Introduction

Although once united under the single term “natural philosophy,” for over a century scholars within the biological sciences striving to understand and manipulate the natural world have seldom interacted with scholars studying culture and society. This is problematic for many reasons, not least the social and cultural consequences of increasingly powerful biotechnology. However, recent developments at the intersection of the natural and social sciences – specifically, theories of cultural evolution, niche construction, and gene-culture co-evolution – have begun to bridge the gap between the study of biology and culture. In this paper we explore how these new interdisciplinary approaches might contribute to the study of agriculture - a topic that straddles the natural-social science divide.

The transition from hunting and gathering to agriculture observed in most human societies is a key event that has radically transformed human societies. For much of its evolutionary history our species practised hunting and gathering, as a few isolated societies still do today (Panter-Brick et al.2001). Beginning around 12,000 years ago, some human populations began domesticating plant and animal species (Fuller et al.2014; Larson et al.2014). The adoption of agriculture triggered the establishment of small permanent settlements and as populations expanded cities, kingdoms, and states. It allowed the creation of new political institutions and forms of social organization and stimulated an upsurge in scientific and technological innovation. It also brought many problems, such as the spread of new diseases and increased social inequality. Agricultural knowledge and technologies have continued to advance at an increasing pace particularly in the last century. The discovery of the rules of genetics by Mendel (Mendel 1866) and their rediscovery around 1900 (Corcos and Monaghan 1990) resulted in the application of plant breeding technologies from the 1930s onwards (Carlson 2004; Koornneef and Stam 2001; Heslop-Harrison and Scwarzacher 2012). The “green revolution” in several developing countries during 1950–1970, which utilized new high-yield crops together with fertilizers and pesticides, was an important landmark in agricultural plant breeding, yet still based on traditional Mendelian breeding methods (Farmer 1986). The era of molecular breeding, including marker-assisted selection (MAS), from 1983 onwards (Ben-Ari and Lavi 2012; Smith and Simpson 1986) was followed by widespread genetic engineering/modification of crop plants (Gasser and Fraley 1989), and more recently by genome editing technologies (Bortesi and Fischer 2015; Sander and Joung 2014). Molecular breeding has transformed agricultural practices worldwide although it often faces strong public and political opposition.

Despite the importance of agriculture to our species’ history and recent rapid advances in molecular breeding technologies, there remain disagreements over which theoretical framework offers the best understanding of the origin, spread, and ongoing transformation of agriculture. Several recent debates and exchanges have revealed a tension between, on the one hand, interpretive, humanities-oriented frameworks that focus on culture and agency on the part of agriculturalists and the socio-political contexts within which agriculture is practised, and, on the other hand, neo-Darwinian approaches that use tools such as optimal foraging theory derived from behavioural ecology to understand agricultural decisions, assuming that human decision-making has genetically evolved to maximise inclusive genetic fitness (e.g., Cochrane and Gardner 2011; Gremillion et al.2014). The former approaches are laudable in their attempt to situate agriculture within the rich socio-cultural contexts that they demand, yet often lack rigorous scientific methods and sometimes suffer from the general malaise within the humanities of being politically motivated, agenda-driven, and disconnected from the natural and behavioural sciences (Barkow 2005; D’Andrade 2000; Slingerland and Collard 2011). The latter approaches are often limited in their theoretical assumptions, and, we would argue, do not fully incorporate the role of culture as more than a proximate mechanism (Laland et al.2011; Mesoudi et al.2013).

Here we follow others (O’Brien and Laland 2012; Rowley-Conwy and Layton 2011; Zeder 2015) in arguing that the study of agriculture can benefit from being situated within a set of new evolutionary approaches to human behaviour – cultural evolution, gene-culture co-evolution, and cultural niche construction – that attempt to incorporate cultural change and individual agency within a rigorous scientific and multidisciplinary evolutionary framework. We highlight several ways in which the study of agriculture can benefit from these frameworks. We also highlight ways in which a consideration of agriculture yields new insights into cultural evolution, gene-culture co-evolution and niche construction. Specifically, we argue that (see also Fig. 1):

  • Changes in agricultural knowledge and practices are a prime example of cumulative cultural evolution (CCE), where beneficial ideas and inventions are selectively preserved and accumulate in number and effectiveness over successive generations. We apply the large body of modelling and experimental insights already obtained for CCE generally to agriculture. This illuminates the recent rapid advance in agricultural knowledge in the last two centuries, and also highlights the role of intentional versus non-intentional modification.

  • Agriculture is a prime example of gene-culture co-evolution (GCC), where culturally transmitted practices affect a species’ genetic evolution, and vice versa. However, this is not just (as frequently argued previously) the case of culturally transmitted agricultural practices changing human genes, but also changing non-human genes contained within domesticated and genetically modified organisms.

  • Agriculture is associated with extensive cultural niche construction (CNC), where agricultural practices transform the environment and those environmental changes alter the selection pressures on agricultural CCE. We argue that agriculture can modify (i) the abiotic environment (e.g., water, salinity, soil composition), (ii) the biotic environment (e.g., domesticated species, pests including insects, fungi, and weeds), and (iii) the social environment (e.g., social norms, regulation, markets), and focus in particular on the latter.

