Ecological-cultural inheritance in the wetlands: the non-linear transition to plant food production in the southern Levant

The paper discusses a multi-proxy archaeobotanical dataset from the published macrobotanical and microbotanical research of 19 Epipalaeolithic sites over a period of 13.5 ka (ca. 25-11.5 ka cal bp) in the southern Levant. The archaeobotanical record includes over 200 phytolith samples extracted from sediments of 11 sites, macrobotanical evidence from seeds, plant tissues and wood charcoal from 11 sites and other microbotanical data from starches and starch spherulites from three sites. Phytolith assemblages show that Epipalaeolithic hunter-gatherers relied on wetland plant resources, and the macrobotanical evidence demonstrates the cultural importance of wetland resources to Late Epipalaeolithic foodways. The paper assesses the archaeobotanical data, using niche construction theory (NCT) to argue that human impacts in the wetlands over generations increased wetland productivity and created long-lasting ecological and cultural inheritances that allowed for the evolution and proliferation of a wetland-based system of wild plant food production. This paper proposes the idea that wetlands provided Epipalaeolithic people with a ‘domestication laboratory’ that allowed them to interact and modify their environment and also to experiment and gain new knowledge to build the long-lasting ecological-cultural inheritances necessary for the transition to plant food production. Importantly, the ecological-cultural inheritance of wild plant food production in the wetlands enabled the non-linear transition to cereal-based wild plant food production and domestication that took place in the southern Levant.


Introduction
Phytoliths (microscopic silica 'casts' of plant cells) are now bridging the gaps in the archaeobotanical record and providing a useful long-term dataset. The phytolith evidence from Epipalaeolithic sites in the southern Levant emphasizes the importance of wetland-based plant resources (Rosen 2013;Ramsey et al. 2017). This paper reviews the available Epipalaeolithic phytolith research and evaluates the data against other macrobotanical and microbotanical evidence from this region (Figs. 1 and 2), within the framework of niche construction theory (NCT, see below) to argue that ecological-cultural inheritance in the wetlands facilitated a non-linear transition to plant food production (non-linear developments do not have straight or direct relationships between variables).
Epipalaeolithic hunter-gatherers relied on wetland resources and benefited from a self-reinforcing positive feedback loop whereby human use made the wetland plants such as Cyperaceae and Phragmites (reed) increasingly productive (Ramsey et al. 2015). Productive anthropogenic wetlands (under human influence) provided the resources needed to sustain generations increasingly 'in place', or settled. Epipalaeolithic hunter-gatherers living in wetland contexts employed a rational balance of reliable wetland resources which provided a year-round availability of edible plants such as Cyperaceae, Phragmites etc. together with animal resources and water. They also foraged strategically beyond the wetland into the surrounding steppe (xeric grass and shrub vegetation) and parkland (open vegetation with occasional trees) for highly seasonal wild cereals, grasses and other resources (Ramsey et al. 2016, p. 3). This paper hypothesizes that the wetlands provided a 'domestication laboratory', both in the wetlands themselves with useful Cyperaceae and Phragmites but also further afield in the steppe and parkland with wild cereals, which allowed generations of Epipalaeolithic people to interact, modify their environment, experiment and gain new knowledge to build the long-lasting ecological and cultural inheritances necessary for plant domestication.
Unexpectedly, Epipalaeolithic peoples appear to have relied on wetland resources for their increasing use of wild grass and cereal resources. Indeed, the hypothesis put forth in this paper adds to the growing consensus that we need to rethink our view of the transition to plant food production in southwest Asia as a linear development just involving cereals. Expanding our understanding of pre-agricultural plant use by using a landscape-level approach, this paper evaluates the multi-proxy archaeobotanical evidence of foodways, material culture and palaeoenvironment over a long time period to assess other early forms of resource use and potential resource management in the wetlands. It contributes to the understanding of Neolithic plant use by exploring the origins of plant food production, assessing the evidence of wild plant use over the Epipalaeolithic period before cultivation and morphological domestication began during the Pre-Pottery Neolithic . The Epipalaeolithic period includes the Early Epipalaeolithic (EE), Middle Epipalaeolithic (ME) and the Late Epipalaeolithic (LE) (Maher et al. 2012b;Belfer-Cohen and Goring-Morris 2014). These time periods coincide with a range of culture groups defined by tool typology (Muheisen 1988;Olszewski 2001;Nadel 2003;Kadowaki 2013), the most famous being the Natufian culture during the LE (Garrod 1957).
