Risk, Reliability and Resilience: Phytolith Evidence for Alternative ‘Neolithization’ Pathways at Kharaneh IV in the Azraq Basin, Jordan

‘Neolithization’ pathway refers to the development of adaptations that characterized subsequent Neolithic life, sedentary occupations, and agriculture. In the Levant, the origins of these human behaviors are widely argued to have emerged during the Early Epipaleolithic (ca. 23 ka cal BP). Consequently, there has been a pre-occupation with identifying and modeling the dietary shift to cereal and grains during this period, which is considered to have been a key development that facilitated increasing sedentism and, eventually, agriculture. Yet, direct evidence of plant use in the form of macrobotanical remains is extremely limited at Epipaleolithic sites and the expected ‘Neolithization’ pathway has not been robustly demonstrated. However, new direct microbotanical phytolith evidence from the large aggregation site of Kharaneh IV, in the Azraq Basin, suggests that increasingly settled occupation was not the result of wild grass and cereal use, but rather the result of a typical hunter-gatherer balance, based on the use of mostly reliable resources supplemented by some risky resources. Moreover, and illustrating this balance, the direct botanical evidence emphases the importance of the wetlands as an under-recognized reliable plant resource. Significantly, the use of these reliable wetland plant resources at Kharaneh IV represents an unexpected ‘Neolithization’ pathway.


Introduction
The complex dynamics of people-plant interactions intrinsic to hunter-gatherer adaptations were central to the development of the 'Neolithic' lifestyle. Many researchers have characterized Epipaleolithic foragers as the link between hunter-gatherers and the first Neolithic farmers (ca. 11,500-10,500 cal. BP). In particular, the Late Epipaleolithic (Natufian period) (ca.

Late Pleistocene Environment at Kharaneh IV
The Azraq wetland is fed by several springs, which are filled by surface runoff and groundwater [44]. The springs have a low discharge-to-replenishment ratio, with a residence time of 20,000 years [45,46]. In the past, prior to modern water demands, the springs provided a secure supply of water, even under arid conditions [35,44,47]. At present, the much-diminished modern wetland is located approximately 40 km northwest of the site. However, in the past Kharaneh IV was situated on the periphery of the large Pleistocene wetland, surrounded by semi-arid steppe and parkland [35,44,[48][49][50][51][52][53]. The wetland would have been a reliable resource-rich environment relative to the surrounding steppe and parkland [54], particularly during arid periods. This characteristic of the Azraq Basin has attracted human groups to the region since the Lower Paleolithic [55][56][57][58]. These dynamics are not unique to the Azraq Basin and can be seen in other localized wetland contexts in the Levant, including for example the wadi systems in the northwestern Negev Desert dunefields containing Epipaleolithic sites [59]. Similar to the Late Pleistocene water bodies in the Negev Desert dunefield, the high water table characteristic of the Azraq wetland during the Early Epipaleolithic does not necessarily reflect a 'humid' phase (i.e. increased precipitation). In the case of the Negev, the development of ancient water bodies was the result of accumulating aeolian sand deposits blocking the wadis, or dune-damming, not increased precipitation [59]. While in Azraq, the aquifers supplying the wetland are fed mainly by groundwater discharge with a millenial scale recharge rate [35], meaning that most of the water accumulating in the Late Pleistocene wetland was a result of increased preciptation thousands of years prior to the LGM.
Recent geoarchaeological analysis by Jones, Maher [35] demonstrates that the wetlands adjacent to Kharaneh IV date to between 23 and 19 ka years ago. At the base of the Early Epipaleolithic occupation (Area B) the wetland deposits are interstratified with the earliest occupation layers. Therefore, it is clear that a rich wetland was located in direct proximity to the site when it was first established. Jones, Maher [35] note that there is little sedimentary evidence from which to reconstruct the environment during occupation of the site. However, they do suggest that the sustained occupation of the site indicates the wetland continued to be a favorable locale for a further 1200 years. Subsequently, the deposition of windblown (loess) deposits and the establishment of an erosional phase between 19 ka and 4 ka BP suggests there was a substantial drying of the wetland and the surrounding landscape [35,52,53].
