We retrieved 90 microremains during the 12 environmental control samples carried out during the dental calculus analyses (Supplementary Table 1). Our results based on this procedure summarize as follows: synthetic and plant fibers and hairs, fungal spores and hyphae, conifer and palm pollens, insect debris, a very low number of phytoliths and some unidentified small starch granules (Supplementary Figure 1). These tests control have allowed us to characterize the flow of contaminations through seasons, suggesting the possible relation with environmental causes. The residues more commonly found were fibers, fungal remains and pollens. Starch granules were very rare in the control samples, and the morphotypes described below were never recovered, supporting the archaeological nature of the microfossils identified within dental calculus.

The state of preservation of dental calculus was moderate, with most of the samples showing weak bands of tartar or small patches on the buccal and lingual surfaces of the teeth. The weight of the dental calculus samples ranged between 0.1–34 mg (mean= 5.1; median= 2.8) (Supplementary Table 2). A total of 194 starch grains were retrieved within dental calculus samples. This number can be considered quantitatively low given the number of individuals analyzed. The relatively low presence of starch grains could be explained by the small size of the calculus samples, because only small amounts of dental calculus could be sampled from most individuals. In fact, there is a significant positive correlation between the size of the samples and the number of starch grains identified by individual (p = 0.01, rho = 0.55), which suggests that the preservation conditions of dental calculus determine the number of starch grains found in our archaeological samples. However, our results can be considered satisfactory since it has allowed the recovery of microresidues in such small quantities.

Of the total number of 21 individuals who preserved dental calculus on their teeth and, therefore, were analyzed, the vast majority (n = 18) provided starch grains in greater or lesser quantities (Table 2). Only the Chalcolithic/Middle Bronze Age individuals from El Toral III failed to discover starch grains. We have identified 47 starch grains in Late Neolithic individuals, 20 starch grains in Chalcolithic individuals, 81 starch grains in Early Bronze Age individuals and 46 starch grains in Middle Bronze Age individuals (Table 2).

From the 194 starch grains identified, 152 granules were taxonomically addressed to a tribe level, and they correspond to Triticeae, Paniceae and Fabeae tribes. Triticeae grains were the most found in the sample and all periods, followed by Paniceae grains and the testimonial presence of Fabeae grains. The other 42 starch grains were assigned as undetermined because their small size prevents a more precise taxonomic identification. Based on their morphometric characteristics, three different morphotypes of starch grains were identified among calculus samples (Supplementary Table 2; Figure 2). Morphotype I represented 72.4% of the identified grains (n = 110), and it is characterized by a bimodal distribution typical of most grasses of the Triticeae tribe (Figure 3). Large grains (≤ 15 μm), defined as type A, are round to oval in 2D shape, lenticular in 3D, with a central or slightly sunken hilum and few lamellae. Smaller grains (< 10 μm), defined as type B, are almost spherical and have a central hilum. The mean size of these starch grains was 21.1 μm. The length of the Triticeae grains ranges between 4.7 and 63.4 μm. The size by sites was: El Hondón (14.6-35.6 μm), El Toral III (23.3-30.8 μm), Los Avellanos I (13.1-52.9 μm), Los Avellanos II (16.1-24.2 μm), La Castañera (4.7-35.3 μm), La Fragua (21.3 μm), El Mirón (6.4-56 μm), El Agua (9.5-49.5 μm), Kaite (8.3-48.4 μm), and Palomera (9.3-63.4 μm) (Supplementary Table 3). Morphotype II described 25% of the identified starch grains (n = 38). They have a 3D polyhedral shape, with a central hilum and fissures radiating from it. The extinction cross is visible, with straight arms, while lamellae are not identified. This type of grain is consistent with those of the Paniceae tribe (Figure 4). The mean size for all sites was 19.8 μm, while the length by site was: El Toral III (11.9-20 μm), Los Avellanos I (25.4 μm), Los Avellanos II (17.1-27.3 μm), La Fragua (13.9-16.9 μm), Palomera (23.8-33 μm), El Mirón (17.9-30.8 μm), and El Agua (8.2-12.5 μm) (Supplementary Table 3). Morphotype III represented 2.6% of the identified grains (n = 4) and has an oval/reniform morphology, with a Maltese cross -x- shaped, whose lamellae are visible and have fissures inside the hilum. These characteristics are usually common in legume species of the Fabeae family (Figure 5). The mean size of them was 31.8 μm. El Toral III Fabeae grain, whose length is 17.7 μm, is fractured and may have been processed. Los Avellanos I grain was greater and measured 52.9 μm, similar to the El Mirón grains, whose dimensions were 34 μm and 44.2 μm, respectively (Supplementary Table 3). The length of this starch grain was 9.3 μm (Supplementary Table 3). No other diagnostic morphotype of starch grain characteristic of domesticated plants or modern wild species of nuts or fruits was identified within the dental calculus samples (Figure 6).

