Corpus-based analysis of genetically modified seed discourse

Data

A corpus was constructed to allow for comparative analysis of two sides of the debate surrounding GM seed (i.e. pro and anti). The corpus was constructed using text from web pages. Web pages were queried and identified manually using the Google search engine with search terms related to GM seed. Some search terms were generic so as to capture perspectives from both sides of the debate (i.e. ‘gm seed’, ‘gm seed AND seed saving’, ‘gm seed debate’). Other search terms were targeted toward a specific side of the debate (i.e. ‘gm seed resistance’, ‘gm seed opposition’, ‘gm seed advantages’, ‘gm seed benefits’). Finally, another generic search was included to identify cultural dimensions (‘gm seed AND culture’). The urls for each page were collected and the text from the associated webpages extracted with a custom script written in the Python language.

The raw text corpus was split into two subcorpora representing anti- and pro-GM seed perspectives. Pages were each qualitatively identified as representing either an anti- or pro- stance on GM seed. Pages that were ambiguous or were of a ‘pros and cons’ nature were omitted from further consideration, though it was noted that the vast majority were easily identifiable as pro or con. Genres include news articles, op eds, blogs, academic articles, reference material, interviews, transcripts, and reports. Another important genre was advocacy statements.

After noise removal, pre-processing and normalization using the Natural Language Toolkit (NLTK) package, the size of the anti-gmo subcorpus was about 135,000 words while the size of the pro-subcorpus was about 101,000 words.

Analytic procedure

Given the many dimensions of the topic, it is appropriate to approach GM seed discourse through multiple perspectives and viewpoints. This paper proposes multilevel discourse analysis that integrates frameworks including cultural discourse analysis (Carbaugh, 2007), sociocognitive approaches (van Dijk, 2014, 2015), as well as critical and ecolinguistic approaches (Stibbe, 2013). The multilevel framework integrated four levels of analysis: (1) ecological; (2) cultural; (3) socio-economic; and (4) cognitive.

Corpus methods were employed in a way that aimed to avoid confirmation bias. While the analysis used primarily qualitative methods, many findings are supported by quantitative evidence. Moreover, each step in the data analysis was applied consistently to both sub-corpora.

Ecological level

Although the human understanding of nature is refined through the scientific method, nature is also culturally framed by worldviews (Weltanschaung) (Dahl, 2016: 217–228). How people understand and communicate about gene editing technology will depend on worldviews with which they make sense of biological and ecological systems. These worldviews may be heavily influenced by natural science, but also by ethnographically variant factors such as approaches to complexity, unpredictability, and the unknown. Of interest in the ecological level of analysis is whether and how the different sub-corpora reflect different worldviews or ways of understanding nature.

Different understandings of nature were first approached by contrasting how specialized terminology is used in the subcorpora. The presence of scientific terminology was determined by querying the corpora with a dictionary of specialized terms. Glossaries of molecular biology terms (Lyons, 2017), herbicides (University of Kentucky, 2018), and pesticides (Wood, 2018) were used to determine frequencies in the subcorpora. Results indicated the pro-GM corpus has five times the frequency of agrochemical terminology as well as a greater variety of agrochemicals (37 vs 7 in the anti-corpus). These results are of interest insofar as they may reflect different communities of practice and perhaps even epistemological orientations toward the natural world. To further investigate this possibility, concordances of the words ecological and biological were used to provide insight into how living systems are referred to in context.

The anti-GM sub-corpus returned 107 lines containing ecological, compared to only 14 for the pro-sub-corpus. These numbers further point to epistemic differences. Whereas the pro-GM corpus is focused more at the molecular level (as indicated by the greater presence of agrochemical terms), the anti-GM corpus seems more concerned with the higher levels of the organization of life and the interrelations between organisms and their environment. One way to examine how the word ecological functions in the corpus is to look at the nouns it modifies. In the concordance lines, nouns modified by the adjective ecological include relationships, processes, systems, complexity, and cycles. In the context of worldviews, these concordances suggest natural phenomena are understood as holistic, dynamic interactions. By contrast, in the pro-GM concordances, ecological modifies farming, agriculture, component, and impacts. Concordances in the pro-corpus suggest more reductive and mechanistic (as opposed to holistic or organicist) approaches to nature. Figure 1 contains a random sampling of concordances for each subcorpus.

