Developing a Reference Collection for Starch Grain Analysis in Early Neolithic Western Temperate Europe


 While we know that cereals played an important role in the diet of Linearbandkeramik (LBK) and Blicquy/Villeneuve-Saint-Germain (BVSG) populations in the Paris Basin, many questions remain to be answered as to the real contribution of other plants. To assess this topic, the recovery of other lines of data beyond macrobotanicals is crucial: starch grains have the potential to reveal additional information regarding past plant use. However, in Western Europe, in particular, for the Neolithic period, there is a significant lag in the development of the discipline. We, therefore, present how our current reference collection (composed of nearly 100 taxa spread across 35 families) was established, the reasoning behind our plant selections, and where the material comes from. Overall, our work shows that even though not all the selected plant organs produce diagnostic starch grains, it may be possible to broaden the spectrum of plants likely consumed by Early Neolithic (and beyond) populations in the Paris Basin, in particular concerning the use of wild plants and specific plant parts, especially underground storage organs (tubers, rhizomes, roots, bulbs, etc.). We believe our research will help guide future scholars in the creation of their own starch grain reference collection and to carry out such analyses on archaeological material from this region by consulting our image database. We conclude by providing a brief summary of what the starch grain record in the Paris Basin tells us to date on ancient plant use.


Introduction
Plant macroremains (those visible to the naked eye) are some of the most direct evidence of the Neolithic diet, especially when they are found charred in a domestic context. Nevertheless, charred plant remains alone permit only a limited knowledge of the diversity of plants and plant parts that were consumed, as fire also damages the most fragile plant parts (Fritz & Nesbitt, 2014). This statement is applicable to the Early  Neolithic period in the Paris Basin where our knowledge of plant use is primarily based on the archaeobotanical record of sites ranging from the Aisne Valley in France to Hesbaye in central Belgium ( Figure 1) (Bakels, 1984(Bakels, , 1999(Bakels, , 2009Berrio, 2011;Dietsch-Sellami, 2004;Hamon, Salavert, Dietsch-Sellami, & Monchablon, 2019;Salavert, 2010Salavert, , 2011. The Linearbandkeramik (LBK) culture, which originated in central Europe and expanded rapidly westwards reaching the Paris Basin around 5100 BC (Salavert, 2017), was subsequently replaced around 4900 BC by another group known as the Blicquy/Villeneuve-Saint-Germain (BVSG).
While we know that cereals played an important role in the diet of these LBK and BVSG populations (Bakels, 2014;Hamon, 2008;Salavert, 2011), many questions remain to be answered regarding the real contribution of other plants, as no single dataset can provide the entire spectrum of plants used in the past (Colledge & Conolly, 2014). For example, what other plants were processed and consumed in western LBK regions, and more specifically the Paris Basin? Did Neolithic populations consume fewer wild resources than those of the Mesolithic (or none, see Bogucki, 2000, p. 204)? What was the function of different tools such as grinding stones and ceramics? To answer these questions, the recovery of other lines of data is crucial, and this is where starch grains have the potential to reveal additional information regarding past plant use.
