Hindsight check: LL2-19 (Ch 19 Hungry for innovation: pathways from GM crops to agroecology)#

Source section: EEA, Late lessons from early warnings: science, precaution, innovation (EEA Report No 1/2013), Ch 19, by David Quist, Jack A. Heinemann, Anne I. Myhr, Iulie Aslaksen and Silvio Funtowicz (pp. 458–485; text pp. 458–477). There are no panels or replies. Check window: publication (2013) to late September 2026. Checked: 25–26 September 2026.

Method note. - General web search was unavailable for this pass because the session’s search budget was exhausted. I retrieved sources by fetching primary repositories and databases directly: - Europe PMC, Crossref, OpenAlex and Semantic Scholar, for peer-reviewed abstracts and metadata (several papers read at abstract level only; I say so where it matters); - the EU Publications Office (Cellar document store and SPARQL endpoint), for the Official Journal texts of EU regulations, directives, Commission decisions, merger-decision summaries, a Court of Justice judgment and European Parliament resolutions; - the European Parliament Legislative Observatory (OEIL), for procedure timelines; - US court and agency sites (Ninth Circuit opinion PDF; US Supreme Court docket; Cornell LII for the Bowman syllabus; EPA’s dicamba page); - the International Herbicide-Resistant Weed Database (weedscience.org), whose tables I parsed and counted myself; - FAO (SOFI 2026 online edition; HLPE 2019 report; the 10 Elements page), CIMMYT, EFSA (Eurobarometer 2019 report), the European Patent Office, the Horizon Europe AGROECOLOGY Partnership site and the CGIAR site; - AgbioInvestor’s Global GM Crop Area Review 2025 (May 2026), downloaded in full. - Several publishers and agencies refused automated retrieval (ScienceDirect, Springer, Nature, Taylor & Francis, OUP full text, OECD, USDA ERS data pages, EUR-Lex). Where I relied on an abstract, a docket entry or a secondary report instead of the full text, I say so. - Annex 3 does not apply. GM crops/agroecology is a new chapter in the 2013 volume, so there is no Annex 3 update. - I read only the section digest, the section’s source extract and web sources. Page numbers are report pages of the 2013 volume. - Source-interest flags. - Industry-linked data: ISAAA [S1] (the chapter itself calls its figures “industry-derived”, p. 463) and AgbioInvestor [S2, S4] (a commercial consultancy whose trait-cost study surveyed the four largest trait developers). I use them for area statistics, which are not seriously disputed, and flag them where they carry an argument. - Developer-partner studies: the African MON 87460 efficacy study [S46] includes authors from the public–private TELA/WEMA partnership. - Advocacy or NGO sources: IPES-Food/Biovision Money Flows [S69]. - Journalism: Retraction Watch [S36], used only for the reasons given in a retraction notice that I could not open directly. - Chapter authors’ own later work: Heinemann co-authored a 2014 paper on US Midwest agriculture [S22] and signed the 2015 “No scientific consensus on GMO safety” statement [S32]. These are the authors’ positions, not independent checks.

Overview#

The chapter is less a single late-lessons case than an argument about two innovation “pathways”. It makes four kinds of claim. Thirteen years on, they have fared very differently.

1. The political economy of the “top-down” pathway: largely borne out, in places more strongly than the chapter predicted. - Narrow delivery persisted. After thirty years of commercial GM crops, herbicide tolerance, insect resistance and their stacks still account for essentially all GM area. In 2025, GM crops covered 216.0 million ha in 30 countries. The same five countries as in 2011 (US, Brazil, Argentina, Canada, India) held about 89%, against 91% then [S2]. Drought-tolerant wheat, virus-resistant beans, Bt eggplant and Bt cowpea exist but together occupy well under 0.1% of GM area [S2]. - The herbicide treadmill accelerated. The weed database now lists 62 glyphosate-resistant species; 34 of them were first recorded in 2013 or later (my count) [S5]. The stacked-trait “fix” (dicamba- and 2,4-D-tolerant crops) was followed within a few years by auxin-resistant and glufosinate-resistant Palmer amaranth [S6]. It also caused a drift disaster: about 3.6 million acres of US soybeans were damaged in 2017. In 2020 a federal appeals court vacated the registrations. Among other failings, it held that EPA had ignored anti-competitive effects, citing evidence that farmers were planting tolerant seed defensively to protect themselves from neighbours’ spraying [S7]. EPA re-approved over-the-top dicamba for 2026–2027 only, under much tighter limits [S8]. - Concentration happened. The 2017–2018 mergers (Dow/DuPont, ChemChina/Syngenta, Bayer/Monsanto) went ahead. The European Commission itself found that the industry had consolidated to “only five global R&D integrated players” and that Dow/DuPont would “significantly diminish innovation competition” [S40, S41]. The US Supreme Court confirmed in Bowman v. Monsanto (2013) that patent rights extend to saved seed [S39]. - EU law has since written the treadmill logic into statute. The 2026 regulation on new genomic techniques excludes herbicide-tolerant plants from its lighter regime. It does so on the ground that such cultivation can breed resistant weeds or raise herbicide use “regardless of the breeding technique” [S12].

