Nascent agricultural extensification and distinct household cultivation strategies supported first population centers in Neolithic southeastern Europe

Introduction

Early Neolithic farming systems in southeastern Europe were based on a cohesive “package” of domesticated cereals and legumes, including emmer wheat, barley, and peas, managed using some form of labor-intensive manuring and weeding regime characteristic of smallholder family farming (Bogaard, 2004; Bogaard et al., 2013, 2016). By the Late Neolithic, some settlements had expanded dramatically, reaching up to 140 ha in extent and housing hundreds, if not thousands, of inhabitants (Gyucha et al., 2019; Marić et al., 2021; Osztás et al., 2016). This demographic and spatial growth likely reconfigured farming practices, encouraging more extensive land-use strategies and reduced labor investment per field in response to rising subsistence demands and increasing constraints on access to arable land (Duffy 2014: 235–250). At these larger settlements, where distances between households and cultivated fields were likely substantial, maintaining regular weeding and intensive manuring may have become increasingly difficult. One effective strategy to reduce labor requirements was the use of animal traction to achieve a form of “extensive cultivation,” enabling the planting and harvesting of larger fields, albeit with lower productivity due to reduced per-area labor inputs (Halstead, 1995, 2014; Styring et al., 2017a).

Differences in access to land, labor, and resources between households may have contributed to emerging inequalities in ancient societies (Filipović et al., 2023; Hayden and Cannon, 1982; Twiss et al., 2024), which may have been expressed in part through contrasting modes of intensive and extensive agriculture (Bogaard et al., 2019; Styring et al., 2017a). Intensive cultivation is typically characterized by small, closely managed plots, high labor inputs, and sustained investment in soil fertility through practices such as manuring and weeding, which typically rely on household labor. In contrast, extensive agriculture emphasizes the expansion of cultivated area rather than labor intensity. It generally involves lower per-area inputs, greater reliance on land availability and traction animals, and is more easily scaled beyond the household level. Along these lines, the advent of extensive land cultivation may have contributed to the amplification of social inequality, as the capital required—particularly traction animals—was unlikely to have been uniformly accessible across households. Livestock, like land, could be accumulated and inherited, facilitating the creation of relational debts through loans and the reproduction of inequities across generations (Bogaard et al., 2019; Halstead, 1995; Kohler et al., 2017). Although cattle appear to have been used for traction as early as the 6th millennium BCE at Çatalhöyük and in settlements located in the Western Balkans by 6000 BCE (Gaastra et al., 2018; Kamjan et al., 2022), there is currently little evidence for the use of animal traction in ard- or plow-based agriculture in southeastern Europe. In the absence of animal traction, potential pathways toward agricultural extensification would have been to plant more crops, albeit in less productive plots or with reduced fertilizer or labor inputs, or reorganize available human labor pools from small, household-centered units to more coordinated labor groups operating across multiple households linked by kinship.

As settlements became more densely populated and spatially nucleated (Duffy et al., 2025), securing arable land within close reach of settlements may have been increasingly difficult on the Great Hungarian Plain, where large swathes of the landscape were inundated with water for months at a time (Gyucha et al., 2011; Lóczy, 1988). Before 19th-century river regulation, the Berettyó River flowed in large meanders from the Apuseni Mountains onto the Great Hungarian Plain, dispersing into the Nagy-Sárrét marshes (Gyucha et al., 2011; Plásztán et al., 2017). This valley formed an extensive, poorly drained swamp system, described as the largest wetland of the eastern Great Hungarian Plain.

Here, we explore the relationship between agricultural regime intensity and the organization of household production on the Great Hungarian Plain, where some of Europe's earliest major population centers, such as Polgár-Csőszhalom, Szeghalom-Kovács-halom, and Csökmő-Kapasztás-domb, emerged during the Late Neolithic by the mid-fifth millennium BCE (Gyucha et al., 2019; Raczky et al., 2002; Riebe et al., 2023). Our analysis focuses on the large tell settlement of Berettyóújfalu–Herpály (Herpály) (Figure 1), located within the land-constrained wetlands of the Nagy Sárrét, where we document variation in household-level cultivation strategies, including manuring and the management of field placement under wetland conditions, through an integrated approach that combines cereal morphometric, weed ecology, and carbon (δ¹³C) and nitrogen (δ¹⁵N) analyses of charred cereals.

Berettyóújfalu–Herpály

Herpály is a 0.3-ha tell site, with two-story buildings flanked by 12 ha of flat, outlying settlement areas (Figure 2) (Hofmann et al., 2019; Kalicz, 2001; Raczky and Anders, 2014; Riebe et al., 2023). The site is attributed to the Tisza–Herpály–Csőszhalom culture, extended from the upper Tisza to the Maros–Tisza confluence near the Serbian border (Kalicz and Raczky, 1987). The Herpály subgroup, defined by distinct painted ceramics, the burial of infants near houses, and a higher importance of wild animals contributing to the subsistence economy compared to neighboring groups, is best documented at this eponymous site (Kalicz and Raczky, 1984; Kalicz et al., 2011; Riebe, 2021); however, an additional 35 settlements are known, mostly along the Berettyó and Sebes Körös Rivers. Situated within a fluvial landscape of meadows, swamps, and forests (Frolking, 2021; Gyucha et al., 2011; Kiss et al., 2015), arable land around Herpály was limited. The landscape surrounding Herpály was wet, with the river in the Berettyó valley forming an extensive, poorly drained perennial swamp system (Figure 2). The floodplain was characterized by frequent overbank flooding, oxbow lakes, backswamps, and abandoned meanders, producing a complex mosaic of wetlands interspersed with slightly elevated ridges in the Berettyó–Körös interfluve. Agricultural productivity within this saturated portion of the plain would have been limited, supporting an estimated population of only 100–300 people in a village setting in the absence of plow agriculture (Duffy, 2014: 235–253). Today, the Herpály region is approximately 500–600 mm of annual precipitation, characterized by relatively dry winters and wet summers (Bihari et al., 2018). Climate modeling suggests that during the Late Neolithic, summers were slightly warmer and precipitation was higher (Magyari et al., 2024), probably ranging between 600 and 650 mm annually in the driest parts of the region.