Triticeae experimental starch grains were the tribe that provided the most homogeneous results between species (Supplementary Table 5). Although all of them showed great variability and were relatively common in terms of mean sizes, T. aestivum experimental grains were larger than others, ranging between 4.9-45.6 μm (mean size of 16.7 μm). This evidence is supported by the Pairwise Wilcoxon test, which highlights that T. aestivum starch measurements are significantly different from those obtained for all Triticeae sampled species (Supplementary Table 6). T. dicoccum size was lower than T. aestivum, ranging between 3.4-72 μm (mean size of 11.4 μm), but some starch grains reported the highest maximum values of the taxa analyzed. The rest of the Triticeae experimental species were very similar in size and did not report (or with very little significance) statistically significant differences (Supplementary Table 6). T. monococcum between 3.3-31.5 μm (mean size of 9.6 μm). H. vulgare between 3.6-34.7 μm (mean size of 9 μm). T. dicoccoides between 4.1-32.5 μm (mean size of 8.8 μm). And S. cereale between 2.9-50.3 μm (mean size of 8.5 μm) (Supplementary Table 5; Figure 7A).

With regard to Paniceae/Andropogoneae experimental starch grains, the size of wild species was smaller than that of domestic ones. S. verticillata ranged between 6.5-17.9 μm (mean size of 10.4 μm), and S. viridis ranged between 5.7-13.3 μm (mean size of 9.1 μm). Among domestic millets, S. halepense was slightly greater than S. italica and P. miliaceum, ranging between 4.7-26.5 μm (mean size of 13.9 μm). S. italica ranged between 5.5-33.2 μm (mean size of 13.6 μm), being larger than P. miliaceum, which ranged between 6.6-19.9 μm (mean size of 11.4 μm) (Supplementary Table 5; Figure 7B). These results were confirmed through statistical comparison, with their lengths being significantly different, except for S. italica and S. halepense (Supplementary Table 6).

Fabeae experimental species showed the greatest variation among tribes. L. culinaris ranged lengths between 8-50.2 μm (mean size of 26.5 μm), V. faba ranged between 13.1-51.5 μm (mean size of 29.1 μm), V. sylvatica ranged between 4.7-45.2 μm (mean size of 20 μm), and P. sativum ranged between 3.4-26.9 μm (mean size 11 μm). L. culinaris and V. faba experimental grains were similar in size, and they had lengths larger than those of the wild native V. sylvatica grains and domestic P. sativum grains, especially in their minimum values (Supplementary Table 5; Figure 7C). In fact, L. culinaris and V. faba were the only species which did not show significant differences between them (Supplementary Table 6). Concerning the two Aveneae/Poeae species experimentally analyzed, both taxa had small sizes. A. sativa starch grains ranged between 4.4-16.3 μm (mean size of 8.9 μm), while A. fatua ranged between 3.3-11.6 μm (mean size of 7.3 μm). Thus, A. sativa reported slightly greater lengths than A. fatua (Supplementary Table 5; Figure 7C), being statistically different (Supplementary Table 6).