The Epipalaeolithic in the southern Levant, the LE in particular, is considered to be important to understanding the origins of agriculture in the region (Bar-Yosef and Belfer-Cohen 1989;Belfer-Cohen 1991;Valla 1995;Bar-Yosef 1998). Diverse wild plant resource use and broader human-environment interactions during the period produced the ecological and cultural foundations for domestication. Yet, previous macrobotanical research in the regions focused narrowly on the intensification of cereal resources (Kislev et al. 1992;Weiss et al. 2004a, b). Only recently have researchers adopted a broader perspective on wild plant use (Caracuta et al. 2016a;Arranz-Otaegui et al. 2018a, b). Moreover, most of the archaeobotanical research has employed a site level approach, producing localized models over short timescales from sites such as Ohalo II, Kharaneh IV and Shubayqa 1. The transition to plant food production cannot be understood by examining any single archaeological site. As the macrobotanical syntheses of Fuller (2012, 2013) and Arranz-Otaegui et al. (2018a, b) show, providing detailed regional archaeobotanical data is essential. Building on this, researchers have recognized the need to take a landscape level approach, which seeks to understand plant use at a broader scale, the scale of plant communities, over longer timescales (Fuller et al. , 2016Allaby et al. 2022). Archaeobotanists need to cross the on-site and off-site boundaries, and examine the foodways, material culture and palaeoenvironmental trends, in order to build comprehensive and comparable botanical and environmental datasets. This paper will evaluate the state of the archaeobotanical record and interpret the evidence within the framework of NCT (niche construction theory) to assess the hypothesis of a non-linear shift from hunter-gatherers to wild plant food producers based on their use of wetland resources. It will conclude with some future directions for research.

Background
Changing scales and our understanding of the transition to agriculture: Niche construction theory and a landscape level approach Humans are the ultimate niche constructing species (Smith 2007). Our ability to shape the environment precipitated the Anthropocene. Niche construction (Lewontin 1983;Odling-Smee et al. 2003) recognizes that organisms modify their ecosystems through their metabolism, activities and choices, affecting the course of their own evolutionary development (Laland et al. 2000(Laland et al. , 2001Odling-Smee et al. 2003). Uniquely, in shaping their environments, humans leave behind an ecological and cultural inheritance such as Jomon cultivated environments in Japan (Bleed and Matsui 2010), technological innovation (Collard et al. 2011), the development of plant food production in pre-European contact southern Ontario, Canada (Crawford 2014), the engineered landscape of Balinese rice terraces, Indonesia (Lansing and Fox 2011), prehistoric 'conservation measures' through the restructuring of resource communities to enhance resilience (Smith 2009), 'wild' plant and animal management or ecosystem engineering in small scale societies (Smith 2011a), or the evolutionary consequences of food production (Wollstonecroft 2011), that archaeology is well situated to uncover (Laland and O'Brien 2010).
Ecological inheritance, following Odling-Smee and Laland (2011), occurs when the niche constructing activities of an organism produce an ecological effect or 'legacy effect' (Cuddington 2011) in the environment that changes the selective pressures for subsequent generations with evolutionarily significant consequences. Beaver dams are an often-used example of this phenomenon in the animal world.
In the human context, the concept of 'evolutionarily significant consequences' is more complex, including both genetic and cultural evolution (Richerson andBoyd 2005) Mesoudi (2017) notes "[t]he emergence of this second evolutionary process [cultural evolution] saw an unprecedented Thornton 2018). Therefore, the process of accumulating ecological-cultural inheritances makes NCT relevant for archaeological studies.
The application of NCT has been steadily growing over the past decade as an approach to understanding humanenvironment interactions and in particular plant use and domestication in archaeology (Smith 2007(Smith , 2011bRowley-Conwy and Layton 2011;O'Brien and Laland 2012;Crawford 2014;Florin and Carah 2018). But the number of criticisms of NCT has also grown, with some arguing that it is only useful for description and does not provide a testable explanatory mechanism (Wallach 2016;Spengler 2021, also for detailed literature review). This criticism is often levelled by optimal foraging proponents who are particularly interested in discovering the prime movers for resource selection and agricultural origins. Academic discourse has tended to pit resource depression, the prime-mover posited in optimal foraging models against resource abundance, the primemover that has tended to be posited by NCT models (Smith 2011b(Smith , 2015Gremillion et al. 2014;Zeder 2015Zeder , 2018. The author does not view NCT in opposition to optimal foraging. Rather, it fits neatly as a part of human behavioural ecology which takes an evolutionary approach to human behaviour. NCT is complementary to a range of other concepts, including historical ecology (Crumley 1994;Balée and Erickson 2006) and human ecology (Butzer 1982). Findings interpreted by NCT can be usefully applied to adjust local optimal foraging models or plant food production models (Mohlenhoff et al. 2015;Mohlenhoff and Codding 2017;Piperno et al. 2017;Florin and Carah 2018;Ready and Price 2021). This nuanced and flexible application of NCT for interpreting archaeological data has led some to suggest that it is less than straightforward or even 'ununified' (Spengler 2021). This criticism highlights the need for a broader discussion about how NCT should be applied usefully to archaeological data. Is it, for example, simply a new term for a range of concepts we already employ such as management, cultural ecosystems and human activities? Does NCT have any explanatory mechanism or testable utility, a middle-range theory application (theories linking human behaviours and natural processes to physical remains in the archaeological record)? Or is it descriptive in use and only helpful for generating and organizing ideas like historical ecology? These are some of the questions we need to think about when using NCT for interpreting archaeological data. This paper argues that NCT is useful in archaeology and in particular for understanding changes and transitions in plant use over long time periods. While different plant management strategies such as vegetation burning, coppicing, weeding, etc. are part of a range of human niche construction or ecosystem engineering behaviours (synchronic), they do not in isolation result in evolutionarily significant extension of genetic evolution by allowing organisms to adapt more rapidly to, and more powerfully create and shape, their environments". Examples of evolutionarily significant cultural evolution include major phenomena such as the transition to plant food production, the start of agriculture, the rise of urban centres and the development of civilizations.