2. The “late lesson in the making” on health: the harm leg weakened sharply; the institutional leg was partly vindicated. - The harm leg. Séralini et al. 2012, the chapter’s main evidence of health “indications” (p. 469), was retracted in November 2013 [S24]. An EU-funded two-year replication on the same maize found no adverse effects [S26]. So did a French six-month study [S27]. NASEM (2016) found “no substantiated evidence” of different health risks from commercialised GE crops [S3]. - The institutional leg. Many of the chapter’s criticisms of risk-assessment practice were taken up or vindicated: - the EU made 90-day whole-food studies mandatory in 2013, and required power analysis and testing of herbicide-treated material [S28]; - it legislated study notification, proactive disclosure and publicly funded verification studies (2019, applying from 2021) [S29]; - the Court of Justice required authorities not to give applicant studies “preponderant weight” [S30]; - a 2016 analysis found industry ties in 40% of Bt-crop efficacy papers, associated with 50% more favourable outcomes [S31]; - in December 2025 a widely cited 2000 glyphosate safety review was retracted over undisclosed Monsanto ghostwriting [S36]. - The hazard dispute persists. The IARC-versus-EU dispute over glyphosate carcinogenicity is unresolved in science, but settled in regulation: the EU renewed glyphosate to 2033 without a qualified majority of member states [S33–S35].

3. The agroecology evidence: the yield claims weakened; the ecosystem-service and smallholder-benefit claims held up in a more modest form. - Larger meta-analyses put the average organic yield gap at about 19–25% [S53, S54], and US commercial data at about 33% [S55]. Badgley and colleagues’ own 2026 retrospective accepts that later studies found yield gaps similar to, or somewhat larger than, theirs [S51]. - The chapter’s reading of Badgley 2007 (“same or superior yields”; mature conversions “consistently out produced industrial operations”, p. 474) overstates even the original paper’s abstract [S51]. - What has held up is more modest: - diversification practices shrink the gap substantially [S54]; - diversification boosts ecosystem services without compromising yield [S57]; - most studies of agroecological practices in low- and middle-income countries report positive food-security outcomes [S58]. - Scaling remains largely undemonstrated: the flagship Indian programme’s randomised evaluation was still at protocol stage in the sources I found [S62]. An abrupt, top-down national switch to organic inputs (Sri Lanka 2021) cut production [S61].

4. Recommendations: partial, then reversible, uptake. - FAO adopted the “10 Elements of Agroecology” (2019) [S63]. The UN food-security expert panel (HLPE) endorsed agroecological approaches while stressing large evidence gaps and chronic underfunding (1–1.5% of relevant budgets) [S23]. - The EU’s Farm to Fork strategy (2020) promised an agroecology research partnership, 25% organic land and a 50% cut in pesticide use and risk [S64]. The partnership launched in 2024 [S67]. - But the pesticide regulation was rejected by Parliament and withdrawn (2023–2024) [S65]. The Commission’s 2025 Vision for Agriculture and Food does not use the word “agroecology” and backs new genomic techniques [S66].

Net weight. - The chapter is strongest where it analyses mechanisms: - appropriability steering what gets developed; - single-tactic resistance treadmills; - consolidation narrowing the pipeline; - developer-controlled evidence. All of these have been borne out, some dramatically. - It is weakest where it: - reads early harm signals from a single contested study; - accepts uncontrolled agroecology project syntheses at face value; - makes categorical predictions (“will always serve last”, “increases food insecurity”). - Its smallholder pessimism needs qualifying. Randomised and panel evidence now shows real smallholder gains from some publicly developed insect-resistant crops [S21], alongside failures and reversals elsewhere [S15, S20].