To create our reference collection, we began by consulting the existing macrobotanical record of our region, but as mentioned, it is limited by preservation issues. Thus, to amplify our reference collection, we turned to archaeobotanical data from neighboring countries in temperate Western Europe, notably the Netherlands, Switzerland, Germany, Denmark, and Poland for both the preceding period (Mesolithic, c. 8500-3500 BC) but also for slightly later periods, notably the Bronze Age (c. 2200 BC). We also consider ethnobotanical reviews of plants used today across Europe to provide a wider perspective of possible plant use. By doing so, we were able to consider the potential and limits of the methodology: what starch grains can we expect to find and identify in the archaeobotanical record? Which domesticated plants, recovered in macrobotanical form, can we identify to species? Which ones are truly diagnostic? Which plants will we be unable to document through this methodology? We thus share our reference collection in the form of an image database¹ in the hope that similar analyses will multiply in the Early Neolithic contexts of temperate Europe, broadening our perception, and understanding of the exploitation of plants for this period and region. (Mercader et al., 2018b). Many studies focusing on the Neolithic, and in turn implying the beginnings of sedentary lifestyles and agriculture, have made use of the potential of starch grains. In particular, they have made it possible to establish the importance of certain resources, such as tubers, on Chinese grindstones and ceramics (Liu, Duncan, Chen, Liu, & Zhao, 2015;Liu, Kealhofer, Chen, & Ji, 2014;Wan et al., 2012;, Yao et al., 2016. Other work on Neolithic contexts has been carried out in North Africa (Lucarini, Radini, Barton, & Barker, 2016), Israel (Nadel, Piperno, Holst, Snir, & Weiss, 2012), South East Asia (Barker & Richards, 2013, Barker et al., 2007, but on sediments), Oceania (Denham, Haberle, & Lentfer, 2004;Horrocks, Bulmer, & Gardner, 2008), and Japan (Shibutani, 2017), but also on the American continent (Dickau, Ranere, & Cooke, 2007;Inomata et al., 2020;Iriarte et al., 2004;Pearsall, Chandler-Ezell, & Zeidler, 2004;Piperno, Ranere, Holst, & Hansell, 2000). Beyond analyses on grinding stones and ceramics, dental calculus trapping starch grains and other microremains has made it possible to carry out analyses of the Early Neolithic in Sudan (Madella, García-Granero, Out, Ryan, & Usai, 2014;Out et al., 2016) and Iraq (Scott Cummings, Yost, & Sołtysiak, 2018).
Western Europe is therefore a real gap in the development of the discipline, in particular for the Neolithic period. Few works exist in our geographical area, one example is the work by Bosquet (2013, 2017) who studied grinding stones from Remicourt "En Bia Flo II" in Belgium, an LBK site. Beyond our immediate area, starch grain analysis has provided insights on the Neolithic period in other parts of Europe. At Tiszasziget (Hungary), starch grains have been extracted from ceramics dating to the Late Neolithic (5000-4500 BC, Pető, Gyulai, Pópity, & Kenéz, 2013). In Northern Germany at the site of Neustadt, Saul et al. (2012) extracted starch grains from charred residues, also collected in ceramic vessels dating between 4600 and 3700 cal BC. The presence of tubers was identified based on the collection of parenchyma fragments from inside ceramics from Late Mesolithic and Early Neolithic sites in the Nether-lands³ (Kubiak-Martens, Brinkkemper, & Oudemans, 2015;Raemaekers, Kubiak-Martens, & Oudemans, 2013), but also from Neolithic domestic sites, causewayed enclosures, and megalithic graves in Germany and Denmark (4000-1800 BC) (Klooss, Fischer, Out, & Kirleis, 2016).
We sought to address this gap by beginning a preliminary starch grain study on a range of Early Neolithic tools and ceramics from across the Paris Basin⁴ (Hamon et al., 2021). To carry out a successful study, however, a solid reference collection is necessary. Unlike other parts of the world, where detailed plant lists and in some cases images of starch grains have been produced, this does not exist for Western Europe. In this paper, we present the characteristics of the reference collection we assembled (how and why the plants were included), to facilitate future analyses in this area in other Early Neolithic contexts in Western Temperate Europe.

Selecting Species for the Reference Collection
In this section, we consider how our current reference collection was established, the reasoning behind our plant selections, and where the material comes from.
Fruits are widespread recovered in the archaeobotanical record of the Neolithic period, whether in LBK contexts in the Paris Basin (Berrio, 2011;Salavert, 2011) or at sites in western-Central Europe spanning between 4400 and 2400 cal BC (Colledge & Conolly, 2014). While it is known that ripened fruits in general tend to be relatively poor in starch grains (with for example the exception of tropical fruits such as the banana or avocado), we still tested some fruits (Crataegus monogyna Jacq., Prunus spinosa L., Malus sylvestris (L.) Mill., Sambucus sp.). The presence of hazelnuts (Corylus avellana L.) and acorns⁵ from deciduous oak (Quercus sp.) are reported from LBK sites in our region of interest and were therefore included (Berrio, 2011). We also included the starch-rich fruits of water caltrop (Trapa natans L.), which have been reported from Mesolithic sites in the Netherlands and at Schwarzenberg Lake in the Czech Republic (Divišová & Šída, 2015), and from Neolithic sites (3500-3000 cal. BC) in Slovenia (Tolar, Jacomet, Velušček, & Čufar, 2011).