In NCT, cultural evolution is transmitted through 'cultural inheritance', which Odling-Smee and Laland (2011, p. 226) define as "that subset of niche construction that is the expression of culturally learned and transmitted knowledge". This as a definition is, however, somewhat problematic from an environmental archaeology perspective. Following this definition, any interaction between humans and their environment that produces evolutionarily significant phenomena stemming from culturally learned or transmitted knowledge is considered cultural rather than ecological inheritance. In studying plant remains, for example, the archaeobotanist would be required to ascertain whether or not any interaction causing landscape modification such as landscape burning and the subsequent ecological succession, or Cyperaceae or Phragmites collection and the subsequent expansion of the wetland margin (phytolittoral zone), was the result of cultural learning, which would be difficult to prove. Plant domestication is a clear example of a process that required culturally transmitted learning. However, the line of knowledge transmission is less clear the further back in time we go from morphological domestication. In many contexts it is inconsequential to the broader questions whether the environmental signature is an ecological or a cultural inheritance. Moreover, Odling-Smee and Laland (2011, p. 227) view cultural inheritance as a subset of ecological inheritance, and therefore the two are closely interrelated. For these reasons, the author sees very little need to separate these two concepts in the context of understanding Epipalaeolithic plant use and the evolution of plant food production. Accordingly, in this paper, these concepts are mentioned together as ecological-cultural inheritance.
Odling-Smee and Laland (2011) recognize that the buildup of niche construction as an ecological inheritance over long time periods strongly affects the development of constructor populations, such as coral reefs or shell mounds. This time dimension to niche construction is particularly significant in human niche construction (Sterelny 2012) as ecological-cultural inheritance, and is therefore key to exploring how humans use and are affected by niche construction. This paper contends that as successive generations build on the ecological-cultural inheritance of their ancestors, their ecological and cultural innovations become more powerful, increasing the impact of their niche construction practices and exerting evolutionarily significant phenomena, which is cumulative cultural evolution (Mesoudi and surface gradient, water flow, water depth, etc. (Fig. 3). A landscape level approach requires comparative diachronic archaeological datasets using appropriate scales such as landscape/population vs. taxa/individual. This approach to NCT avoids the pitfall of being 'allencompassing' (Spengler 2021) and favours no single assumption from first principles regarding the prime mover behind human resource selection, for instance resource depression vs. resource abundance, and emphasizes a bottom-up data-led approach. NCT requires the evaluation of the appropriateness of the datasets being used, an appraisal all researchers should make regardless of their preferred theoretical framework. In this paper, diachronic archaeobotanical evidence provides insights into Epipalaeolithic ecological-cultural inheritances and helps in understanding the development of broader cultural changes, like the transition to agriculture. It will help us refine and better apply middle-range theory, that directly links human behaviour to physical remains, such as optimal foraging theory and plant food production models.

Non-linear interactions between people and plants, and plant food production models
It is important to have a long time perspective of human plant use to be able to understand the foundations for the origins of agriculture. Ecological-cultural inheritance played a crucial part in this transition. Indeed, the accumulation of ecological-cultural inheritances and associated ecological and cultural feedbacks, which can radically alter the patterns of plant use and their changes, increases the likelihood of non-linear, or irregular, developments. Linear developments tend to be additive, x leads to y. Straight-line thinking is typical when modelling relationships, and simplification is necessary. Non-linear developments are not based on proportional cause and effect outcomes. Interactions between humans and their environment and with plants are co-evolutionary dynamics in which causes and their effects are not necessarily proportional. Co-evolution is one of the major mechanisms for generating non-linear interactions (Holland 2014). Indeed, as Freeman et al. (2015) point out "the feedbacks caused by niche construction… [are] known to cause non-linear trajectories of evolution". Most relationships or interactions are non-linear. In the context of longterm people-plant interactions, many of the developments are non-linear. Therefore, rather than applying a linear retrospective focused on cereal development, a data-led bottomup approach is necessary to understand plant use during the Epipalaeolithic (Fig. 4a). The potential for non-linearity makes it important to integrate a NCT perspective in the application of plant food production models (Fig. 5;Ford 1985;Harris 1989Harris , 1996Smith 2001). phenomena, either genetically or culturally. There are two primary variables to consider when determining the utility of NCT in archaeological contexts. First, is it possible to identify and track evidence of ecological-cultural inheritance, the long-term generational impacts of niche construction (diachronic)? Second, can the investigation inform understanding of evolutionarily significant phenomena?