Claim-by-claim#

Claim 1: Herbicide-tolerant crops’ advantages are “disappearing” as glyphosate-resistant weeds evolve rapidly, putting farmers on a “treadmill” and pushing some back to tillage and “(possibly more toxic)” herbicides; stacking extra tolerance genes will further undermine sustainable weed management (Box 19.1, pp. 462–463)#

Original claim (pp. 462–463). - HT crops simplified weed control through “near exclusive reliance on a single agrochemical product”. Their advantages “are, however, disappearing”, because extensive glyphosate use “has led a rapid evolution of glyphosate-resistant weeds” (p. 462). - Farmers “have entered into a treadmill where overuse of a single product leads to tolerance and tolerance is overcome with more product”. Some are “returning to tilling and using other (and possibly more toxic) herbicides” (p. 462). - “Stacking additional herbicide tolerance genes … is not an alternative to IWM”, because the new traits will encourage “continued neglect of public research and extension in integrated weed management” (p. 462, quoting Mortensen 2012).

Subsequent developments - Spread of glyphosate resistance. - As of September 2026, the International Herbicide-Resistant Weed Database lists 62 weed species resistant to glyphosate (HRAC Group 9), in 393 species-by-country/state entries across 31 countries [S5]. - By my count of the table, 34 of those species were first recorded in 2013 or later, and 186 of the 393 entries date from 2013 onwards [S5]. - Palmer amaranth alone has 56 entries, including populations resistant to up to five or six modes of action (Kansas 2015 and 2021; Arkansas 2016; North Carolina 2025) [S5]. - The stacked-trait systems of the mid-2010s met resistance quickly. - Dicamba-tolerant (“Xtend”) soybeans were planted on about 1.7 million acres in 2016, 27 million in 2017 and, with cotton, 56 million acres in 2018 [S7]. - Auxin (dicamba/2,4-D-group) resistance in Palmer amaranth is recorded in Kansas (2015, 2018, 2021), Tennessee (2020), Arkansas (2023) and North Carolina (2025), and in waterhemp in Illinois (2016) and Iowa (2021) [S6]. - Glufosinate resistance, the third stacked tolerance, appears in Palmer amaranth (Arkansas 2020; North Carolina 2022) and waterhemp (Illinois 2025) [S6]. - Drift damage. - The Ninth Circuit recorded that in 2017 state agencies were investigating 2,708 formal complaints. University weed scientists estimated about 3.6 million acres of soybeans in 24 states (about 4% of US soybean acreage) were damaged by off-field dicamba movement [S7]. - An EPA official confirmed “more than 3.6 million acres” and said reported damage was likely an underestimate [S7]. - Regulatory and judicial response. - In National Family Farm Coalition v. EPA (3 June 2020), the court vacated the 2018 registrations of XtendiMax, Engenia and FeXapan [S7]. It held that EPA had “substantially understated” risks it acknowledged, and “entirely failed to acknowledge” three others: - the likelihood that label restrictions would not be followed; - “anti-competitive economic effects”; - the risk that dicamba “would tear the social fabric of farming communities”. - The court documented that farmers were planting dicamba-tolerant seed “as a defensive measure against damage from neighbors”. It found a “substantial … risk that DT soybeans, and possibly DT cotton, will achieve a monopoly or near-monopoly” [S7]. - [Background, not retrieved: a federal district court in Arizona vacated the 2020 re-registrations in February 2024.] - EPA’s current page (updated 6 February 2026) approves three dicamba products for over-the-top use on cotton and soybeans for the 2026 and 2027 seasons only [S8]. The conditions include: - a maximum of 1 lb/acre/year; - a ban at 95°F or above; - doubled volatility-reduction agent; - a 240-foot downwind buffer; - a review after two seasons. - Herbicide-use trends. - NASEM (2016): total kilograms of herbicide per hectare fell when HT crops were first adopted, but “the decreases have not generally been sustained”. It added that kilograms applied is a poor proxy for risk [S3]. - Perry et al. (2016), using plot-level data for 1998–2011, found that adopters of glyphosate-tolerant soybeans used 28% more herbicide than non-adopters. For both maize and soybeans, GT adopters used increasingly more herbicide relative to non-adopters over time, “consistent with the emergence of glyphosate weed resistance” [S9]. - Benbrook (2016): global glyphosate use rose almost 15-fold after 1996; HT crops account for about 56% of global glyphosate use [S10]. - Schulz et al. (2021), weighting 381 pesticides by toxicity to eight non-target species groups, found rising applied toxicity to terrestrial plants in HT soybeans since about 2010 [S11]. That fits the chapter’s parenthetical “possibly more toxic”. - Legislative uptake of the mechanism. Regulation (EU) 2026/1388 on new genomic techniques (17 June 2026) excludes herbicide-tolerant NGT plants from the lighter “category 1” regime. Its reasoning (recital 19) is that HT cultivation “can lead to the development of weeds resistant to those herbicides or to the need to increase the quantity of herbicides applied, regardless of the breeding technique” [S12]. HT traits are also barred from the regulation’s sustainability incentives [S12].