Like all other Neolithic settlements located within 100 km of Herpály, agriculture focused primarily on emmer wheat (Triticum dicoccum) and barley (Hordeum vulgare), followed by einkorn wheat (Triticum monococcum) (Bogaard et al., 2007a; Füzes, 1990; Gyulai, 2010a, 2010b, 2012; Hartyányi and Nováki, 1974; Hartyányi and Sz. Máthé, 1980; Hartyányi et al., 1968; Moskal-del Hoyo and Lityńska-Zając, 2016; Moskal-del Hoyo et al., 2018; Nagy et al., 2014). Peas (Pisum sativum) and lentils (Lens culinaris) were also grown, while wild fruits, such as cornelian cherry (Cornus mas), plums (Prunus cerasifera, P. domestica, P. spinosa), wild strawberry (Fragaria vesca), and occasional grape (Vitis vinifera subsp. sylvestris), were also gathered. Livestock management systems at Herpály relied primarily on cattle (28% of the total faunal assemblage), although details of the proportions of wild and domestic varieties require publishing of the full assemblage (Bökönyi, 1986; Raczky et al., 2020). Suids were also heavily exploited (38%), with a large proportion likely representing wild boar (Bökönyi, 1986). A similar pattern is also seen at the nearby tell site of Polgár-Csőszhalom, where wild boar account for 13% of the identified species (Raczky et al., 2002). Red deer were intensively hunted (21% of large mammals), and roe deer were also occasionally hunted (Raczky et al., 2020).

Excavations at Herpály were conducted from 1977 to 1982 in two areas (Trenches I–II and Trench III, Figure 2); details on excavation and site chronology are provided in Supplementary Material Data Sheet 1. Five main building horizons and corresponding chronological phases have been distinguished within the 300 cm thick Late Neolithic layer sequence of the Herpály tell (Supplementary Material Data Sheet 1, Figure S1). Level 8, featuring the so-called “large burnt horizon” at a depth of 265–340 cm, represents a continuation of the occupation established in level 9 (Kalicz and Raczky, 1984). The complete state of destruction by fire and the deposition of thick, unburned layers of clay on burned remains to “cap” the structure suggest that these houses were deliberately burned as part of a collective ritual practice, a common practice in the Late Neolithic (Chapman, 1999; Kalicz et al., 2011; McPherron and Srejović, 1988; Raczky et al., 2020; Stefanović, 1997; Tringham, 2005). Crucially, the fire preserved the houses along with their internal arrangement and furnishings, including charred seed remains that likely represent the remains of stored grains and pulses kept within the houses.

Nine closely spaced, two-story tripartite houses—most approximately 10–12 m in length and 4–5.5 m wide—were uncovered (level 8: houses 5, 5/a, 6, 7, 8, 9, 10, 11, and 12; Figure 3). Radiometric determinations measured from charcoal recovered from the floor and rubble contexts of these houses show strong overlap, suggesting that these structures were contemporaneous in use (4634 to 4541 cal BCE, 68.2%) (Raczky et al. 2020, Table 1). Within house 11, “house 11/a” and “house 11/b” described different strata: the upper stratum “a” is interpreted as the collapse of the second-story floors of these structures, while the stratum “b” represents the ground floors. Note that houses 5 and 5/a represent two adjacent structures.

Table 1. Berettyóújfalu–Herpály sample context information.

Locations found in Figure 3.

Excavation documentation labels some samples as layer 7, but subsequent analysis and interpretation now classify them as layer 8 (Raczky et al. 2020).

Hearths and other internal features were rebuilt multiple times (e.g., house 9; Raczky et al. 2020: 30–31), unlike other Neolithic structures in nearby Romania that were built only with the intention of single-use burning (Gheorghiu, 2009). It is unclear whether all houses from level 8 were burned at the same time. However, the close proximity of houses 6–10, somewhat separated from closely packed houses 5a, 5b, 11, and 12, suggests that burning took place in a maximum of two events (Figure 3). The fire could not have been contained to a single house in a cluster, given their close spacing and the high flammability of their presumably thatched roofs.

Inferring ancient cultivation regimes through stable isotope analysis and functional weed ecology

Agricultural production is shaped in part by labor accessibility and resource availability, both of which influence cultivation practices and in turn configure intensive or extensive agricultural regimes. Intensive cultivation is closely associated with abundant household labor and agricultural strategies aimed at maximizing agricultural output per unit area (Bogaard et al., 2019). Practices such as tilling, weeding, and manuring are highly labor-intensive but enhance crop yields by improving soil fertility and suppressing weedy competitors (Bogaard et al., 2016; Halstead, 1995). In contrast, land-limited strategies can increase agricultural production by expanding the area under cultivation while reducing labor inputs in intensive agricultural practices (Jones et al., 2000). Although yields per unit area decline under extensification, absolute yields can increase due to the larger cultivation land areas (Bogaard et al. 2016). Reflecting lower labor inputs, extensively managed fields are typically characterized by reduced soil disturbance, lower fertility, and decreased productivity per unit of area (Bogaard et al., 2022). Intensive garden agriculture is characterized by high labor demands, typically drawn from the household, while extensive agriculture involves a lower labor investment, freeing part of the population from routine fieldwork. However, placing larger areas under planting also increases the risk of harvesting bottlenecks that require the short-term mobilization of far more people than is needed for most of the year.

Crop-growing conditions, cultivation practices, and associated labor investments can be reconstructed through stable isotope analysis of cultivars and weed ecologies from the fields in which those crops were grown (Bogaard et al., 2007b, 2016; Fraser et al., 2011). The isotopic composition of charred crop seeds reflects local environmental conditions, in particular water availability and soil fertility (Bogaard, 2004; Bogaard et al., 2013, 2016; Styring et al., 2016b, 2017b). Carbon isotope ratios of C₃ cereal crops are strongly influenced by water availability (Flohr et al., 2019; Jones et al., 2021). Under water-stressed conditions, C₃ plants partially close their stomata, reducing stomatal conductance and intercellular CO₂ concentrations, which results in the enrichment of ¹³C in growing plant tissues (Araus et al., 1997). Carbon isotope ratios measured from charred seeds have successfully informed on light intensity (the “canopy effect”), irrigation, waterlogging, and the spatial separation of plots (Araus et al., 1997; Fan et al., 2018; Ferrio et al., 2005; Flohr et al., 2019; Riehl et al., 2008). Nitrogen isotope ratios of cereal crop plants reflect the nitrogen isotopic composition of the soils in which they grow (Högberg, 1997; Robinson, 2001). Natural influences on soil N include climate, topography, waterlogging, soil salinity, and mycorrhizal fungal associations (Szpak, 2014). In agricultural landscapes, field fallowing and fertilization are important modifiers of the soil nitrogen cycle. The addition of animal manure results in the enrichment of ¹⁵N in crop plants due to the preferential volatilization of ¹⁴N in ammonia from soils (Bogaard et al., 2007b; Choi et al., 2017; Fraser et al., 2011). Manuring can produce significant increases in crop δ¹⁵N values. Experimental field studies in temperate Europe and the semi-arid Levant demonstrate that cereals grown in the absence of manure exhibit very low δ¹⁵N values (below ∼2.5‰), while crops consistently manured over the long-term yield markedly higher δ¹⁵N values of ∼6‰ or more (Antolín et al., 2025; Bogaard et al., 2013).

