Risk, Uncertainty, and Resiliency in the Face of Ancient Climate Change: The Case for Legumes

Abstract

Continuing improvements in our understanding of ancient climate change renders it necessary to expand our toolkit for exploring human responses to climatic shifts. Currently, archaeological methods for exploring the resilience of ancient human agricultural systems—in addition to strategies for managing risk and/or uncertainty—are frustratingly limited in comparison to the rich ethnographic record of how humans have navigated climatic stressors. This article proposes that legumes might provide a new, albeit woefully understudied, vector for potential analyses, especially given their central role in traditional agricultural systems as a buffer against environmental stress. The peculiar agronomic character of legumes, especially among the widely cultivated varieties that are toxic in their unrefined state, could allow for robust hypotheses about agricultural strategies to be tested against our paleoclimate record. Importantly, these hypotheses could be tested against a wide variety of models of human–plant and human–environment interaction, as they could be based on labor costs rather than assumptions of ancient cultural preference. Legumes, however, present particular difficulties as objects of analyses, and therefore some methodological cautions are in order. Consequently, instead of proposing and testing hypotheses, this article seeks instead to inspire future research in relation to our constantly improving data.

1. Introduction

Past decades have seen dramatic improvement in our knowledge of ancient climate change (see articles in this special edition), though archaeological tools for identifying past human responses to climate change remain frustratingly limited in comparison to the ethnographic record for human adaptations to climatic stress. Archaeological studies have looked for correlation between climatic disruption and the disaggregation, dispersal, or general movement of populations as direct indicia for human abandonment of unfavorable environments and/or socioeconomic arrangements [1], though timing and causal relationships often remain hotly debated [2]. Other studies have focused on water management strategies [3], but the difficulty of dating water systems renders this proxy highly specific in application. Given improvements in archaeobotanical data, it has, however, recently become possible to assess regional and local impacts of climate change through the exploration of agricultural strategies that may have been employed to mitigate risk and/or uncertainty. This mode of analysis presents a potentially rich toolkit for studying human response to climatic stressors, though current methods are confined predominantly to comparing patterns in cereal agriculture and/or the relative exploitation of fodder versus pasture among agropastoralists [4,5,6]. To enrich our toolkit, I propose that legumes—and especially toxic legumes—provide a highly useful, albeit woefully understudied, proxy for identifying agricultural strategies that ensure yield under less-than-ideal circumstances. While we must always recognize that a farmer’s selection of crops is subject to manifold cultural, agronomic, and ecological considerations, the peculiar character of many leguminous species means that they offer unique potential for hypotheses that are testable against our paleoclimate record.

For this article, I will focus on leguminous crops associated with the core group of domesticates that originated in southwest Asia, species that were eventually utilized in agricultural systems across Africa, Europe, and Asia. Despite long-standing recognition of the centrality of legumes within early farming systems [7], they have received only a fraction of the attention granted to cereals. Partially, this stems from overemphasis on classic concepts like the so-called “Mediterranean Triad” [8], but at another level there are intrinsic difficulties in the archaeobotany of legumes that render other crop species more straightforward as objects of inquiry (see Section 3). Legumes, however, possess many characteristics of great interest to archaeologists exploring how ancient food production might have been adapted to the challenges posed by climate change. The ethnographic record of traditional farming practices is replete with evidence for the special role of legumes in resilient farming strategies, including how multiple species might be planted simultaneously to exploit their individual strengths as a buffer against environmental stress (see Section 4). Furthermore, ancient and traditional agricultural systems have simultaneously exploited legumes that range in levels of toxicity, wherein the higher labor required to process toxic species safely was counterbalanced by their general hardiness vis-à-vis other, less toxic variants. This is of special interest in relation to past human responses to climate change, as such species might provide insight into how food producers managed risk through crop diversity.

Legume toxicity is central to this article, as it does not necessarily prevent exploitation, but rather requires special knowledge and/or additional processing requirements to provide critical nutrition for both humans and livestock. For those in possession of such knowledge and the wherewithal to undertake more laborious processing, these toxic species can form a cornerstone of strategies for mitigating both short-term uncertainties alongside providing longer-term insulation against the effects of macroscale forces like climate change. Importantly, by focusing on labor requirements rather than modern assumptions that certain species were more or less desirable to an ancient individual—information which we generally lack and often rests instead on the faulty projection of our own etic sensory assumptions as normative [9]—this method has the potential to be more testable against a wider variety of models for human–environment and human–plant interaction.

