Selective culinary uses of plant foods by Northern and Eastern European hunter-gatherer-fishers

TL;DR

Northern and Eastern European hunter-gatherers selectively used plant foods in pottery, combining them with animal ingredients, as revealed by a multi-method analysis of foodcrusts.

Contribution

A combined analytical approach reveals selective plant food use in hunter-gatherer diets, beyond what lipid analysis alone can detect.

Findings

Foodcrusts contained plant tissues like wild grasses, legumes, fruits, and herbaceous plant parts.

Hunter-gatherers selectively used specific plant species and parts, often in combination with animal ingredients.

Lipid residue analysis alone under-represents the diversity of plant processing in pottery.

Abstract

Carbonised food deposits preserved in pottery vessels, often termed ‘foodcrusts,’ are frequently encountered on hunter-gatherer-fisher (HGF) pottery throughout Northern and Eastern Europe. While lipid residue analysis is frequently employed to determine their composition, this technique favours the identification of animal products. In this study, we present a combined analytical approach, including high resolution microscopic analysis (Digital Microscopy and Scanning Electron Microscopy) together with molecular and isotopic analysis of lipids (GC-MS and GC-C-IRMS) and bulk isotope analysis (EA-IRMS) to further understand the composition of foodcrusts. Eighty-five pottery vessels with foodcrusts were analysed from 13 archaeological sites dating from the 6th to the 3rd millennium BC, of which 58 have allowed for identification of plant tissues, such as wild grasses and legumes, fruits, and the roots, tubers, leaves and stems of herbaceous plants. The results demonstrate that the choice of plant foods was remarkably selective, with hunter-gatherers favouring certain plant species and even their parts over others and combining these with specific animal ingredients. The results also reveal that our knowledge of plant processing in pottery is likely to be grossly under-represented by relying on lipid residue analysis alone.

Figures

Map of study area and example of a foodcrust.Map showing sites included in this study and example of pottery sherd with preserved foodcrust. The map was drawn using GIS software (QGIS 3.3) based on the ASTER digital terrain model (https://earthexplorer.usgs.gov/), which is in the public domain and has no restrictions on re-use. Lakes and river centrelines are from naturalearthdata.com and also in the public domain.

Digital and SEM images showing Amaranthaceae plant parts and detail of cell-tissues.A) Foodcrust fragment extracted from vessel from Serteya, N1075, under digital microscope (magnification 100x, scale bar 500µm); B) Detail of Amaranthaceae calyx and seed under digital microscope (magnification 300x, scale bar 250µm); C) SEM micrographs showing detail of Amaranthaceae seed; D) SEM micrograph showing detail of stem tissue representative of Amaranthaceae family species; E) SEM micrograph showing detail of leaf/stem epidermis tissue representative of Amaranthaceae family species. See individual SEM micrographs working distance and magnification details.

Digital and SEM images showing guelder rose (Viburnum cf. opulus) berry embedded in foodcrust from vessel D9-EBK-26-I from Dąbki and detail of pericarp, mesocarp and endosperm tissue layers of seed.Top left: image of pottery sherd with visible foodcrust and guelder rose berry highlighted (photo taken with macro photographic camera, scalebar 1 cm). Bottom left: Digital microscope image of guelder rose berry extracted from foodcrust (magnification 100x, scalebar 1 mm). Top right: SEM micrograph showing detail of guelder rose berry. Bottom right: SEM micrograph showing detail of guelder rose berry tissues. See individual SEM micrographs working distance and magnification details.

SEM micrographs showing Amaranthaceae plant parts and parenchyma tissues identified in foodcrusts X6344 and X6342 from Syltholm II.A and B show inflorescence tissues with calyces and seeds from foodcrust X6344 (A: highlighted seed in microstructure; B: detail of seed under SEM). C and D show detail of two types of parenchymatous tissue identified in foodcrust X6342. C: potential sea beet parenchyma tissue under SEM; D: potential sedge tuber parenchyma with measurements of individual cells under SEM). See individual SEM micrographs working distance and magnification details.

Plot of stable carbon isotope values of C16:0 and C18:0 fatty acids extracted from HGF pottery analysed in this study.The data are plotted against reference ellipses (68%) derived from modern authentic reference products corrected for the Suess effect [32]. A. Distinguishes vessels that also have full suites of aquatic lipid biomarkers. B. Distinguishes vessels with plant macro-remains in the foodcrusts.

Plot of atomic C:N ratios and δ15N values obtained from foodcrusts adhering to HGF pottery analysed in this study.

Correspondence analysis biplot summarising the occurrence of foods identified in the foodcrusts by region; the two dimensions here capture 39.6% of the variation in the data.