Finally, some wild fruit species were also statistically analyzed. C. avellana experimental starch grains were the greatest and most variable species, ranging between 6.7-56 μm (mean size of 23.3 μm). Q. ilex ranged between 5.7-66.2 μm (mean size of 18.4 μm), Q. robur ranged between 5.4-30.7 μm (mean size of 14 μm), P. pinea ranged between 4.7-45.6 μm (mean size of 11.5 μm), C. sanguinea ranged between 4.1-11.4 μm (mean size of 7.1 μm), P. spinosa ranged between 6.9-16 μm (mean size of 11.8 μm) and A. unedo ranged between 4.3-19 μm (mean size of 7.3 μm) (Supplementary Table 5; Figure 7D). All of them reported statistically significant differences (Supplementary Table 6).

A variety of other plant structures of probable non-dietary origin were also found entrapped with the tartar, including, from largest to smallest quantity, fibers, fungal spores, phytoliths, charcoals, pollen granules, and wooden tissues (Supplementary Table 4; Supplementary Figure 2). They may have been related to a wide range of non-dietary activities (Radini et al., 2017). The presence of plant fibers can be associated with different craft activities such as cordage, textile, basketry or net-making in which the mouth is used as a third hand (Cristiani et al., 2018). Although fungal spores may become useful food items, such as mushrooms and fungi, the ubiquity of fungal spores in the environment suggests that they may well result from accidental ingestion or inhalation (Hardy et al., 2016). On the other hand, phytoliths might have also been naturally present in the environment, water, and soil, suggesting the inclusion of phytoliths in dental calculus for non-dietary reasons (Norström et al., 2019). Charcoals may also have reached the mouth by accidental ingestion through food and/or breathing due to exposure to hearths (Hardy et al., 2016). And pollen granules are most likely the result of inhalation, reflecting the environmental conditions in which this individual lived (Hardy et al., 2018). Finally, the wood inclusion in calculus can also vary from the use of a toothpick to crafting activities (Cristiani et al., 2021).

Triticeae grains were recovered at all sites, individuals and cultures, being quite frequent in all the samples. This trend suggests a broader cultural pattern in which Triticeae species were cultivated across regions, both in coastal and inland landscapes, and all individuals consumed these cereals. The individuals with more Triticeae starch grains were dated in the Early and Middle Bronze Age, while Late Neolithic and Chalcolithic individuals provided fewer quantities of granules. These differences in the number of starch grains recovered in each individual could be explained by different consumption among the members of these communities, or instead, by the state of preservation of the dental calculus of the oldest individuals. Concerning Paniceae grains, they are abundant in the samples, but quantitatively lower than Triticeae grains, suggesting that these crops were of lesser importance in the diet of these groups. They were also documented in all cultures and geographic locations, but only in eight of the ten sites and in ten of the 18 individuals, indicating a greater variability among individuals and sites. There are sites where these grains have not been preserved in all individuals (i.e. Los Avellanos I), and even sites without any evidence of their consumption (i.e. Abrigo de La Castañera). At the individual level, most of the individuals show an equal, low proportion of Paniceae grains, except for Individual 2 from Kaite, an adolescent female with signs of hypoplasia and caries. On the other hand, Fabeae starch grains were identified in three individuals and sites: El Toral III and Los Avellanos I (Late Neolithic) and El Mirón (Middle Bronze Age). However, they were recovered in a very small proportion, which prevents us from interpreting why they appear indistinctly in some individuals and specific chronological periods.

The FAMD distinguished two groupings of the different archaeological starch grains (Figure 8). The first one is represented by the sites analyzed in this work and dated from the Late Neolithic to the Middle Bronze Age. Triticeae starch grains are more commonly grouped with these sites and periods, although they are located in the center of both dimensions, suggesting that these cereals were widely cultivated across cultures. The second one is characterized by an assemblage studied previously by us and dated to the Late Bronze Age (González-Rabanal et al., 2022). Paniceae starch grains are predominantly associated with this site, indicating that these species are more highly associated with the Late Bronze Age and supporting the turning point in the Late Prehistory of the Cantabrian Region. Instead, Fabeae is located at the upper right side of the FAMD but is more associated with the first grouping. However, their small sample size in the record can constitute a comparative bias.