While both our paleoclimate and archaeobotanical knowledge are constantly improving, if the desire is to conduct large-scale regional and cross-regional assessments of how factors like climate change might impact human–plant interactions, or even if we wish to explore at a local scale how individuals might strategize on the basis of crops available within their given milieu, a great deal of methodological care is required in relation to known gaps or biases in our current data. Legumes present particular difficulties in this respect (see Section 3). Consequently, this article predominantly seeks to highlight their potential and inspire future hypothesis generation such that we might consider new avenues for identifying risk and uncertainty management strategies, alongside long-term agricultural resiliency in relation to climatic conditions. The potential for legumes—toxic or otherwise—to provide proxy evidence for ancient adaptive agricultures intrinsically lies at the intersection of several distinct models for exploring human–plant interaction, some of which are often considered to be incommensurate. To elucidate such a framework, this article is structured as follows. First, I will discuss general issues related to resiliency in food production systems, risk management, and the decision-making processes behind the selection of cultivars (Section 2). I transition to the difficulties inherent to leguminous species as archaeobotanical data and their subsequent neglect within the literature (Section 3), which is followed by a discussion of their agricultural benefits in risk management (Section 4) and a characterization of legume toxicity among the species affiliated with the southwest Asian family of domesticates (Section 5). Finally, I offer some preliminary comments on potential ways to employ legumes in an exploration of our evolving dataset (Section 6). Additionally, the Supplementary File S1 supplies a glossary of terms that may be of use for the reader. While the focus here is on leguminous species from the orbit of southwest Asian domesticates, the framework presented here might prove transferrable to other species or regions.

2. Climatic Fluctuation, Resiliency, and Risk Management

At a certain level, exploring the role of legumes within food production—whether they served as food or fodder—requires a degree of theoretical pluralism. Indeed, even if the stated objective of this article is to inspire hypotheses for future testing against the ever-improving archaeobotanical data, the theoretical pluralism employed here merely indicates the frameworks within which such hypotheses might be tested. The usage of concepts like “risk management” and “resilience” already implies certain modes of thought; thus, it is prudent to pause and provide definitional clarity. Risk management, as employed here, operates predominantly within the understanding developed within Human Behavioral Ecology (HBE), wherein risk is understood as probabilistic variance in returns [4]. This is to be separated from uncertainty, which is to be understood as incomplete knowledge and therefore unpredictability of outcome probabilities. The two can, especially within archaeological application, be difficult to separate in practice, though the key distinction for risk management lies in the decision-maker’s knowledgeable tailoring of strategies on the basis of known past outcomes [10]. Both concepts are intrinsically short-term in their interpretive object (i.e., the strategic efforts of individual food producers), though this does not necessarily detract from their centrality in exploring how food production systems were used to mitigate the impact of climate change—one need only contemplate how long-term climatic cycles might have been experienced within individual lifespans as so-called “creeping normalcy” [11]. Resilience, in contrast, provides the more macroscale framing through which to understand how ecological, social, and social–ecological adaptive systems linked to food production might change in relation to longue durée forces like climate change, though it is necessary to separate resilience from a teleological approach that simply renders it as a contrasting state to collapse [5,12]. Applying the above language to the case of especially toxic legumes, such species might form a part of short-term risk management strategies among farmers, whereas their presence/absence within broader regional agricultures might play a key role in the resiliency of said system in the face of ecological processes derived from anthropogenic or non-anthropogenic causes [13].

While the language above is clearly indebted to fields like HBE, theoretical pluralism here requires engagement with approaches often treated as contradictory to the core models of HBE. For instance, there has been ongoing dispute over the utility of tools derived from HBE for the evaluation of food production, with Optimal Foraging Theory (OFT) receiving the greatest degree of attention [14,15,16,17]. This has coalesced around Niche Construction Theory (NCT) as a potentially stronger explanation for the adoption and elaboration of systems of human food production [18]. While the issue is more acute in debates about the origins of agriculture [13], the ready integration of NCT into resilience thinking makes it of special interest here. For instance, the role of legumes in nitrogen fixation and cover cropping—regardless of if farmers fully understood the benefits as such—can be viewed in light of the constructed niche wherein soil fertility becomes part of the ecological inheritance for subsequent users of agricultural land. While legume toxicity certainly can factor into both HBE- and NCT-based approaches, especially with respect to the interplay between nutritive returns and diversification as a method of risk management, other elements require attention—namely the material affordances of legume species, the transmission of practices related to their effective exploitation, and even their affective qualities (sensory or otherwise).