OPEN IN VIEWER

Figure 1.

Random sample of concordances of ‘ecological’.

Concordances of biological gave similar results. The anti-GM corpus contains a higher frequency of biological concordance lines (94 vs 10). Nearly half of all concordances in the anti-GM sub-corpus contain the collocate biological diversity. Other nouns modified include evolution, balance, processes, and hotspot. As with the ecological concordances, biological is used in the context of complex, holistic systems. Modified nouns in the pro-GM concordances include resource, solution, methods, screening, and controls. Again, the notion that life processes are manageable, reducible, and analyzable suggests a different epistemological orientation to nature.

Cultural level

Discussions of culture in relation to language are fraught with ambiguities and risks of essentialism (Sharifian, 2015). Although a detailed discussion of cultural essentialism is beyond the scope of this paper, the current study takes an ordinary language or meaning-as-use approach to the notion of culture, which asserts that cultural categories implied in everyday speech are as good an analytical starting point as any (Frayne, 2017). Based on this premise, to consider the cultural context in the present study, the word culture was examined in context within the sub-corpora so that concordances of the word culture could provide insight into this context as well as meanings of the term within the corpus. There were 215 mentions of culture found in the corpus. Nearly all (211) of these mentions were in the anti-GM sub-corpus.

Concordances indicate how culture stands in relation to other discourse themes and, ultimately, to the overall theme of seed. For example, the close relation between food and culture stands out in lines such as ‘. . .food as part of culture and identity’; ‘Culture without food is not culture’; and ‘. . .unique food systems and culture’. These associations indicate that anti-GM discourse views seed in relation to food production, consumption, and cultural identity. Generally, the concordance lines point to culture as something to be preserved. Collocated words like customs, traditional, ancient, and preserve are all indicative of the idea of culture as something of deep meaning and value. While most uses of ‘culture’ are positive or affirming, it is also used in a pejorative sense. For instance, phrases ‘consumer culture’, ‘capitalistic culture’, and ‘today’s culture’ indicate that modern culture is held in critical view.

Both keywords and concordances point to the interrelations between ecology and culture. The tradition of seed saving is thousands of years old and traces back to the origins of agriculture and even human society itself. In modern times, however, the use of commercially patented and owned seed often precludes or prohibits collecting, saving, planting, harvesting, and exchanging seed. Viewing seed discourse in ethnographic terms reveals rich symbolic associations between ecology and culture. In the anti-GM corpus, we see how natural and biological processes work as connotations for how people understand themselves as living beings that are parts of a holistic natural order. By contrast, much of the pro-GM discourse frames the topic in more literal or denotative terms. Based on this observation, it is conceivable that disagreements concerning GM seed stem from a failure to acknowledge cultural associations and connotations.

Concordances of culture suggested that indigenous identity is prominent in the corpus. To see how this identity is split between the sub-corpora, frequencies of indigenous were calculated for each. The anti-GM corpus has 784 unique mentions of the word indigenous while the pro-corpus has none. This result is further indicative of a vastly different cultural context between the anti and pro corpora.

The concordances give an idea of how culture, in a general sense, functions in the corpus. However, the lines say little about which cultures are represented. To gain a better idea of different cultures in the corpus, the distribution of geographic entities in the data was considered. Specifically, city and country names were extracted and analyzed as proxies for possible cultural variation. Despite the obvious limitations of reducing cultural variation to place names, there are several reasons for investigating geographic distribution in this way. Data might reflect how perspectives in the GM-seed debate diverge along national lines. Some argue that small-scale farmers in the Global South are the losers in the shift toward GM seed, while corporations and farmers in the Global North are those who benefit (Nu and Dev, 2009; Parfitt, 2013). Accordingly, one might expect the pro- and anti-GM sub-corpora to reflect more Global North and Global South sources, respectively.