Wild/weedy plants were also considered. For the most part, these were collected at the MNHN and completed with seeds from the collection at the Archaeobotany lab of ArScAn (UMR 7041, Arch. Env) in Nanterre, France. Seeds found commonly in Early Neolithic Paris Basin contexts include goosefoot  (Bakels, 1999;Berrio, 2011;Dietsch, 1996;Salavert, 2011). The wild carrot (Daucus carota L.), in seed form, has been recovered from the region, albeit in small quantities (Berrio, 2011;Dietsch, 1996). It is unclear which part(s) would have been consumed, but roots (raw) and young stewed leaves are today consumed in some parts of Spain (Tardío, Pardo-de-Santayana, & Morales, 2006). We were able to test both the wild carrot's root and seed. The common reed (Phragmites australis (Cav.) Trin. ex Steud.), a completely edible plantin particular the young stems and rhizomeswas also included in our collection as it has been reported from Neolithic sites in central Europe (Colledge & Conolly, 2014) but also from Mesolithic contexts (Kubiak-Martens, 1999). Also, in the Poaceae family, and found in small quantities in the Paris Basin archaeobotanical record, are cockspur grass (Echinochloa crus-galli (L.) Beauv.) and annual meadow grass (Poa annua L.), which we included (Bakels, 1999;Berrio, 2011). Hairy crab-grass (Digitaria sanguinalis (L.) Scop.) does not seem to be mentioned in the archaeobotanical record of the Paris Basin, yet the seeds can be ground into flour (Simkova & Polesny, 2015), and it was present along the Mediterranean for potential use by the Neolithic populations between the early sixth millennium to the late third millennium BC (Delhon, Binder, Verdin, & Mazuy, 2020).
We have also considered plants that are reported to have only appeared or become important in the region both earlier and later. For example, the earliest gold of pleasure (Camelina sativa (L.) Crantz) has been reported in Switzerland around 4000 BC, becoming more common only between 1800 and 1200 BC in south-eastern and central Europe (Zohary et al., 2012, cited by Larsson, 2013. Although there is some limited evidence of Camelina in France in the Middle Neolithic, it is argued that cultivation in Western  5 The acorns (Q. ilex) used in this study are not native to the Paris Basin, but are from the Mediterranean region instead.
France likely began in the Late Bronze Age (Toulemonde, 2010). The same can be said about the sweet chestnut (Castanea sativa Mill.), whose introduction to southern France dates to probably the Roman period (Buonincontri, Saracino, & Di Pasquale, 2015).
One of the major problems in trying to establish the actual spectrum of plant use is to a certain extent due to the lack of USOs, recovered in macrobotanical form, in part due to the way they are consumed or prepared (fresh or boiled), making it unlikely that they will end up carbonized (Scheel-Ybert, 2001). While exceptions exist regarding the presence and probable use of USOs for the entire Neolithic period from contexts across Northern, Central, and Western Europe, there is a real gap of information on the presence of USOs at Early Neolithic sites in the Paris Basin. The little data that we have beyond the Paris Basin come in the form of tubers of lesser celandine (Ficaria verna Huds.), which have been widely reported from Early Neolithic (4000-3400 BC) contexts in northern Germany and Denmark (Klooss et al., 2016), while wild garlic (Allium ursinum L.) is reported as having been consumed by the Neolithic populations living near the Chalain Lake in the Jura, France (Dommelier, Bentrad, Paicheler, Pétrequin, & Bouchet, 1998citing Pétrequin & Pétrequin, 1988. Turnip⁶ (Brassica rapa var. rapa) has been reported from waterlogged contexts (4400-2400 cal BC) in Central Europe (Colledge & Conolly, 2014) and in seed form from Stare gmajne (Slovenia) dated between 3500 and 3000 cal BC (Tolar et al., 2011). Other remains include a Liliaceae bulb from Early Neolithic (5240-4990 cal BC) contexts at the site of Taï near the Mediterranean (Bouby, Durand, Rousselet, & Manen, 2019), and tuber oat-grass bulbs (Arrhenatherum elatius subsp. bulbosum) from Late Neolithic Germany (3500-2800 cal BC) (Kirleis, Klooß, Kroll, & Müller, 2012). In fact, more information concerning the use of USOs comes from European Mesolithic sites in Poland, the Netherlands, and Scotland⁷ (see full references in Kubiak-Martens, 2016), where Scanning Electron Microscope (SEM) techniques have made it possible to identify parenchyma. Archaeological research at these hunter-gatherer sites has yielded a rich collection of starchy foods in the form of knotgrass rhizomes (Polygonum sp.), tubers of arrowhead (Sagittaria cf. sagittifolia), and horsetail (Equisetum sp.). Moreover, a potential sedge family (Cyperaceae) corm/stem base along with Schoenoplectus lacustris (L.) Palla was also recovered. In Neolithic contexts, a few mentions are made for the recovery of Scirpus, Carex, Cyperus, and Bolboschoenus (Kirleis et al., 2012), although not all species will necessarily be related to food consumption (e.g., Scirpus lacustris, Dietsch, 1996). Bolboschoenus maritimus (syn. Scirpus bolboschoenus) charred tubers were recovered from Late Neolithic contexts in the Netherlands (Kubiak-Martens et al., 2015); the stem bases, nutlets (achenes) and tubers can all be consumed (Kubiak-Martens, 1999). It was interpreted that tubers of a related taxon (Bolboschoenus glaucus) were used to produce flat bread-like products at Shubayqa 1, a Natufian huntergatherer site in northeastern Jordan (Arranz-Otaegui, Carretero, Ramsey, Fuller, & Richter, 2018).