Full text

Introduction

Foraging for wild plants was an inherent component of prehistoric hunter-gatherer-fisher (HGF) subsistence strategies, but direct evidence of this practice, details of the types of foraged plants, and their specific uses, is often obscure. Given suitable recovery techniques, traditional archaeobotanical analysis has been successfully employed to identify the procurement and presence of plants at HGF occupation sites [1–4] but has provided little insight into culinary habits. Moreover, our knowledge of HGF subsistence tends to focus on the analysis of zooarchaeological assemblages derived from hunting and fishing activities [e.g., 5–7] or stable isotopic analysis of human bone collagen [e.g., 8–10], which preferentially reflects protein-rich animal products. These biases limit our appreciation of HGF foodways and our knowledge of forager plant management strategies, which may be key to understanding the eventual transition to crop cultivation and food production. The appearance of pottery in Northern/Eastern European HGF communities provides a rare opportunity for a fresh evaluation of how plant foods were selected, manipulated, prepared and consumed by these communities and is the focus of this study.

The topic of HGF foodways began to attract archaeologists’ attention during the second half of the 20^th^ century. The majority of these studies, however, focused on resource procurement and subsistence activities, rather than on culinary practices [11]. Compared to subsequent periods, social and cultural aspects of HGF culinary life were often ignored. Similarly, the initial proliferation of studies on HGF food procurement clearly focused on animal resources, and it was only after David Clarke (1976) highlighted the potential importance of plants during the European Mesolithic that studies began in earnest [12]. These early studies placed high-caloric plant resources, such as nuts, roots and tubers at the centre of HGFs diet [12,13]. Other plant parts, such as charred remains of vegetative tissues and seeds, were recognised once archaeological sampling and processing techniques advanced (e.g., flotation [14]).

Largely due to preservation and taphonomic issues, early studies throughout Europe tended to focus on questions related to procurement and the early management of plant resources by HGF communities [15]. It was only in the late 1990s, with the proliferation of archaeobotany and palaeobotanical studies, that the reasons behind the presence of plants in the archaeological record began to be investigated. These investigations directed their attention to the most frequently encountered samples, i.e., plant macro remains. More complex amorphous carbonised deposits derived from cooked foods, although frequently encountered both as isolated finds and adhering to pottery (commonly termed ‘foodcrusts’), were largely disregarded by archaeologists, due to the challenges posed by their analysis and a lack of consensus regarding a methodological approach. Despite this, pioneering microscopic studies of foodcrusts often encountered on HGF pottery produced encouraging results: Layers of grass (Poaceae) leaf remains together with fish tissues, including fin rays and scales from cod (Gadus morhua) were identified in Ertebølle vessels from Tybrind Vig [16] and mistletoe (Viscus album) leaves and pollen from ribwort plantain (Plantago lanceolata) and common plantain (P. major) were identified in two sherds from the nearby settlement at Ronæs Skov [17], both in Denmark.

In the last decade, more systematic research on foodcrusts has been undertaken. For example, light microscopic analysis and Scanning Electron Microscopy (SEM) of Mesolithic and Neolithic pottery vessels from the Lower Rhine basin and northern Belgium [4,18–21] have revealed cellular structures preserved in foodcrusts, allowing a variety of plant foods to be identified, ranging from roots and tubers to domesticated cereals during the later Neolithic period. Some promising results have been obtained on starch [22,23] and phytoliths [24] from the western Baltic, including pottery vessels used by HGFs. In the last decade there has been an increase in studies using SEM to investigate formation processes attributable to culinary activities, although these have so far focused on isolated food remains, i.e., amorphous charred fragments [25–28].

In parallel, major advances have been made in the chemical analysis of foodcrusts on European prehistoric pottery, using both molecular and isotope-based approaches [29], often in tandem with direct radiocarbon dating through AMS. Applied to HGF vessels, lipid residue analysis has been remarkably effective in revealing the presence of animal products, such as marine and freshwater fish, terrestrial animals and dairy products [30–34]. In contrast, plant products have been harder to identify [35]. Lipids derived from plants, such as very long-chain fatty acids in addition to triterpenoids and sterols [36] are now beginning to be more systematically studied, but they provide low taxonomic resolution. Combining lipid residue analysis with microscopic analysis of plants offers a complementary and comprehensive approach but has only been undertaken relatively recently and not extensively [19,25,37,38].

This study aims to provide new insights into the contribution of plant resources to food preparation and cooking, by focusing on HGF ceramic vessels from Northern and Eastern Europe. We sought to identify whether there were any differences in the use of plants by region, different ecological settings and how plants were used in conjunction with other types of food resources. With these insights, we aim to evaluate the potential of pottery to provide more detailed knowledge about HGF plant food exploitation than previously possible. The study focuses on HGF pottery from 13 sites with good organic preservation across a range of landscape settings, including the Upper and Middle Don, Upper Volga, Western Dvina and circum-Baltic (Fig 1).

Map of study area and example of a foodcrust.Map showing sites included in this study and example of pottery sherd with preserved foodcrust. The map was drawn using GIS software (QGIS 3.3) based on the ASTER digital terrain model (https://earthexplorer.usgs.gov/), which is in the public domain and has no restrictions on re-use. Lakes and river centrelines are from naturalearthdata.com and also in the public domain.