Faunal evidence demonstrates that the inhabitants at Kharaneh IV had available a wide variety of animal resources, including water dependent species such as equids and aurochs, and smaller animals including tortoise and waterfowl [60]. However, they relied most heavily on gazelle (80% NISP in the midden, pit, cache and hearth contexts of area B and 90% in area A) [60,61]. Based on the skeletal-part profile of the assemblage the authors suggest that hunting occurred close to the site [60]. The consistent long-term hunting practices evident from the faunal assemblage indicates that throughout the sites occupation, the wetland and surrounding steppe and parkland environments were rich in game and provided a dependable supply of animal resources for both food and material manufacture.
suggests the site was visited by large numbers of people, staying at the site for long periods of time [26,60], on a multi-seasonal or possibly year-round basis [64].
Kharaneh IV's lithic assemblage is significantly different from the assemblage at Jilat 6, while the smaller nearby site of Ayn Qasiyya located in the central Azraq Oasis, features parallels with both lithic assemblages [35,60,65,66]. These varied lithic assemblages suggest that different groups of people from different regions of the Levant converged in the rich wetland environment. These large aggregation settlements may have been established as part of a strategy to claim distinct territories, legitimized through persistent occupation and the construction of socially meaningful places. Supporting this idea, Kharaneh IV features the earliest documented hut structures (of which there are currently three identified) in Jordan [36, 60, 63], a possible subfloor burial, a rich faunal assemblage, worked bone objects, a groundstone assemblage, red ochre and marine shell beads [26,60,61,67]. These material remains provide evidence for increasing sedentism, the use of complex trade networks, sophisticated food processing, personal adornment practices and symbolic behaviors.
Garrard and Price first surveyed the site in the 1970s. Small test excavations were subsequently undertaken in 1981 and 1985 by Muheisen, who excavated three areas totaling~15 m 2 (Area A, Area B and a small trench to the north of Area B). Muheisen [68] documented four occupational phases in total, labeled A-D. These phases included two later Geometric Kebaran phases (C and D) and two earlier Kebaran phases (A and B). In 2008 the Epipaleolithic Foragers in Azraq Project (EFAP) renewed excavations at the site [69][70][71][72]. EFAP excavations focus mainly, but not exclusively, on two areas of the site, an Early Epipaleolithic occupation (Area B, equivalent to Muheisen's Kebaran Phase A and B) and a Middle Epipaleolithic occupation (Area A, equivalent to Muheisen's Geometric Kebaran Phases C and D). Excavation of these occupations provided the samples for this study (Fig 2).

Modeling the Balance between Risky and Reliable Plant Resource Use at Kharaneh IV
Models of hunter-gatherer adaptation tend to share one basic assumption-that humans are rational decision makers [41][42][43]. Decision-making can be rooted in social as well as evolutionary 'reason. ' Yet, in the case of economic decisions (subsistence and settlement), hunter-gatherer scholars have tended to favor one of two explanations for human rationale: either the maximization of optimality, characteristic of optimal foraging theory, or the maximization of fitness, characteristic of human behavioral ecology (HBE) more generally [73]. However, there is growing sentiment supporting the idea that hunter-gatherer decisions were not always driven by maximization goals. Rather, they had only to be good enough to satisfy the range of overlapping interests and needs within the community [73], to 'satisfice' rather than to 'optimize' [74].