Late Mesolithic datasets (c. fifth mill. BC), reported from Tybrind Vig and Halsskov, both lacustrine areas in Denmark, testify to the presence of sea beet (Beta vulgaris ssp. maritima), whose roots are rich in starch and sugar, and pignut (Conopodium majus (Gouan) Loret.) tubers (Kubiak-Martens, 1999, 2002; we were unfortunately unable to get a hold of these latter two plants. Seeds of waterlilies (Nuphar and Nymphaea) have been reported from Mesolithic and Neolithic contexts, indicating that they were probably consumed (see Bouby, Dietsch-Sellami, Martin, Marinval, & Wiethold, 2018;Dietsch, 1996;Kirleis et al., 2012;⁸ Kubiak-Martens, 2002, 2016Raemaekers et al., 2013). However, it is known that waterlily rhizomes are also edible, and their consumption is widely attested in the ethnographic record (Kubiak-Martens, 2016). The Mesolithic archaeobotanical record is also a good source of information on other plants besides USOs. For example, a charred caryopsis of floating sweetgrass (Glyceria fluitans (L.) R. Br.) was recovered at Tybrind Vig (Kubiak-Martens, 1999). Ethnobotanical studies in the Czech Republic indicate that these seeds can be ground into flour (Simkova & Polesny, 2015). We were unable to get a hold of this species but tested a relative-G. maxima-instead. Cattail reeds (Typha sp.) have a long history of use, as far back as the Upper Palaeolithic, where hunter-gatherers at the Bilancino site in Italy prepared flour from the  6 Unclear if seed or bulb. 7 For example, Całowanie (Poland), Yangtze Harbour (The Netherlands), and Staosnaig (Scotland) (see Figure 1). 8 Also recovered charred storage tissue of waterlily (p. 235). starchy rhizomes (Aranguren, Becattini, Lippi, & Revedin, 2007). They are also reported in Mesolithic contexts from northern Netherlands and Poland (Kubiak-Martens, 1999;Perry, 1999). Two species are noted as likely being present at the time, T. angustifolia and T. latifolia, and we were able to include the latter in our reference collection.
The presence of a rare celery (Apium graveolens L.) schizocarp⁹ from Parkhaus Opera, a Neolithic site on the shore of Lake Zurich, Switzerland (3176-3153 BC), indicates that this plant may have been used as a bread condiment (Heiss et al., 2017). We, therefore, included this plant, along with wild celery (Apium graveolens var. graveolens) in our reference collection. Other members of the Apiaceae family, although not reported in the literature to date, were considered: bulbous chervil (Chaerophyllum bulbosum L.) and celeriac (Apium graveolens var. rapaceum), both of which were purchased from local markets in France. Raemaekers et al. (2013) cogently noted that species recovered in seed form from Early Neolithic sites in the Netherlands produce edible stems or fleshy shoots. These include glasswort (Salicornia europaea L.), sea aster (Aster tripolium L.), stinging nettle (Urtica dioica L.), greater burdock (Arctium lappa L.), and chickweed (Stellaria media (L.) Vill.), to mention but a few. Remains of Arctium minor and stinging nettle are reported from central European Neolithic sites, dated between 4400 and 2400 cal BC (Colledge & Conolly, 2014), although it is unclear which parts of the plants were recovered. The authors do note that for Arctium minor, the entire plant is edible, while the stinging nettle is usually used for its leaves and oil. We were unable to find any archaeobotanical evidence of sea aster or of glasswort. However, the leaves of the former are well known for their edibility, while the young stems of the latter are consumed. Seeds of chickweed are reported from the Paris Basin in small quantities, but not necessarily from the Early Neolithic (see Bakels, 1999).