The geotext Python package (Palenzuela, 2018) was used for geotagging and entity recognition of country and city names. Results indicated 38% of countries in the anti-GM corpus are from the Global South compared to 13% in the pro-GM corpus (Figure 2). Using concepts from information theory (Shannon and Weaver, 1949), it was also shown that the diversity and evenness of country frequencies is about twice as high in the anti-GM data. This suggests that the anti-corpus contains a greater variety of countries and, in turn, might reflect a greater diversity of cultural perspectives.

OPEN IN VIEWER

Figure 2.

Python output from geographic entity frequencies.

Entity extraction methods were also used to quantify the temporal horizons in the corpus. All years mentioned in the corpus were collected by querying by matching years in the text and manually verifying and cleaning results. While most dates fall in recent decades (since the 1990s), plotting suggests dates from the anti-GM corpus span a larger range and are more evenly dispersed (Figure 3). The entity extraction of place names and dates gives evidence to the claim that geographic and temporal contexts of the anti-GM corpus are more diverse and expansive.

OPEN IN VIEWER

Figure 3.

Comparison of years in the corpus.

Socio-economic level

Divisions in the GM-seed debate are deeply related to socio-economic imbalances and differing conceptions of economic growth. The 1950s and 60s saw the development of Green Revolution technologies (synthetic fertilizer and pesticides, irrigation) followed by genetically modified seed in the 1970s. These developments coincided with a neo-liberal policy framework in the 1970s and 80s, which (many argue) gave way to consolidation of agricultural supply chains (Kaditi, 2013). This included policies in the U.S., Canada, and other developed countries that moved seed biotechnology from the public sector to the private seed industry (Pray and Fuglie, 2015). Vertical integration in agricultural supply chains accompanied horizontal consolidation of intellectual property rights for seed biotechnology (Fulton and Giannakas, 2001). Rapid developments in genetic engineering and biotechnology took place in this economic context, leading to the situation today, where a handful of corporations control the majority of the world’s seed markets and patents (Howard, 2009).

The controversy over GM-seed has coincided with neoliberal economic reforms. Free Trade Agreements (FTAs) have often granted legal intellectual property rights to international seed companies (GRAIN, 2018). The prohibition on seed saving can apply not only to patented varieties, but to any seed varieties that have not been registered or pre-approved. For instance, the case could be made that the benefits of GM seed fall disproportionately to the largest customers (i.e. farmers with access to large acreages, machinery, credit, and subsidies) (Azadi et al., 2016). Also, seeking the largest return on investment, market-driven R&D will naturally emphasize major crop varieties and sales volume, rather than specialty crops and fresh produce that suit local circumstances and small-producers. Economic analysis might, therefore, suggest that divisions over GM seed are not necessarily a consequence of biotechnology itself, but the economic context in which it has developed. Consequently, those resisting the use of commercial seed varieties might be doing so as part of practical efforts to feed their families and earn a living. In other words, resistance may be based on lived realities faced by small-scale farmers in addition to cultural, ecological, or other factors.

Of interest in the present analysis is how the economic context is reflected in the corpus. The key terms analysis suggested that income is a much more prominent concept in the pro-GM data. Frequency data further support this view. The word income appears in the anti-GM corpus with a frequency of 0.22 per thousand words. In the pro- corpus the frequency is over eight times higher at 1.78. The collocate farm income is common with 174 mentions in the pro-GM corpus, compared to only 2 in the anti-corpus. This contrast is notable because the term farm income is used in national and international policy discourse. It is fair to say the term farm income is associated with larger-scale commercial farming. By contrast, in the anti-GM corpus, the term income is more likely to occur in the context of smaller-scale, household economics. Collocates community income, household income, or the possessive farmers’ income are all prevalent. Concordance lines refer to income in the context of local produce sales, and subsistence activities. Whereas the pro-GM corpus focuses on income gain and positive impacts, the anti-GM corpus refers more to losses, poverty, and scarcity (Figure 4).