Materials and methods

Archaeological materials

A total of 85 pottery sherds with substantial amounts of foodcrusts were chosen from sites in different settings (coastal, lagoonal, riverine and lacustrine) to examine whether ecological variables, particularly in plant availability, translated to culinary practice (Table 1). As foodcrusts are not present on all sherds in these assemblages, it is important to note that the sample is selectively drawn and therefore biased to food preparation activities that results in the formation of extensive carbonised deposits. Where suitable botanical records exist [1–3,39–43], the local availability of plant resources at each site was evaluated.

Table 1: Summary table of regions, sites and number of specimens chosen for the current study.

Foodcrusts selected for analysis were gently dislodged from the pottery sherds and subsampled for microscopic and organic residue analysis. Based on initial low-powered and digital microscopic assessment of the foodcrusts, areas of the foodcrusts with visible inclusions were favoured for further SEM analysis.

All necessary permits were obtained for the described study, which complied with all relevant regulations. Permits for subsampling, exporting and analysis of the materials were obtained with all relevant institutions, i.e., permissions to study materials from sites in Russia were granted by The State Hermitage Museum, Ivanovo State University and the Institute for the History of Material Culture RAS. Materials from sites in Poland were accessed with permission of the Institute of Archaeology and Ethnology, Polish Academy of Sciences in Poznań. Lithuanian Institute of History and Lithuanian National Museum granted access to samples from sites in Lithuania.

Creation of experimental reference materials

Cooking experiments were designed with the objective of investigating the type and nature of potential food mixtures and preparations made by HGFs in the study region. These were informed by the archaeobotanical assemblages recovered from these sites and preliminary results from lipid residue analysis and microscopy. Experiments focused on the preparation of two main plant species: (a) guelder rose (Viburnum opulus) berries, and (b) a series of species of the Amaranthaceae family, i.e., beet (Beta vulgaris), goosefoot (Chenopodium album) and saltbush (Atriplex sp.). For (a), berries were gathered in late September 2020 from two locations in the South of England (Brentford and Richmond upon Thames) and frozen immediately after collection. Cooking experiments were undertaken on an open fire in replica pottery vessels of known composition and manufacturing conditions (see Bondetti et al. 2020). Berries were boiled in 100 ml of water both on their own and in a mixture (1:1 ratio) with freshwater fish, i.e., carp (Cyprinus carpio), with the purpose of replicating food mixtures identified through microscopic examination of archaeological foodcrusts. For the second set of experiments (b) stems and flowers containing seeds from two species of Amaranthaceae plants (Chenopodium glaucum and Beta vulgaris) were gathered in July 2021 from the aforementioned locations. Following identical methods as for (a), these were stewed both alone and in combination with fish. Temperature of the food was controlled using a thermocouple and maintained at 120–160° C. Cooking vessels were transferred closer or further from the open flames and active embers depending on required temperature. S1 Data summarises the details about temperature, cooking times and main physical attributes of the food products and formed foodcrusts.

High-resolution microscopy of foodcrusts

Archaeological and experimental pottery sherds were first scanned under a stereomicroscope to identify the most suitable area of the foodcrusts for further analysis. From there, a small portion of the foodcrusts (ranging from 0.5 cm to 4 cm depending on their extent) was gently removed from the ceramic surface using a clean scalpel for detailed study. The foodcrusts were then cleaned using a scalpel and brush to ensure that any remaining clay particles from the pottery surface were removed.

Selected foodcrusts were analysed following a three-step microscopy approach which involved: (i) initial observation using a low-powered binocular microscope (Leica MZ APO) at magnifications between 8x to 50x; (ii) subsequent study using a VHX-5000 Keyence digital microscope at magnifications from 20x to 200x to describe the foodcrusts’ microstructural appearance and the physical attributes in more detail; (iii) targeted SEM analysis (using a variable pressure Hitachi S-3700N Scanning Electron Microscope) was undertaken on sub-samples which were seen to contain organic inclusions including presumed plant remains such as cell-structures, plant tissues or seeds. For these analyses, we considered diagnostic plant and animal particles as those with good-enough preservation and large-enough and articulated cell-tissues to allow plant part/family/genus/species identification.

During SEM observation, images of foodcrusts’ matrices were captured, at 16–20 mm working distance (depending on the fragment size of the foodcrusts). Standard 50x magnification images were captured to allow an in-depth study of their internal microstructure. Visible organic particles and inclusions were also photographed at standard magnifications (100x, 200x, 300x, 500x, 800x and 1000x) when required and measured using the measurement tool available within the SEM instrument software. These were later identified using published reference guides [44], plant reference collections housed at the UCL Institute of Archaeology, Museum of London Archaeology (MOLA) and University of York and the experimental reference materials described above. The latter were particularly important to monitor the impact of cooking and the formation of foodcrusts on visible attributes of the plant structures.