I have argued that theories of practice, materiality, and sensory archaeology form viable tools for exploring why individuals might or might not exploit toxic legume species—or any plant species for that matter [19]. While it has been repeatedly noted that OFT provides a hypothesis against which the data can be tested—thereby highlighting deviations from the expectations of the model—rather than a rigid explanatory covering law [14], to my knowledge there have been no proposed mechanisms to grapple with how OFT might highlight plant exploitation strategies that were potentially unconcerned with nutritive value and, instead, were concerned with some other quality of the plant-based comestible—e.g., how staples like cereals could have been emphasized for their affective qualities as a fermented drink, as discussed recently by Paulette [20]. In short, emic values might result in outcomes predictable via OFT, though the actual causal structure leading to the outcome might be fully independent of the assumptions of OFT. In such cases, should archaeologists desire highlighting something other than the prime mover of basic evolutionary fitness, it is incumbent to employ more malleable—though less intrinsically testable—tools from the broader suite of social theories concerned with reciprocal construction of agents and their sociocultural and material worlds [13]. Certainly, aspects of each aforementioned framework shed some light on why farmers employed certain legume species over time. Importantly, however, the use of legumes to better understand past agricultural strategies would be testable in some capacity within the confines of all the above frameworks.

3. The Problem of Legumes

Despite the fact that the domestication of various leguminous species seems to have been roughly coeval with that of cereals in the so-called “Fertile Crescent” [21,22], the adoption, spread, and elaboration of legume cultivation remains poorly understood. Though the discipline has largely abandoned concepts like the “Neolithic package,” “founder crops,” or the “Neolithic revolution” as oversimplifying and rendering teleological the diverse species, regions, timelines, and processes that characterize the development of food production [23,24,25,26], the cultivation of legumes—including toxic varieties—seems to have formed a key component of the diverse modes of food production that spread from southwest Asia to regions as far afield as Iberia and northwest Africa [27]. While the importance of legumes is evident, the general lack of attention given to them is not casual neglect. Rather, it stems from complications derived from a confluence of factors, namely leguminous botanical characteristics, the interplay between ancient agricultural and culinary practices with site formation processes, and biases introduced through archaeological recovery. Collectively, within the region of interest, this renders legumes difficult to study, difficult to recover, and—from a paleoethnobotanical standpoint—difficult to interpret.

For early periods especially, a major issue is that it is nearly impossible to differentiate domesticated legumes from their wild predecessors. The characteristics commonly associated with legume domestication—indehiscence (i.e., resistance to separation of the seed/pod from the plant), free-germination (i.e., lack of seed dormancy), and a smooth testa—are typically not observable in archaeologically recovered legumes [26,28]. This contrasts with the well-known signifiers of domestication in archaeologically recovered cereals [25]. Consequently, within the region of interest, early legume cultivation is often only identified via higher densities of legumes in association with clearly domesticated cereals or, alternatively, the presence of legumes outside of their ancient wild range [26]. The former intrinsically renders the study of legumes subsidiary to that of cereals, whereas the latter is stymied by the fact that legumes—including domesticated types—can also appear as weeds of cultivation in relation to other crops, as is the case with bitter vetch in modern-day Egypt [29,30]. Consequently, our capacity to discern why food producers might adopt legume cultivation remains poorly articulated. This pertains not just to their initial adoption by the earliest food producers but also why later food producers might—or might not—include foreign legumes among their cultivars as part of a resilience strategy. The issue is rendered more acute by anthropogenic and species-specific issues that suppress archaeological recovery of legumes.