OPEN IN VIEWER

Figure 4.

Sample of concordances of ‘income’.

The previous cultural-level analysis of country entities in the corpus suggests that a higher proportion of countries mentioned in the anti-corpus are from the Global South. This result points to possible economic differences between the two sub-corpora. To further explore this hypothesis, GDP data was used. All countries mentioned in the corpus (including city mentions) were ranked according to GDP (nominal) per capita. Among all countries, the average rank, average GDP, and percentages in the top and bottom quartiles (according to GDP), were determined. Results suggest that the anti-GM corpus does indeed reflect lower income countries, though the significance of this difference is debatable. The average country rank (by GDP) is 11% lower in the anti-corpus; the average GDP is 8% lower; the percentage of countries in the top quartile by GDP is 9% lower; and the percentage of countries in the bottom quartile is 19% higher than in the pro-GM corpus.

Cognitive level

The cognitive level aims to understand and compare the mental representations of language users in both sub-corpora. Elements of cognitive analysis of discourse, as outlined by van Dijk (2014, 2015), include topics, implications, presuppositions, local coherence, and lexical meanings/connotations. Some of these elements have been covered, albeit not explicitly, in the other levels. In addition to elaborating on topics and connotations, this section looks at implication and presupposition. Coherence is examined through the notions of interdiscursivity and conceptual blending.

Implication pertains to cognitive analysis insofar as it contains socially shared knowledge that is inferred from explicit semantic contents of discourse (van Dijk, 2000). Implicature (Grice, 1975) is inferential and context dependent, that is to say, it relies on knowledge domains or mental schema. Sperber and Wilson (1995) make the distinction between implicature and explicature, where (in explicature) assumptions are explicitly stated. It may be the case that implications are understood more or less universally. It is also possible that implications rely on knowledge that varies across different cultures, communities of practice, or other groups. In such cases, implicature can be a reflection of ‘cultural scripts’ (Wierzbicka, 1985, 2003).

To obtain a representative sample of implications in the corpus, uses of the word means were considered (i.e. A means B). Although this relationship between A and B may be explicit, the assumptions underlying this relationship may not be. There are 81 and 43 such uses in the anti- and pro-GM sub-corpora, respectively. The lines were manually and qualitatively analyzed and written in symbolic notation with a right arrow (→) used to signify implication between propositions (e.g. A → B → C). The numbered segments below contain selected examples of implications in the pro-GM sub-corpus, together with summarized implication relationships.

(1) Easier farming MEANS more food which, in turn, MEANS less expensive food. [Easier farming → more food → cheaper food]

(2) Decreased use of pesticides, MEANS less pesticide production demand and also less energy use on the farmers’ end, too.

[Decreased pesticide use → less energy use]

(3) Many plants are designed to use less pesticides and chemicals to grow, which MEANS less exposure to these potentially toxic substances for farmers and consumers.

[Less pesticides → less exposure to toxins]

(4) Many GMOs are tailored for specific environmental conditions, which MEANS saving water in drought-prone areas and less use of chemicals.

[GMO seed → saving water AND less chemical use]

(5) GM [foods] have improved flavor and texture, as well as delayed ripening. This MEANS produce will stay fresh for longer periods of time.

[GMO seed → delayed ripening → produce stays fresh longer]

The five examples above express relationships among different variables, such as food supply and food prices (1); pesticide use and energy use (2); pesticide use and toxins (3); genetic modification and water/chemical use (4); and genetic modification and preservation of freshness (5). Of interest, is the extent to which these relationships are explicit and quantifiable. Moreover, the logic of the relationships is more-or-less self-explanatory as, for instance, the relation between food supply and food prices. In fact, in each example, the reader could conceive of a mathematical function depicting the relationship in question. Moreover, this relationship is often two dimensional; that is, between two variables (e.g. seed type and preservation time). Overall, little additional context is necessary to explain the relationships in question. Now, consider some examples from the anti-GM corpus:

(6) farmers. . .quickly lose control over seed management, production and eventually their land. This MEANS they lose their food sovereignty. . ..