The concept of a landscape level approach is borrowed from ecology and was most clearly outlined for archaeological applications by Karl Butzer (1982). Butzer's approach emphasized a diachronic understanding of the 'landscape context' (1982, p 38). Landscapes are complex systems, they are multi-component, consisting of biotic elements such as animals and plants, and abiotic elements such as soil, water, surface gradient, etc., and these are constantly interacting, changing and affecting subsequent ecological processes (Newman et al. 2019) in the sense of niche construction and ecological inheritance. The concept of a 'landscape level' approach ('landscape framework' according to Allaby et al. 2022) has recently been applied to understanding the origins of plant domestication and human plant use with the aim of moving beyond site-level analyses (Fuller et al. , 2016Allaby et al. 2022). In the context of tracking trends of Epipalaeolithic human plant use, this paper employs a landscape level approach that shifts focus from one plant of interest such as Schoenoplectus (club rush), an approach typically applied in macrobotanical analysis, to viewing a range of plant taxa (plant communities) in their biotic and abiotic contexts; for wetlands, their biotic components are plants; and their abiotic components the water chemistry, Fig. 3 Simple illustration demonstrating how biotic and abiotic components can interact to change the wetland; a, shallow surface gradient produces a larger phytolittoral zone; b, steep surface gradient produces a smaller phytolittoral zone Plant food production models map the transition from hunting and gathering to agriculture. They are necessarily simplifications, but context and complexity must be added conditions before the Bølling-Allerød, which featured rapid climatic amelioration to warm and wet, before the start of the Younger Dryas, another cold/dry period (Robinson et al. 2006, for summary of climatic periods/events). These trends are now being confirmed at the local scale with studies that combine archaeobotanical remains with stable isotope analysis (Caracuta et al. 2016b;Belli et al. 2021) (El Wad and Raqefet caves, Table 1).
The climatic changes would have caused shifting environmental foraging opportunities. Warm/wet periods were characterized by the expansion of woodlands which were rich in various nuts and fruits. Cold/dry periods were characterized by the expansion of steppe and parkland which had a range of edible wild grasses and cereals growing there (Rosen and Rivera-Collazo 2012). Steppe and parkland varied greatly in primary productivity between cold/dry and warm/wet periods (Gibson 2009), with grasses and parkland resources being more productive during warm/wet periods (Ramsey et al. 2016, p. 8, Fig. 3). This variation means that the wetland environments that occurred in the region in places such as Azraq basin or Hula marsh would have been particularly important during cold/dry periods such as the LGM, when the steppe and parkland would probably have been less productive . Wetland environments provide reliable perennially available plant resources, specifically plants with edible roots like Cyperaceae, which actually increase in nutrient quality during dry periods of low growth (Wrangham et al. 2009). The phytolittoral zone, the vegetated edges typical of marshes and other shallow water environments, are identified as the most productive part of the wetlands Keddy 2000). Some Cyperaceae (sedges etc.) and Phragmites sp. (reeds) which would have been key plant taxa of great economic and subsistence value to humans, grow in such shallow wetland environments.
Regarding subsistence, Hillman et al. (1989) outlined the processing methods and dietary uses of Cyperus tubers worldwide. Of the species available in the southern Levant, C. rotundus rhizomes are high in carbohydrates and are shown to have a calorific return of 4.435 kcal/hr gathering time, higher than most cereals (Wright 1994). Experimental work on another of the Cyperaceae available in the southern Levant, Bolboschoenus maritimus (sea club-rush), suggests that grinding and pulverizing of the roots would have softened the tissues and increased the bioaccessibility of nutrients through cell rupture and separation (Wollstonecroft et al. 2008) Hillman (2000 also conducted processing experiments demonstrating that Bolboschoenus sp. seeds can be roasted and ground into a palatable flour. Ethnobotanical studies from the Great Basin, USA, provide a useful ethnographic analog, further emphasizing the potential of wetland resources, such as those found in the southern Levant, for in their application. In the 'middle ground' between hunting and gathering and agriculture, a range of strategies for using plants has been outlined. Ford's (1985) model identifies cultivation as the main stage between foraging and domestication. Harris' (1989) model is more nuanced in identifying wild plant food production and cultivation, with domestication marking agriculture. This model was revised in 1996, removing the hard boundary for domestication and focusing instead on the shift from wild plant food procurement to wild plant food production (Harris 1996). Smith's (2001) model identifies two categories between hunting-gathering and agriculture, which are low-level food production either without domesticates or with them (Fig. 5a). These authors all mark the differences between their model categories based on the use of certain management practices and their related outcomes such as domestication. Harris' (1989) model is particularly descriptive regarding the kinds of activities for plant use that mark each category. For example, wild plant food procurement includes burning vegetation and protective tending; wild plant food production includes transplanting, weeding, replacement planting/ sowing, storage and irrigation/drainage, while cultivation includes land clearance and systematic soil tillage.
NCT should be applied at the macro level of analysis topically (taxa vs. landscape) and temporally (synchronic vs. diachronic). While the plant use activities listed above are examples of ecosystem engineering, they are not in isolation evidence of evolutionarily significant niche construction (Odling-Smee 2010). Rather, these activities of plant use need to accumulate through time with repeated and/ or group activity, potentially over generations to exert an ecological-cultural inheritance of evolutionary significance. Accordingly, for the Epipalaeolithic in the southern Levant it is critical to take a long term diachronic view of the environmental and archaeological evidence to construct appropriate archaeological and archaeobotanical interpretations of regional transitions in plant food production.