Given the ancient culinary techniques for processing dried legumes—namely soaking, boiling, and/or pulverization—there is relatively little opportunity for their preservation via common vectors like charring [31]. Additionally, the oft-forgotten possibility that certain varieties—wild or domesticated—were simply consumed fresh as a green vegetable would also lower probabilities of preservation [32]. Consequently, it is typically only the rare instances of sudden preservation that allow for substantive legume remains to enter the archaeological record in ways visible to the most historically common collection strategies. The issue is exacerbated by foddering since the whole plant is valuable as livestock feed [33], thereby reducing the probability that portions of leguminous plants will enter the archaeological record as crop processing waste. Similarly, experiments have shown that legumes contribute little to the phytolith record [34], further reducing their archaeological visibility. Bioarchaeological evidence, either in the form of isotopic data [35] or studies of dental calculus [36], can demonstrate the human consumption of legumes, but neither possesses sufficient resolution to explore what species were consumed beyond the level of genus. Other microbotanical evidence, such as that derived from starch and palynological analyses, presents similar difficulties [37]. The formation of the archaeobotanical record is already a highly reductive process, and legumes are more susceptible to pre-depositional loss compared to other seed-bearing species from the pool of southwest Asian domesticates. They are also resistant to archaeological recovery, as charred legumes are prone to post-excavation loss even with high-resolution recovery techniques like flotation. Given past tendencies to only sample densely visible organic deposits preserved in situ within features like silos, taxa like cereals and their accompanying weeds are typically over-represented compared to legumes [19]. Indeed, it has recently been demonstrated that high numbers of soil samples directly correlate with larger and more taxonomically diverse legume assemblages [38]; therefore, it is only with relatively recent systematic archaeobotanical sampling that we could begin to appreciate the role of legumes within food production systems. Hence, it seems most appropriate that we develop hypotheses to be tested, rather than prematurely testing them against highly dynamic data.

4. The Benefits of Legume Exploitation

Legumes have become a central theme of modern agronomic literature, wherein the hardiness and nutritive value of many leguminous species has been noted as key to future resilience and sustainability in the face of modern climate change [39]. While exact characteristics vary by species, it is possible to generalize about their resistance to even extreme fluctuations in temperature, moisture, and soil conditions, such that legumes are often a failsafe within traditional farming systems wherein their performance vis-à-vis other crops—including other legumes—grants them the flexible status of food, fodder, or both [4,28,33,38]. Moreover, toxicities present in various legume species, the subject of Section 5, also lend resistance to pests—e.g., as with L-canavanine [40]. Substantial work has also been conducted on the nutritive complementarity of legumes and cereals and the capacity for legumes to improve soil nitrogen stocks—either through fixing atmospheric nitrogen or as a component of green manure [41]. Collectively, the above mark legumes as a cornerstone of many traditional and ancient food production systems, lending them a special place within risk management strategies and ecosystem engineering practices irrespective of the degree to which an individual farmer or farming culture employs direct knowledge of the underlying causative principles for these benefits [19]. Despite these generalizations, however, certain legume species exhibit variable performance under different environmental conditions, wherein the co-cultivation of multiple types might indicate agronomic strategies grappling with the effects of climate change.

The characteristics of individual species that are relevant to their viability in the face of changing and/or volatile climatic conditions stem predominantly from plant morphology, growth habit, seed size/composition, and phytochemistry. In food production systems favoring monocropping, these differences allow for strategies of temporal diversification, where each species is planted sequentially according to the seasonal conditions to which it is best adapted, with a general progression in traditional agriculture systems being from those species that better tolerate dryer conditions, such as the lentil (Lens culinaris Medik.), bitter vetch (Vicia ervilia L.), grass pea (Lathyrus sativus L.), and common vetch (Vicia sativa L.), to those which are more temperature- and moisture-sensitive, like the garden pea (Pisum sativum L.), chickpea (Cicer arietinum L.), and fava bean (Vicia faba L.) [28,33]. Notably, the ethnographic literature from southwest Asia, the Mediterranean basin, and north/northeast Africa also provides alternative legume diversification strategies based on mixed cropping, including methods like relay planting (i.e., broadcasting the seed of one species into beds filled with another, already-mature species), intersowing (i.e., planting different species at the same time but harvesting separately), or broadcast mixed sowing (i.e., maslins wherein the resulting mixture of species can be harvested simultaneously or separately depending on growing conditions and maturation rates). In these cases, the individual character of each species means that the farmer has insurance against less predictable weather conditions, an especially desirable strategy during periods of climatic transformation. The conditions that manifest within a given growth season and, therefore, the subsequent differential success of each sown species can provide a flexible crop of food, fodder, or both, depending on the surviving quantity and desirability of each harvested species. In difficult seasons, hardier legumes produced in such mixed fields that are considered less desirable and therefore typically grown as fodder might become an important failsafe in human diets [4,28,33]. Indeed, the ethnographic literature on traditional food production in Ethiopia revealed that informants intentionally cultivated the less desirable grass pea in legume co-agricultures specifically due to its resilience in the face of conditions deleterious to other legumes [42,43]. Consequently, archaeologists should pay close attention to evidence for legume co-cultivation in relation to known instances of climatic events, as it may be possible to see how food producers responded to changing environmental affordances. This may be especially relevant in relation to farmers who chose to include in their repertoire toxic legume species that require additional steps to exploit effectively, as these are among the hardiest legumes cultivated regularly.