The decision to exploit certain resources is made relative to the other available options. Accordingly, resource selection and ranking decisions were determined at the local level in relation to the unique ecological mosaic (resource aggregation, productivity and predictability) of a region, as well as technological and social factors [75]. Changing climate in the Late Pleistocene would also have greatly impacted the availability of local resources. As such, Rosen and Rivera-Collazo [76] have suggested that Epipaleolithic foragers used an adaptive strategy that cycled in predictable ways from warm/wet phase focus on forest resources to dry/cold phase adaptations which targeted steppic resources-shifting easily back and forth in a cyclical manner from one to another. Recent phytolith work at Ohalo II builds on this model and suggests that wetland resources may also have been particularly important during dry/cold phases [37]. This cyclical adaptation suggests a great resilience, but also important variability in Epipaleolithic responses to these major climatic changes. Phytolith Evidence for Alternative 'Neolithization' Pathways at Kharaneh IV In the Southern Levant, phytolith evidence attests to the local nature of hunter-gatherer plant-use practices [38]. Habitats were targeted and the resources selected based on their appeal relative to the other options. Consequently, hunter-gatherer lifestyles and the exploitation of specific resources or environments (i.e. wetlands) varied according to the local resource and environment options (wetlands, parkland and steppe). Some earlier Epipaleolithic huntergatherer sites provide evidence for a burgeoning Neolithic lifestyle [23]. The unique local ecological factors and historical trajectories of these sites facilitated this transition. The Azraq wetland was an essential part of the ecological setting for Late Pleistocene hunter-gatherers and, therefore, should be a serious consideration when modeling regional plant-use (but see [77]). Indeed, Savard, Nesbitt and Jones ' [40] have suggested that reliable valley bottom plant resources were central to 'Neolithization' .
Wetlands can provide reliable and perennially available plant resources, specifically aquatic roots, which actually increase in nutrient quality during dry, low-growth periods [78]. However, wetlands should not be construed as lush plant food producing oases or 'Edens' [79]. Indeed, swamps, marshes, bogs, fens, wet meadows and shallow water are all defined broadly as 'wetlands' , although each have their own unique characteristics determined mainly by the transition between terrestrial and aquatic habitats [80]. The phytolittoral zone, the vegetated littoral typical of marshes and the edges of some shallow water environments, are identified as the most productive part of the wetlands [80,81].
As noted by Ramsey and Rosen [38], plants that thrive in this phytolittoral zone include some sedge varieties (Cyperaceae), cattails (Typha sp.) and reeds (Phragmites sp.). All of these plants are of great economic and subsistence value to humans, for the fauna they attract, as well as their own nutritional and favorable ecological qualities [81]. However, compared to other plant resource types, the roots of these aquatic species provide the lowest return rate at ca. 182 kcal/h, compared to terrestrial roots (ca. 2,267 kcal/h), nuts and acorns (ca. 832 kcal/h), and small seeds (ca. 364 kcal/h) ( [82], and references therein). Consequently, wetland plant foods might be shown to provide what people need (reliable, but low calories and water), rather than what people want (high calories, variability and flavor) [38].
Steppe environments in the Near East include a variety of economically important wild cereals and grasses, and can be seen to exist between two extremes, forest steppe (parkland) and desert steppe, canopy cover to sparse grass cover [83]. Situated on a transition controlled by precipitation levels, steppe environments can fluctuate greatly in terms of primary productivity. With increased precipitation, forest steppe (parkland) will support fruit, nut and mast 'orchards' . Importantly, reduced aridity will increase the reliability and length of harvest for these favored resources. Arid steppe environments have a very low primary productivity. In contrast, less arid steppe environments can have relatively high primary productivity [83].
The contrasting opportunities found in the wetland versus the steppe/parkland in the Azraq Basin offered hunter-gatherers clear resource options, which forced them to prioritize resources and the levels of risk they were willing to tolerate. While the resource potential of the favored steppe/parkland would have varied, given a sensitivity to changes in climate, the lessfavored resources in the wetlands would have provided a reliable fallback (Fig 3). Lee [84] found that the! Kung would eat as much vegetal food as they needed (meeting nutritional needs), but ate as much meat as possible. This pattern is expected according to Lee wherever two or more foods are available. Humans are apt to focus on the more reliable resource (i.e. wetland resources), but still prize the less reliable alternative (i.e. steppe/parkland resources). The push and pull between what people need (predictable foods) and want in their diet (preferable foods) is central to questions about plant-use in the past. This tension means hunter-gatherers do not necessarily choose to remain in the wetland until forced to do otherwise (through exhaustion of resources, social and/or climatic factors).