The composition of weed species in archaeobotanical assemblages provides an indirect yet complementary measure of the growing conditions of ancient cereals (Bogaard et al., 2016; Kroll, 1997). Weeding is a labor-intensive cultivation strategy that enhances crop growth by reducing competition for water and soil nutrients. The distribution of weedy taxa in a field that is plowed, seeded, and subsequently left with minimal human intervention differs markedly from that of intensively managed gardens due to differences in soil disturbance that favor annual, fast-growing, disturbance-tolerant taxa in intensively managed plots, and perennial or low-disturbance adapted taxa in more extensively cultivated field (Bogaard, 2002; Charles et al., 1997; Jones et al., 1999).

Materials and methods

We analyzed charred seeds and botanical remains preserved in situ by burning in level 8 (houses 5, 5/a, 6, 7, 8, 9, 10, 11, and 12; Figure 3), excavated under the direction of Nándor Kalicz, who died in 2017. Seed remains were identified and recovered from 33 contexts by the excavators (and presumably bucket-floated off-site at a later time). The material was analyzed by Helmut Kroll in the 1980s and is listed in Table 1 (see also “Samples” in the Supplementary Material and the R code), using the comparative collection at the Institute for Prehistory and Protohistory at the University of Kiel. Helmut Kroll also died in 2024, and the archaeobotanical material is published here for the first time. The seed remains are currently archived at Eötvös Loránd University in Budapest.

Cluster analysis was used to structure archaeobotanical assemblages for crop-specific weed ecological [Functional Interpretation of Botanical Surveys (FIBS)] analysis. Generally, contexts containing ≥60% emmer or barley remains were classified as emmer- or barley-dominated, respectively, while the remainder were designated as mixed. Although this cutoff is somewhat arbitrary, there is precedent for the approach (Filipović et al., 2019). In practice, most contexts at Herpály are heavily skewed, with one taxon dominating, often comprising over 80% of represented taxa (see Results). Wild and weed taxa from emmer- and barley-dominated contexts were therefore assumed to derive primarily from the cultivation of the dominant crop within each household. Although a shared origin cannot be demonstrated with certainty, this approach reduces depositional mixing and allows weeds to be analyzed by both crop type and household.

To evaluate interhousehold variation in stored resources, flotation contexts, each representing distinct localized collection areas from each house structure, were comparatively assessed for taxonomic patterning (i.e., wheat- or barley- dominant) and for species diversity using hierarchical clustering analysis (HCA): the Bray–Curtis dissimilarity method (Bray and Curtis, 1957), followed by visualization using Ward’s (1963) method. All contexts and taxa were included. Flotation contexts of similar taxonomic makeup from a given house (e.g., “wheat” cluster contexts from houses 11/a and 11/b) were combined for interhouse comparison using HCA to compare the relative abundance of the main domestic species. The taxonomic consistency of crop cereal–legume assemblages across hierarchical clusters (i.e., determining whether houses were more or less similar in their taxonomic representation) was assessed by testing the homogeneity of multivariate dispersion among clusters using PERMDISP (Anderson et al., 2006) based on Bray–Curtis dissimilarities assessed from relative abundances of selected taxa (Triticum and Hordeum grains, Pisum, and Lens). The crop processing stage was determined according to crop–chaff–weed ratios and was quantified according to Van der Veen (1992) for 17 assemblages. Each contained more than 400 individual plant part remains, allowing the stage of processing represented by the sample to be determined.

Cereal morphology

Cereal grain morphometrics reflect both genetic background, growing conditions, and environmental factors, including water availability and soil fertility (Araus et al., 2014; Fuller et al., 2017; Motuzaite Matuzeviciute et al., 2021). As a result, changes in seed size have been successfully used to isolate nascent Neolithic experimentation with cultivation, track the subsequent establishment of large-seeded domesticated grains and pulses (Fuller, 2007; Fuller et al., 2017; Liu et al., 2016; Ritchey et al., 2021; Sun et al., 2024; Willcox, 2004), differentiate between cereal species (Colledge, 2001; Reed et al., 2019), and infer cultivation practices linked to variation in water availability and soil fertility regimes (Antolín et al., 2024; Araus et al., 2024; van Bommel et al., 2021). Experimental work demonstrates that barley grains exhibit a non-linear response to increased watering, with later landraces exhibiting greater plasticity under improved moisture conditions than their Near Eastern progenitors (Araus et al. 2024). Although charring impacts grain shape and size, grain length and breadth remain reliable proxies for relative grain weight, even under substantial mass loss from burning (Ferrio et al., 2004; Larsson and Bergman, 2023).

Here, length measurements were used to estimate grain size in terms of volume (Larsson and Bergman 2023). We chose cereal samples based on completeness (for measurements) and degree of burning (for reliable isotopic values). Measurements were taken from 82 out of the 104 grains, including length, breadth, breadth to base, height, and front of shoulder dimensions (Supplementary Material Data Sheet 1, Section 2.1, Figure S2). The mean values and standard deviations of archaeological seeds were compared with those of modern varieties grown in a single season and field to assess whether the archaeological specimens were likely grown in a single plot and within a single season (Larsson and Bergman 2023). Grains were photographed in dorsal and lateral views using a stereomicroscope equipped with a digital camera, and measurements were taken using ImageJ v.1.8.0, an open-source image analysis program (photographs on Zenodo, https://doi.org/10.5281/zenodo.19487993).

Seed selection for stable isotope analysis

Single seeds were selected for isotope analyses to better capture the range of variation in environmental inputs associated with cultivation practices and to assess interhousehold differences in those practices (James et al., 2025). A total of 104 cereal grains were selected from four houses at Herpály (houses 6, 7, 10, and 11). We aimed to select seeds of different taxa for isotope analysis from the same contexts. This was possible for emmer and einkorn, but only once with barley and wheat, as most contexts containing wheat lacked barley, and vice versa. Each grain was visually assessed for degree of burning, and a sample was subjected to attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) to identify exogenous contaminates including carbonates, nitrates, and humic acids that could alter the original isotopic composition of charred seeds (Supplementary Material Data Sheet 1, Section 2.2). ATR-FTIR spectra from the study sites was compared with reference spectra generated by Vaiglova et al. (2014) (Supplementary Material Data Sheet 1, Section 2.2, Figure S3, Supplementary Material Table 1, raw data). Spectra revealed no peaks associated with nitrate, carbonate, or humic acids, and therefore, charred seed specimens were not pretreated before mass spectrometry. No charring offsets were applied to measured isotope ratios, as isotopic effects are minimal (<0.32‰) (Stroud et al., 2023), and cereals from the Herpály assemblage were chosen with similar degrees of charring.