Complications - NASEM found HT crops “contribute to greater yield where weed control is improved”. It found “little evidence” that weed-species shifts had caused agronomic harm. It said integrated weed management can delay resistance, “especially in cropping systems not yet exposed to continuous glyphosate applications” [S3]. The chapter’s “disappearing” advantages are therefore a matter of degree, and they are not uniform. - I did not find post-2013 primary data on reversion to tillage, so that sub-claim is unchecked. - The chapter’s reference to glyphosate health harms rests on Greenpeace 2009 and Séralini 2012 (p. 462). That thread is assessed under Claim 4.

Verdict: strengthened. Resistance spread widely after 2013, and resistance to the stacked replacements appeared within a few years of their launch. The stacked systems produced an off-target drift crisis serious enough for a federal court to vacate registrations. EU law now states the treadmill mechanism in terms. The “defensive adoption” dynamic the court described goes beyond what the chapter anticipated. An externality can itself force uptake and entrench a product.

Implications for weight. This is the chapter’s best-evidenced mechanism and can carry substantial weight: single-tactic control of an adaptive system breeds resistance, and “stacking” tactics buys time rather than escaping the treadmill. The strong form of the lesson should be stated with NASEM’s caveats: yield and management-flexibility benefits are real where weed control improves, and total herbicide mass is a poor risk metric.

Claim 2: After nearly 20 years of commercialisation only two traits (herbicide tolerance and insect resistance) matter commercially; five countries account for 91% of GM area; the 159 million ha of 2011 is “just 3%” of world agricultural land (p. 463; Figs 19.1–19.2, pp. 464–465)#

Original claim (p. 463). - “Despite more than 30 years of research and development and nearly 20 years of commercialisation … surprisingly only two traits have been significant in the marketplace — herbicide tolerance and insecticide production.” - The US, Brazil, Argentina, India and Canada “account for 91 % of the global GM crop production”. The 159 million ha in 2011 is “seemingly a large figure, but in reality is just 3 % of the world’s agricultural land.”

Subsequent developments - Traits. - NASEM (2016): “the only characteristics that have been introduced through genetic engineering into widespread commercial use are those that provide insect resistance and herbicide resistance”. Virus resistance and non-browning traits existed, but “on a relatively small number of hectares” [S3]. - ISAAA (2019) reported new approvals (drought- and salt-tolerant soybeans, non-browning apples, insect-resistant sugarcane). Stacked HT/IR traits covered 45% of GM area [S1]. - AgbioInvestor’s 2025 review shows the narrowness persisting [S2]: - soybean, maize, cotton and canola made up about 99% of the 216.0 million ha; - HB4 drought-tolerant (and glufosinate-tolerant) wheat covered 55,250 ha in Argentina, about 0.8% of its wheat, grown in a closed-loop system; - GM sugarcane about 0.1 million ha; - virus-resistant beans in Brazil about 5,000 ha; - Bt eggplant in Bangladesh 1,789 ha; - Bt cowpea about 24,000 ha in Nigeria. - Golden Rice was commercialised in the Philippines in 2022 (about 38,000 ha) but “removed from the market in 2024” [S2]. - In the US, stacked varieties rose from 1% of maize area in 2000 to 84% in 2025 [S2]. - Geography. - ISAAA 2019: 190.4 million ha in 29 countries; the same five countries held 91% [S1]. - AgbioInvestor 2025: 216.0 million ha in 30 countries [S2]: - US 34.8%, Brazil 32.4%, Argentina 11.8%, Canada 5.4%, India 4.9%, together about 89%; - China entered the top ten with commercial GM maize and soybean (approved 2023) and grew 79% in 2025, to 6.2 million ha; - Burkina Faso stopped growing GM cotton after 2015; - Europe’s GM area shrank to 70,653 ha, almost all MON 810 maize in Spain; six other EU countries stopped planting between 2008 and 2016; - new adopters are mostly small (Kenya, Nigeria, Ethiopia and Ghana, each well under 0.1 million ha). - The “3%” denominator. - NASEM put the 2015 GE area (almost 180 million ha) at “about 12 percent of the world’s planted cropland” [S3]. The chapter’s 3% therefore depends on counting permanent meadows and pastures, where no GM crop is grown. - By 2025, GM varieties covered 28.2% of the combined area of the eleven crops for which they exist, including 76.7% of world cotton and 73.9% of world soybean [S2].