Fig. 1
figure 1

A schematic illustration of three approaches for understanding agriculture and plant breeding. (a) Cumulative cultural evolution (CCE) occurs as beneficial modifications are accumulated over time via repeated innovation and social learning, with an increase in some measure of improvement (e.g. crop yield and quality). (b) Gene-culture coevolution (GCC) typically describes the interaction between human genes and agricultural practices (an example of CCE), to which we add the additional interaction with non-human genes of domesticated animals and plants. (c) Cultural niche construction (NC) describes how agricultural practices may shape the abiotic, biotic and social environment, with those changes feeding back to shape agricultural practices

Full size image

The following sections address each of these points in the context of selected examples of plant breeding via new molecular tools. We apply these insights to two case studies: golden rice and stress tolerance. We conclude by highlighting outstanding questions that arise from our attempt to place agriculture within these frameworks.

Agriculture as Cumulative Cultural Evolution

For most of the twentieth century, the study of cultural change remained largely separate from the biological sciences. From the 1970s, scholars began developing a formal theory of cultural evolution, in which cultural change is viewed as an evolutionary process that shares key characteristics with, but differs in important ways from, genetic evolution (Boyd and Richerson 1985; Cavalli-Sforza and Feldman 1973, 1981; see Mesoudi 2011a, 2017 for reviews). This approach incorporates cultural change and variation into a theoretical framework that is consistent with the evolutionary sciences. Central to this approach is the idea that cultural change constitutes an evolutionary process in its own right: it is a system of inherited variation that changes over time, as Darwin defined evolution in The Origin of Species (Darwin 1859). ‘Culture’ is defined here as learned information that passes from individual to individual via social learning processes such as imitation, teaching, and spoken or written language. Social learning therefore provides the inheritance system in cultural evolution, paralleling genetic inheritance in genetic evolution.

Recognising this parallel, we can borrow and adapt tools, concepts, and methods from the biological sciences to study cultural change (Mesoudi et al.2006). These include mathematical models (Boyd and Richerson 1985; Cavalli-Sforza and Feldman 1981), phylogenetic analyses (Gray and Watts 2017), lab experiments, archaeological data and field research (Mesoudi 2011a). Importantly, this research does not unthinkingly import genetic models of change and apply them to cultural change without considering the unique aspects of the latter. For example, we can incorporate multiple pathways of inheritance: not just from parents to offspring like genetic evolution, but also transmission from non-parents and between peers (Cavalli-Sforza and Feldman 1981). Psychological processes such as conformity work to favour common behaviours, while prestige bias spreads behaviours associated with high status individuals (Boyd and Richerson 1985). There may be Lamarckian-like transformation such that novel cultural variants are not blind with respect to function (Boyd and Richerson 1985) but may be intentionally created by individuals to solve specific problems. This allows agent-based decision-making forces to be incorporated into an evolutionary framework (Mesoudi 2008).

One interesting property of human cultural evolution is that it can be cumulative (Tennie et al.2009). Other species exhibit social learning, and this is sometimes powerful enough to generate between-group behavioural traditions. For example, chimpanzee communities across Africa exhibit group-specific tool use profiles (Whiten 2017). Yet only humans appear able to accumulate and recombine behavioural modifications over time via social learning, generating complex cultural traits that could not have been invented by a single individual alone (Dean et al.2014; Tennie et al.2009).

Agriculture is a prime example of cumulative cultural evolution (Fig. 1). Other species practice agriculture in a sense, most famously leaf-cutter ants of the genera Acromyrmex and Atta that cultivate a type of fungus (Schultz and Brady 2008). However, the adaptations responsible for this are genetic, not cultural. Human agriculture is the result of repeated behavioural innovations that spread, accumulate, and recombine via social learning through and beyond communities. This allows for great flexibility, often involving the simultaneous use of multiple domesticated species, and more rapid change over time, on the order of thousands, hundreds, or tens of years rather than millions as in the case of ant-fungus genetic evolution (Schultz and Brady 2008). In humans, agricultural knowledge, practices, and technologies are culturally evolving traits that often show a cumulative increase in scope and complexity over time (Fig. 2). Typically, these traits are sequentially linked, with prior inventions necessary for the emergence of subsequent ones. Key innovations include irrigation by controlling water flow via canals and other waterways, the invention of different types of plough, the conversion of gaseous nitrogen to inorganic nitrogen fertilizers to enhance crop yields, the industrial mechanization of a variety of agricultural processes, and the discovery of the principles of genetics that allowed classical plant breeding. Recent CCE has resulted in new agricultural and computerized technologies, e.g., drip irrigation (Camp 1998) and precision agriculture (Mulla 2013), and the application of novel molecular tools for breeding of crops and farm animals, such as the use of in vitro procedures for plant propagation (Loberant and Altman 2010), fertility control, and genetic modifications in farm animals (Hasler 2003; Xu et al.2006) and molecular markers for selection (Smith and Simpson 1986, Ben-Ari and Lavi 2012), genetically-modified (GM) plants (Gasser and Fraley 1989; Harfouche et al. 2019) and genome editing of crops (Bortesi and Fischer 2015; Sander and Joung 2014). As expected for a historically contingent, culturally evolving process these various innovations occurred in stops and starts, showed different trajectories in different societies, and were sometimes lost, reintroduced, or recombined (Fuller et al.2014). Agriculture therefore fits several ‘extended criteria’ of CCE specified by Mesoudi and Thornton (2018): not just repeated improvement as a result of individual and social learning, but also sequential dependence of innovations, branching lineages, and recombination across lineages.