The recovery of ferns is rare in the archaeobotanical record, yet some examples do exist and document their use as food or medicines (see for example Fiorin et al., 2018 who studied dental calculus from people in Medieval Majorca). In the case of bracken fern (Pteridium aquilinum (L.) Kuhn), this plant has been widely used around the world, especially for its rhizomes and young fronds (Divišová & Šída, 2015). Mesolithic period charred parenchyma studied by Kubiak-Martens (2008) was shown to belong to bracken and likely male fern (Dryopteris filix-mas (L.) Schott). Bracken fern was also recovered from Late Bronze (905-869 BC) settlements in the French Alps (Bouby & Billaud, 2005).
As there is a lack of data on USOs in the archaeobotanical record, we turned to ethnobotanical reviews that mentioned roots, tubers, and corms as being consumed to expand our reference collection. These studies for example mentioned the use of consuming raw rampion bellflower (Campanula rapunculus L.) roots,¹⁰ as well as their shoots and leaves (Mattalia, Quave, & Pieroni, 2013;Simkova & Polesny, 2015). Other USOs consumed include rhizomes of Lords-and-Ladies (Arum maculatum L.), used as a flour or boiled, and tubers of bulbous chervil, which are noted as being used in boiled dishes (Simkova & Polesny, 2015). The rhizomes of common comfrey (Symphytum officinale L.) are also noted as being eaten. The consumption of the tubers of the tuberous pea (Lathyrus tuberosus L.) seems to be common in parts of Italy, France, and the Netherlands, but more rarely in Spain (Mattalia et al., 2013;Tardío et al., 2006). We also included the tubers of great pignut (Bunium bulbocastanum L.) as these are reported as being consumed in the Western Italian Alps (Mattalia et al., 2013). The bitter roots of the great yellow gentian (Gentiana lutea L.) are used to prepare a digestive liquor (Abbet et al., 2014). Dandelion (Taraxacum officinale (L.) Weber ex F.H. Wigg) leaves and flowers are used in multiple ways (Abbet et al., 2014), and so are the roots, which can be consumed either raw or cooked. Horseradish (Armoracia rusticana G.Gaertn., B.Mey. & Scherb.) is another plant that could have been gathered for its roots (Saul et al., 2012), even though to date no archaeobotanical remains have been recovered suggesting its use in ancient times. Parsnip (Pastinaca sativa L.) is a commonly consumed root but not well-known in the archaeological record: according to Zohary et al. (2012) it has been recovered from Roman sites in Europe.
 9 A type of dry fruit that when ripe, splits into single-seeded parts known as mericarps. 10 Also reported in Bronze Age contexts (905-869 BC, see Bouby & Billaud, 2005).  To date, we have collected and tested 99 species that cover 35 families ( Table 1). Out of those, we tested fruits (n = 11), seeds (n = 69), underground storage organs (n = 22), and stems (n = 2). For three taxa (Daucus carota, Rumex acetosa, and Schoenoplectus lacustris), we tested both the seeds and the USOs, while we also tested acorns at different stages of maturity. Thus, we have a total of 103 samples. Besides those obtained from the herbaria, most of the plants were either purchased from local markets (turnip, chestnut, bulbous chervil) or nurseries (e.g., lesser celandine, common comfrey). A small percentage of the taxa were directly gathered from the wild (e.g., dandelion, wild carrot, wild garlic).