Organic residue analysis

Lipid residue analysis was carried out 74 foodcrust samples using the acidified methanol protocol (S3 Data). Where sufficient foodcrusts were available (n = 64), bulk δ^13^C and δ^15^N isotope analysis were performed. Full methodological details including instrument parameters for lipid residue and bulk isotope analysis of foodcrusts are described in supplementary methods (S4 File). The majority of the data have been previously reported [30–34]. Newly generated lipid and bulk isotope data from the site of Serteya II are presented in the S3 Data

Data analysis

The presence or absence of identifiable classes of plant and animal ingredients, namely tissues of Viburnum opulus, Amaranthaceae, Poaceae, Fabaceae, tissues belonging to plant underground storage organs, other plant macrofossils, fish macrofossil remains, were combined in a contingency table with the presence or absence of lipid biomarkers (aquatic, terrestrial animal and plant). We undertook an exploratory ordination of these data using correspondence analysis, and tested the relationship between food remains, geography and pottery technology using the Mantel test of Jaccard distance matrices. The Jaccard distance was used to characterise the similarities between vessels solely on shared traits, rather than shared absences, which are not meaningful for archaeological materials. The R package vegan [45] was used to undertake these analyses, using 999 random permutations of the distance matrices to calculate test scores.

Results and discussion

Identification of plants

Plant types, including wild grasses and legumes, fleshy fruits or berries, green vegetables and roots/tubers from plant species native to the wider ecological area of research were identified in a total of 58 of the 85 foodcrusts through microscopic analysis. Detailed results are presented by region in the following sections below.

HGF plant food resources

Upper–middle Don river Basin (Cherkasskaya 5, Cherkasskaya 3, Lipetskoe ozero, Kopanische and Staroe Torbeevo).

Foodcrusts from sites in the Don River basin indicate that HGF communities utilised the seeds of several herbaceous plants, specifically wild legumes and wild grasses [see 14]. Fragments of Poaceae seeds were frequently encountered, with positive identification of kernel pericarp cell tissues, such as transverse and longitudinal cells (bran) and aleurone layers. Although visible, bran cells were scarce and poorly preserved, which meant that identification to genus or species levels was not possible. However, the presence of only single-celled aleurone layers excludes several Poaceae genera, such as wild oats (i.e., Avena strigosa), some species of brome grass (Bromus spp.) and barley species (Hordeum spp.) [44]. Additionally, two foodcrusts from the site of Cherkasskaya 3 showed an abundance of small wild legume seeds, with distinct seed coat (testa) patterning and clearly visible palisade and hourglass cells, embedded in their carbonised matrices. Although identification of these seeds to species was not possible, the size and morphology of the seeds indicate they are likely derived from one or more species of clover (cf. Trifolium spp.). Site-specific archaeobotanical investigations in this area and period are scarce, and therefore little is known about the plant species which would have been available to HGFs.

Upper Volga and Dnieper-Dvina area (Sakhtysh 2a, Zamostje 2, Rudnya Serteyskaya and Serteya II).

Results from this region showed that guelder rose berries (Viburnum opulus) and the leaves, stems and inflorescences derived from the Amaranthaceae family were frequently cooked in pots.

Several studies have already noted the presence of Viburnum berries embedded in foodcrusts on pottery sherds from Zamostje 2 [1,2,46]. Additionally, biomolecular analysis has demonstrated the presence of plant biomarkers such as α-Amyrin, *β-*Amyrin and β-Amyrone terpenoids, associated with the presence of fleshy fruits [31]. Further microscopic analysis and subsequent comparison with experimentally created food products confirmed the identification of Viburnum berry epidermis and flesh parenchyma tissues. Experimental materials were also compared with similar remains present in foodcrusts from other sites in the study region, confirming the presence of Viburnum berries also in these pots. The state of remains varied depending on the degree of processing/cooking and charring or carbonisation, ranging from the preservation of whole berries with intact internal tissue layering and preserved seeds, to some samples that only contained small, degraded fragments of parenchyma. Plant parts belonging to species of the Amaranthaceae family were also found embedded in the foodcrusts from the sites of Serteya II and Zamostje 2.

The identification of a small number of almost complete seeds within their calyces was occasionally possible. However, in most of the cases, Amaranthaceae plants were identified through the presence of fragmented seeds or patches of visible cell structures or tissues attributed to different epidermal layers of the seeds, calyces and stems (Fig 2). In some instances, further SEM exploration of the foodcrusts revealed cross-sections of seeds with visible divisions amongst epidermal layers: outer epiderm, composed of isodiametric cells, middle layer of parenchyma and inner epiderm [as per 76: 329]. Previous archaeobotanical studies have shown Viburnum and a few species of the Amaranthaceae family, particularly Chenopodium spp., to be common plant species in this region, specifically those from two species of goosefoot: Chenopodium glaucum/rubrum and Chenopodium album [1,2,41,43]. Based on the results obtained from the microscopic measurements of the visible seed remains within the foodcrusts from these sites and comparison with reference collection seeds and experimentally cooked foods, the Amaranthaceae species which are most likely represented in the pots are those with smaller seeds (<1mm) such as Chenopodium glaucum/rubrum and a number of Atriplex species (e.g., Atriplex prostrata). Large-seeded species such as fat hen (Chenopodium album) and beets are excluded. The role of Amaranthaceae species and Chenopodium spp. specifically as a food component throughout prehistory has long been debated amongst archaeobotanists. Several archaeobotanical studies have suggested their use by Mesolithic and Neolithic communities in Europe [47,48], intensified thanks to the possibilities of pottery vessels [2]. However, their identification as plant ingredients in a culinary context was not verified until recently [19].