Fig. 2
figure 2

Key evolutionary events in agriculture and general biotechnology. Schematic illustration of cultural evolution of the major agricultural niches and sub-niches and accompanying technological and biotechnological innovations are depicted (bold line) as a relative skills and research index vs. the timeline (from the accepted start of agriculture, domestication, to present). The parallel evolution of general key biotechnological events is also depicted (standard line). Several major events in agricultural evolution are indicated with arrows pointing to the approximate time. The resulting major agricultural sub-niches are indicated in a series of encircled bold numbers above the timeline (1 to 7), at the approximately corresponding period: initial plant domestication resulted in small-scale horticultural food production (1); With further domestication, large-scale agricultural food production took place as a result of trial-and-error plant trait selection and agronomic improvements (2); As excess quantities of food became more available, people started to extend the shelf life of fresh food by preservation via drying, salting, smoking and other technologies, some of which were practised already by hunters-gatherers (3) and by fermentation (Nummer 2002) (4); Three key events further enhanced food quantity and quality from the thirteenth century (5): (a) long distance travelling and discoveries of new countries resulted in imports and exports of new plants between countries, which allowed for new gene combinations, global gene exchange and domestication of new species, (b) introduction of agricultural machinery during the industrial revolution, foremost the steel plough, cotton gin, seed drills, and later tractors as well as (c) chemical synthesis of ammonia that resulted in massive use of nitrogen fertilizers and large increase in crop production (Erisman et al.2008). Discovery of Mendel’s laws of genetics and its rediscovery later, allowing revolutionary intentional science-based traditional breeding (Hallauer 2011) (6). This was followed by molecular breeding using genetic engineering, and more recently by genome editing (7)

Full size image

Viewing agriculture as CCE allows us to draw on the large body of formal models and experiments that have explored the factors that allow, facilitate, and constrain CCE and apply these insights to agriculture. CCE is thought to depend on high fidelity social learning, which is required to faithfully preserve beneficial innovations across generations and over time (Lewis and Laland 2012). This social learning also needs to be selective, either selectively preserving successful practices, or selectively learning from successful individuals (Laland 2004; Mesoudi 2011b). In the context of small-scale agriculture, this may involve the observation of, or teaching by, expert plant and animal breeders. Since the emergence of formal systems of science, one-to-one transmission has been replaced by the transmission of knowledge in publications such as journals, books, and patents, which greatly increase the fidelity of social learning.

Equally important to mechanisms of social learning are aspects of demography. In order to support continued CCE, populations must be large enough to sustain the repeated transmission of knowledge (Henrich 2004; Powell et al.2009), and they should also ideally be partially connected, e.g., via migration, such that different innovations can emerge in different groups and then become recombined, rather than the entire population fixating too soon on a single suboptimal solution (Derex and Boyd 2016). The recombination of beneficial traits can generate exponential increases in knowledge, as seen in the patent record (Youn et al.2015) and in the need for aplying macine learning (Harfouche et al. 2019).

Finally, the type of innovation can affect the dynamics of CCE. Miu et al. (2018) found, in a computer programming tournament, two classes of innovations: small, incremental ‘tweaks’ that were common but unlikely to lead to major increases in performance, and rarer ‘leaps’ that made bigger changes to existing knowledge, were more likely to fail, but had a small chance of a major improvement. These rare innovative leaps may play a disproportionate role in CCE (Kolodny et al.2015) (see Fig. 2).

An interesting question is whether innovation is intentional or not. In genetic evolution, there is no foresight. Genetic mutations arise randomly with respect to their adaptive effects; beneficial mutations are no more likely to arise when they are needed than when they are not. In cultural evolution, however, innovation may be intentionally directed in ways that make adaptive variants more likely to occur. Clearly, people are not omniscient (Mesoudi 2008), but this intentionality may speed up CCE compared to random modifications, as suggested by models of ‘guided variation’ (Boyd and Richerson 1985) and ‘iterated learning’ (Griffiths et al.2008). On the other hand, major innovative leaps in CCE often arise by accident, suggesting that randomness can be useful; classic cases include the discovery of penicillin and x-rays (Simonton 1995). Of course, real cases of innovation may involve both chance and intention. The issue of intentionality in the emergence of agriculture has been debated extensively (Abbo et al.2014; Fuller et al.2012; Kluyver et al.2017), often in oppositional terms. Cultural evolution models, such as those of guided variation, permit the inclusion of both intentional and non-intentional factors, to compare their combined effects on the speed and form of agricultural CCE. Recent GM technology represents, however, the ultimate in intentional modification, with agricultural CCE no longer dependent on random genetic mutation and recombination to create superior breeds.

Agriculture as a Driver of Gene-Culture Co-Evolution

Gene-culture co-evolution incorporates CCE, but focuses on those cases where cultural inheritance causes changes in gene frequencies, which feeds back on cultural evolution, forming a co-evolutionary dynamic (Feldman and Laland 1996; Laland et al.2010). Several classic cases of human gene-culture co-evolution involve agriculture, given the growing evidence that agricultural practices have left indelible signatures on the human genome over the last 12,000 years (Laland et al.2010; Richerson et al.2010). O’Brien and Laland (2012) discuss two classic cases: first, the spread of lactose tolerance alleles from around 7500 years ago in central European populations as a consequence of the cultural practice of dairy farming (Gerbault et al.2011; Itan et al.2009); and second, the spread of sickle-cell alleles in West African populations that confer resistance against malaria, which increased in prevalence following the clearing of forests for yam cultivation, creating pools of standing water where mosquitoes breed (Wiesenfeld 1967). In both cases, there is clear archaeological, anthropological, and genetic evidence that cultural practices came first, followed by genetic responses that continue to affect behavioural variation across contemporary human populations.