[GM seed → loss of control over seed management → loss of food sovereignty

(7) Monsanto (and other companies) own the rights to the modified DNA in their seeds. This MEANS farmers would have to buy seeds from them each year, and maybe more than once. [companies own seed rights → farmers have to buy seed annually]

(8) they will cause reduced genetic diversity of plants and animals in the environment. What this MEANS is that the DNA, which codes for proteins in an organism, will become more similar between individuals of a species.

[ reduced genetic diversity → loss of biodiversity]

(9) And if Paraguay is so dependent [on foreign companies] for such a basic thing as food. . .it MEANS that this is a subordinate country.

[food dependence → national subordination]

(10) the nature of the promoter MEANS that the inserted genes are liable to be unstable and move out again.

[promoter genes → instability in inserted genes → plant instability and variable gene functioning → unintended effects]

(11) just three companies sell more than half the seeds on the market. . .this MEANS that the biological diversity of crops is declining, making our food supply less likely to adapt well to climate change.

[three companies control seed market → declining biodiversity → less adaptive food supply]

In comparison with the pro-examples, the implications above are not quantifiable. In many, cases the propositions cannot be expressed as variables; rather, the relationships involve implicit contextual or subjective factors that often evade explicit representation. Consider the following relationships: GM seed and food sovereignty (6); food dependence and national subordination (9); and, biodiversity and adaptability of the food supply (11). In these cases, the relationship would be very difficult (if not impossible) to quantify. Moreover, an understanding of these relationships demands context based on an implicit set of factors/assumptions. Examples 6 and 8, invoke cascading sets of political/economic consequences induced by the adoption of GM seed by farmers, and culminating in loss of national sovereignty. In other cases, the relationships invoke biological complexity. In 10, the antecedent proposition (promoter gene) leads to unintended consequences via genetic instability. Similarly, in 11 the connection between seed market concentration, biological diversity, and ability to adapt to climate change depends on a complex set of social and ecological factors. Together, the implications observed in the corpus reinforce what was observed in the ecological level of analysis. Compared to the pro sub-corpus, the anti-GM discourse rests on mental models characterized by systems and conceptual schema that encompass holistic interrelations at the interface of human and natural systems. Pro-GM discourse, by contrast, tends toward more explicit and quantifiable relationships.

The claim that anti-GM discourse encompasses more holistic interrelations is not meant to imply it is somehow more truthful or accurate. Rather, it is an attempt to characterize the conceptual and mental space within which the discourse operates. It is suggested that interactants in anti-GM discourse construct meaning by combining and mapping concepts from different mental spaces in ways that are not common in pro-GM discourse. Blending Theory (Turner and Fauconnier, 2002), tells us that this combining and mapping of concepts gives rise to meaning as an emergent structure that is beyond the sum of its parts (Evans and Green, 2006: 403). While pro-GM discourse evidently employs conceptual blending as well, the results of preceding analyses suggest that the emergent structure of the anti-GM discourse differs in the size and number of input spaces that contribute to an emergent structure of meaning.

To pursue the basic idea that concepts are combined to form emergent meaning, we can consider interdiscourse as a particular manifestation of Blending Theory. Interdiscourse refers to the relations that a discourse has to other discourses. Drawing from Bullo (2017), we can view interdiscourse as part of a process of conceptual integration and sense making. As an example, we can examine an excerpt from the anti-GM sub-corpus. The following excerpt is from the ‘Maize Manifesto’ released January 15, 2013 by the National Union of Autonomous Regional Peasant Organizations (UNORCA) in Mexico.