The starch grains most documented in this study have been morphologically assigned to the Triticeae tribe (Henry and Piperno, 2008) (Figure 2). This morphotype was already common among hunter-gatherers of Balkans and Italy (Cristiani et al., 2016; 2018; 2021; Nava et al., 2021; Oxilia et al., 2021; Carra et al., 2023). Likewise, this tribe is the most identifiable starch morphotype in the dental calculus analyses conducted in Iberian farmers up to now (Bucchi et al., 2019; González-Rabanal et al., 2022; Salazar-García et al., 2022; Coutinho et al., 2024), suggesting that the cultivation of different species of wheat and barley was key to the economy of these farming societies from the Neolithic onwards. Previous experimental analyses in species of the Triticeae tribe have revealed that wild starch grains (Aegilops genus) show a larger size distribution than domestic ones (Triticum and Hordeum genera) (Cristiani et al., 2021). Our experimental measurements support the morphological features of the archaeological starch grains and have revealed a wide range of sizes for the different Triticeae domestic species, avoiding their taxonomic identification to species level (Supplementary Table 5; Figure 3). This finding aligns with previous experimental analyses (Bocanegra and Sáez, 2012; Cristiani et al., 2016). In fact, there is no homogeneity in the existence of statistically significant differences between the length of the archaeological and experimental starch grains of the Triticeae tribe, varying the presence and/or absence of significance between species and sites (Supplementary Table 7). However, the morphology and size obtained for archaeological starch grains type A are compatible with those of larger starch grains of wheat and barley species (Figure 7A; Figure 9A).

Wheat and barley were the most widely cultivated cereals in Iberia during Late Prehistory (Buxó and Piqué, 2008). Carpological evidence for the Cantabrian region supports the presence of H. vulgare (naked and hulled varieties), T. dicoccum, T. monococcum, T. aestivum and T. durum in several Cantabrian sites such as Arangas, El Mirón, Los Gitanos, Arenaza, Kobaederra, Lumentxa or Pico Ramos (Zapata, 2002; López-Dóriga, 2016) (Table 3). In fact, direct dating of cereal seeds from El Mirón, Kobaederra and Pico Ramos demonstrates the early cultivation of these cereals since the second half of the fifth millennium cal BC (Zapata et al., 2004; Peña-Chocarro et al., 2005; Zapata et al., 2007). Such an early date is less likely for such species as S. cereale, which did not reach the peninsula until the Iron Age (Seabra et al., 2023). Thus, our methodological approach to the identification of starch grains in dental calculus corroborates the archaeobotanical data previously obtained for this region. Similarly, stable isotope evidence supports this attribution, so all the individuals analyzed here show a C₃ terrestrial diet based on the consumption of cereals, meat and dairy products (Jones et al., 2019; González-Rabanal et al., 2020; González-Rabanal, 2022).

The combination of naked and hulled wheat varieties such as emmer (T. dicoccum), einkorn (T. monococcum) or hard wheat (T. durum) and bread wheat (T. aestivum), and hulled and naked barley (H. vulgare var. vulgare and H. vulgare var. nudum) was possible in some periods of the Late Prehistory of the Cantabrian Region (Zapata and Pena-Chocarro, 2005). However, hulled species were more common during the Early Neolithic, while free-threshing wheats were almost absent, reflecting peculiar features for agricultural practices, suggesting that hulled species were primarily cultivated due to ecological factors. Hulled wheats are resistant to poor soil conditions and fungal diseases, and barley also tolerates poorer soils (Zapata et al., 2004). Similarly, they demonstrate a good adaptation to wet and mountainous conditions. As part of the Eurosiberian region, the Cantabrian Region is a mountainous area with an oceanic climate with generally mild temperatures and high precipitation, and no notable summer droughts. Only in this particular region of Iberia is emmer still cultivated using traditional methods due to its favorable yields under these conditions (Peña-Chocarro and Zapata, 1998). The late adoption (5^th millennium cal BC) of a farming economy in this region (Fano et al., 2015) would be in line with this hypothesis. The first cereals cultivated in this geographical area would have taken some time to acclimate to the new ecological conditions, influenced by the geographic characteristics of the short, narrow valleys that descend from the Cantabrian Cordillera to the Atlantic shore (Peña-Chocarro et al., 2005), a landscape more suitable for pastoralism in the initial stage of the neolithization process (López-Merino et al., 2010).