In regards to climatic factors, it is critical to acknowledge that such trends would have manifest differently in different regions. The southern Levant is characterized by a plethora of micro-habitats. However, archaeologically, we have tended to lump these varied micro-habitats together when considering human adaptation. Indeed, the 'risky' environment of the Western Levant during the Last Glacial Maximum may not have been so 'risky' in Azraq. Hunter-gatherer decisions are determined at the local level, and accordingly, the way in which climate trends impacted local micro-habitats should be considered.
The human preference for preferred foods, a varied diet and, importantly, satisfaction [42,85], means that during periods of climatic amelioration and increased reliability of the local landscape (warm/wet phases), people should have been more willing to take calculated risks, choosing to exploit plant resources outside of the wetland to a greater degree as the region of lower risk foraging expanded (Fig 4). Therefore, during periods of increased aridity and decreased reliability of the surrounding Azraq landscape (cold/dry phases), hunter-gatherers should have relied to a greater extent on the relative productivity and reliability of the wetlands (Fig 4). To paraphrase Bettinger [41]-don't take chances unless you have to. Wait until the odds are in your favor. This understanding of hunter-gatherer adaptation in combination with direct botanical evidence suggests an alternative 'Neolithization' pathway at Kharaneh IV.

Methods
All necessary permits were obtained for the described study from the Director General, Department of Antiquities of Jordan (Dates: 2008 Ref No. 12-5-2852Ref No. 12-5-2404, which complied with all relevant regulations. All archaeological phytolith slides produced as part of this study are stored at the University of Toronto, Toronto, Canada, freely accessible upon request from the corresponding author. The modern comparative collection phytolith slides are freely accessible upon request from A. Rosen and are housed in the University of Texas at Austin's Environmental Archaeology Laboratory. Lastly, the data that informs this paper is available in S1 Table.

Sediment Sampling
In order to investigate some of the principles outlined above, the senior author analyzed phytolith data from the Epipaleolithic site of Kharaneh IV (Geographical coordinates: 31°4 3' 24 N, 36°27' 15 E). A total of 31 sediment samples were analyzed for this study (Fig 2,  Table 1). The samples were taken from a variety of on-site contexts, including occupation Phytolith Evidence for Alternative 'Neolithization' Pathways at Kharaneh IV surfaces and features (hearths, dumps and post-holes). Excavations were conducted on a 1 x 1 m grid. However, these squares were further divided into 50 x 50 cm quads or 25 x 25 cm subquads when finer control was needed. The site was excavated following natural and cultural layers, which were further subdivided into 5-10 cm arbitrary levels. All of the excavated sediment was subject to flotation with additional samples collected for micro-artifacts, micro-fauna, micromorphology, and sediment analysis [71]. The phytolith samples for this analysis were collected from the sediment samples taken during the 2008, 2009 and 2010 excavation seasons.

Laboratory Methods
Phytoliths were extracted from the sediments following Rosen's [86,87] protocol, which employs a series of techniques to remove carbonates, clays and organics, before extracting the phytoliths. First, the sediment was sieved though a 0.25 mm mesh to remove the coarse sediment fraction. A sub-sample of approximately 800 mg was weighed and taken for analysis. The sample was treated with 30 ml of 10% HCI to remove the carbonates. To disperse the clays, a solution of sodium hexametaphosphate (lab grade Calgon and distilled water) was added to the sample. The clays were removed from the sample by decanting, after settling the fine sands and silts in an eight cm column of water for one hour. This process was repeated until the suspense was clear. Organic matter was removed by dry ashing the samples in a muffle furnace for 2 hours at 500°C. The phytoliths were then extracted from the remaining fraction using heavy density separation. A sodium polytungstate (SPT) solution (with distilled water) calibrated to 2.3 specific gravity was used to separate the phytoliths from the heavier minerals. The phytoliths were then poured off into a clean centrifuge tube, washed in distilled water, dried, weighed and then mounted on microscope slides in Entellan. The phytolith slides were counted at 400x magnification using a transmitted-light microscope (Nikon Eclipse E200). A minimum of 300 single-cells and 50 multi-cells (whenever possible) were counted on each slide. The results are expressed as number per gram of sediment. The absolute counts (number per gm sediment) for each phytolith type were calculated using a modified method outlined by Albert, Lavi [88]; Albert, Weiner [89] see Power, Rosen [90] for details.

