an IDC (D90/D10) rather than Gini coefficients, as it provides more detailed information about inequality distribution within societies (50). Following Shapiro's normality tests, we implemented nonparametric approaches using Kruskal–Wallis and Dunn's tests to compare distributions across chronological periods, sex, and locations. All analyses were performed using R statistical software (packages: ggplot2, ggpubr, rstatix, dplyr, ggrepel) (75), with significance levels denoted at *P < 0.05, **P < 0.005, and ***P < 0.0005. Sex ratios were presented using a logarithmic scale, where 0 indicates equal representation, and positive/negative values show male or female predominance, respectively.

Detailed chronological distributions, comprehensive statistical analyses, and additional methodological information can be found in the Supplementary Material (SI1).

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Funding

This work was supported by the HORIZON EUROPE Marie Sklodowska-Curie Actions (H2020-MSCA-IF-2019 897565, Project acronym AIDE) for Rozenn Colleter, and the European Research Council (ERC-2018-StG 803676, Project acronym ARCHEIS) for Klervia Jaouen.

Author Contributions

Rozenn Colleter (Conceptualization, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review & editing), Klervia Jaouen (Formal analysis, Funding acquisition, Investigation, Visualization, Writing—original draft, Writing—review & editing), Dominique Garcia (Supervision, Funding acquisition, Project administration), and Michael P. Richards (Supervision, Funding acquisition, Writing—original draft)

Data Availability

All data used in this study are available in the Supplementary material. Datasets S1–S4 contain the complete datasets analyzed in this work. Dataset S5 compiles all bibliographic references cited in the article together with the associated metadata.