Processing the Modern Plant Samples and Recording the Starch Grains
Fresh plant material was first cleaned and peeled (when dealing with tubers, roots, and fruits). Otherwise, material from herbaria or already dried was directly processed. The sample (seed, fruit, USO) was then gently crushed using a mortar and pestle. To avoid damaging the starch grains in an extremely hard seed (e.g., cereals) they were placed for 1-2 h in distilled water, before cutting off a small piece with a clean scalpel and then gently rubbing it against a clean microscope slide (using a clean toothpick when necessary). A drop of 1:1 glycerine:water solution was added before covering with a coverslip. The glycerine solution allows for the starch grains to be more easily rotated and viewed in both cross-polarized and transmitted light. Although the sample will eventually dry, it can be re-hydrated before observation using this afore-mentioned solution. The reference slides were then examined at 600× magnification using a Nikon E600 POL microscope and starch grains were measured using NIS Elements software, which provided statistical data including the mean, standard deviation, and range of sizes. The number of grains counted by different specialists when establishing their reference collection varies greatly, from at least 50 grains (Hart, 2014;Li, Pagán-Jiménez, Tsoraki, Yao, & Van Gijn, 2020;Musaubach, Plos, & Babot, 2013;Piperno, Weiss, Holst, & Nadel, 2004), to at least 300 (Mercader et al., 2018b). We measured 50 simple starch grains whenever possible, and these were randomly chosen on the slide. Photographs were taken under transmitted and cross-polarized light, and attributes such as the shape of the grain, the type and position of the hilum as well as the presence or absence of facets and fissures, were described (see Supplementary materials for photos and descriptions). Whenever possible, we kept materials (stored separately) in case new slides need to be prepared. Some species in our reference collection (e.g., Trapa natans, Iris sibirica) were obtained from Dr Laura Longo (Università Ca' Foscari), who prepared ultrapure starch pellets extracted using a sequential water/ethanol protocol.¹¹ In this case, a small amount of the resulting powder was placed on slides, where a drop of 1:1 glycerine:water solution was then added, before adding a coverslip.

Results and Discussion
We tested 103 different plant parts, but since we tested immature and mature acorns, our calculations are based on a total of 102 different taxa. Out of these, we found that 69 produced starch grains (68%), with a large majority of USOs producing starch grains (77%), followed by seeds (68%). There is a 50/50 chance of recovering starch from fruits or stems, but this probability may well vary if we tested a wider number of stems. It is immediately clear that compared to Hart (2014), who found that only 10 out of 64 species present across Southwest Asia produced starch grains, we found a much greater number of species that do produce these microbotanical remains. This is of course a positive outcome, but we must also keep in mind that even when starch is produced, not all starch grains are diagnostic to genus and even less to species. Moreover, some species will produce exceedingly small grains that will be hard to effectively see using a stereomicroscope. When the grains produced are below 5 μm, these are difficult to characterize with a light microscope, even at 600×. This is the case for species included in our collection such as sea beet, goosefoot, chickweed, as well as some taxa in the grass family (Festuca, Poa). In this case, additional higher resolution microscopy such as that obtained by SEM may help to further identify these starch grains (Jane, Kasemsuwan, Leas, Zobel, & Robyt, 1994). The fruits we investigated for most part did not contain starch, except for acorns, water caltrop, chestnut, and horse chestnut. It has been reported that unripe fruits will contain more starch than ripe fruits, as these will convert the starch to sugar as the fruit ripens (Gott et al., 2006). We tested both green and mature acorns, and while our results do not suggest that there is a significant difference, more tests involving other taxa are necessary.
We also found that the presence of starch in seeds is variable. Cereals, notably those in the Triticeae tribe (e.g., wheat and barley), have a bimodal distribution (grains come in two sizes), and produce abundant quantities of starch grains, with the larger ones being lenticular in shape. The morphology of some wild grasses has also been explored in depth (Hart, 2014;Piperno et al., 2004;see Yang & Perry, 2013 for a specific focus on species found in China). We found starch grains in all the grass seeds we collected, except for greater sweetgrass. The smallest grains were produced by cockspur, stipa, timothy-grass, and Poa, followed by Digitaria and Setaria; with barren brome seeds producing the largest starch grains among the wild grasses sampled.
The Fabaceae family produces very recognizable starch grains with a distinct longitudinal cleft. We extracted starch from all the domesticated taxa tested (peas, lentils, beans grass pea, and bitter vetch), and from both wild Vicia species. We discovered that not all members of this family produce starch grains, this was the case for species of Lotus, Medicago, and Lupinus.