Mass spectrometry

Each grain was cut in half, broken into chunks, and weighed (∼2.5 mg) into tin capsules for stable carbon and nitrogen isotope analysis (Reed, 2024). The samples were then crushed within the tin capsule during folding. Stable carbon and nitrogen isotope ratios were measured using an isoprime visION continuous-flow isotope ratio mass spectrometer coupled with a vario PYRO cube elemental analyzer (Elementar Analysesysteme GmbH, Langenselbold, Germany) at the Archaeology Stable Isotope Laboratory (ASIL), Kiel University. Stable carbon and nitrogen isotope ratios were calibrated against the VPDB (δ¹³C) and AIR (δ¹⁵N) scales using glutamic acid standards USGS40 (δ¹³C −26.39 ± 0.04‰, δ¹⁵N −4.52 ± 0.06‰) and USGS41a (δ¹³C + 37.63 ± 0.05‰, δ¹⁵N + 47.55 ± 0.15‰) (Qi et al., 2003, 2016). Measurement uncertainty was monitored using two internal standards: millet (δ¹³C −13.75 ± 0.12‰, δ¹⁵N + 1.76 ± 0.04‰, n = 74) and wheat (δ¹³C −27.38 ± 0.08‰, δ¹⁵N + 3.01 ± 0.10‰, n = 76). The reported stable isotope ratios for these internal standards represent long-term averages calibrated to VPDB and AIR with USGS40 and USGS41a.

Analytical uncertainty was calculated following Szpak et al. (2017). Precision [u(Rw)] was ±0.09‰ for δ¹³C and ±0.12‰ for δ¹⁵N based on repeated measurements of calibration standards, check standards, and sample replicates (n = 8 pairs). Accuracy [u(bias)] was ±0.12‰ for δ¹³C and ±0.11‰ for δ¹⁵N based on the difference between the observed and known δ values of the check standards and the long-term standard deviations of these check standards. The total analytical uncertainty was ± 0.14‰ for δ¹³C and ±0.16 for δ¹⁵N. All measurements with replicates are summarized in Supplementary Material Table 2, and duplicates were averaged using the R script.

Weed ecology

We apply functional weed ecology analysis by comparing the taxonomic composition of the assemblage from Herpály with that of modern fields subjected to different cultivation regimes. Environmental inputs that define cultivation regimes, such as watering and soil disturbance, can be inferred from measurable functional ecological traits of weedy taxa, such as leaf area and maximum canopy height (Bogaard, 2001; Charles et al., 1997). This approach, called FIBS, reconstructs past crop management strategies by comparing the functional trait composition of archaeobotanical weed assemblages with those from modern fields grown under known cultivation conditions. Traits such as maximum canopy height, leaf area, and life cycle (i.e., annual, biennial, perennial) provide insights into soil disturbance, moisture, and competition, ecological variables that are impacted by cultivation intensity and field management practice.

Here, we use plant attributes provided by Stroud et al. (2024) to characterize the Herpály weed assemblage (Supplementary Material Table 3, “Weeds”). Flowering periods of weeds in Hungary were estimated from https://floraveg.eu/taxon/ and https://pfaf.org (Supplementary Material Table 3, “FLOWPER”). Discriminant analysis was then applied to differentiate farming regimes based on functional ecological traits and compared with experimentally grounded models that characterize weed assemblages associated with low-labor, extensive field cultivation and high-labor, intensive garden cultivation (Bogaard et al., 2016). Model 1 compares weeds and their functional traits identified in extensively managed cultivation fields located in Haute Provence, France with those associated with small, intensively managed cereal crop plots in Asturias, Spain (Charles et al., 2002; Jones et al., 1999, 2000). The two agricultural regimes were distinguished by differences in soil fertility and disturbance, identified based on canopy height, canopy diameter, leaf area per node/leaf thickness, mean specific leaf area, and length of the flowering period (Stroud et al, 2024). Model 2 uses Asturias, Provence, Evvia (Greece), and Morocco to distinguish between high- and low-intensity regimes that prioritize fertility rather than disturbance. Traits characterizing Model 2 are the same as those of Model 1, but exclude the length of the flowering period. A summary of the attributes for these models can be found in Supplementary Material Data Sheet 1, Table S1.

Weed ecology was analyzed using the R package WeedEco Model 2 version 1.0.0 (Bogaard et al., 2016, 2018; Stroud et al., 2024). Samples from Herpály containing weed remains were analyzed by house floor (house 10 and 10/a and house 11/a and 11/b) rather than by the entire house for Models 1 and 2. The small number of weed remains in any individual flotation context precluded analyzing by localized context, but the distinct floor assemblages in two houses allowed analysis by floor, increasing the sample number from two (houses 10 and 11) to four. This was done to keep the number of weed species per sub-assemblage at four or more and to prevent erratic results (Stroud et al., 2024).

Interpreting stable isotope values

Inferring the application of manure to soils in ancient societies depends in part on an understanding of the distribution of nitrogen isotopes in the ancient floral biome for a given landscape. Here, we estimate nitrogen isotope ratios for floral growth in landscapes minimally impacted by human management using published nitrogen isotope values for wild herbivores recovered from seven Neolithic sites roughly contemporaneous with Herpály, located within a 100 km radius of the site (Supplementary Material Data Sheet 1, Section 2.4, Table S2) (Styring et al., 2016a, 2024; Vaiglova et al., 2023). Collagen δ¹³C values of wild herbivores from Herpály averaged −20.6 ± 0.8‰, indicating these animals browsed in open landscapes (Gillis et al., 2020a). Herbivores exhibited nitrogen isotope values averaging 6.1 ± 0.8‰ in δ¹⁵N and, after taking into account a trophic level fractionation factor of 3–5‰ (Hedges and Reynard, 2007; Vaiglova et al., 2023), suggests low nitrogen isotope values for vegetation in the Herpály region, approximately 2‰ in δ¹⁵N. We further referenced archaeological charred seed δ¹⁵N values measured in this study against manuring bands derived from the nitrogen isotope values measured from cereal crop seeds grown under present-day farming regimes; these values are plotted against the natural log of annual rainfall levels (Styring et al., 2017) (see Supplementary Material Data Sheet 1, Section 2.4 for details). The interpolated “threshold” value for “heavy manuring” at Herpály was 8.1 ± 1.0‰.

We corrected for changes in the carbon isotopic composition of atmospheric CO₂ (δ¹³C_air) during the Holocene and calculated Δ¹³C from δ¹³C_air and charred seed carbon isotope ratios (δ¹³C_p) according to Farquhar et al. (1982):δ¹³C_air was calculated according to a smoothed curve of δ¹³C in atmospheric CO₂ (Ferrio et al., 2005; White et al., 2015), where past levels of δ¹³C_air were obtained from the NOAA ESRL Carbon Cycle Cooperative Global Air Sampling Network database (https://gml.noaa.gov/aftp/data/trace_gases/co2c13/flask/README_surface_flask_co2c13.html). The mean atmospheric CO₂ value for the occupation period is −6.29‰. All charred seed samples were recovered from a tightly constrained chronological horizon.