What is less often recognised in discussions of gene-culture co-evolution is that agriculture also causes genetic change in non-human species. Many definitions of agriculture require there to be human-induced genetic changes in the domesticated plant or animal (Rowley-Conwy and Layton 2011). This non-human genetic changes may be the result of intentional or unintentional artificial selection for traits that increase yields, or the side effects of such selection. The entire package of genetic changes in a domesticated species is sometimes called the “domestication syndrome” (Larson et al.2014). There is extensive evidence, particularly since the advent of gene sequencing, for sustained genetic changes in domesticated species of both plants and animals (Zeder 2015). In plants the domestication syndrome may include larger seeds, synchronous germination, or fruit ripening that makes sowing or harvesting easier, and reduction in chemical defences (Fuller et al.2014). In animals, the syndrome includes increased docility, changes in body shape and size, and altered reproduction patterns (Larson and Fuller 2014). In some cases, non-human genetic change coincides with human genetic change, as in the case of lactose tolerance genes in humans and corresponding changes in cattle genes (Beja-Pereira et al.2003). Genetic modification by conventional and molecular intentional breeding represents further genetic change as a result of agricultural practices, and is covered in more detail below.

Agriculture as Niche Construction

As O’Brien and Laland (2012) have argued, agriculture is also a prime example of cultural niche construction. Niche construction is the general biological principle that organisms do not just passively adapt to their environments. Often they actively construct their environments, with those modifications in turn affecting their own and other species’ evolution (Odling Smee et al.2003). These modified environments may be inherited via what is termed ecological inheritance. Cases of non-cultural niche construction occur in numerous species; examples include earthworms’ burrowing and mixing activities, which alter soil nutrient content, and beaver dam building, which creates standing water. These activities have evolutionary consequences: for example, earthworms have retained their freshwater kidneys rather than adapt to the terrestrial environment because the mixed soil they create allows easier absorption of water.

Cultural niche construction occurs when the behaviours that modify environments are at least partly socially learned, and the consequences potentially affect subsequent cultural evolutionary dynamics (as well as, potentially, genetic evolutionary dynamics; this would be a case of GCC) (Kendal et al.2011; Laland et al.2000). The ‘environment’ here can be physical or abiotic (e.g., soil composition or climatic conditions, both of which strongly affect plant development), biotic (composed of other species; in the case of domesticated plants this would include phytopathogenic fungi, bacteria, insects and viruses) and social (composed of other individuals of the same species, e.g., competition between neighbouring plants at the root level).

Despite romantic notions of the “noble savage” living passively in an unaltered environment, hunter gatherers frequently engage in cultural niche construction by modifying their environments through cultural practices such as controlled burning of vegetation (Boivin et al.2016; Rowley-Conwy and Layton 2011; Smith 2011). Large-scale agriculture brought about cultural niche construction orders of magnitude more extensive (O’Brien and Laland 2012; Rowley-Conwy and Layton 2011). Agriculture caused huge changes to physical environments, including the clearing of forests, the irrigation of previously arid environments, the dispersal of domesticated plants and animals, and the introduction of new parasites and pests. Agriculture also brought about huge changes to human social environments, including increased population density and new forms of social organisation (e.g., new forms of hierarchies). Finally, the accumulation of agricultural knowledge and practices shaped environments in which further accumulation of agricultural practices was made more likely (CCE). In fact, large-scale agriculture, which produces the majority of the food consumed worldwide (e.g., rice, corn, wheat, canola, soybean) is generally a monoculture (i.e., a single type of plant species that is cultivated in large land areas as a crop for human consumption), unlike home gardens, natural savannahs, pastures and forests, which contain many species. Agriculture therefore results in modified niches compared with the natural vegetation, with clear effects on ecosystems (Matson et al. 1997).

A consideration of how agricultural practices shape, and are shaped by, social environments allows us to consider the mutual dynamics among agriculture and the social norms, regulations, and markets that often determine whether a particular technology or practice spreads or not. A good example of this is the acceptance and rejection of GM foods (see below).

Case Studies: Biotechnology

Most previous discussion of GCC in the context of agriculture concerns deep human history and prehistory (e.g., lactose tolerance and dairy farming; O’Brien and Laland 2012). In our case studies we instead focus on recent biotechnology and molecular breeding to illustrate the points raised above and demonstrate the relevance of these theoretical frameworks to contemporary issues. Moreover, studying recent scientific discoveries and technologies offers richer data for testing theories of cultural change compared to the ancient events of early domestication, which can only be studied indirectly via historical or archaeological methods.

Following the Neolithic agricultural revolution and initial crop domestication, and all subsequent agricultural improvements including traditional breeding methods based on Mendelian genetics, a new agricultural phase started in the middle of the twentieth century: the era of molecular breeding, genetic engineering, and in vitro biology (Fig. 2). While some scholars refer to these as ‘revolutions’ (or at the extreme, a single ‘agricultural revolution’), they are clearly all a process of CCE, with each major advance dependent on previous advances. Molecular breeding and genetic engineering could not have been invented without existing knowledge of Mendelian genetics. Yet, there are differences. The Neolithic agricultural period, i.e., plant and animal domestication, as well as other technological improvements in agriculture and biology (e.g., the use of irrigation and fertilizers) are more protracted and evolved sequentially over a period of hundreds or thousands of years (Fig. 2). In contrast, the time span of adopting and applying molecular plant breeding technologies and in vitro biology has been much shorter. Such technologies emerged far more rapidly, and became a working reality only within the last few decades. The molecular structure of DNA was first published in 1953 (Watson and Crick 1953), and the first genetically modified (GM) or transgenic plant (i.e., produced via incorporation of recombinant DNA), tobacco, was first created in the laboratory in 1982 (De Framond et al.1983; Gasser and Fraley 1989; Tepfer 1984; Zambryski et al.1983). Farmers began to plant GM crops in 1996, and in 2017, the 21st year of commercialization of biotech crops, 189.8 million hectares (a ~112-fold increase) of biotech crops were planted by up to 17 million farmers in 24 countries, which makes GM crops the fastest adopted crop technology in recent times (Altman and Hasegawa 2012a, b; Farre et al.2010; ISAAA 2017; Harfouche et al. 2019; Moshelion and Altman 2015). Molecular genetics, including genetic engineering of crops and the use of molecular marker-assisted selection, as well as novel gene editing technologies like the CRISPR/Cas9 system and synthetic biology (Baltes and Voytas 2015; Bortesi and Fischer 2015; Zong et al.2017) and other in vitro procedures, such as in vitro propagation (micropropagation) (Khayat 2012; Loberant and Altman 2010), are currently modifying the breeding opportunities of domesticated and cultivated plants globally (Altman and Hasegawa 2012a, b; Farre et al.2010; Moshelion and Altman 2015; Harfouche at al. 2019; Potrykus and Ammann 2010). This is also true for in vitro and molecular genetic procedures in farm animals and humans, e.g., in vitro fertilization (Bavister 2002).