Phytolith Analysis
Phytoliths are microscopic silt-sized particles of opaline silica. They form when plants take up soluble silica from the ground water. The silica is then deposited in and around the intracellular and extracellular spaces, creating durable inorganic silica 'casts' of the plants' cells. This process is genetically and environmentally determined [91,92]. Grasses, sedges and palms (monocotyledons) readily produce phytoliths, often distinctive to plant family, genus and more rarely, species. Woody trees and other herbaceous dicots also produce phytoliths, although far fewer and with more irregular forms [93]. Indeed, grasses produce 20 times more phytoliths than dicot wood and 16 times more than dicot leaves [94].
In grasses and other monocots, silica is deposited passively and actively in the cells of the plant. Therefore, phytoliths can form in individual cells, producing single-cell phytoliths, or as a suite of attached adjacent cells producing multi-cell forms, known also as silica skeletons. By studying the anatomical morphology of these fossilized sections of plant tissue it is possible to make identifications down to the plant genus or species level. Single-cell monocotyledon phytoliths are identified according to the ICPN classification system where possible [95]. Key phytolith microfossils employed in this study were identified by the criteria described in Table 2.
Modern comparative examples are pictured in Fig 5.

Results and Discussion
The phytoliths throughout the site appear to be well preserved with the presence of delicate morphotypes such as hairs and some large multi-cells, suggesting favorable preservation conditions. Two common monocot phytolith types, 'psilate long cells' and 'rondel'-shaped short cells, dominate all of the samples. Psilate long cells are found in all grasses and sedges and they have limited diagnostic utility, except as indicators of stems (also described as culms) rather than inflorescences, which, are indicated by dendritic long cells (primarily in pooid grasses). Phytolith Morphologies (single-cell unless described as multi-cell)

ICPN alternative References to identification criteria/comments
Psilate long-cell G~E longate psilate margin Most frequently found in grass stems [96,97].
Echinate long-cell M Elongate echinate General of monocots. Of particular importance as a morphology that is found in Phragmites (reed) culms (Fig 5F) Dendritic long-cell G Elongate dendritic Found primarily in pooid grass husks and are characterized by finely branched processes [98,99]. See the ICPN schematic drawings [95]. (Fig 5H).
Saddle short-cell G Generally chloridoid grasses [100], but also appears in Phragmites.

Wild grass husk G (multi-cell)
Generally pooid grass. Dendritic long cells, with papillae and short cells (mainly rondel). Cork cells are sometimes silicified (for a more detailed discussion of husk identification methods please refer to [99]) (Fig 5H; Fig 6A).
Bulliform G Found in the leaves of grasses, also known as motor-cells [96]. (Fig 5B).
Juncus-type M See (Fig 108 and 114A in [106]). Characterized by small, linear stacks of uniform oval to cube shaped cells.
Blocks DM Found mainly in dicot wood, cf. to parallelepiped block forms [89]. Phytolith Evidence for Alternative 'Neolithization' Pathways at Kharaneh IV However, the dominance of 'rondel' short cells is important for the reconstruction of local environmental conditions (Fig 6).