Epipalaeolithic climate and environmental foraging opportunities
Isotope data from the Nahal Soreq cave provides a longterm palaeoclimatic record for the region. Dated by uranium/thorium and expressed as calendar years, the δ 18 O (calibrated for rainfall) and δ 13 C (calibrated for vegetation, C 3 vs. C 4 ) values are used to show changes between warm/ wet and cold/dry periods in the region ( Fig. 2; Bar-Matthews et al. 1997Ayalon 2003, 2004). The data show that during the Epipalaeolithic period there were pronounced climatic changes, with an extreme cold and dry period during the Last Glacial Maximum (LGM) followed by a less extreme period of fluctuating cold/dry    Fuller 2012, 2013;Arranz-Otaegui et al. 2018a, b). The published microbotanical phytolith record discussed here currently includes over 200 samples extracted from the sediments of 11 sites. Macrobotanical evidence from seeds, plant tissues and wood charcoal is available from 11 sites and other lines of microbotanical evidence (starch and starch spherulites) from three sites are also included. The archaeobotanical results are provided as a qualitative summary, with references to the published data, to provide an important reference list and summary of plant use. It was not possible to compare the results between all the sites quantitatively, because the varied data sets are often either not comparably sampled and/or the raw data are not available in the publications. Moreover, using this approach to compare results quantitatively requires close control of contexts, such as that achieved at Ohalo II (Ramsey et al. 2017). However, comparing plant remains qualitatively between sites is a powerful tool for understanding trends in plant use, but issues of differential taphonomy between archaeobotanical remains and processing discrepancies must be ruled out or acknowledged as possibly contributing to the patterns in the archaeobotanical assemblages. The results in this paper are discussed qualitatively at two scales, first in broad groupings at the landscape level, or as ecozone types (wetland, steppe and parkland, and woodland), generally shown by phytolith data following the phytolith groupings in , Table 2; Fig. 6, and secondly in more detail at the level of taxa according to the available macrobotanical and other archaeobotanical evidence. Phytolith analysis is particularly suited to a landscape scale approach, for although phytoliths are not very taxonomically specific, they are generally more taphonomically robust than other plant remains such as seeds. This allows easier comparison between sites, and in future studies, between on-site and off-site contexts.

Wetlands
The phytolith evidence suggests that Epipalaeolithic people relied heavily on wetland resources for craft uses such as hut construction as well as for food, with plenty of evidence for Cyperaceae and Phragmites throughout the entire 13.5 ka period (Fig. 6, Tables 1 and 2). This interpretation is also supported by more recent archaeobotanical investigations, particularly macrobotanical evidence of Cyperaceae recovered from the hearths at Shubayqa 1, along with bread-like remains that included Bulboschoenus sp. tubers (probably prepared as flour), and starch evidence of Cyperus sp. from cupstone mortars at Raqefet cave, all of which demonstrate the use of wetland resources for food. The archaeological evidence shows the use of a grinding stone in the preparation of wetland plant resources and calls food, providing a wealth of evidence for how wetland edge plant resources may have been employed during the Epipalaeolithic (Ramsey et al. 2017). For example, Phragmites sp. seeds may have been dried and ground into a flour, the young shoots and leaves boiled as pot herbs, or the stems may have been dissolved in water to make a sweet drink (Ebeling 1986). These ethnographies emphasize the use of stone tools including hand grindstones, pestles and mortars, both portable and in bedrock, for preparing wetland plant materials for food.
The climatic and environmental setting of the southern Levant during the Epipalaeolithic period made wetland resources an important part of hunting, gathering and potentially, plant food production. However, the use of wetland resources is often underestimated in the region when evidenced only from macrobotanical remains, as demonstrated by Ramsey et al. (2017) in an analysis comparing the phytolith and macrobotanical assemblages from Ohalo II. This paper emphasizes a multi-proxy archaeobotanical approach for reconstructing foodways, material culture and palaeoenvironment. These approaches are standard practice in palaeoecology, however until recently, they were less well developed in archaeobotany (García-Granero et al. 2015).
In the past decade this has changed with the effective combination of macrobotanical and microbotanical studies at the site scale (Delhon et al. 2008;Dickau et al. 2012) and at the scale of individual food remains (Heiss et al. 2017;Arranz-Otaegui et al. 2018a). This paper outlines a multiproxy approach on a regional scale.

Archaeobotanical evidence: constructing a broad picture of epipalaeolithic plant use in the southern Levant
The published macrobotanical and microbotanical evidence from 19 Epipalaeolithic sites (Figs. 1 and 2; Table 1) is reviewed here to explore plant use on a regional scale in the southern Levant, mainly Israel and Jordan, although the site of Abu Hureyra, Syria, is included. Previous syntheses of plant use from the broader region and/or different time periods were based on macrobotanical remains only Cyperaceae 'cones' (Ollendorf 1987(Ollendorf , 1992, 'fanshaped bulliforms' (cf. reeds), reed culm and leaf, Phragmites sp. culm and leaf (Ryan 2011) Woodland Platelets, honeycomb, coarse verrucate, polyhedron (Bozarth 1992), all irregular dicots Steppe/parkland, grasses Dendritic long-cells, papillae, cereal straw (Rosen 1993), all husk multi-cells (Rosen 1992) grass resources) were used, and probably played as important a role in the diet as pre-domestic wild cereals (Rosen 2010). It is clear from the macrobotanical remains that there is a range of plant taxa from steppe and parkland environments that would have been important contributors to Epipalaeolithic diets, including Chenopodiaceae and Fabaceae. This suggests that plant use through the Epipalaeolithic period was not based increasingly on cereals. Rather, a range of wild plant taxa was used according to what was available in the local environment. This is particularly evident with the changing environmental opportunities that accompanied shifts in climate through time. For example, phytolith evidence suggests that during periods of increased aridity, evidence of steppe and parkland resources, particularly wild grasses and cereals, often increases in importance along with wetland resources at sites like Ohalo II, Kharaneh IV, Eynan, Hayonim terrace and Hilazon Tahit (Fig. 2).