Digital and SEM images showing Amaranthaceae plant parts and detail of cell-tissues.A) Foodcrust fragment extracted from vessel from Serteya, N1075, under digital microscope (magnification 100x, scale bar 500µm); B) Detail of Amaranthaceae calyx and seed under digital microscope (magnification 300x, scale bar 250µm); C) SEM micrographs showing detail of Amaranthaceae seed; D) SEM micrograph showing detail of stem tissue representative of Amaranthaceae family species; E) SEM micrograph showing detail of leaf/stem epidermis tissue representative of Amaranthaceae family species. See individual SEM micrographs working distance and magnification details.

Baltic region (Dąbki 9, Syltholm II, Daktariškė 5 and Kretuonas 1B).

Lipid residue analysis has previously shown marked differences in the use of pottery throughout the Baltic region [32]. Freshwater animal fats were a major component of the foodcrusts adhering to eastern Baltic Narva culture pottery, as well as pottery from Dąbki in Poland, while marine-derived lipid biomarkers predominated in Ertebølle ceramics from the western Baltic [32]. Ertebølle ceramics also had a greater proportion of terrestrial animal resources, including ruminant products, compared to those in the eastern Baltic [33]. Microscopic analysis also revealed a higher frequency of freshwater fish in pottery from the Lithuanian sites in contrast with a more varied range of animal products identified through lipid residue analysis from the site of Syltholm II in southern Denmark.

This is consistent with the identified plant remains within foodcrusts from all four sites. As plant types varied across the region, this likely reflects a combination of the local environmental and cultural choices. Fragments of cf. Poaceae seeds were frequently encountered, with visible kernel cell tissues, in particular aleurone layers in foodcrusts from Daktariškė 5 and Kretuonas 1B. At Dąbki, Viburnum berries were frequently found in the foodcrusts with high concentration of freshwater fish biomarkers, implying a targeted selection of ingredients (Fig 3). The absence of Viburnum berries and seeds in the archaeobotanical record from Dąbki suggests that this species was targeted for a specific purpose which required pottery vessels for their preparation, rather than being consumed regularly. In contrast, macrobotanical analysis of plant remains from archaeological deposits at Dąbki indicate the gathering, preparation and consumption of starch-rich plant parts such as roots and rhizomes of sea beetroot (Beta vulgaris subs. maritima) and flowering-rush (Butomus spp.) [3,49]. No remains of these taxa were found in any of the analysed foodcrusts from Dąbki, however.

Digital and SEM images showing guelder rose (Viburnum cf. opulus) berry embedded in foodcrust from vessel D9-EBK-26-I from Dąbki and detail of pericarp, mesocarp and endosperm tissue layers of seed.Top left: image of pottery sherd with visible foodcrust and guelder rose berry highlighted (photo taken with macro photographic camera, scalebar 1 cm). Bottom left: Digital microscope image of guelder rose berry extracted from foodcrust (magnification 100x, scalebar 1 mm). Top right: SEM micrograph showing detail of guelder rose berry. Bottom right: SEM micrograph showing detail of guelder rose berry tissues. See individual SEM micrographs working distance and magnification details.

Remains of rhizomes were, however, identified in food residues from the site of Syltholm II in Denmark, with parenchymal and vascular tissues being identified, indicating cooking and consumption of plant underground storage organs. Detailed study of the characteristics displayed by vascular and parenchyma tissues and measurements of parenchyma cells identified in the Syltholm II foodcrusts indicate the use of roots from the Amaranthaceae family (i.e., cf. Atriplex spp., cf. Chenopodium spp. cf. Beta vulgaris subs. maritima) and potential sea club-rush tubers (cf. Bolboschoenus maritimus). In the first case, visualisation of secondary growth of anomalous vascular and storage parenchyma tissue in transverse section in the form of concentric rings of vascular tissue with broad bands of storage parenchyma in between them are consistent with sea beet or similar species [see [50–52]] although caution is needed as only partial tissues are present. This anomalous growth is characteristic of roots from plants belonging to the Amaranthaceae family and differs considerably from the microstructure observed in other plant storage organs. Remains of sea club-rush tubers were identified by the solid parenchyma tissue with cells measuring between 25–35µm, combined with randomly placed vascular bundles throughout it. Among the plants whose rhizomes and tubers could have been exploited is a variety of beets, reeds, sedges and waterlilies. This fits with previous archaeobotanical investigations in the area, which identified roots, tubers and bulbs as potential food resources including sea beet (Beta vulgaris ssp. maritima), charred wild garlic bulbs (Allium ursinum) and pignut tubers (Conopodium majus) [47,53].