Vegetation Trends
The grass sub-families of pooids (with 'rondel' morphotypes) are formed in C 3 grasses and their prevalence indicates a prevailing cool or temperate climate. Chloridoids (indicated by 'saddle' short cells) generally form in C 4 grasses, and panicoids (indicated by 'cross, ' 'biolobe' and 'polylobe' short cells) generally form in C 4 grasses. Relative ratios of these sub-families are employed as a general proxy for temperature and level of aridity (Fig 7). High ratios of pooid to total pooid, chloridoid and panicoid grasses support the interpretation that inhabitants of Early Epipaleolithic Kharaneh IV were exploiting plants from the cooler more temperate micro-habitats that we expect were more prevalent around the site during the LGM. Given chloridoid grasses tolerate arid conditions better than panicoid grasses, the ratio of chloridoid to chloridoid and panicoid grasses is employed as a proxy for aridity, with higher ratios indicating plants from micro-habitats which were more arid, and lower ratios indicating plants from less arid zones [97,113]. At Kharaneh IV, there is a clear shift from a higher chloridoid ratio to Phytolith Evidence for Alternative 'Neolithization' Pathways at Kharaneh IV a lower chloridoid ratio from the Early to Middle Epipaleolithic (Fig 7). The evidence suggests that inhabitants of Kharaneh IV were exploiting plants from less arid zones in the steppe and parkland surrounding Kharaneh IV during the Middle Epipaleolithic occupation. This is in keeping with warming and wetting trends, which prevailed in the Southern Levant after the LGM. The finding is important because the primary productivity of steppe environments can fluctuate greatly based on the level of precipitation (Fig 3). It is possible this shift in vegetation reflects the increase in rainfall during the Middle Epipaleolithic, which resulted in the development of a more productive steppe/parkland in Azraq. This increase in the productive potential of the steppe/parkland should result in the expansion of lower-risk foraging into the wider steppe. A shift in foraging strategy should be evident in the phytolith assemblage through increasing use of steppe/parkland and woodland plant resources during the more humid Middle Epipaleolithic (warm/wet phase) when compared to the comparatively arid Early Epipaleolithic (cold/dry phase).
While human adaptation and plant use is constrained by plant availability, and therefore reflects environmental opportunities, given the assemblages are from on-site contexts, their composition is determined by human behavior. Yet, other on-site botanical assemblages, including the wood charcoal assemblage at El-Wad Terrace in Israel [114] have been employed effectively to reconstruct off-site vegetation. Moreover, as previously noted, at Kharaneh IV contemporaneous off-site sediments are rare and, so far, the on-site contexts provide our only avenue for reconstructing the local vegetation.

The Balance Between Risky and Reliable Plant Resource Use at Kharaneh IV
To develop a broad understanding of the foraging strategy employed at Kharaneh IV, the phytoliths are grouped to identify wetland, steppe/parkland grasses and woodland ecozone-types ( Table 3, Fig 8). These categories do not necessarily conform to traditional vegetation zones, but rather provide a general picture of plant-use categories. See Ramsey and Rosen [38] and Ramsey, Rosen [37] for a full discussion concerning the use of these categories.
From the distribution of ecozone-type phytoliths it is clear that wetland-type resources predominate. However, while phytolith evidence shows that wetland plant resources (reeds and sedges) were employed extensively, without starch evidence from the edible plant parts (seeds or roots) or evidence from contexts that point to consumption, such as groundstone residues, it is not possible at this time to identify how Early and Middle Epipaleolithic peoples at Pooid to pooid, panicoid and chloridoid grass ratio is a proxy for temperature (higher ratio indicates cooler conditions). Chloridoid to chloridoid and panicoid grass ratio is a proxy for precipitation (higher ratio indicates drier conditions).