Complications - Narrowness is partly explained by forces the chapter did not stress: regulatory time and cost, and litigation. - An industry-sourced survey (AgbioInvestor 2022, for the four largest developers) put time to market for a new trait at 16.5 years, up from 13.1. The regulatory phase took 51.1% of that time and 37.6% of cost [S4]. - Court challenges halted Golden Rice in the Philippines and GM food crops in Kenya [S2]. These are industry or secondary accounts. But they mean narrowness is not only an artefact of IP-driven incentives (see Claim 10).

Verdict: held up. The two-trait concentration and five-country dominance persisted almost unchanged for thirteen more years, despite some diversification of crops, countries and public-sector products. The “3%” framing was misleading when written and remains so. The chapter’s valid point was about uneven commitment across countries (Fig. 19.1), not global scale.

Implications for weight. The lesson that broad early promises narrow to what can be packaged and sold in the largest markets (pp. 463, 466) is well supported. The rhetorical use of a flattering or unflattering denominator is itself a lesson: headline scale depends on the choice of reference base.

Claim 3: GM benefits have been overstated: the highest yields come from conventional breeding rather than GE traits; even mature adopters (US-south cotton) saw no net economic benefit; pesticide reductions are overstated (Box 19.1, p. 463; Box 19.2, p. 468)#

Original claim. - “The highest yielding varieties of GM crops are so because of ongoing and intensive genotype improvement through traditional breeding, rather than through the development of genetically engineered traits” (p. 463). - “Even in the most mature GM agroecosystems, such as cotton plantations in the US south, GM-farmers have not enjoyed a net economic benefit” (p. 463, citing Jost 2008). - Benefits “that may have been overstated are the reduction in pesticide use …, higher yields … and farmer income” (p. 468).

Subsequent developments - Yields. - NASEM (2016): “Although the sum of experimental evidence indicates that GE herbicide resistance and insect resistance are contributing to actual yield increases, there is no evidence from USDA data that the average historical rate of increase in U.S. yields of cotton, maize, and soybean has changed” [S3]. Bt traits reduce losses where target pests are damaging [S3]. - Klümper & Qaim (2014; 147 studies) found average yield gains of 22%, larger for insect-resistant than herbicide-tolerant crops and in developing countries. They note that several source studies did not report sample sizes or variance [S13]. - Pellegrino et al. (2018; 21 years of field data on GE maize) found yields 5.6–24.5% higher than near-isogenic lines and lower mycotoxin levels [S14]. - The UN food-security expert panel (HLPE, 2019) summarised this literature. It warned that “significant methodological limitations prevent assigning measured benefits to GM traits rather than to other factors” [S23]. - Farmer economics. - NASEM: GE soybean, cotton and maize “have generally had favorable economic outcomes for producers who have adopted these crops, but there is high heterogeneity in outcomes” [S3]. - Klümper & Qaim: average profit gains of 68% [S13]. - I found no post-2013 primary source testing the specific US-south cotton claim (Jost 2008). The national-level synthesis runs against it. - Pesticides. - Insecticide reductions from Bt crops are well documented: NASEM [S3]; Perry et al. found 11.2% less insecticide in IR maize, falling further over time [S9]. - Herbicide reductions were not sustained (Claim 1) [S3, S9, S11]. - Long-term Bt cotton records. - India. Kranthi and Stone (2020; a Perspective) argue that yield increases track fertiliser and other inputs rather than Bt adoption [S15]. They find initial pesticide reductions reversed: with pink bollworm resistance and non-target pest surges, “farmers now spend more on pesticides today than before the introduction of Bt” [S15]. Qaim published a critical reply, and the authors responded [S16], so this remains contested. - Independent entomological work documents field-evolved pink bollworm resistance to Bollgard II Bt cotton. It also shows rising insecticide use and insecticide resistance in the same pest [S18]. Tabashnik et al. (2023) count 26 cases of “practical resistance” to Bt crops by 2020 in 11 pest species, including pink bollworm in India [S17]. - India still grows Bt cotton on about 95% of its cotton area (10.6 million ha) [S2]. - Pakistan. Pakistan still plants mostly single-gene (Cry1Ac) Bt cotton on about 95% of cotton area. Seed companies are “reluctant to commercialise new GM technology … due to lack of IP protection and regulatory uncertainty” [S2]. Field studies report highly variable toxin expression, and evidence of practical resistance in H. armigera [S19]. - Burkina Faso. Burkina Faso, once Africa’s showcase, began a complete phase-out of GM cotton in 2016, citing “the inferior lint quality of the GM cultivars” [S20]. HLPE adds high seed prices, sector governance problems and secondary pests [S23]. No GM cotton has been grown there since 2015 [S2]. - The chapter authors’ later work. Heinemann et al. (2014) compared the US Midwest with Western Europe. They concluded that the US system “is not exceptional in yields or conservative on environmental impact” [S22].