Δ13C=δ13Cair−δ13Cp(1+δ13Cp1,000)

Many researchers use crop water-status bands described as poorly watered/moderately watered/well-watered, defined by experiments linking watering condition to crop carbon isotope values (Flohr et al., 2011, 2019; Wallace et al., 2013), to interpret Δ¹³C values. Most of these experimental study locations were carried out in semi-arid environments where water is scarce (e.g., Jordan, Syria, Spain). The wetland environment of the Herpály site suggests that drier cultivation plots were likely rare and more coveted than the wetter ones. We therefore avoid using these common Δ¹³C thresholds because of the likely different nature of watering in the Herpály context.

Statistical analysis and figures

Maps were produced using ArcMap 10.8 and Adobe Illustrator 2025. Levine's tests and equal-variance tests were performed before parametric and non-parametric analyses, and specific correlation analyses were chosen based on sample size. Statistical analyses and figures were generated using R v4.5.0 software (R Core Team, 2025). The R script is available on Zenodo (https://doi.org/10.5281/zenodo.19487993).

Results

Assemblage and contexts

More than 33,000 carbonized plant remains were identified from 32 out of the 33 flotation contexts selected from layer 8 (summarized in Table 2, described in Supplementary Material Data Sheet 1, Section 3, and by context in Supplementary Material Table 4). Emmer wheat was the most commonly cultivated cereal (T. dicoccum, 48%), followed by barley (H. vulgare, 21%), einkorn (T. monococcum; 10%), peas (P. sativum; 12%), and lentils (L. culinaris; <1%). Wild/weed taxa and fruits and nuts were not well represented (<1%). A small proportion of wheat grains exhibited greater length-to-width ratios than emmer and einkorn and also demonstrated relatively flat or slightly concave ventral sides, a morphological trait associated with Triticum timopheevii (Kohler-Schneider, 2003). It may be that these larger grains bearing distinctive morphologies from Herpály belong to this taxon; however, this assemblage was analyzed in the 1980s before T. timopheevii was recognized within the prehistoric record (Jones et al. 2000). A close morphological reassessment of the grains and chaff is needed to further resolve the taxonomic assignment of the wheats, as the whole assemblage was not assessed for an exact proportion.

Table 2. Summary of archaeobotanical remains per house unit at Herpály.

The relative proportions of all cereal and legume taxa for all flotation contexts are illustrated by house in Supplementary Material Data Sheet 1, Section 3.1, Figure S4. When wheat or barley is present, one of the two dominates the assemblage (Figure 4). Flotation contexts lacking both wheat and barley tend to have more uncommon species such as cornelian cherry or pulses. HCA indicates that flotation contexts from different houses can be classified into three distinct types: a wheat group (cluster 2), a barley group (cluster 3), and an “other” group, which generally contains few or uncommon taxa (cluster 1) (dendrogram in Supplementary Material Data Sheet 1, Section 3.1, Figure S5).

The different clusters are easily recognizable in Figure 4, which excludes flotation contexts with fewer than 100 cereal remains from the filtered dataset (wheat, barley, pea, and lentil). The PERMDISP permutation test (999 permutations) showed no significant difference in dispersion among the clusters [F_(2,23) = 1.87, p = 0.185], indicating comparable levels of internal variability across groups. Within Cluster 2 (wheat) a PERMANOVA analysis, which compares group centroids rather than dispersion, was conducted to assess whether macrobotanical assemblages varied significantly between houses. The analysis indicated no significant difference among houses [F_(3,11) = 0.77, R² = 0.173, p = 0.703], suggesting that the variation in taxonomic composition within this cluster is not structured by household. A similar finding holds for cluster 3 [barley: F_(2,3) = 0.91, R² = 0.377, p = 0.467].

Overall, the glume-to-grain ratio at Herpály is low, indicating that most contexts held dehusked and sieved grains with little to no chaff and few weed species (Van der Veen, 1992) (Table 2), strongly suggesting that most grains analyzed here were from storage contexts. The exception is sample 23 from house 10, where the rachis-to-barley-grain ratio is high, indicating a stage of crop processing.

Carbon and nitrogen isotopes

Summary statistics for carbon (δ¹³C) and nitrogen (δ¹⁵N) isotope values measured from charred seeds recovered from houses 6, 7, 10, and 11 are presented in Table 3, along with corresponding biplots shown in Figure 5. Carbon and nitrogen isotope values exhibit a significant positive correlation in emmer wheat, whereas no such relationship is observed in einkorn or barley (emmer: r = 0.38, p = 0.015; einkorn: r = 0.07, p = 0.695, barley, r = −0.17, p = 0.375).

Table 3. Summarized results of the stable isotope analysis on charred grain per house at Berettyóújfalu–Herpály.

For the Herpály charred seed assemblage, barley averaged −23.8 ± 0.7‰ in δ¹³C and is depleted 1.3‰ and 0.9‰ in ¹³C relative to emmer and einkorn, respectively, while no carbon isotopic difference was observed between emmer and einkorn wheats (Kruskal–Wallis: χ² = 33.86, df = 2, p < 0.001; Dunn's post-hoc test, Supplementary Material Data Sheet 1, Section 3.3, Table S3). All taxa exhibited similarly high nitrogen isotope values, averaging between 7.7‰ and 8.7‰ (Table 3), with einkorn yielding significantly higher δ¹⁵N values, on average 1.0‰, compared to barley (Kruskal–Wallis: χ² = 10.75, df = 2, p = 0.005; Dunn's post-hoc, Supplementary Material Data Sheet 1, Section 3.3, Table S4).

On average, the carbon isotopic composition of emmer wheat varied between houses, whereas einkorn wheat and barley showed more uniform carbon isotope values across houses. Specifically, emmer exhibited differences in a one-way ANOVA (F = 5.57, df = 3, p = 0.003, Tukey’s post-hoc, Supplementary Material Data Sheet 1, Table S5). Emmer wheat grains from houses 7 and 11 were on average depleted 1‰ in ¹³C compared to those from house 6 (p.adj = 0.019 and 0.005, respectively). Barley showed no significant variation in its carbon isotopic composition across households Kruskal–Wallis test (χ² = 3.31, df = 2, p = 0.192). Barley from both houses 7 and 11 exhibited a wide range in carbon isotope values (–25.0 to −21.8‰ and −25.5 to −23.2‰, respectively) compared to that from house 10, which exhibited low carbon isotopic variation (Figure 5, Table 3). Einkorn showed a significant Kruskal–Wallis result (χ² = 8.40, df = 3, p = 0.038); however, Dunn's post-hoc test revealed no significant pairwise differences between houses, indicating that the variation was diffuse rather than concentrated between specific houses (Supplementary Material Data Sheet 1, Supplementary Table S6). Specifically, wide carbon isotopic variation was visible in einkorn grains from house 7 (–25.9‰ to −21.9‰) and, to a lesser extent, house 10 (−23.6‰ to −21.8‰) compared to grains from houses 6 and 11 (−22.9‰ to −21.7‰ and −23.6‰ to −22.6‰, respectively).