Woodland
Phytolith evidence suggests that woodland resources often increase in importance with climatic amelioration, such as the increased rainfall immediately after the LGM (Last Glacial Maximum) at sites like Kharaneh IV, and immediately before the Younger Dryas climatic event at sites like Eynan and Hayonim terrace (Fig. 2, Table 1). Macrobotanical into question previous links between increasing evidence of grinding stones (Wright 1991(Wright , 1994 and increasing cereal use (McCorriston and Hole 1991;Willcox 2005;Edwards 2007), to the exclusion of other resources in the region. It is also important to note that Cyperaceae roots and tissues are incredibly delicate, and they are almost certainly under-represented and under-identified in archaeobotanical assemblages ("missing food", Arranz-Otaegui et al. 2018b). With many of the Epipalaeolithic sites displaying evidence for increasing sedentism (more settled life), and increasing use of grinding stones at, or close, to wetland environments, it is certainly the case that wetland plant resources formed a kind of 'natural food storage' (Lee and DeVore 1968), a reliable resource that enabled settlement for longer periods. Moreover, wetlands would have provided more than plants. Indeed, water itself would have been a key attraction for humans, along with a range of large animals and birds to hunt and fish to catch.

Steppe and parkland
Phytolith and macrobotanical evidence suggests that the role of pre-domestic wild cereals, wheat and barley, in Epipalaeolithic diet has been overstated. Phytolith evidence from the Early Epipalaeolithic through to and during the Natufian period suggests that small-seeded grasses (non-cereal wild Fig. 6 Images of phytoliths recovered, organized by environmenttype: 1. wetland-type phytoliths, 1a stacked 'fan-shaped bulliform' (cf. reeds) (Kharaneh IV), 1b sedge cones cf. Scirpus (Ohalo II), 1c Phragmites culm (Ohalo II); 2. woodland-type phytoliths, 2a coarse verrucate (dicot leaf) (Kharaneh IV), 2b platelet (dicot leaf) (Wadi Madamagh), 2c scalloped (dicot leaf) (Kharaneh IV); 3. parkland-type phytoliths, 3a wild grass husk (Ohalo II), 3b cereal straw (Ohalo II), 3c wild grass husk (Kharaneh IV)  beads, ochre and other desirable items (Muheisen 1988;Nadel 1994;Byrd and Monahan 1995;Hershkovitz et al. 1995;Simmons and Nadel 1998;Bar-Yosef Mayer 2005;Martin et al. 2010;Richter et al. 2011Richter et al. , 2012Maher et al. 2012bMaher et al. , 2016Pedergnana et al. 2021). The Natufian period provides the earliest published evidence for recognizably modern food innovations in the form of bread-like remains at Shubayqa 1, and potentially fermented beer-like drinks at Raqefet cave, although it must be acknowledged that there are ongoing questions about the identification of fermentation damage in archaeological starches. The processed food remains at Shubayqa 1 were made with a combination of Bolboschoenus tubers and wild cereals (Arranz-Otaegui et al. 2018a). The potentially fermented drinks at Raqefet cave also incorporated wetland plants and wild cereals (Liu et al. 2018). These studies demonstrate the crucial relationship between wetland plants and cereals in Epipalaeolithic diet, poetically referencing both the past and future of human cuisine. Importantly, these socio-economic advances hold the key to understanding the development of plant food production, which was a protracted non-linear process, the beginnings of which can be traced back to the Early Epipalaeolithic (~ 23 ka cal bp) Fuller 2012, 2013). Indeed, there was a period of over 15,000 years before agriculture started, when hunter-gatherers moved in and out of various types of subsistence and settlement which, when they became established later, became hallmarks of the Neolithic ). The archaeobotanical evidence shows that the Epipalaeolithic hunter-gatherers were using a wide variety of types of subsistence and settlement and were not knowingly on a path towards increasing cereal use, cultivation or inevitable domestication. Accordingly, in the absence of a clear linear progression of increasing cereal use, increasing sedentism and increasing social complexity (Fig. 4a), the transition to plant food production in the southern Levant needs to be re-examined.