5. Legume Toxicity

Archaeologists have a tendency to treat toxic legumes as highly dangerous, with their consumption potentially resulting in permanent disability or death [44]. A literary survey of paleoethnobotanical studies shows frequent statements that certain legume species—especially of the genera Vicia or Lathyrus—are unfit for human consumption, relegated either to famine food or fodder [32]. Multiple authors have, however, noted that legume toxicity has been overstated across species and, importantly, that toxins can be mitigated using a variety of techniques—indeed, many such species are dietary staples among modern populations who exhibit no ill effect from their frequent consumption [45,46,47,48,49,50]. The situation has been further aggravated by the recycling of references, including a rather confusing history of the epidemiological literature wherein Greek and Roman texts were often uncritically paired with preliminary—and often flawed—phytochemical analyses [19]. Importantly, nearly every leguminous species contains deleterious compounds, more commonly known as anti-nutritional factors (ANFs), with the relevant difference among them simply being the required mitigation techniques and consequences of unmodified consumption.

Regarding commonly exploited legumes, this article will focus on the six species most often recovered archaeologically within the regions that adapted food production from the southwest Asian pool of domesticates: lentil, garden pea, chickpea, fava bean, grass pea, and bitter vetch. Of these, the ANF content of the first four is predominantly composed of protease inhibitors (especially trypsin inhibitors), phenols, tannins, phytic acid, lectins, saponins, and oligosaccharides. While the bioactivity of each is different, the effects largely range from the inhibition of nutrient uptake to moderate gastrointestinal discomfort [51]. More importantly, however, even in an unprocessed state, the collective post-consumptive impact—either from a nutritive or health/wellbeing standpoint—is minimal, and even the most basic cooking techniques (e.g., exposure to heat and water) are more than sufficient to neutralize bulk ANF content across all four species [51,52,53,54,55,56,57]. In short, the agronomic and nutritive benefits of these four species can be realized with little additional labor input—simply cooking them was more than sufficient, and, therefore, they should occupy a separate conceptual category from more toxic legume species. The only possible exception is the fava bean, which has been subject to grave, albeit misplaced, archaeological anxieties about its association with favism—a genetic disorder wherein consumption of fava beans (or even exposure to fava pollen) leads to hemolytic anemia [37]. It should, however, be noted that favism has been overstated in the literature, as attested not only by the commonness of fava beans in modern foodways but also because the compounds in fava beans connected with favism—vicine and convicine (as precursors to divicine and isouramil)—can be dramatically reduced by the simple act of cooking [58,59]. Consequently, these four species can be regarded as relatively straightforward additions to food production and/or strategies of crop diversification, possibly even as interchangeable elements wherein we should expect relatively little resistance to their adoption as foreign cultivars provided another similar leguminous species was already familiar to the farmer in question. As such, they are unsurprisingly some of the most commonly recovered legume species in regions that adopted crops derived from the pool of southwest Asian domesticates.