Emmer wheat grains from house 11 were significantly depleted 1.8‰ in ¹⁵N relative to those from houses 6, 7, and 10, which were similar in their nitrogen isotopic compositions (Kruskal–Wallis χ² = 17.57, p < 0.001; Dunn's post-hoc all p’s < 0.01, Supplementary Material Data Sheet 1, Table S7). Einkorn wheat grains from house 11 were also significantly depleted 1.1‰ in ¹⁵N, compared to those from house 10 (ANOVA, F = 4.92, df = 3, p = 0.017; Z = –2.76, p.adj = 0.035, Tukey's HSD test in Supplementary Material Data Sheet 1, Table S8). Barley grains exhibited similar nitrogen isotopic compositions across households (Kruskal–Wallis, χ² = 4.55, p = 0.103). Charred cereal grains from house 7 also exhibited high variation in their nitrogen isotopes compared to those from all other houses (Figure 5).

Cereal grain size

Individual metrical data are provided in Supplementary Material Table 5 and summary cereal metric data are provided in Supplementary Material Data Sheet 1, Section 3.4, Table S9. Standard deviations for grain lengths measured for all three cereal taxa at Herpály ranged from 0.4 to 0.6 mm. Length was positively correlated with breadth in all species (emmer = 0.952, einkorn = 0.914, barley = 0.756, all p’s < 0.0001, Supplementary Material Data Sheet 1, Supplementary Table S10), confirming that length is an appropriate proxy for volume. Differences in grain size between houses were observed for einkorn and emmer, but not for barley (Supplementary Material Data Sheet 11, Table S11, Figure S5). Seed length was positively correlated with both δ¹³C and δ¹⁵N values for emmer wheat (δ¹³C, r = 0.62, p < 0.001; δ¹⁵N, r = 0.31, p = 0.080), but not for einkorn wheat or barley (Figure 6).

Weed ecology

Eleven weed species were represented in the Herpály assemblage (n = 444, including Bromus arvensis, Chenopodium album, and Galium spurium, Supplementary Material Table 3, “Weeds”). Each house floor assemblage contains 4–7 weedy taxa (Figure 7). Discriminant function model 1 (Figure 8a) separates low-input and high-input farming regimes based on differences in soil fertility and disturbance, while model 2 (Figure 8b) separates these two regimes only according to soil fertility (Bogaard et al., 2016).

The house 10/a floor assemblage, characterized by a concentration of barley grain and chaff from an early stage of crop processing, falls at the upper boundary of the low-input range in both models (Figure 8). In contrast, wheat-dominated assemblages from house 11/1 lie at the lower end of the high-input range. Meanwhile, the wheat-dominated assemblage from house 10 displays traits consistent with a high-input, high-disturbance regime. These discriminant function values, associated with each house floor, are plotted against mean carbon and nitrogen isotope values for each cereal crop to evaluate the relationship, if any, between cultivation practices that impact soil fertility and disturbance as reflected in the weed ecology and those that modify plant water use and soil nitrogen (Figure 9). Kendall's tau test indicates a significant correlation (T = 0.815, p = 0.032) between δ¹³C values and the discriminate function values for both models 1 and 2, revealing that low δ¹³C values are associated with low fertility and low disturbance, while high δ¹³C values are associated with high fertility and high disturbance. A positive correlation between δ¹⁵N values and discriminate function values is also visible, although it is not significant (T = 0.593, p = 0.080).

Discussion

Early Neolithic farmers in central Europe practiced a form of “garden cultivation” characterized by small-scale, highly managed, and permanent cultivation plots placed close to settlements and intensively managed as part of household-level production strategies (Bogaard 2007). On the Great Hungarian Plain at the early Neolithic settlement of Ecsegfalva (5800–5500 BCE, Körös culture), agriculturalists also engaged in labor-intensive garden cultivation focused on emmer and einkorn in disturbed, nutrient-rich soils maintained through manuring (Bogaard et al., 2007a, 2013, 2016). The suite and proportions of crops at Herpály are consistent with those at Middle and Late Neolithic sites in the region (Supplementary Material Data Sheet 1, Section 3.5). Fifteen hundred years after the arrival of the first farmers, however, agro-production on the Great Hungarian Plain began to diverge from some of the labor-intensive agricultural practices they had brought with them.

Intensive manuring across households at Herpály suggests equitable access to dung

Manuring was a centrally important agricultural practice at Herpály, as indicated by generally high nitrogen isotope values averaging ca. 9‰ in both emmer and einkorn wheats in houses 6, 7, and 10. These values fall within or exceed the estimated range for heavy manuring in temperate European environments (Bogaard et al., 2007b; Fraser et al., 2011), implying sustained, labor-intensive inputs of animal dung to arable fields. Such intensive regimes would have required not only sufficient livestock but also the coordinated availability of labor for dung collection, transport, and field application, suggesting that households were not constrained by labor availability in maintaining soil fertility. Furthermore, the relatively low nitrogen isotopic variation among emmer and wheats from these houses points to consistent manuring, with dung evenly spread across cultivated plots. This pattern implies that dung was readily accessible to these households, either sourced from their own animals or obtained through communal pooling.

Although intensive manuring associated with concentrated garden cultivation reminiscent of earlier Neolithic agriculture systems was practiced at Herpály, variation between houses and occasionally large ranges in isotopic values in a given house suggest occasions in which cereals were grown in less productive soils, either due to the amount of manure added in a given year or to the nature of the soil itself. The wide nitrogen isotopic variation among both emmer and einkorn wheats in house 7 (6.6‰ – 10.7‰, and 6.9‰ – 10.1‰, respectively), suggests less intensive and more extensive practices were applied to some plots. House 11 exhibits significantly lower mean nitrogen isotope values in emmer and einkorn than other houses (6.9‰ and 7.9‰, respectively), and the functional weed ecology indicates intermediate labor inputs. It is also possible, however, that some plots were simply located in regions of lower productivity already, such as newer fields established later in the history of the settlement that had not yet accumulated the richness of several generations of manuring. House 10 barley δ¹⁵N values average 7.0‰, low by comparison with other houses, and also point toward a burgeoning extensification in some households or fields farther away characterised by less rich topsoils. The degree to which household access to manure shaped extensification decisions remains less explored. It may be that these two households had sufficient labor available but lacked reliable access to the quantities of dung required for intensive garden cultivation; unfortunately, no archaeological traces of dung were recovered. In such cases, extensification may not reflect a deliberate shift toward lower investment strategies, but rather a pragmatic response to constraints on key agricultural inputs. Where dung was scarce or irregularly available, households may have compensated by cultivating large plots with lower per-unit inputs, resulting in greater isotopic heterogeneity and lower mean δ¹⁵N values in crops. This would imply that emerging differences in livestock ownership, herding arrangements, or participation in communal manure-sharing networks had tangible downstream effects on crop management strategies, even within an otherwise broadly intensive agricultural regime.