In this paper, it is argued that the use of wetland resources during the EE at sites like Ohalo II and Kharaneh IV laid the groundwork for a non-linear transition to plant food production (Figs. 4b and 5b -c; Ramsey et al. , 2017. This view echoes that of Savard et al. (2006) who suggested that reliable valley bottom plant resources (such as wetland resources) were central to Epipalaeolithic diet. Indeed, the resource stability provided by the use of wetlands facilitated increasing sedentism and the development of an ecological-cultural inheritance of teaching, social learning, tools and food preparation that endured beyond site occupation. As shown at Eynan, Hilazon Tahit and Hayonim terrace, subsistence using the reliable wetland plant and animal resources was reinforced during increasingly arid and environmentally unpredictable periods such as the Younger Dryas, as originally argued by Rosen evidence of woodland use includes Amygdalus sp., Pistacia, Quercus (oak, acorns), Olea (wild olive) and Vitis (grape), and indicates that when these were available they were used. Poor macrobotanical preservation combined with differential phytolith records, along with the very limited number of starch residue studies, makes fully evaluating the overall importance of these resources problematic. The records are differential because dicot woodland taxa do not produce as many or as diagnostic phytoliths as monocots such as Poaceae.

Discussion
The shift from wild plant food gathering to production: ecological-cultural inheritance and the non-linearity of Epipalaeolithic plant use Phytolittoral (water edge) resources respond positively with increased growth to disturbances such as fire, trampling or gathering (Ramsey et al. 2015). This characteristic of wetland plants produces a positive feedback loop, encouraging further human use and impact, and results in a productive self-reinforcing ecological-cultural inheritance. Therefore, the use of wetland environments and the ecological inheritance produced by human disturbance facilitated increasing sedentism (the author views sedentism as spanning a mobility spectrum from more to less sedentary) by increasing the availability of reliable perennially available wetland plant resources (Figs. 4b and 5c;Ramsey et al. , 2017. This dynamic has particular significance for the sites of Ohalo II, Kharaneh IV, river Jordan Dureijat, Eynan and Shubayqa 1. These sites are all located in the direct vicinity of a marsh or lake environment with a wealth of perennially available wetland plant resources such as roots, as well as other critical animal resources. The rich animal resource availability at Ohalo II is discussed by Steiner et al. (2022), and at other sites includes, fish (Ohalo II and river Jordan Dureijat) (Zohar et al. 2018;Pedergnana et al. 2021), wetland birds (Ohalo II, Shubayqa 1) (Simmons and Nadel 1998;Yeomans and Richter 2016) and gazelle (Kharaneh IV) (Martin et al. 2010;Spyrou et al. 2019). The sites also display overwhelming archaeobotanical evidence of wetland use (Rosen 2004(Rosen , 2010(Rosen , 2011Marder et al. 2015;Pedersen et al. 2016;Ramsey et al. , 2017Arranz-Otaegui et al. 2018a, b;Sharon et al. 2020). Several of these sites have strong archaeological evidence for increasing sedentism in the form of huts made of brushwood, then stone (Valla 1988;Nadel and Werker 1999;Maher et al. 2012a;Haklay and Gopher 2015;Richter et al. 2017;Ramsey et al. 2018) and strengthening social ties to a particular place, evidenced by on-site burial customs and a rich material assemblage that includes shell irrigation/drainage. This paper proposes that in the southern Levant the shift from wild plant food gathering to wild plant food production using wetlands may be detected through the introduction of new management activities, seen in the changing ratios of phytolith taxa and charcoal records in off-site cores; and the changing cultural significance of resources, detected with the introduction of new foodways, seen archaeological via processed food remains and starch spherulites. A long-term perspective is required to build a picture of change and to detect when such a change in scale occurred, as part of the multi-proxy archaeobotanical evidence of foodways, material culture and palaeoenvironment (Fig. 5c).
The Epipalaeolithic archaeobotanical evidence from the southern Levant, strongly supports the interpretation that wetland environments were chosen for long-term occupation and the resources from them were brought back to the settlements for a range of material and food purposes. The analysis of an off-site section in the Azraq basin demonstrates that Epipalaeolithic disturbances in the wetland environments would have increased productivity of critical wetland plants (Ramsey et al. 2015). The phytolith evidence shows the scale to which wetland environments were used compared to other environments (Ramsey et al. , 2017. The macrobotanical and microbotanical evidence is now beginning to demonstrate the range of ways in which wetland resources were incorporated into diet, including the culturally transformative introductions of 'bread' and potentially 'beer' during the Late Epipalaeolithic (Arranz-Otaegui et al. 2018a;Liu et al. 2018). This evidence, demonstrating the scale and cultural importance of wetland plant resource use, marks the first step in constructing an argument for the development of a wetland-based wild plant food production in the southern Levant during the Epipalaeolithic period (Fig. 5b, c).
A strong multi-proxy archaeobotanical dataset centred on evidence of foodways, and critically palaeoenvironment, is required to further assess this hypothesis. Building the evidence for the use of wetland resources for food is an important step and should include revisiting previous macrobotanical assemblages at sites like Ohalo II to search for unidentified plant food remains, including parenchymatous tissues and processed food remains.Identifications may be aided through the use of scanning electron microscopy (SEM) and new micro-CT (micro computed tomography) X ray imaging technologies. The capacity for studies of artefact residues and signs of use wear should also be developed in the region. Securely dated palaeoenvironmental reconstructions of the immediate areas around wetland-based sites, charting trends in microcharcoal (disturbance/management) and phytoliths (local environment), in conjunction with broader regional coring investigations including and Rivera-Collazo (2012). The use of reliable wetland resources by Epipalaeolithic people helped make possible the socioeconomic advances usually associated with cereal cultivation before the transition to a wild plant food production based on cereals. Indeed, wetland resources arguably underpinned the increasing use of less reliable but perhaps more desirable cereal resources through the Later Epipalaeolithic (Fig. 4b).