Polyhedral starch grains (morphotype II) found in the sample may be morphologically attributed to different Paniceae/Andropogoneae plants, such as millet species (Figure 2). The identification of starch grains from wild grasses of the Paniceae and Andropogoneae tribes was found among foragers (Cristiani et al., 2016; 2021; Nava et al., 2021; Oxilia et al., 2021). Based on their average size and morphological features this morphotype was also identified within the dental calculus of farmers in the Chalcolithic-age farmers from El Mirador cave in the Sierra de Atapuerca (Burgos, Spain) (Bucchi et al., 2019), in the Chalcolithic/Bronze Age site of Grotta dello Scoglietto (Tuscany, Italy) (Mariotti Lippi et al., 2017), in the Bronze Age site of Laderas del Castillo (Alicante, Spain) (Romero et al., 2024) and in the Late Bronze Age site of El Espinoso (Asturias, Spain) (González-Rabanal et al., 2022). The experimental starch grains from Paniceae species studied in this work exhibit similar morphological features as the archaeological starch grains, excepting S. halepense (Andropogoneae tribe), which showed different morphological traits than the archaeological ones (Supplementary Table 5; Figures 4, 7B). However, their length suggests that there are notable differences in grain size distributions between experimental and most of the oldest archaeological samples such as the Late Neolithic individuals from El Toral III, Los Avellanos I and Kaite, the Chalcolithic individual from Los Avellanos II, the Early Bronze Age individuals from Palomera and the Middle Bronze Age individual from El Mirón (Figure 9B). These differences in size are corroborated by the Pairwise Wilcoxon test, finding significant differences with all the experimental Paniceae species analyzed, except in cases in which the archaeological sample size bias the results (Supplementary Table 7). For these reasons, we can consider as aff. Paniceae to this set of starch grains. Instead, the Paniceae starch grains from the Early/Bronze Age individuals from La Fragua and El Agua are similar in size to the large starch grains of S. italica and P. miliaceum. In this case, the statistical comparisons showed no significant differences between the length of the Paniceae grains of these sites, and the S. italica grains, similar to what happens with the El Espinoso Late Bronze Age site (Supplementary Table 7). With respect to the experimental species analyzed, S. italica can have occasional oval grains and a centric hilum traversed by fissures which vary in form. In contrast, starch grains from P. miliaceum have a mostly polyhedral shape with a centric hilum where fissures are less common (Yang et al., 2012). Experimental starch grain size differs slightly between both species, being marginally smaller and statistically different the grains of P. miliaceum (Supplementary Table 6). Wild millet species, such as S. viridis or S. verticillata are even smaller, with a more characteristic spherical morphology and short fissures (Madella et al., 2016; González-Rabanal et al., 2022), which allows us to discard them as the source of the archaeological starch grains. The duality mentioned above could indicate that the larger starch grains of the aff. Paniceae tribe of the oldest individuals would belong to different, still underexplored species than those smaller Paniceae grains documented in some Bronze Age individuals. A great variety of Paniceae and Andropogoneae species grow in water environments (Cristiani et al., 2021), and a mixture of species from such genera might likely have been used by Cantabrian human groups for different purposes, while the domestic millets could be the source of the smallest Paniceae archaeological starch grains. This hypothesis could explain the differences in size between some archaeological Paniceae starch grains and the experimental ones.