Complications - The balance of evidence splits along trait lines: - Bt traits have delivered real yield-protection and insecticide benefits in many settings, though often eroded by resistance where refuges and stacking were weak [S3, S13, S14, S17]; - HT traits have delivered management convenience more than yield or pesticide reductions [S3, S9]. - The chapter’s single-source claims (Gurian-Sherman 2009, Jost 2008) are directionally right about yield potential (the NASEM finding), but wrong as blanket statements about farmer economics.

Verdict: partly held up. - Held up: yield potential comes mainly from breeding; herbicide-use reductions were overstated. - Qualified by resistance: initial insecticide reductions from Bt were real, but were later eroded by resistance in some places (India, Pakistan). - Weakened: the claims of no net economic benefit and overstated income gains, which NASEM and meta-analyses contradict on average. Heterogeneity is large, and some national cases (India long term, Burkina Faso) are contested or reversed.

Implications for weight. Use the chapter’s scepticism as a caution about which benefit is being claimed and over what time horizon. Do not use it as evidence that the technology delivered nothing. The strongest transferable point is temporal: benefits measured in the first years of adoption can erode once pests and weeds adapt and the surrounding input system shifts.

Claim 4: GM crops may be “a late lesson case in the making”, with “early indications of harm … just emerging”, including adverse health effects appearing only after 120 days (Séralini 2012); risk assessment is structurally unable to detect harm (developer data, confidentiality, 90-day trials, surrogate proteins, broad comparators, weak monitoring) (Box 19.2, pp. 468–470)#

Original claim (pp. 468–470). - The literature is “accumulating indicators both of inflated benefit claims and of evidence of adverse effects” (p. 468). - “Research has indicated the importance of life-time studies for health affects where indications of adverse health impacts only manifested after 120 days (Séralini, 2012)” (p. 469). - The Box lists obstacles: - industry restrictions on material; - risk research funding at about 1% of USDA biotech research; - career risks for dissenting scientists; - developer-supplied testing and “the funding effect”; - confidentiality; - underpowered studies; - surrogate proteins; - 90-day trials only; - reference lines that inflate background variation; - weak post-release monitoring (pp. 468–470). - Conclusion: “The critical late lesson that may be emerging from GM crops is not the evidence of harm — the early indications of harm are just emerging — but the persistence of the same institutional patterns” as asbestos, benzene and BSE (p. 470).