The molecular breeding technology described above is clearly a case of CCE, building on what went before (e.g., Mendelian genetics) and far exceeding what any single individual could achieve alone. The exponential accumulation of knowledge is a well-recognized characteristic of CCE (Enquist et al.2008). There are many potential explanations for this, including the recombination of an increasing number of traits (Enquist et al.2011; Youn et al.2015) or the enhancement of innovation and discovery as a result of CCE products such as scientific instruments (Enquist et al.2008; Mesoudi 2011b). Molecular breeding is also a case of GCC, where the genes of other species are directly and intentionally modified using culturally evolving scientific techniques. These genetic modifications in turn demand new and more powerful scientific techniques and knowledge. Finally, molecular breeding involves extensive CNC, in terms of major changes to the abiotic, biotic, and social environments.

Case Study 1: Golden Rice

Rice, originally domesticated in East Asia around 8-9kya, is a major staple food for billions of people worldwide, supplying the majority of energy and carbohydrate requirements in addition to other nutritional factors (Wing et al.2018). Historically, rice is thought to have played a role in human GCC by driving the selection of alcohol dehydrogenase alleles in rice-farming populations in which rice was used in fermentation of food and beverages (Peng et al.2010). In addition to this long history of traditional breeding, rice has more recently been subject to some of the first molecular breeding.

Rice is generally consumed in its “polished” refined form by removing its outer layers. As a result, the edible part of rice grains consists of the endosperm that contains starch granules and protein bodies. However, this part lacks several essential nutrients that are more abundant in the outer layers of the grain, such as the carotenoid pro-vitamin A (β-carotene), which is converted in the body to vitamin A. Thus, reliance on polished rice as a primary staple food, which is an example of culturally evolving culinary traditions, results in vitamin A deficiency, a serious public health problem that is the primary cause of blindness and other diseases in new-borns in many developing countries (Srikantia 1975).

Conventional breeding of rice to increase vitamin A content is impractical due to the lack of appropriate rice cultivars that produce pro-vitamin A in the grain. Research into the β-carotene biosynthetic pathway resulted in the ability to defeat vitamin A deficiency by genetically transforming commercial rice varieties using two daffodil genes and one bacterial gene, resulting in vitamin A-rich rice (Burkhardt et al.1997). This genetically engineered, polished, fortified “golden rice” can supply sufficient pro-vitamin A for the body to convert into vitamin A (Potrykus 2001). Subsequent molecular breeding is leading to “green super rice,” that has a lower ecological footprint (Wing et al.2018).

The continual modification and accumulation of GM rice breeds, from traditional rice to golden rice to green super rice, represents a case of CCE where we see continual improvement in multiple criteria of yield, nutritional quality, fit to local agricultural practices, and ecological sustainability. The genetic changes in rice brought about with domestication and selection have been succeeded by traditional breeding and recently by direct, intentional genetic modification, representing a case of GCC between human agricultural scientific practices and rice genomes (as well as human genes, in the case of alcohol dehydrogenase).

Rice has also been responsible for extensive CNC. This involves not only the modification of abiotic and biotic environments, but also social environments. One key feedback between agricultural practices and social environments has been oppositional. Like many other GM crops, the adoption of golden rice, despite its health benefits, has been delayed considerably due to legislation, socioeconomic issues, and public concerns. Compared to non-GM rice varieties, the adoption and deployment of golden rice was delayed for more than 14 years by the demanding GM-regulation process. The first scientific procedure was published in 1997. Under regular processes golden rice could have reached farmers’ fields in Asia by 2002, but in fact was not officially approved for human consumption, except for planting by selected farmers, until 2013–2014 (Potrykus 2010). While regulation is needed to establish public safety, many hurdles existed not because of scientific problems or safety regulation, but rather due to the negative political climate surrounding GM-technology and the activities of anti-GM activists, the lengthy Intellectual Property (IP) rights approval, the lack of financial support from the public domain, and GM-regulation procedures that required several technological solutions (Potrykus 2010). These delays created a situation where no public institution could deliver GM products because of the high expenses of large-scale production, which resulted in the de facto monopoly of a few potent commercial industries that supplied high-priced seeds to farmers. Since then, GR2E Golden Rice, a provitamin-A biofortified rice variety, received its third positive food safety evaluation by the United States Food and Drug Administration (US FDA) in May 2018, following earlier approvals by Food Standards Australia New Zealand (FSANZ) and Health Canada, all based on the principles of the World Health Organization (WHO), the Food and Agriculture Organization (FAO) of the United Nations, and other international agencies (IRRA 2018).