The first carpological evidence of millet exploitation in Iberia dates to the Middle Bronze Age (Buxó and Piqué, 2008). However, millets were not systematically exploited until the Late Bronze Age and Iron Age (Tereso et al., 2016). Millets are poorly represented in the archaeological record of the Cantabrian Region. Macroremains of S. italica were identified in Kobaederra (Level 1) and Arenaza (Layer 9), chronologically assigned to the Chalcolithic and Early Bronze Age, respectively (Zapata, 2002) (Table 3). However, those seeds have not been directly dated; thus, we cannot exclude the possibility that they might be intrusive. Moreover, there is a notable absence of P. milliaceum seeds in the carpological records of the Cantabrian Region. By contrast, the carpological record of Northwest Iberia (Galicia and Northern Portugal) is richer, with both species being abundant in Middle and Late Bronze Age settlements (Bettencourt, 2000; Tereso et al., 2016; Figueiral et al., 2017; Jesus et al., 2020). Anyway, both species might have been consumed in a mixed way since millet poly-cropping has been suggested ethnographically (Moreno-Larrazabal et al., 2015; Teira-Brión, 2022). On the other hand, S. halepense can be discounted as a potentially consumed species since the morphology of its starch grains is quite different from those of S. italica and P. miliaceum and sorghum did not reach Iberia until the Middle Ages (Pérez-Jordà et al., 2024). No carpological evidence of the wild native species (Panicum repens L., S. viridis, or S. verticillata) has been found in the Holocene sites of northern Iberia, although today, they are widely distributed across central and southern Europe, but are almost unknown in Atlantic Europe (Pyankov et al., 2010).

The Late Neolithic/Chalcolithic individuals of El Toral III, Kaite and Los Avellanos I and II have reported some sparse Paniceae grains, but their belonging to domestic millets is discounted since they are larger than experimental millet species and a prior arrival of these species is not supported by carpological research, and they probably belong to the group widely known as “forgotten millets” or other Paniceae/Andropogoneae taxa (Lucarini et al., 2016). More likely is this association in the Early Bronze Age individuals of La Fragua and Palomera, although the archaeobotanical evidence for this period is very doubtful, and both sites provided differences in starch grain size. Finally, the starch grains from the Middle Bronze Age individuals, at least those from El Agua, offer a higher probability of belonging to domestic millets because in this period there is carpological evidence of millets. If we pay attention to the stable isotope evidence of Late Neolithic/Chalcolithic and Early/Middle Bronze Age individuals, the carbon isotope values do not support the consumption of C₄ plants in the Cantabrian Region (Arias, 2005; Sarasketa-Gartzia et al., 2018; Jones et al., 2019; González-Rabanal et al., 2020). In any case, if millets were already cultivated at this time, their consumption may have been occasional because the long-term bone collagen record does not register the carbon enrichment typical of eating millets. Thus, the agriculture of millet species was not fully developed until the Late Bronze Age, as supported by stable isotopes and dental calculus from El Espinoso cave (González-Rabanal et al., 2022). In that research, in addition to the high carbon isotope values, we identified 74 starch grains, of which 43 belong to the Triticeae tribe and 29 to the Paniceae tribe. As millets allow for two growing seasons per year, other crops such as those of the Triticeae tribe could also have been cultivated alongside them annually. These findings reveal a more diversified and mixed agriculture during this time, marking the primary shift in the farming economy of the Cantabrian Region during Late Prehistory.

The existence of starch grains within dental calculus identified as morphotype III suggests that some legumes were also part of the diet of human groups during the Holocene (Figure 2). However, the sparse evidence of their starch grains prevents calculating the real weight in their diet. All Fabeae grains are large granules that fall within the margins observed in experimental species of L. culinaris and V. faba (Supplementary Table 5; Figures 5, 7C, 9C), whose maximum lengths also coincide with this size. In fact, there are no significant differences between the archaeological Fabeae starch grains and the experimental ones (Supplementary Table 7). V. sylvatica and P. sativum experimental starch grains were statistically smaller; therefore, the archaeological grains likely belong to a domestic legume species. Wild species of the family Fabaceae are well-represented in the Early Mesolithic and Late Mesolithic individuals from the Danube Gorges area of the Balkan peninsula (Cristiani et al., 2016; 2021). In Iberia, a substantial quantity of starch grains was recovered from the Early Neolithic individuals at La Dehesilla cave (Cádiz) (Coutinho et al., 2024).