Subsequent developments: the health-harm evidence - Séralini 2012 retracted. - Food and Chemical Toxicology retracted the paper in November 2013 (notice in vol. 63, January 2014) [S24]. As widely reported, the notice found no evidence of fraud but judged the results inconclusive. The main criticisms were the small group sizes and the tumour-prone Sprague-Dawley strain. [I could not open the notice itself; wording as reported in secondary sources.] - The authors republished an amended version in Environmental Sciences Europe in June 2014 [S24]. The retraction itself was criticised by some environmental-health scientists [S25]. - EU-funded replication. - The G-TwYST consortium ran two 90-day trials and a two-year combined chronic toxicity/carcinogenicity trial of the same NK603 maize, with and without Roundup, in Wistar Han rats. It concluded “no adverse effects related to the feeding of the NK603 maize cultivated with or without Roundup for up to 2 years were observed” [S26]. - France’s GMO90+ six-month study of NK603 and MON810 (with metabolomics and transcriptomics) found no biologically meaningful GM effect [S27]. - NASEM (2016). NASEM found “no substantiated evidence of a difference in risks to human health between currently commercialized genetically engineered (GE) crops and conventionally bred crops”. It explicitly recognised “the inherent difficulty of detecting subtle or long-term effects”, and found no pattern of health differences between North America and Western Europe after GE foods were introduced [S3]. - Dissent persists. A 2015 paper by 15 authors, including chapter author Heinemann, argued there is “no scientific consensus on GMO safety” [S32]. This is a minority position against NASEM and the EU-funded feeding studies.

Subsequent developments: the institutional critique - EU rules on feeding studies (2013). Commission Implementing Regulation (EU) No 503/2013 (3 April 2013) made a 90-day whole-food rodent feeding study mandatory for single-event GM plants “to improve consumer confidence”, although EFSA considered it justified only case by case (recital 11) [S28]. The same regulation: - requires a power analysis to size studies; - requires testing of material from HT plants “exposed to the intended herbicide”; - prefers historical background data to reference varieties; - requires a systematic literature review [S28]. These answer several of Box 19.2’s methodological points (low power, herbicide-treated material, variance-inflating comparators). - Article 12 required the Commission to review the 90-day requirement in light of the GRACE project by 30 June 2016. The only consolidated version in the Publications Office catalogue is the original, so the requirement appears unamended as of 2026 [S28]. - Transparency Regulation. Regulation (EU) 2019/1381 (applying from 27 March 2021) amended the General Food Law [S29]. It: - requires business operators and laboratories to notify EFSA of all studies commissioned to support an application, on pain of inadmissibility (Art. 32b); - adds public consultation on submitted studies (Art. 32c); - allows the Commission, “in exceptional circumstances of serious controversies or conflicting results”, to have EFSA commission verification studies (Art. 32d); - mandates proactive disclosure of supporting data (Arts. 38–39e). Recital 66 acknowledges “public concerns about the Authority’s assessment … being primarily based on industry studies” [S29]. - Court of Justice. In Blaise (C-616/17, Grand Chamber, 1 October 2019), a reference arising from the prosecution of activists who damaged cans of glyphosate weed killer in French shops, the Court held that applicant-supplied studies are not unlawful as such [S30]. But authorities “are of necessity bound to take into account relevant evidence other than the tests … submitted by the applicant”, and must “not … give in all cases preponderant weight to the studies provided by the applicant” [S30]. - The funding effect in GM research. Guillemaud et al. (2016), analysing 672 papers on Bt-crop efficacy and durability, found conflicts of interest in 40%. A conflict of interest was associated with a 50% higher frequency of outcomes favourable to the company [S31]. This is direct post-2013 evidence for the chapter’s “funding effect” point (p. 469), in the benefit literature rather than the safety literature. - Glyphosate: hazard dispute and evidence integrity. - IARC classified glyphosate as “probably carcinogenic to humans” (Group 2A) in March 2015 [S33]. - EFSA (2015, 2023) and ECHA’s Risk Assessment Committee (2017; 30 May 2022) concluded it does not meet the criteria for classification as carcinogenic [S34, S35]. - The EU renewed approval on 28 November 2023 for ten years. Neither the Standing Committee nor the Appeal Committee delivered an opinion (no qualified majority either way), so the Commission adopted the act itself. The renewal notes that indirect biodiversity effects “could not be excluded” and that assessment methods for them are still lacking [S35]. - In December 2025, Regulatory Toxicology and Pharmacology retracted Williams, Kroes & Munro (2000), a glyphosate safety review cited about 614 times. The retraction followed litigation-disclosed evidence that Monsanto employees co-wrote it without acknowledgement, and the review’s reliance on unpublished Monsanto studies [S36]. In June 2026, Taylor & Francis was investigating two further Critical Reviews in Toxicology papers (2013, 2015) on similar grounds [S36]. - On 25 June 2026, the US Supreme Court (7–2) reversed and remanded a Missouri judgment for a Roundup plaintiff in Monsanto Co. v. Durnell, on whether federal pesticide law pre-empts label-based failure-to-warn claims where EPA has not required the warning [S37, docket only; I did not read the opinion]. - Continuing political contestation in the EU. - The Publications Office catalogue lists 63 European Parliament resolutions (2015–2026) on draft Commission decisions authorising GM food and feed. These are the vehicle for Parliament’s non-binding objections [S38]. - A December 2025 example objects to a sugar beet tolerant to glyphosate, dicamba and glufosinate. It notes that member states again delivered “no opinion”, and that “the Commission continues to authorise GMOs” despite this [S38].