doi:10.1371/journal.pone.0164081.g007 Kharaneh IV employed these resources in their diet. To rectify this research gap more refined context-specific microbotanical analysis and starch analysis of ground stone samples from Kharaneh IV is on-going. In spite of a lack of direct evidence, indirect evidence including ethnographic data from hunter-gatherers subsisting in analogous environments [115][116][117][118][119][120][121][122][123] and experimental evidence demonstrating the nutritional potential of processed sea club-rush roots (Bolboschoenus maritiumus (L.) Palla.) [124][125][126][127], a wetland sedge that has been recovered from ancient sites in the Levant and Anatolia [54], points to the importance and reliability of wetland plant-use in the region. Consequently, the prevalence of low-risk wetland-type resources throughout the site's use demonstrates that the inhabitants of Kharaneh IV employed a risk adverse wetland-oriented adaptation. This is not unexpected [38] and fits with the model of risky and reliable hunter-gatherer resource use described above. During the Early and Middle Epipaleolithic occupations, steppe/parkland grasses-type resources were also regularly exploited. This suggests that the inhabitants of Kharaneh IV were willing to tolerate higher risk foraging when supported by the reliable resources from the wetland. Based on the presence of dendritic long-cells (grass husks) (Fig 9) it appears that husks were prevalent on some occupational surfaces (loci 030 and 080), but were concentrated particularly in the feature contexts. In the Early Epipaleolithic contexts the features include dumps and pit fills and in the Middle Epipaleolithic contexts the features of interest are hearths (loci 080, 034, 065). The specialized nature of these contexts suggests that the husk remains are the result of wild grass and cereal processing refuse and/or disposal practices. Phytolith Evidence for Alternative 'Neolithization' Pathways at Kharaneh IV Comparing psilate long-cells to dendritic long-cells provides an indication of the level of grass and cereal use at the site (Fig 9). It appears that wild grasses and cereals were employed to a lesser degree at Kharaneh IV (5%) than at Ohalo II (14%). This trend noted by Ramsey and Rosen [38] is attributed to the lower levels of risk associated with foraging in the mosaic parkland landscape surrounding Ohalo II. Yet, at Kharaneh IV when increasing productivity in the steppe/parkland may have expanded the lower risk foraging options during the Middle Epipaleolithic occupation, the hunter-gatherers choose not to increase their exploitation of steppe/ parkland-type grasses. Rather, steppe/parkland grasses-type resource exploitation remained largely unchanged, while the level of woodland-type resource exploitation increased (Fig 8). This trend towards woodland resource use with the expansion of lower risk foraging, the results of a more humid Middle Epipaleolithic climate, is in keeping with the Epipaleolithic adaptive cycles proposed by Rosen and Rivera-Collazo [76].
Reviewing the dicot leaf and wood trends (Fig 10), it appears that the inhabitants at Kharaneh IV adjusted their foraging strategy in the Middle Epipaleolithic to include more dicot Phytolith Evidence for Alternative 'Neolithization' Pathways at Kharaneh IV resources. Considering there is no evidence for a change in wetland-type plant use trends (Fig  8) (i.e. shift from wetland to steppe resources-push factors), it is possible that this shift was facilitated by the increasing productivity of the surrounding parkland (pull factors), but it is unclear if this trend reflects increasing use of dicot resources for fuel, food or material purposes. The use of dicot resources as fuel is supported by the increase of dicot wood. Yet, this increase may also reflect a more general increase in dicot resources for fuel, materials and food. Ongoing starch analysis and continued investigation of on-site phytolith remains in conjunction with an expanded dicot reference collection will help clarify this trend.
It is important to understand why dicot resources and not wild grass and cereal resources, were prioritized with the possible expansion of lower risk foraging opportunities during the Middle Epipaleolithic occupation of the site. Like Kharaneh IV, Ohalo II also features a wetland-oriented lifestyle [37]. However, it is clear from both the microbotanical (phytolith and starch) [37, 128,129] and macrobotanical (for example see [130][131][132]) assemblages at Ohalo II that wild cereals and grasses were an important component of the diet. One explanation for the more muted use of grass resources at Kharaneh IV might be related to the types of grass resources available. It is possible that the rich mosaic environment near Ohalo II on the Sea of Galilee hosted a more preferable variety of grasses, including wild cereals, than the expansive steppe and parkland in the Azraq Basin. Starch and macrobotanical analysis may provide clarification. Even with the expansion of lower risk forging opportunities, the inhabitants at Kharaneh IV may have had no interest in gathering more grass resources.