This negative feedback in the form of oppositional social norms and increased regulation has prevented the timely adoption of an available solution to vitamin A deficiency, and similar situations exist for other GM crops. Together with other technologies, GM crops have the potential to help ameliorate many of the world’s most challenging problems, including hunger, malnutrition, disease, and poverty. However, this potential cannot be realized if the major barriers to adoption - which are largely socio-cultural rather than technical - are not overcome (Altman and Hasegawa 2012a, b; Farre et al.2010).

Social norms, culinary preferences, and legal regulations are themselves culturally evolving systems that co-evolve with scientific knowledge and technological practices. Consequently, the acceptance and spread of agricultural practices and products may vary cross-culturally. For example, while large global commercial companies tend to invest mainly in major world staple crops (e.g., soybean, corn, canola, wheat, and rice), many other local plants remain “orphan crops.” This is why the government of India, where eggplants are an important part of the diet, embarked on a mission to produce GM insect-tolerant Bt brinjal (eggplants), which were adopted rapidly and commercialized despite some legislative problems and concerns that were later raised (Kolady and Lesser 2012; Medakker and Vijayaraghavan 2007).

An appreciation of the social environment within which agricultural practices are situated, as follows from a CNC approach, has much in common with social science approaches that stress the embeddedness of new plant crops within socio-political contexts, not just performative qualities such as potential yield (Stone and Glover 2017). Indeed, recently demand has been growing for heirloom rice, traditional rice breeds that have lower yield than Green Revolution rice, but which are marketed as socially and environmentally responsible products embedded in local cultural traditions (Stone and Glover 2017).

Case Study 2: Plant Stress Tolerance/Resistance

Major advances in molecular breeding have resulted in the genetic modification of crops to improve biotic stress resistance, including resistance to pests like insects, phytopathogenic fungi, viruses, nematodes, weeds, and others (Ceasar and Ignacimuthu 2012; Gurr and Rushton 2005; Scholthof et al. 2011; Suzuki et al.2014; Vidavsky and Czosnek 1998), and abiotic stress tolerance, including tolerance to drought, salinity, extreme temperatures, heavy metal toxicity, and others (Hirayama and Shinozaki 2010; Vinocur and Altman 2005; Zhu 2016). The two specific examples discussed here are herbicide and insect resistance.

Herbicide resistance was developed to combat weeds. With the intensification of agriculture, weeds became a serious economic threat to farming, resulting in increased agricultural production costs and yield loss of cultivated crops. This is especially the case with intensively grown and irrigated plants that enhance weed growth in addition to the desired crop. This problem has been dealt with traditionally either by labour-intensive manual weeding, which is usually performed in less developed countries by women, by tillage, or by heavily spraying fields with large amounts of toxic herbicide chemicals that pollute the environment (Christensen et al. 2009; Griepentrog and Dedousis 2009; Melander et al.2005). To avoid these costly solutions, weed management was simplified and manual work was reduced by genetically modifying crops to be herbicide resistant. This allows the use of considerably smaller amounts of broad-spectrum herbicides since they kill only the weeds and not the crop (Bonny 2016; Gressel 2009a, b). For example, herbicide-tolerant GM crops were created that express a soil bacterium gene that produces a glyphosate-tolerant or glyphosate-degrading form of an enzyme, resulting in glyphosate-tolerance (Castle et al.2004) and resistance to commonly used glyphosate herbicides. This cannot be achieved by traditional breeding. Currently, herbicide-resistance is the dominant trait deployed globally in soybean, maize, canola, cotton, sugar beet, alfalfa, and other crops, and is being adopted increasingly rapidly by farmers, comprising about 53% of the 180 million hectares of all GM crops in 2015/16 (ISAAA 2017).

Insect resistance provides crops with defences against herbivorous insects. Over the centuries farmers have selected plant varieties that are more resistant to insect pests. As for herbicide resistance, traditional breeding for insect resistance was not very successful, and was followed from the 1940s by widespread spraying of fields with chemical insecticides. This had several drawbacks, including environmental pollution and damage to other non-pest organisms (Newton 1988; Weston et al.2011). The biotechnological solution involved genetic modification of cultivated crops resulting in insect resistant plants that kill specific pests when digested. Insect tolerant GM cotton, potato, canola, corn, and other crops were developed through the introduction and expression of the soil bacterium Bacillus thuringiensis (Bt) cry genes, resulting in production of the endotoxin cry protein crystals that selectively kill target insect larvae eating the leaves (de Maagd et al.1999). This technology has several limitations, and improved methods have been developed recently, including genome-editing technology and “gene stacking,” i.e., the introduction and expression of multiple genes that create several toxic proteins (e.g., Gatehouse 2008; Lombardo et al.2016).

The successive inventions and discoveries that led from traditional breeding and use of chemical pesticides to genetically modified herbicide and insect tolerant plants constitutes another case of CCE. Each step is dependent on earlier innovations, and measures of improvement have increased, from crop yield and quality to reduced environmental harm. With our expanded definition of GCC to include non-human genes, the genetic modification of crops to incorporate bacterial genes to improve tolerance are also cases of GCC, given the culturally-driven changes in non-human genes.