As expected, seeds with high oil/fat content, and used for extracting oil such as opium poppy and hazelnuts, did not contain starch grains. As noted by Gott et al. (2006), in these plants, the main storage is lipid, and therefore, little to no starch will be produced or stored. We did, however, find starch grains in the seeds of flax, but only in the ultrapure pellets prepared by Laura Longo and Elena Badetti. Even then, these starch grains remained very tightly packed within the rest of the seed matrix. For this reason, it may be harder to detect them in the archaeological record if they are simply gently pounded or ground. For this initial reference collection, we focused on seeds and USOs, and set aside the study of the leaves. In the latter, starch grains, known as transient or transitory starch, have been determined to be small (less than 7 μm; Gott et al., 2006;Haslam, 2004), and therefore probably not diagnostic. However, leaves will have to be included in the future as, more recently, the presence of larger (non-transient) starch grains in the leaves and stems of domesticated and wild plants have been noted (Liu, Wang, & Levin, 2017;Yang et al., 2014). Neolithic populations likely had access to a large range of plants they could have used for their greens (leaves, stems): for example, stinging nettle, wild carrot, dandelion, and dock/sorrel (Mattalia et al., 2013;Simkova & Polesny, 2015). It will be of interest to test these in the future to determine whether the leaves of such species produce diagnostic starch grains. After testing an important number of seeds from wild plants, we found that starch grains were present in a majority of these. However, based on their tendency to be on the smaller side (between 2 and 11 μm), the probability of correctly identifying the ones we recover in the archaeobotanical record is rather slim, unless an SEM is used.
One issue that is clearly at the forefront of starch grain analysis is its utility in identifying underground storage organs in the archaeobotanical record. However, knowing which taxa to include in a reference collection has been problematic when we rely on solely macrobotanical remains. For the Neolithic, few taxa have been reported (but see Klooss et al., 2016;Kubiak-Martens, 2016), yet it is evident that multiple plants could have been gathered by these populations. Hart (2014) already made a similar observation for studies of USOs in Southwest Asia. We believe that our biggest contribution here is to provide a list of putative plants, besides those already found at other sites in the Neolithic, whose USOs may have been consumed by LBK/BVSG populations in Paris Basin. It should be noted that not all USOs contain starch grains, in fact, other reserve carbohydrates can be present (either in place of or in addition). Both monocotyledons and dicotyledons have fructan-containing species, and their families include Poaceae, Campanulaceae, Amaryllidaceae, Iridaceae, Asteraceae, and Liliaceae (Ranwala & Miller, 2008). In our case, we were unable to find starch in the roots of wild garlic (Amaryllidaceae) or dandelion (Asteraceae); however, the Siberian iris (Iridaceae), dogs' tooth-violet (Liliaceae), and the common reed (Poaceae) were all rich in starch. We, therefore, believe that it is important to also test species in these families. Overall, we found that a vast majority of the USOs we collected contained starch grains, for the most part, the grains are typical of USOs: elongated with eccentric hila, but other forms of starch grains exist. Taxa that did not produce starch grains in their USOs include members of the Apiaceae family (wild carrot, wild celery, celeriac), but this is not a fast rule as other members in the same family (bulbous chervil and parsnip) have starch grains (although neither produces particularly diagnostic grains).