Alongside underground storage organs (roots/tubers), the Syltholm II foodcrusts also exhibited well-preserved remains of other plant parts (i.e., inflorescences and seeds) belonging to the Amaranthaceae family, as seen from sites in the Upper Volga region and Upper Dnieper area (Fig 4). Again, measurements of the identified calyces and seeds indicate the use of small-seeded species such as saltbush (Atriplex spp.), previously recorded in great quantities in archaeobotanical research in the study region [47,53]. Plant components were found in mixtures with animal resources, both visible under SEM and in their lipid compositions. Identified animal resources include aquatic products, such as freshwater fish, alongside a few samples for which lipid residue analysis have shown the presence of dairy products, most likely acquired from adjacent farming groups [33].

SEM micrographs showing Amaranthaceae plant parts and parenchyma tissues identified in foodcrusts X6344 and X6342 from Syltholm II.A and B show inflorescence tissues with calyces and seeds from foodcrust X6344 (A: highlighted seed in microstructure; B: detail of seed under SEM). C and D show detail of two types of parenchymatous tissue identified in foodcrust X6342. C: potential sea beet parenchyma tissue under SEM; D: potential sedge tuber parenchyma with measurements of individual cells under SEM). See individual SEM micrographs working distance and magnification details.

Organic residue analysis

Overall, it was possible to directly compare foodcrusts from 85 vessel units that were subjected to microscopic, lipid residue and bulk isotopic analysis, providing an opportunity to comment on the complementarity of both suites of techniques.

In addition to the widespread microscopic evidence for plants, nearly all the foodcrusts had lipid distributions typical of degraded animal fats. Carbon isotope measurements of the principal fatty acids (δ^13^C_FA_) and the presence of distinctive lipid biomarkers show that these were predominantly derived from freshwater fish or shellfish (Fig 5A). The presence/absence of plant macro remains also shows little correspondence with the δ^13^C_FA_ values (Fig 5B), suggesting that they did not make a significant contribution to the overall lipid content of the residue. Nevertheless, plant lipid components, including triterpenoids (α- and β-amyrins) and phytosterols, were identified as minor constituents of the extracted residue in 52 out of the 85 analysed foodcrusts [54]. These have little diagnostic value and are produced by a wide range of plant parts and types (i.e., fruits, leaves, bark, wood, and resins). Only ca 60% of foodcrusts with plant derived lipids contained evidence of plant macro remains, with no association with the specific types of plants identified microscopically (see S2 Data). However, 50% of the samples with macrofossils attributable to Viburnum spp*.* also contained either α- or β-amyrin, whilst these compounds were less frequently identified in samples with other plant macrofossils (S2 Data). The co-occurrence of Viburnum berries and amyrins has been noted at Zamostje 2, with a drastic decrease in later Lyalovo culture comparing to the earliest stage represented by the Upper Volga culture [32] and is corroborated here through the analysis of a broader range of samples. In contrast, 92% of the foodcrusts which contained fish particles (i.e., scales, fragmented bones) also contained lipids derived from heating aquatic products, compared to 69% of foodcrusts without visible fish fragments. Overall, the two approaches provide complementary evidence: fatty animal tissues and dairy products dominate the overall lipid content but plant tissues, being microscopically identifiable, are fairly ubiquitous.

Plot of stable carbon isotope values of C16:0 and C18:0 fatty acids extracted from HGF pottery analysed in this study.The data are plotted against reference ellipses (68%) derived from modern authentic reference products corrected for the Suess effect [32]. A. Distinguishes vessels that also have full suites of aquatic lipid biomarkers. B. Distinguishes vessels with plant macro-remains in the foodcrusts.

A further approach involves considering foodcrust nitrogen isotope values (δ^15^N) and atomic carbon to nitrogen (C:N) ratios. This approach may provide a better, albeit crude, quantitative estimation of the plant versus animal contribution to the foodcrust as a whole, as it is sensitive to other non-lipid macromolecules such as proteins and carbohydrates [29,55]. These data are plotted in Fig 6 and show that the δ^15^N values and atomic C:N ratios are exponentially negatively correlated. Samples with lower δ^15^N values and higher atomic C:N ratios are likely to have a higher plant contribution by total mass. As many of the samples have relatively low atomic C:N ratios and higher δ^15^N values, we conclude that they are composed mainly of aquatic protein-rich tissues, in agreement with the lipid results. Nevertheless, plant macrofossils are still present in these samples. For the few samples where most of the bulk foodcrust appears to have been derived from plant tissue (e.g., atomic C:N ratio >10), the incidence of plant-macrofossils is no greater. We conclude that (a) the presence of identifiable plant fragments does not reflect their quantitative contribution to the foodcrusts, and (b) that their preservation is variable and perhaps influenced by the degree of charring.