Maslin cropping supported Neolithic households at Herpály

Maslin cropping, the mixed cultivation of cereal species such as wheat, barley, or rye, is an important strategy for ensuring crop stability under pressure from pests, variable rainfall, or fluctuating soil fertility (McAlvay et al., 2022). Einkorn tends to produce slightly higher yields than emmer under poor or marginal conditions, while emmer has slightly higher yield potential under better-than-medium conditions. Maslin cropping exploits interspecific differences in growth traits across crop species and potentially offers yield stability, weed suppression, and disease resistance (Jones and Halstead, 1995; Snyder et al., 2020). The practice, however, also involves tradeoffs that may sacrifice ease of processing and maximum yield in favor of resilience by exploiting interspecific differences in growth traits of each crop species, thereby broadening the responsiveness of crops to fluctuating growing conditions (Jones and Halstead, 1995; Snyder et al., 2020).

The sorting of the archaeobotanical assemblage into emmer-einkorn contexts that exclude barley (with one exception) and barley contexts that exclude wheats provides strong evidence that the wheats were grown together, while the barley was grown separately. Einkorn and emmer, moreover, are consistently found in similar proportions, with emmer dominating (ca. 80%) and einkorn comprising the rest. If the wheats had been grown separately, it is unlikely that they would have appeared in storage contexts in such similar proportions as seen in most of the samples. The weedy taxa associated with barley-rich assemblages are associated with low-disturbance, low-fertility soils, while wheat-dominated assemblages are characterized by taxa associated with higher disturbance and greater fertility (Figure 9).

The close similarity in the carbon isotope values of einkorn and emmer in houses 6, 10, and 11 suggests they were cultivated together under a shared watering regime rather than in spatially or seasonally segregated plots. Comparable patterns of carbon isotopic similarity between einkorn and emmer have been documented at other European Neolithic sites, including Vaihingen an der Enz and Vráble, interpreted as evidence for mixed cropping (Bogaard et al., 2013; Gillis et al., 2020b). At Herpály, einkorn and emmer from these three houses, exhibiting δ¹³C values averaging between −24.0‰ and −23.5‰, indicate that wheats were well watered. For house 7, the substantially wider variation in einkorn δ¹³C values relative to emmer, together with the >3‰ range in emmer δ¹⁵N values, suggests that emmer and einkorn were not intercropped but instead derived from several different fields or different growing seasons. It is relevant that all wheat isotope values for house 7 come from context 7, dominated by barley, while most of the other contexts sampled from this house are dominated by pulses, fruits, and nuts (Figure 4). As these are the only wheat samples taken from a flotation context not dominated by wheat, we expect that the explanation lies in differences in field location or growing season.

The similarly high nitrogen isotope values for einkorn (δ¹⁵N_mean = 8.7 ‰) and emmer (δ¹⁵N_mean = 8.2 ‰) observed for the entire Herpály assemblage, as well as a positive (if weak) correlation between nitrogen isotope values and grain size observed in emmer (R = 0.31, p = 0.080), further suggests that these two species were in general mixed and grown together to benefit grain size in emmer, dominated by wheat. Seed length is positively correlated with both δ¹³C and δ¹⁵N values in emmer wheat, but not in einkorn wheat or barley (Figure 6). Experimental work demonstrates that manuring increases grain size in emmer wheat and barley, but not in einkorn, reflecting adaptive responses of einkorn to low nutrient availability (Larsson and Bergman 2023). Einkorn wheat is well adapted to nutrient-poor soils and, while moderate fertilization can positively influence yields and components, very high soil nitrogen levels have less effect on grain yields and may increase water lodging (Maneva et al., 2015). Emmer and einkorn wheats were likely grown together as maslin crops at Herpály as part of a risk reduction strategy. It is unlikely that barley was grown as a maslin alongside both emmer and einkorn. At Herpály, barley Δ¹³C values are on average 1.0‰ – 1.5 ‰ higher than those seen in emmer and einkorn, suggesting that barley was grown under different water availability conditions (Figure 10). A comparatively higher Δ¹³C value for barley can be expected due to barley's earlier grain infilling period (Araus et al., 1997; Ferrio et al., 2005). In semiarid environments characterized by spring season rainfall, earlier maturing barley exhibits Δ¹³C values up to 1‰–2‰ higher than wheats planted at the same time (Wallace et al., 2015). However, in the more temperate environments of the Great Hungarian Plain, where modest spring rainfall is followed by summer rains, later-ripening wheats reasonably could have received similar amounts of rainfall as barley. If so, cultivation plots for barley may have been placed on wetter soils located further away from the settlement (Figure 2). Einkorn and emmer wheat perform best on well-drained to moderately drained soil, as both cultivars are sensitive to prolonged waterlogging, particularly during early growth and tillering (Sundgren et al., 2018). Barley, however, is marginally more resilient to poorly drained and episodic flooding than wheat (de San Celedonio et al., 2014). At Herpály, ideal plots closest to the settlement that received the most frequent labor inputs may have been reserved for wheat, while barley was preferentially grown in low-lying or seasonally wet plots farther from the settlement. The lack of correlation between carbon values and seed length in barley suggests that water availability did not affect seed growth at Herpály (Figure 6), at least during the season in which the recovered assemblage was grown.

Lower nitrogen isotope values distinguishing barley and wheat recovered from house 7 and especially house 10 indicate that barley was less intensively manured than wheat and further corroborates separate cultivation of barley from emmer or einkorn (Figure 11). For house 11, similar barley and einkorn δ¹⁵N values, coupled with an ∼1‰ depletion in δ¹³C in barley relative to einkorn, suggests that maslin cropping cannot be entirely ruled out. The clear sorting of flotation contexts into wheat and barley storage areas makes it unlikely (Figure 4), but ethnographic accounts of sowing, harvesting, and processing maslins together, followed by hand-sorting them for storage, do exist (McAlvay et al. 2022: 5).