The idea that wild cereals and grasses were a desirable resource for Epipalaeolithic people is based on several related factors. First, ethnographic studies demonstrate that taste is one of the primary factors that determine food choice by hunter-gatherers (Heim and Pyhälä 2020). While flavour is a subjective characteristic, archaeobotanical evidence strongly supports the idea that hunter-gatherers put time and effort into gathering and processing wild cereals, both in the Levant (Piperno et al. 2004;Weiss et al. 2004a;Arranz-Otaegui et al. 2018a, b) and beyond (Aranguren et al. 2007;Mariotti Lippi et al. 2015;Cristiani et al. 2021). Second, cereals and grasses can play a critical role in many processed foods (Valamoti 2011;Arranz-Otaegui et al. 2018a) and have favourable biophysical properties. For example, adding cereal flour with gluten to processed starchy foodstuffs helps create an elastic dough (Shewry 2019). Third, processed foods may have been used as luxury food items for status acquisition (Hayden 2003). For example, analysis demonstrates that the flour used to make the 'bread-like' remains at Shubayqa 1 was ground very finely, requiring meticulous processing, much finer than later 'staple' breads (Arranz-Otaegui et al. 2018a), which suggests that it may have been an early luxury food item.
Epipalaeolithic hunter-gatherers relied on wetland resources and the resulting human effects upon them caused long lasting ecological-cultural inheritances, which as the paper argues, likely led to an unexpected and non-linear shift to wild plant food production using wetland plants prior to wild grasses and cereals (Fig. 4b). The transition to a wetland based wild plant food production in the Levant can be identified by viewing resource use at a landscape level (Fig. 5c).
Hunter-gatherers often manage a range of several resources in their environments to encourage their productivity (Peacock and Turner 2000;Anderson 2005;Turner 2005). Accordingly, archaeobotanical attention needs to shift from individual taxa to a group of taxa in a landscape, with a multi-proxy archaeobotanical approach to identify when a particular environment was being used. In Harris' (1989) model, he marks the transition to wild plant food production by naming key management activities. However, many of these are difficult to identify archaeologically and are not relevant for all plant taxa, such as transplanting, weeding, replacement planting/sowing, storage and Acknowledgements The work represents the build-up of ideas and influences over the author's doctoral and post-doctoral research. These ideas started to form under the expert supervision of Arlene Rosen and the late Karl Butzer and benefited from the insightful guidance of Michèlle Wollstonecroft. They have been further developed with colleagues in Cambridge, including Matthew Collins, Graeme Barker and the Cambridge Archaeobotany Group. Many conversations with Anna Florin were particularly rewarding. The participants of the Oxford Symposium on Neolithic Plant Management Strategies in Southwest Asia helped mould the paper. Thanks are due to the symposium organisers, Jade Whitlam and Alexander Weide. The comments of the two reviewers saved the author from many inconsistencies and omissions and were essential for improving the paper. The paper's errors and shortcomings are all the author's. The writing of the paper was supported by a Marie Sklodowska-Curie Actions Individual Fellowship

Conflict of Interest
The author has no relevant financial or non-financial interest to disclose.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/. pollen analyses should be used to track the scale and intensity of Epipalaeolithic resource management practices. To achieve regional scale palaeoenvironmental reconstructions, new machine learning algorithms are being developed to use artificial intelligence to identify phytoliths (such as Berganzo-Besga et al. 2022) and amass the large amount of archaeobotanical data needed. These lines of archaeobotanical evidence would show the impact of human management on the wetlands, and the range, scale and cultural importance of wetland plant use on-site, and when combined with archaeological evidence for increasing sedentism and cultural complexity, would demonstrate the change from wild plant food gathering to a wetland-based wild plant food production.

Conclusion
This paper provides the first multi-proxy (macrobotanical and microbotanical) synthesis of the published archaeobotanical evidence for the Epipalaeolithic period in the southern Levant. This dataset demonstrates that Epipalaeolithic hunter-gatherers relied heavily on wetland resources for materials and food. Viewed within the framework of niche construction theory (NCT), this paper argues that human impacts in the wetlands over generations increased wetland productivity, creating long-lasting ecological and cultural inheritances. This feedback is hypothesized to have promoted critical socioeconomic advances, including increasing sedentism, strengthening social ties to particular places and the rise of recognizably modern food innovations such as bread-like foodstuffs, which all contributed to changing Epipalaeolithic practices of plant use, and importantly, the evolution and proliferation of a wetland-based wild plant food production . The wetlands provided Epipalaeolithic people with a 'domestication laboratory' which allowed succeeding generations to interact and modify their environments, experimenting and gaining the necessary experience to build the long-lasting ecological-cultural inheritances needed for the transition to plant food production. Importantly, the ecological-cultural inheritance of a wetland-based wild plant food production enabled the non-linear transition to cereal-based wild plant food production and their later domestication in the southern Levant. Wetland-based innovations supported increasing use and experimentation with cereal resources over the 13.5 ka Epipalaeolithic period. This hypothesis should guide future research directions.