The situation, however, changes dramatically with respect to the other two species of interest—the grass pea and bitter vetch. While both largely exhibit the same ANF content of the four species discussed above, they contain additional compounds of greater toxicity. Notably, these compounds do not allow for the casual use of these species as fodder plants, as is commonly assumed [60], but rather, they can be lethal in relatively small concentrations to monogastric livestock species and, furthermore, must be proportionally controlled in the diets of more resilient ruminants in order to avoid negative outcomes [47,61,62,63]. Consequently, special knowledge is required regardless of whether the crop in question is destined for human or animal food. Sustained consumption—or rather overconsumption—of grass pea has been associated with a condition variably known as lathyrism or neurolathyrism, a toxic myelopathy resulting in paralysis and death that stems from the presence of BOAA (β-N-oxalyl amino-L-alanine)—analogously ODAP (β-oxalyl-α,β-diaminopropionic acid). While the exact mechanisms of the disorder are not fully clear, the manifestation of the disease in modern populations that regularly consume grass pea is not nearly as severe as has been previously assumed in either the epidemiological or archaeological literature—the disease predominantly manifests under circumstances of sustained, elevated, and often near-exclusive consumption of grass pea [64,65,66]. That so many cultures around the world continue to consume grass pea safely or—as noted in some ethnographic investigation, cautiously—is because many groups exercise either conscious dietary balancing (e.g., mixing grass pea with other legumes or foodstuffs) or more rigorous processing that dramatically reduces the offending neurotoxin—e.g., varying combinations of soaking, heating (both dry and wet), grinding, dehulling, and even fermentation [42,43,67,68,69].

Bitter vetch was, for a long time, erroneously associated with lathyrism [19], though more recent phytochemistry has demonstrated that negative health outcomes in bitter vetch consumption—human or animal—are instead related to the presence of a different toxic compound, L-canavanine [40,47]. For humans and livestock, elevated consumption of L-canavanine—specifically via legumes of the genus Vicia—has been associated with inhibited nutrient uptake, fetal malformation, neurotoxic disturbances, hallucinogenic effects, hair loss, diarrhea, paralysis, cirrhosis of the liver, hypoglycemia, and arrythmia [46,68]. Much like the case with the grass pea, the chemical properties of L-canavanine and its variable presence in different portions of the bitter vetch seed mean that its reduction is only possible through a combination of processing methods that might variably include grinding, dehulling, soaking, heating, drying, and, potentially, fermentation [19,31,40,60,70,71]. Importantly, despite these phytochemical issues in both bitter vetch and the grass pea, as well as the additional labor required in rectifying them, ethnographic evidence has demonstrated that both legumes are valued within food production systems specifically for their hardiness in relation to a wide variety of extreme agricultural conditions, such that—while they may not always be the preferred food source—their cultivation as part of a diverse repertoire of legumes and grains serves to manage risk and insulate from uncertainty [42,47]. Consequently, it is inappropriate to dismiss these species as low-quality famine foods, but rather, we should explore them as potential indicia for strategies in relation to navigating various stressors of ancient agricultural systems—including but not limited to climate change.

6. Future Directions in the Archaeology of Toxic Legumes

Considering the difficulties for recovering legumes archaeologically (see Section 3), the treatment of legume cultivation over the longue durée in relation to any research question is difficult beyond discussions of presence/absence. Notably, however, we cannot always extrapolate from instances of presence whether the legume was wild, a weed of cultivation, or intensively exploited. Similarly, it remains difficult to assess whether absence is genuinely meaningful or if it is simply a consequence of the myriad biases outlined in Section 3. For instance, despite earlier arguments that bitter vetch cultivation likely never occurred west of Anatolia in during the Neolithic [72], subsequent data collection has demonstrated a near-continuous arc of early Neolithic cultivation of bitter vetch across southern Europe, into the Danube River Valley, and connecting Iberia with northwest Africa [27,73]. For this reason, it has been argued—especially for regions like Neolithic northeast Africa where archaeobotanical studies remain relatively rare—that we must exhibit extreme caution in relation to the “lack of balance between data and theories” [74]. It is necessary, therefore, to consider how we might meaningfully engage with the available data, even as it improves.