Emergent extensive agriculture in early population centers is congruent with long-held intensive cultivation practices

Agricultural management strategies at Herpály were broadly shared across the community, but households also practiced diverse and distinct practices that are suggestive of emerging differences in access to labor and land. In a settlement supporting a large population but situated in a landscape characterized by large areas of standing water for much of the year, households were likely constrained in their land-use options and had to prioritize which plots were cultivated close to the house and which would be located farther away. Households likely had to balance access, labor investment, and risk.

The cultivation practices at Herpály mark a divergence from those found outside the Near East among the first farmers in Europe (Bogaard, 2005). Weed assemblages from LBK contexts in Germany and EN contexts in Hungary consistently cluster at the high-intensity end of the investment spectrum, with only minor deviations toward lower-intensity regimes indicated in FIBS analyses (Bogaard et al., 2016: 70). In contrast, although manuring levels remained high at Herpály, its weed assemblages indicate a pronounced decline in labor investment allocated to weeding compared to Europe's Early Neolithic systems.

The divergence in weed assemblages between barley-dominated and wheat-dominated assemblages indicates that, even at the scale of individual households, crop cultivation was spatially differentiated and managed with different levels of labor investment. Emmer-rich contexts are routinely associated with high-intensity weeding practices, while barley-rich contexts are usually associated with lower-intensity weeding. Ethnographic documentation of smallholder families in Greece and Spain using a labor-intensive, high-yield strategy indicates that 90% of all cultivation takes place within 500 m of the settlement (Charles et al., 2002; Jones, 2005; Jones et al., 1999), while smallholder farmers using traction animals in these same environments do not visit cultivation plots for manuring and regular weeding, but will travel up to 4 km to their plots (Charles et al., 2002; Jones et al., 1999). As Late Neolithic settlements in the Carpathian Basin grew in size, residents likely had to travel further to reach their fields. Barley, the heartier crop, was planted farther away, with less oversight, and, according to carbon isotope values, possibly under wetter conditions.

Expanding agricultural production and European inequality

There is a strong potential for inequalities to develop between households farming fixed inherited plots, as well as differences in access to land or manure (Halstead, 1989). The massive increase in the size of some Neolithic settlements in the Carpathian Basin in the fifth millennium BC is well documented (Duffy et al. 2025), and in a wetland environment such as the Nagy Sárrét, dry land would have been a scarce resource (Duffy 2014). The results presented here provide a unique picture of household variability, indicating that although villagers at Herpály were coping with expanding settlement sizes, some may have been doing better than others. Caring for cattle requires reliable access to productive pastures and a reliable labor pool for their upkeep, and villagers with more animals than others were in a better position to manure their fields and ensure successful harvests.

This household variation at Herpály shows that the community had planted the seeds of differential access to resources, perhaps comparable to the differentiation in food storage demonstrated at the 20-ha site of Vinča Belro-Brdo 300 km to the south (Filipović et al. 2023). Vinča shows evidence of house size variability, incipient craft specialization, and prestige objects and long-distance trade (Borić, 2009; Chapman and Souvatzi, 2020; Tasić et al., 2015). Vinča's occupation ended with the intentional burning of structures containing abundant fruit and seed storage at ca. 4550–4525 BCE, at the time of, or just a little after, the burning of structures at Herpály (Filipović et al. 2023). Vinča, the researchers argue, may have succumbed to weakening of communal ties in the face of emergent socioeconomic differentiation. Unlike Vinča, however, occupation at Herpály carried on for another occupation level, perhaps 50–75 years [4579–4442 BCE (68.2%)] (Raczky et al. 2020: 436). The similarity of crop storage types (wheat/barley), the proportion of preserved taxa, and overall low levels of stored fruit indicate that the household variation in land use at Herpály did not translate into broader inequalities, which may be related to their ability to last through the social transformations at the end of the Neolithic for another generation.

Conclusion

By the mid-fifth millennium BC, agricultural production at Herpály indicates that the transition from intensive garden cultivation characteristic of earlier Neolithic settlements to later extensive field-based systems had begun. The continued use of intensive manuring for emmer and einkorn, but more limited weeding—indicated by the middle to low input labor in the weed assemblages at Herpály—suggests a shift away from labor-intensive practices. At the same time, crops appear to have been cultivated under differing watering conditions, with wheats typically cultivated under higher-input regimes and barley more often in lower-input regimes, perhaps in more distant or less frequently tended fields. This strategy favored labor input in emmer but mitigated the potential loss of emmer in dry years by using einkorn and barley, which were less sensitive to extremes in water availability, waterlogging, and lower levels of fertilizer inputs if they were not available.

Crucially, the absence of large-scale agricultural extensification at Herpály, despite large settlement size and constrained land availability, contrasts with subtle but consistent differences in cultivation practices between houses, indicating household-level differences in their access to manure, land, or labor. Such differences may reflect emerging inequalities in access to agricultural resources at a moment when settlement size and population density were placing new pressures on land management. Yet, unlike contemporaneous Vinča settlements, where socioeconomic differentiation between houses may have threatened the social order (Filipović et al., 2023), the distribution and ratios of stored crops at Herpály remained equivalent across houses, perhaps contributing to the settlement's continuity beyond the horizon of ritual burning of the houses.

When situated within the longer regional trajectory, Herpály becomes a crucial datapoint in understanding how inequalities flickered rather than escalated in the Carpathian Basin. Over five millennia, neither house-size inequality nor mortuary wealth indicators showed sustained upward trends, and episodes of heightened disparity were typically short-lived (Duffy et al., 2025). The potential for inequality increased with agricultural intensification and later traction technology but was frequently countered by leveling mechanisms—mobility, community fission, collective labor projects, and perhaps the routine dissipation of surplus through ritual or monumental investment (Feinman and Neitzel, 2023; Johnson, 1982; Miller, 2021). Continuity in cereal storage and the lack of marked disparities in what and how people grew food align with such dynamics: cooperation remained resilient, even as household practices diversified. Thus, rather than representing the onset of extensified, extra surplus-generating agriculture, Herpály exemplifies how dense Neolithic farming communities accepted marginal differences in access to labor, plots, or manure and negotiated resource pressures while experimenting with flexible, risk-buffering cultivation strategies that varied across households and fields, without committing to runaway inequality.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. ChatGPT 5.2 and Claude Code (Sonnet 4.6) were used in the reviewing, annotating, and modifying of R code, and identifying inconsistencies in the manuscript and Supplementary Materials. All work carried out using AI was checked and revised by the authors.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence, and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fearc.2026.1820137/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Richard Madgwick, Cardiff University, United Kingdom

Reviewed by: Melissa M. Ritchey, Washington University in St. Louis, United States

Magdalena Blanz, University of Vienna, Austria