Finally, traditional and molecular selection for stress tolerance constitutes an extensive example of CNC. Human efforts to genetically modify plants to improve their tolerance to biotic and abiotic stress has allowed the spread of cultivated plants into land and regions where they could not have survived before. This involved the spread of organisms and their genes, either by straightforward domestication of new plant genes (e.g., the potato from Peru-Bolivia, and tomato from Chile to Europe (Diamond 1997). See also Fig. 2 on gene transfer that accompanied European expansion to the New World), by traditional breeding, or by gene transfer from any organism to the GM plants as described above. All of these activities create new agricultural niches that feed back to the agricultural process. The spread of agriculture is also associated with the spread of pests. The use of both herbicide and insect tolerant crops reduces the amount of sprayed chemicals and thus can positively impact the environment, countering some of the negative consequences of the agriculturally constructed niches (Pimentel 1995). It may also reduce the toxic effects of insecticides and other pesticides on human health (Levine and Doull 1992; Nicolopoulou-Stamati et al.2016), including Parkinson’s disease (van der Mark et al.2012).

As in the case of golden rice, the impact on and feedback from the social environment is of great interest and importance. As noted, women are the main work force in planting, weeding, and harvesting agricultural plots in many developing countries (Gressel 2009a, b; Subramanian et al.2010). In reducing the need for time-consuming manual labour, GM herbicide tolerant crops can potentially improve their socioeconomic status, can save many women from long working hours in the field and improve their economic situation and quality of life, as indicated in several cases (Carpenter 2013). Other studies show that biotechnology and the adoption of insect-resistant cotton in India generated more productive employment and greater earning power for women, with a consequent improvement in quality of life (Agarwal 1984; Subramanian et al.2010; Subramanian and Qaim 2010). Similarly, a study in South Africa found that planting of Bt cotton benefitted women in the household (Bennett et al.2003). In Burkina Faso, fewer insecticide applications needed for Bt cotton meant women spent less time fetching water (Zambrano et al.2013), although cultivation of herbicide-tolerant cotton in Colombia resulted in the hiring of fewer women for weeding, traditionally a female task, with potentially negative economic consequences (Zambrano et al.2013). However, there are some indications that, unlike with traditional crops, women in Colombia and the Philippines appear to participate equally with men in the decision-making and supervision of insect tolerant (Bt) cotton cultivation (Yorobe and Smale Yorobe and Smale 2012).

Interestingly, these recent developments relating to gender roles may be reversing the historical effects of culturally evolving agricultural practices on gender-biased division-of-labor. Alesina et al. (2013) provide evidence that the introduction of the plough several centuries ago allowed men to monopolise food production, resulting in the loss of socioeconomic power for women, who had previously participated in food production.

Discussion

In summary, we have argued that new and complementary approaches within the evolutionary human sciences – cumulative cultural evolution (CCE), gene-culture co-evolution (GCC), and cultural niche construction (CNC) (see Fig. 1) – can provide theoretical frameworks for understanding the many impacts that agriculture has had on human societies and on the planet. Unlike prior papers that argue similarly (Heslop-Harrison and Scwarzacher 2012; O’Brien and Laland 2012), we have focused on recent biotechnology rather than the distant past, both to demonstrate that these frameworks are relevant for contemporary issues and events, and to make some novel points not apparent when focusing only on the past.

First, we argue that agriculture is an excellent case of CCE. It involves the sequential improvement over time of agricultural knowledge (both scientific and non-formal knowledge systems) and practices (from small-scale habits and routines to large-scale technology) via the repeated cycle of innovation and cultural transmission. Viewing changes in agricultural practices as an evolutionary process and recognizing the resultant co-evolutionary dynamics and feedbacks facilitates connecting this cultural process with the biological/evolutionary/natural sciences, preventing a false and unproductive nature-culture dichotomy. Agriculture informally exhibits the classic exponential increase in knowledge and practices that is typical of CCE, with recent change seemingly orders of magnitude faster than past rates of change, allowing the large body of work exploring the drivers and inhibitors of CCE to further contribute to agricultural research.

Second, we argue that the standard notion of GCC, where human cultural practices shape human genes and vice versa should be expanded to include culturally driven changes in non-human genes. This includes, by definition, domestication, which entails the traditional breeding of domesticated species. More recently this has involved direct genetic modification with the introduction of GM crops.

Third, agriculture is a prime example of CNC, involving extensive modification of abiotic, biotic, and social environments, and feedback from these environments to agricultural knowledge and practices. Most interesting from our perspective are feedbacks with the social environment. Adoption of golden rice and other GM crops has generated resistance from activist groups, political parties, and regulators due to fears over food safety, genetic contamination, and an aversion to ‘tampering with nature.’ These concerns provoke increased regulation and safety testing within the agriculture industry to ensure that GM products are as safe as possible. While adequate levels of health regulation are of course needed, overly stringent regulation can prevent potentially beneficial innovations from spreading. The ideal outcome would be increased population health and reduced environmental impact as a result of GM crops such as golden rice, green super rice, and herbicide/insect resistant plants, as well as drought and salinity tolerant crops, post-harvest loss of food, use of novel fertility control in farm animals and more. Another positive social feedback is the impact on gender roles, with herbicide tolerant GM crops releasing women from tedious manual labour (weeding) and improving educational and economic outcomes (Fig. 3).

Fig. 3
figure 3

Major agriculture and culture-associated niche construction and plant gene-culture coevolution. The different interacting components of cumulative cultural evolution (CCE), plant-specific gene-culture coevolution (GCC), and environmental/agriculture-associated cultural niche construction (CNC) are schematically represented. Two major components are implicated: the physical environment, i.e., geography, the terrain, climate, and more (Box 1), human cultural factors, including ingenuity, technology and scientific discoveries (Box 2). Both may modify, shape, interact and coevolve with specific genes of domesticated plants (and farm animals) (Box 3). Once a certain selected gene combination has been fixed in a domesticated plant (or a farm animal) it can be again modified by traditional breeding techniques or by employment of novel molecular tools (MAS, GM, Genome editing) to produce novel gene combinations affecting mainly genes associated with modified plant products and metabolites (Box 4) and genes for improving plant survival/ tolerance to environmental stresses (Box 5). The novel plant products or traits can in turn result in the creation of new environmental niches, affect the expression of human genes through consuming those products, resulting in ongoing coevolution of biomes (i.e., the entire complex body of living organisms including plants, animals, and microorganisms), CCE, GCC, and ENC (Box 6)

Full size image

To expand the utility of these theoretical frameworks we propose the following novel research questions:

How Does Agricultural CCE Operate?