Plot of atomic C:N ratios and δ15N values obtained from foodcrusts adhering to HGF pottery analysed in this study.

Site location and pottery technology

A question remains as to whether the use of pots for certain foods, or combinations of specific foods, simply reflects the opportunistic use of local resources, or whether there was a coherent set of social traditions associated with these phenomena. To explore this question, we undertook a correspondence analysis, which calculates the degree of similarity between foodcrusts based on the presence or absence of food remains. No additional down-weighting or data transformation was applied. The results (Fig 7) confirm our expectations that site location plays an important role, with circum-Baltic sites differentiated from those further east and south by the presence of roots and tubers or Viburnum. A complex interaction between the variables differentiates the Don and Volga regions, with foodcrusts from the Dnieper-Dvina region having similar traits to those from the Volga region. We tested the strength of the geographic signal for the earliest phase of ceramic distribution using the Mantel test of distance matrices, confirming a small but significant correlation of 0.25 (two-sided p-value = 0.001) between site location and culinary use. This improves upon previous work [54], which did not recover evidence of a correlation between location and culinary practice at each site based upon absorbed lipid residues alone. The correlation is likely to be driven partly by local vegetation, which obviously influenced the set of plants used by HGFs at any given time. Our study samples material from steppe, forest and surrounding ecotones, and we expect a pattern of spatial autocorrelation in our results due to the different resources available at each locality.

Correspondence analysis biplot summarising the occurrence of foods identified in the foodcrusts by region; the two dimensions here capture 39.6% of the variation in the data.

However, the pots themselves provide independent data about social traditions in the form of traits associated with their manufacture. Drawing upon previously published trait descriptions of these vessels [54], we tested whether there were broad affinities between technological characteristics and their culinary use. These traits describe fabric preparation and modelling technique; conscious decisions made by people. We did not consider sets of features describing morphology, as foodcrusts only form on cooking vessels, nor decoration, as few of the pots analysed here were richly decorated. Using the Jaccard matrix as our proxy for cultural ‘distance’ between pots, the Mantel test indicates a moderate correlation of 0.48 (two-sided p-value = 0.001) between pottery technology and culinary use. When we control for location using a partial Mantel test, thus addressing spatial autocorrelation, the correlation remains statistically significant at 0.20 (two-sided p-value = 0.002). This implies that, among the many potential sources of technological variation, vessels that were technologically different from one another tend to have been put to different uses. This remains true, to a lesser extent, for pots found within the same locality or region. Alternatively, or additionally, the traditions involved in pottery manufacture and use may have evolved between communities along similar gradients due to the mechanisms by which knowledge was shared between communities. Either way, the coupling of technological and usage traits cautiously suggests there was a degree of cultural coherence in the use of pottery by hunter-gatherers, which varied over similar spatiotemporal scales.

Towards an understanding of HGF cuisine

The combined application of microscopy techniques and lipid residue analysis to the study of foodcrusts from HGF pottery vessels has proved a successful approach for the identification of a varied range of food commodities, both plant and animal resources. Large-scale isotopic and molecular analysis of foodcrusts and absorbed residues from pottery vessels from sites across the study area has shown the key role aquatic resources (e.g., fish) played in HGF cuisine which were likely a staple food source for the HGF communities. This point is widely supported by lipid residue analysis of pottery without preserved foodcrusts [30–32,34] as well as numerous archaeozoological and technological finds associated with fishing [42,46,49,56–59].

Microscopic analysis of the pottery shows however that fish was not the sole ingredient of these food preparations, but rather the primary ingredient for more elaborated culinary practices. Based on this new evidence, we can dismiss the possibility that European forager pottery had a predominantly non-culinary role, for example for processing fish oils for other purposes such as fuels, sealants or lubricants [60]. Nevertheless, here we are restricted to the analysis of pottery with foodcrusts; a use-alteration marker which perhaps only forms through specific culinary practices or post-depositional conditions. We note that a wider lipid residue survey of pottery use, including vessels with no foodcrusts, points to greater ‘sub-regional’ variation in pottery use across this study region [32].

Although a range of plant resources featured in the food deposits from across the study area and, in some cases, certain plant resources were present in more than one region, our results show that there was a general tendency towards combining specific foods into distinct preparations and in particular regions (Fig 7). For example, in many instances Viburnum berries often appear to have been the sole plant ingredient mixed with freshwater fish. The preparation of Viburnum berries specifically with fish is represented in foodcrusts from pottery vessels from sites in both the Upper Volga and Baltic regions and appears to be especially important for the site of Dąbki.