If hunter-gatherer decisions had only to be good enough to satisfy the range of overlapping interests and needs within the community [73], it is possible that the reliable supply of plant resources in the wetland and the rich assortment of game, meant the inhabitants at Kharaneh IV achieved a sustainable balance between the use of risky and reliable resources, which suited their needs. While the changing landscape may have expanded the lower risk foraging opportunities and potentially facilitated an increase in the use of dicot resources, plant-use strategies remained largely unchanged. Indeed, the contrasting opportunities found in the wetland versus the steppe/parkland in the Azraq Basin offered hunter-gatherers clear options about what plant resources they wanted to prioritize and what levels of risk they were willing to tolerate, and they appear to have chosen to limit their risk by consistently focusing on lower risk foraging opportunities.

Conclusions: Alternative 'Neolithization' Pathways
'Neolithization' pathway refers to the development of adaptations that characterized subsequent Neolithic life, sedentary occupations and agriculture. However, the appearance of these incipient behaviors during the earlier Epipaleolithic did not inevitably lead towards a Neolithic lifestyle. Rather, the earlier Epipaleolithic could be viewed as a period of 'Neolithic' fits and starts. In this paper, we have presented new direct botanical evidence of one such fit and start. The evidence from Kharaneh IV has been employed to consider how the inhabitants used plant resources and how their selection of risky and reliable resources resulted in a resilient plant-use strategy that may have facilitated their increasingly sedentary lifestyle, and for that reason, represents the origins of an alternative and unexpected 'Neolithization' pathway.
At Kharaneh IV, their increasingly settled lifestyle, a key component of later 'Neolithic' lifestyles, is shown in this paper, based on plant use evidence, to have been born out of a rational balance, typical of hunter-gatherer adaptations [41][42][43], between the use of risky resources (i.e. wild cereals, grasses and other seasonal resources) found in the surrounding steppe/parkland landscape, and the use of reliable resources (i.e. sedges and reeds) found in the wetland landscape beside which the site itself was established. This strategy was resilient because by 'hedging their bets' on the year-round larder of the wetland and foraging strategically beyond the safety that afforded, the inhabitants at Kharaneh IV were able to aggregate in large groups and settle for longer than ever before in one place.
These new findings support Asouti and Fuller [27] in questioning the idea that Epipaleolithic plant-use practices emerged as 'pre-adaptations' to food production and, importantly, lends support to earlier works that have questioned the importance of grasses and wild cereal resources in the development of sedentism [40,133]. Indeed, the increasingly intensive settlement seen at Kharaneh IV, based on the phytolith evidence, was not the result of increasing grass and wild cereal use at all, but rather the result of a typical hunter-gatherer balance, based on the use of mostly reliable resources supplemented by some risky resources. In providing direct botanical evidence illustrating this balance, we have also brought to light an under-recognized reliable plant source, the wetland. For this reason, the use of reliable wetland plant resources at Kharaneh IV represents an unexpected 'Neolithization' pathway. Importantly, this reinforces Savard, Nesbitt and Jones' [40] contention that reliable resources were central to 'Neolithization' .
While, in hindsight, the lifestyle exhibited at Kharaneh IV may represent an incipient step towards 'Neolithization' , ultimately it was a result of the resilient wetland-oriented huntergatherer lifestyle employed by Early and Middle Epipaleolithic peoples. More on-site phytolith analysis and starch analysis in appropriate contexts (focused on features and ground stone tools) is currently underway to further refine how wetland and steppe/parkland starchy resources may have been employed in the diet. These types of analyses should be extended to later-period sites in the Azraq Basin, as we expect the use of reliable wetland plant resources continued to be central to the development of Neolithic life ways and perhaps even facilitated the adoption of cereal cultivation, and eventually agriculture.
Supporting Information S1