Our current work makes it clear that not all taxa will produce diagnostic starch grains and in turn be useful for archaeobotanical studies. If we disregard those that are too small to be clearly visible with a microscope at 600× (species which have already been discussed earlier in the text), we can propose the following observations regarding those taxa that are more likely to be diagnostic or identifiable. Cereals such as wheat and barley are easily categorized due to the presence of lenticular starch grains. However, to differentiate between these taxa, is not as straightforward, even when a large reference collection is available (see for example Bocanegra & Sáez, 2012). Moreover, when age and taphonomic factors are added to this mixture, differentiating wheat and barley may be even more problematic. One taxon that could potentially be confused with wheat or barley is Bromus sterilis; however, the grains are wider when viewed from the side and the extinction cross is not often bilateral. Oats will likely be identifiable if they are still packed into clusters, but we were unable to distinguish between the two species we tested. When the grains are loosened, these could be mistaken for Digitaria sanguinalis. The latter produces slightly more angular grains than those produced by both A. sativa and A. strigosa. Within the Cyperaceae family, the starch grains in the USOs of both Cyperus esculentus and C. rotundus are diagnostic. The same cannot be said for the starch grains in the seeds of Carex hirta and Bolboschoenus maritimus. The grains produced by taxa in the Fabaceae family are extremely diagnostic. Distinguishing between the various species may be possible, with peas, lentils, and Lathyrus sativus deemed especially diagnostic. Within the vetches, we found that V. hirsuta and V. tetrasperma lacked distinct lamellae, which are more visible on V. sativa and V. ervilia. We did find that several taxa produced small (5-8 μm on average), polyhedral or angular starch grains with a centric hilum. These include seeds of Echinochloa crus-galli, Phleum pratense, Bolboschoenus maritimus, and species of Setaria. However, the latter, in particular S. verticillata, will be likely distinguished by the presence of a distinct continuous double border. USOs of various species, as noted earlier, are more likely to be identified as such given their shape and relatively larger sizes. We propose that Brassica rapa var. rapa, Gentiana lutea, Erythronium dens-canis, and Ficaria verna could thus be identified in the archaeobotanical record. Finally, fruits of Aesculus hippocastanum and Quercus are also good candidates for identification. While we were unable to gather all the plants that we had hoped to sample, for example, sea beet, pignut, tuber oat-grass, and tuberous pea¹² (see Table 2), the creation of a reference collection is often a work in progress, and we plan to continue expanding it. We believe our work thus far provides a starting base for starch grain analysis in the region. Besides continuing our search for new plant species that could have been used and present in the region during the Early Neolithic, some additional types of samples need to be considered. For example, we need to collect plants at different stages of maturation but also from different environments, as these factors may affect the size of the starch grains (see references in Gott et al., 2006). Other plant parts could have been utilized, for example, the inner bark tissue of pine and birch, which are starch-rich resources (Gott et al., 2006;Kubiak-Martens, 2016;Sandgathe & Hayden, 2003) and will also need to be considered. Experimental work that comprises mechanical processing (grinding, pounding), as well as thermal exposure (cooking) and fermentation, is essential to gather a broader picture of how starch grains are modified (see for example Cagnato, 2019;Chantran & Cagnato, 2021;Henry, Hudson, & Piperno, 2009;Li et al., 2020;Wang et al., 2017). Preliminary work (Cagnato, Hamon, & Salavert, in press) on domesticated species (cereals and pulses) has already been carried out, but we plan to expand this collection with the processing of wild plants, especially tubers and rhizomes, due to their presence in archaeological samples from the Paris Basin (Hamon et al., 2021).
Overall, our work here shows that through starch grain analysis, it may be possible to broaden the spectrum of plants likely consumed by Early Neolithic (and beyond) populations in the Paris Basin, in particular concerning the use of wild plants and specific plant parts, especially underground storage organs. Now that a large selection of starch-rich species has been identified, future research can focus on determining whether it will be possible to clearly differentiate between them. We hope that our research helps guide future scholars in the creation of their own starch grain reference collection as there is the necessity for a solid database, and this across disciplines.

What Does Starch Grain Analysis Tell Us Thus Far About LBK Food Processing?
To conclude, we provide a brief synthesis of what we know about food transformation to date in the Paris Basin. The application of starch grains has revealed that cereals were not the only foods processed on grinding stones. Notably, 9 LBK grinding stones from Remicourt "En Bia Flo II" in Belgium revealed the presence of wheat, barley, oats, peas, and acorns (Chevalier & Bosquet, 2013. Our research (Hamon et al., 2021), carried out on a large corpus of grinding stones (n = 32) from across LBK and BVSG sites in the Paris Basin, supports the notion that grinding stones were multipurpose as we also found starch grains of wheat, barley, and peas. Cooked (or at least heated) plants were also processed with these stone tools, based on our data. Finally, we also found evidence for the processing of different types of USOs (Hamon et al., 2021;Cagnato, Hamon, Salavert, & Elliott, in prep.), thereby proving that starch grain research can be fruitful and provide a new vision of past plant use.
Funding information: The work was financed by a DIM MAP Ile de France postdoctoral fellowship and the ANR Homes grant (ANR-18-CE27-0011).
Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission. CH, CC, AS, and ME designed the project, and CC prepared the reference collection. CC prepared the manuscript with contributions from all co-authors.