In our country there are more than 60 native races and thousands of local varieties of maize, which instead of representing some kind of risk, carry important virtues thanks to their selection and adaptation by indigenous peoples over more than seven thousand years. Some of these native varieties offer higher yields than the ones manipulated by Monsanto. The imposition of transnational frankenseeds would mean an end to this richness and the loss of the ancestral milpa tradition as a sustainable system of maize production and symbol of the Mesoamerican cultural inheritance.

A high degree of interdiscursivity appears in this text. In this excerpt (and the Manifesto as a whole) the scientific discourse is not dismissed but is hedged as ‘some kind of risk’ suggesting that, despite empirical research, there are unknowns associated with the technology. Scientific and technical aspects are explicitly acknowledged in this way but are also implicitly placed in the framework of ancestral tradition and culture. Transitivity suggests interconnectedness and blurred boundaries between nature and culture. The notion of ‘selection and adaptation’ of seed varieties over thousands of years suggests a natural attunement to complex biological processes, in contrast to an unnatural, hubristic ‘manipulation’ of varieties by a large corporation. Similarly, the colonial, historical images could also work as a biological metaphor. The age-old Mayan milpa tradition of crop rotation and nutrient cycling being lost at the hands of a ‘transnational’ seed is akin to an invasive species threatening an ecosystem. The historical context is also referred to in the use of the geographic term ‘Mesoamerica’ (a cultural and bioregion) as opposed to the more historically recent nation states of the region. Frankenseeds is itself an example of a type of conceptual blending known as compounding, whereby two or more morphemes combine to form a word (Evans and Green, 2006: 415). A subset of the meanings associated with each morpheme combines to give a unique and distinct meaning. The term frankenseeds appears in both sub-corpora and is used among GM critics. Of course, the word invokes Shelly’s Frankenstein, which is a portrayal of the dark side of industry and science as well as romanticism as a reaction to industrialization and Enlightenment disenchantment. Putting these literary themes in the context of GM seed is merely one example of how complex blending of concepts occurs in the excerpt and in GM discourse as a whole.

In addition to blending of concepts within the corpus, we can also consider how situated context and non-discursive practices relate to the text. Such considerations are crucial for the cognitive level in light of the notion of situated cognition, or the premise that knowledge is inseparable from action. In other words, knowledge (and therefore discourse) is bound to social, cultural, and physical contexts (Greeno, 1998). The ‘Maize Manifesto’ is not merely an article; rather, it is a political act that accompanied protests and hunger strikes by indigenous peasants in the Mexican capital. In short, the text is not understood in isolation, but in the situated context within which it was produced. The ‘Maize Manifesto’ passage points to possible differences in extra-linguistic context between the sub-corpora. Specifically, we can consider that the anti-GM discourse takes place in a context closer to situated engagement with the topic, whereas as the pro-discourse is more likely to approach the topic through a third person, objective observer. Theoretical and empirical/scientific claims of pro-GM discourse contrast with the first-person lifeworld perspective of certain actors, such as those who produced the ‘Maize Manifesto’ text.

The essence of GM seed is often framed as a biological object. This is a logical result of a conceptual approach that presupposes a human as subject and the genetically altered organism as object. As such, discursive truth claims ultimately rest with those who possess specialized knowledge of this object relation (i.e. molecular biologists); those who take a third-person position over and against the object of study. By contrast, speakers in the ‘Maize Manifesto’ begin with human experience and confront how GM technology is embedded in a plurality of contexts.

Plurality and language games

The results suggest that the anti-GM discourse embodies a plurality of actors in a way that the pro-GM discourse may not. This plurality is suggested in the distribution of geographic entities in the corpus as well as the cultural context of the anti-GM sub-corpus. In addition, the cognitive analysis indicates that intersubjectivity, connotative meanings, and first-hand experiences with respect to GM seed, are undercurrents in the anti-GM discourse. Intersubjectivity is itself an expression of plurality, insofar as it is the coming together of diverse human subjects. This intersubjective orientation contrasts with the third-person, subject-object perspective common in pro-GM discourse.