As noted, theoretical models and experiments suggest several complementary mechanisms upon which CCE depends, including high-fidelity social learning, selectively biased social learning targeted towards successful traits or individuals, recombination of disparate solutions, innovation that includes large risky leaps, and large (or partially connected) populations. Which of these is responsible for agricultural CCE could be addressed via archaeological and historical records, e.g., by quantifying the frequency and impact of different innovations (cf. Miu et al.2018 for computer code) or the rate of recombination across different domains (cf. Youn et al.2015 for patents). We might expect these mechanisms to change over time, or vary cross-culturally (Mesoudi et al.2016). The cases of recent agricultural breeding technologies discussed here afford the opportunity to study the drivers of CCE in real time, with richer datasets than those available to archaeologists and historians.

One interesting distinction already studied in the CCE literature is between intentional change by individuals (often called ‘guided variation;’ Boyd and Richerson 1985) and unintentional change via the copying of successful traits or individuals (often called ‘direct’ or ‘indirect’ bias). This relates to debates in the archaeological literature over the extent to which domestication developments were intentional or unintentional (Abbo et al.2014; Kluyver et al.2017). Formal modelling of the kind used in the CCE literature may inform this debate, at the least highlighting how both processes can operate together, or vary in importance across different species, historical periods, and societies, and should not be viewed as mutually exclusive. Molecular breeding seems to be under more precise control than traditional breeding due to the fact that only specific genes are targeted rather than whole genomes of two traditionally bred species, but still risks unforeseen consequences especially in its social effects.

Finally, there are interesting questions regarding the ‘fitness’ criteria of agricultural CCE, i.e., the quantity that is being maximised (Mesoudi and Thornton 2018). Two obvious criteria are crop yield (productivity) and nutritional content, but we have raised several additional criteria that may trade-off with these. Golden rice, for example, maximises human health by reducing Vitamin A deficiency beyond simple calorific intake. Green super rice and herbicide tolerant GM crops minimise environmental degradation. Heirloom rice explicitly trades off yield and productivity with local cultural preferences (Stone and Glover 2017), albeit only in small-scale traditional farming communities. In this sense, the cultural fitness criteria that shape CCE are themselves evolving amongst farmers, scientists, and consumers.

CNC within Social Environments

We have argued that the most interesting niche construction dynamics involve feedback between agricultural practices and the social environment, e.g., social norms of consumers, regulatory bodies, and markets. Social norms also culturally evolve, partly according to the psychological biases of members of society that make some ideas or attitudes more likely to be recalled and transmitted than others, known as ‘content biases’ (Mesoudi 2011a). These may well affect moral norms concerning biotechnology (Mesoudi and Danielson 2008). For example, GM foods may violate psychological biases that provide us with ‘folk intuitions’ about the natural world (Atran 1998), including that species have discrete essences that are violated when genes are transferred across species. Similarly, people seem to have general psychological biases to attend to, recall, and transmit disgust-eliciting stimuli (Eriksson and Coultas 2014), and moreover disgust-related taboos are more likely to occur against meat than plant products (Fessler and Navarrete 2003). This fits with evidence that there is more opposition to GM animals than GM plants (Schuppli and Weary 2010). Nevertheless, consumption of GM plants is still debated in many countries, mainly on the basis of health hazard concerns (Altman and Hasegawa 2012a, b; Davison 2010; Echols 1998). Further experimental and observational work integrating the many psychological dimensions of norm transmission can be applied to norms surrounding biotechnology (Mesoudi and Danielson 2008).

There is evidence for cross-cultural differences in acceptance or rejection of GM foods. For example, consumers in the US seem much more accepting than EU consumers towards GM foods (Gaskell et al.1999). Such differences demand explanation in terms of divergent cultural histories. Intriguingly, there is some evidence that agriculture and societal organisation have been co-evolving for millennia. Talhelm et al. (2014) show that in China, historically rice-farming regions are more collectivistic than historically wheat-farming regions. They suggest that the intensive and demanding labour required by rice farming created closer social ties and social interdependence. For example, rice agriculture demands more water and thus greater coordination of irrigation across plots of land; when rice is grown on steep hill slopes, as it often is, the farmers must cooperate and coordinate to ensure adequate irrigation for all plots. Wheat farming, by contrast, requires less irrigation management and therefore less need to coordinate and cooperate across farms. In such cases, we see agriculture shaping social orientations, which may in turn shape the subsequent spread or acceptance of further agricultural practices.

Conclusion

Agriculture has transformed our species and our planet to such an extent that it is one of the primary reasons why some scholars advocate the renaming of the current epoch to the Anthropocene (Ellis 2015; Ellis et al.2018; Lewis and Maslin 2015). The rapid rates of socio-cultural and scientific-technological change over the last century have only increased this impact, sometimes positive and sometimes negative. Here we have attempted to integrate several recent scientific-technological changes in agricultural knowledge and practices with an understanding of agriculture’s impact on environments, including social environments, within novel theoretical frameworks of CCE, GCC and CNC.