Two other food mixes were prevalent in the foodcrust record from the analysed sites. The combination of wild grasses and legumes with freshwater fish (e.g., cyprinids) was widely identified in samples from the Middle-Upper Don basin, whereas green vegetables (Amaranthaceae) and freshwater fish was particularly important role in HGF diets at Serteya II and Zamostje 2. Amaranthaceae inflorescences were also identified in foodcrusts from Syltholm II, here being clearly mixed in with other plant ingredients including roots and tubers alongside fish and dairy, suggesting the use of the pots for the preparation of a wider range of plant and animal food products.

Plant uses and culinary choices

Undoubtedly the variety of plant foods that was available to HGFs is vast, as are the ways in which these were likely prepared and consumed by these communities. Our analysis of carbonised food remains from HGF pottery vessels shows that certain plant species were combined with other specific ingredients. Pottery enabled this practice and indeed the capacity to combine ingredients in specific ways could have been a factor motivating the adoption of this technology. Furthermore, these culinary practices show commonalities over delineated geographical areas. The correlation between certain types of plant and animal commodities and pottery is especially noticeable when contrasting the results from the foodcrust analysis against archaeobotanical and zooarchaeological records. For instance, common fruits, nuts, roots and tubers are more frequently recovered as charred remains during flotation than identified within foodcrusts. This discrepancy might be due to the fact that these foods, amongst the most robust plant parts, were preferentially cooked using ‘aceramic’ methods, e.g., roasting directly on fires or in other structures, such as pits or hearths. Whilst taphonomic biases need to be considered, a lack of complementarity between faunal/botanical assemblages and pottery usage patterns highlights the complexity of HGF culinary practices [61,62].

In some cases, the difference is stark. For example, Viburnum spp., an edible berry native to most of Europe, hardly seems to have been exploited in Western Europe yet is deeply embedded in Eastern European culinary traditions. Today, Viburnum opulus, known as kalyna or kalina, is utilised in a range of food and beverages across Poland, Ukraine and Russia, due to nutritional and medicinal properties [63]. The berries are mildly toxic, if consumed raw, and have a bitter taste and release a strong pungent smell during cooking. At Zamostje 2, Viburnum berries were preferentially cooked with freshwater fish, which would have made them more palatable. A similar dish called Mos’, a sweet jelly made by pulverising dried fish and mixing it with fish skins, a variety of berries and occasionally tubers, is known to have been consumed by the Nivkh people in the Russian Far East [64].

The conscious selection of plant resources was not limited to plant type but also extended to plant parts that were targeted as food components and added to the pot contents. While the macrobotanical records from the sites of Serteya II and Zamostje 2 exhibit high concentrations of seeds from ruderal species, mainly goosefoot (Chenopodium album, Chenopodium glaucum/rubrum), the foodcrusts from these sites contained complete Chenopodium spp. inflorescences, including the calyces, stems and leaves. Additionally, the presence of both green and underdeveloped calyces (more likely to be associated with flowering season during May-June) and ripe calyces with embedded seeds (which appear around early July until late August), it is likely that HGF communities at these sites were gathering these plants throughout the spring and summer, rather than solely when in fruit.

Conclusions

While lipid residue analysis has been extensively conducted on pottery vessels to study culinary practices, this approach has mainly revealed the presence of animal products, with the identification of specific plants presenting a persistent challenge, despite recent advances [65]. Equally, although high-resolution microscopic analysis has been successful at identifying organic tissues in carbonised food residues, it is limited to products and cooking processes that leave identifiable structures within the amorphous mass. A combined approach, as shown in this study, is clearly recommended for the future. More importantly, perhaps, the study highlights what is missing when single techniques are used in isolation. In this case, we were able to re-evaluate the role of HGF European pottery, albeit on a subsample of vessels with preserved foodcrusts, to show unequivocally that plants were incorporated into ceramic culinary practices. We show HGF communities approached plant foods selectively, consciously choosing certain species over others, and combining these with specific animal ingredients. Such practices would have created new aesthetics, tastes, flavours and textures that would be hard to achieve without pottery technology and may have contributed to the motivation for its invention/adoption.

This study has shown the successful application of ORA and high-resolution microscopy to the study of carbonised food remains in pottery vessels. Whilst we are limited by the preservation of foodcrusts, our results show the importance of plant food use in Mesolithic and Neolithic European HGF cuisines. Further experimental and reference studies are needed to deepen understanding on foodcrusts formation during cooking and how different plant and animal resources used might affect this. It is important to also note that preservation biases, post-depositional processes and regional variability on archaeobotanical baselines present additional challenges for identification of plant remains in foodcrusts. Therefore, regional reference collections including both botanical specimens and food products are also vitally important.

Supporting information

S1 DataExperiments.Summary of experiments and parameters carried out as part of food products replication.(XLSX)

S2 DataMicroscopy results summary.Summary of raw data and results from high-resolution microscopy.(XLSX)

S3 DataORA results summary.Summary of raw data and results from Organic residue analysis.(XLSX)

S4 FileProtocol.ORA methodology. Detailed organic residue analysis methodology.(DOCX)