The plurality inherent in GM discourse manifests as a greater emphasis on culture in the anti-GM side of the debate, which points to interculturality as a source of misunderstanding. Rather than taking a view of intercultural misunderstanding as incompatibility between different cultural communities (Ting-Toomey and Oetzel, 2001), it may be more appropriate (at least in this debate) to view the misunderstanding in terms of whether the cultural dynamics are even being taken into account. To put it another way, GM seed discourse consists of multiple culturally embedded language games, but there is a tendency to reduce, simplify, and decontextualize the discursive plurality. This tendency is a manifestation of bias or mental shortcuts the human brain uses to reduce complexity (Tversky and Kahneman, 1974). Uncertainty leads to the stress response in humans (De Berker et al., 2016), so it is natural to seek more secure conceptual ground by replacing ambiguous concepts (like culture) with clear and distinct frameworks (like the language of empirical science). But to paraphrase from Wittgenstein’s (1986/1953) Philosophical Investigations, is an indistinct concept not ‘often exactly what we need?’ (§71).

To refer to culture as an ‘indistinct concept’ is to say it has a cluster of meanings and associations. It is complex, multilayered, and unsurveyable. Culture can refer to a myriad of subjective experiences, artifacts, behaviors, practices, beliefs, values, communication patterns, cognitive structures, etc., so vast and complex as to evade sharp definition. In the face of the complex web of relationships and perspectives, people may attempt to organize and categorize the discourse into familiar categories and frames. Cognitively, this categorization functions similar to stereotype, where information is simplified in order to make sense of an otherwise too complex world (Tajfel, 1981). No doubt, this simplification is inevitable in the face of a complex topic such as GM seed. However, the plurality and cultural context of the anti-GM discourse makes this side of the debate particularly susceptible to reduction and simplification.

Expanding intercultural communication competence

In the face of this complexity and tendency toward bias, one could question how mutual understanding might be achieved or how this analysis might contribute to the broader debate. One outcome is the importance of connotative, non-literal meanings. For instance, there are several examples of the role of metaphor in communicating concepts that could not be made explicit. In the keywords, concordances, and the ‘Maize Manifesto’ excerpt, there is a bidirectional metaphor between culture and nature. In other words, there is an implicitly understood reciprocal relationship between human culture and nature. The diversity and complexity of each is understood in terms of the other. The function of this biocultural metaphor is to capture the internal complexity of the topic at hand and frame the issue holistically.

Another outcome of this analysis is the recognition that misunderstandings may result from different language games being played. For instance, if someone operating within a positivist/scientific language game is confronted with a statement expressing cultural identity, they will need to recognize that the entire framework of meaning has changed. In addition to this recognition, a new set of cognitive structures is necessary to communicate effectively within the ‘new’ language game.

With respect to current relations between science, technology, and business, these findings point to the importance of intercultural competence in ecological debates. One aspect of intercultural competence is cognitive complexity, which has been identified as important in avoiding stereotypes and being perceptive to subtleties in communication. Gudykunst (1995) identifies cognitive complexity as crucial to managing uncertainty and anxiety in intercultural communication. While these aspects certainly apply to the present context, there are other dimensions of cognitive complexity which could be emphasized when it comes to GM seed debates. While intercultural communication has emphasized cognitive complexity as perceptual and communicative skills (e.g. perceiving nuanced differences), the analyses in this paper also points to the need for complexity in terms of abstract mental structures and frames. Cognitive complexity might be elaborated as an ability to navigate language games. To avoid bias, it would involve continuously questioning the assumption that one has the whole picture and context. This notion of complexity is consistent with that in existing intercultural competency literature, in that understanding others requires changing one’s frame of reference (Friedman, 2014). However, it repudiates approaches rooted in the rhetorical tradition, where effective communication is understood as the ability to achieve desired outcomes or the ability ‘to control and manipulate’ one’s social environment to obtain goals (Wiseman, 2001: 209).