Legumes are traditionally poorly represented in archaeobotanical records due to taphonomic problems related to their processing. Although the existence of legumes in the Cantabrian archaeobotanical record is unknown until the Neolithic, some wild legumes could have been present in this region since the beginning of the Holocene. In fact, preliminary anthracological analysis of El Toral III has documented charcoals containing burnt legume remains in several stratigraphic units of the Mesolithic shell midden (Gutiérrez-Zugasti, personal communication). Later, several legumes such as lentil (L. culinaris), pea (P. sativum) broad bean (V. faba), grass and red pea (Lathyrus sativus L. and Lathyrus cicera L.), bitter betch (Vicia ervilia L. Willd.) and common vetch (Vicia sativa L.) were common in the Early Neolithic carpological record of numerous regions of Iberia (Zapata et al., 2004; Peña-Chocarro et al., 2018). In contrast, seeds of these species are absent in the Cantabrian Region, although some Fabaceae vegetal remains have been found in the Neolithic levels of El Mirón, Kobaederra, Lumentxa and Pico Ramos (Table 3). The first taxonomically legume remains are a Vicia sp. seed documented in the Arenaza Bronze Age layer 9 and a P. sativum seed discovered in El Mirón Chalcolithic level 4. Likewise, V. faba and P. sativum were widely identified in several sites of northwest Iberia during the Bronze Age (Tereso et al., 2016); therefore, it would not be surprising if these crops were also cultivated in other northern areas of the peninsula. Therefore, it is possible that some domestic legumes, such as peas, lentils, or beans, were used as food supplements by the Cantabrian communities from the Neolithic onwards.

Our results have not shown any starch grains with key morphological characteristics to be assigned to the Aveneae/Poeae tribe. However, a single, dubious angular starch was discovered in an Early Bronze Age individual from La Fragua. This grain has an irregular morphology, a central hilum and a visible but asymmetric Maltese cross. The lamellae are not observable. This morphology can be compatible with that documented in modern species of the Aveneae/Poeae tribe (Figure 5), but the absence of other morphological traits and its poor preservation suggest caution. For these reasons, it has been considered undetermined. The length of this starch grain was 9.3 μm (Supplementary Table 3), being consistent with the averages obtained in our experimental oat species (Supplementary Table 5, Figure 5, Figure 7C). Several starch granules of the Aveneae/Poeae tribe have already been identified in dental calculus from foragers, especially during the Late Mesolithic in the Balkans (Cristiani et al., 2018; 2021) and within Chalcolithic/Bronze Age individuals from El Mirador (Spain) and Grotta dello Scoglieto (Italy) (Mariotti Lippi et al., 2017; Bucchi et al., 2019). For the Cantabrian Region, carpological references to Aveneae/Poeae species are very scarce, with possible grains of Avena sp. in the A1 level of Los Gitanos (López-Dóriga, 2016) and in layer 9 of Arenaza (Zapata, 2002) (Table 3). The Iberian archaeobotanical record indicates that Avena species were not cultivated until a late stage of Late Prehistory (Buxó and Piqué, 2008). Thus, both archaeobotanical and dental calculus Iberian evidence suggest that, if consumed, these crops, typical of temperate ecosystems, would be cultivated sporadically and in addition to other cereals.

So far, no other starch grains from flax (Linum usitatissimum L.), poppy (Papaver somniferum L.) or additional crops have been identified within the dental calculus of late prehistoric humans in Cantabrian Spain. Similarly, starch grains associated with wild fruits have not been identified (Figure 6), although the smallest starch grains without taxonomical identification could belong to them. Wild nuts and fruits such as hazelnuts, acorns and Rosaceae fruits must have played a major role in the diet of the farming societies, as they were abundant, predictable, and storable food resources in this region (Zapata, 2000). In fact, some species are abundant in the Cantabrian carpological record, mainly hazelnuts and acorns (Zapata, 2002; López-Dóriga, 2016). Moreover, the anthracological analysis of charcoal remains has indicated the exploitation of several woody taxa in the archaeological sites of the region, and their wood was probably used as fuel (Uzquiano, 1995).