Fundamentally, the question relates to the presence/absence of a given leguminous species within the regional assemblage. The intrinsic issues with botanical data, however, mean that rather than simple presence/absence, we must try to understand the degree to which we can assume that an absence from a botanical assemblage represents a genuine condition of antiquity (i.e., total absence from the agricultural system) or whether it is an artifact of something else—site formation processes, collection bias, etc. Conversely, there remains the problem of whether presence indicates intentional exploitation or if, instead, it signifies some other possibility on the spectrum of human–plant relations (weed of cultivation, wild species, etc.). The presence of large caches of cleaned seeds are, at one level, the only definitive indicator of exploitation. No equivalent certainty can be assumed in relation to the absence of a taxon within an assemblage. Instead, both absence and low levels of presence within an assemblage must be treated cautiously with respect to their significance.

We therefore need to ask whether meaningful analyses must be quantitative, semiquantitative, or purely qualitative in character. The direct quantitative comparison of macrobotanical remains within and across regions implies that each assemblage within the overall dataset constitutes an accurate reflection of human–plant interactions in antiquity. Notably, the considerations described in Section 3 indicate that, for legumes especially, this assumption is untenable, and furthermore, the irregular character of the data discourages immediate leaps to tests of statistical significance. Instead, we must consider the target population about which we would like to draw conclusions based on our samples. Archaeologists do not directly sample the sum total of all plants that were in use at a site in antiquity, but rather we sample a volume of space for which the preserved macrobotanical remains are statistics of a sample element rather than the sample itself [75]. Importantly, this means that our target population can only be specified relative to space within a defined volume of sediment, not in relation to some hypothetical (and currently unknown) ancient population of plants [76]. Consequently, each sample within our dataset is a volumetric fraction of a constrained spatial population, thereby requiring both qualitative and quantitative characterization to assess the degree to which that sample might reflect broader agricultural strategies at a site or regional level. Put more plainly, the presence and absence any species within an assemblage does not directly correlate with patterns of human–plant interaction in antiquity but rather is a consequence of what survived pre- and post-depositional processes, what is present within the given volume of sediment accessible to archaeologists, and—in turn—what was actually sampled and subsequently not lost to various processes post-excavation.

Multiple strategies have been proposed to cautiously account for these issues in cross-regional datasets, thereby allowing us to draw larger conclusions about human–plant interaction over the longue durée. While quantitative studies via direct statistical analyses have been conducted, these have had greatest efficacy when applied to smaller regions [77]. More often, either in intra- or interregional discussions, more exploratory techniques have been applied. This has included visual, semiquantitative geographic representations wherein ranges in the number of identified specimens (NISP) of a given taxon are represented by abstract symbols [78], semiquantitative analyses employing the Representativeness Index (RI) [79], as well as semiquantitative applications to characterize regional patterns of presence–absence that combine qualitative sample characterization and distributional data (ubiquity) with diversity indices like the Shannon–Weaver Index and Index of Heterogeneity [19], to name but a few. Regardless of the approach taken, any hypothesis testing related to leguminous crops within and across agricultural systems, toxic or otherwise, will require some degree of accounting for the issues discussed in Section 3. This could take the form of constraining analyses to sites wherein systematic collection procedures were employed, but, ideally, any proposed method would possess the ability to engage meaningfully with less-than-ideal datasets from earlier periods of scholarship.

7. Conclusions

While this discussion is clearly preliminary, it is hoped that it will offer a potentially useful avenue for future inquiries. Collectively, only a few paleoethnobotanical tools have thus far been proposed for the identification of strategies related to the management of risk and/or uncertainty or the degree of resiliency that existed within ancient agricultural systems. Though legumes present problems still in need of resolution, they provide an additional, novel possibility for identifying how food producers strategized in relation to dynamic environmental conditions. Rapid improvements in the quality and quantity of archaeobotanical data offer increasing hope for higher resolution interrogation of ancient agricultural strategies in relation to shorter- and longer-term stressors, especially those related to climate. Given that our climatic data is also consistently improving, the types of legumes exploited—and the combinations in which they were exploited—could conceivably form a powerful piece of proxy evidence to be tested diachronically against regional and local climatic swings. Similarly, they provide promise in exploring how food producers navigated more localized landscape affordances. Certainly, the variables dictating an individual agent’s selection of a given cultivar are myriad, but by focusing those species requiring higher labor requirements and/or specialized knowledge for their effective exploitation, it will be possible to test our modern assumptions within a diverse array of frameworks for understanding human–environment and human–plant interaction.