Complications - The EU-funded long-term studies were designed partly to test the chapter’s claim that 90-day trials miss chronic effects. For the product in question they did not find a late-emerging effect [S26, S27]. That supports NASEM’s conclusion, not the chapter’s “early indications”. - But these studies were exactly the kind of independent, publicly funded testing the chapter called for (pp. 468–469). Their existence is partly a vindication of its institutional argument. - The glyphosate ghostwriting retraction and the Bt conflict-of-interest analysis show that the chapter’s worry about evidence integrity was well founded [S31, S36]. That remains true whatever the eventual verdict on glyphosate’s hazard.

Verdict: partly held up. - Health-harm leg: weakened. The one study offered as an early harm signal was retracted, and a two-year EU replication found nothing. No substantiated food-safety harm from commercialised GM crops has emerged in the mainstream assessments I checked. - Institutional leg: held up and in part strengthened. The funding effect, the dependence on developer data and the lack of transparency have all been documented since 2013, and all have become the target of EU legislation and case law.

Implications for weight. Use Box 19.2 for its institutional diagnosis, not for its harm signal. The chapter explicitly says the lesson is institutional (p. 470), and that part has aged well. Its tendency to treat single contested studies as “early indications” should count against its evidential judgement.

A useful refinement: “Is the evidence-generating process trustworthy?” and “Is the product harmful?” are separable questions. They can have different answers, and independent replication can settle the second while reforms address the first.

Claim 5: IP-led “top-down” innovation will keep feeding a loop that concentrates wealth, knowledge and influence in the seed and agrichemical sector; firms lose the incentive to breed conventional varieties; seed saving becomes incompatible with IP (pp. 467, 476–477)#

Original claim. - “The patenting of germplasm is concentrating IP rights-based control of the seed supply under a very small number of multinational corporations”, and the top-four concentration ratio (CR4) had “breached a critical threshold” (p. 467). - Quoting UK POST (2011): seed companies “may have reduced incentives to develop conventional varieties”. In US soybeans, conventional breeding is “now mainly left to universities and to small seed producers” (p. 467). - “Seed savings and exchanges have become incompatible with these more severe IP instruments” (p. 467). GM crops “destroy local seed savings and exchange practices” (p. 476). - Top-down approaches “contribute to a feedback cycle that concentrates resources, knowledge, and influence as witnessed in the seed and agrichemicals sector” (p. 477).

Subsequent developments - Seed saving and patents. Bowman v. Monsanto Co., 569 U.S. 278 (13 May 2013, unanimous) held that “patent exhaustion does not permit a farmer to reproduce patented seeds through planting and harvesting without the patent holder’s permission” [S39]. - Consolidation. - Dow/DuPont (Case M.7932, approved 27 March 2017 with conditions) [S40]. The Commission found the industry “already characterised by oligopolistic innovation competition”. “Following successive waves of consolidation there are now only five global R&D integrated players”. The merger “would be likely to significantly diminish innovation competition”, through reduced incentives to continue existing lines of research and to develop as many new products in the long term [S40]. - ChemChina/Syngenta approved 5 April 2017 [S42]. - Bayer/Monsanto (Case M.8084, approved 21 March 2018) [S41]. The Commission found competition concerns in seeds, traits (including “innovation competition” in broad-acre traits and herbicide-tolerance systems), crop protection and related farm-advisory services. It noted that the deal “would strengthen Monsanto’s dominant position in corn HT trait market” [S41]. Approval required divesting a broad package of seeds, traits, crop-protection and related assets to BASF, plus Bayer’s global vegetable-seed business [S41]. - The result was four integrated firms (Bayer, Corteva, Syngenta Group and BASF). Clapp (2021) sets out how concentrated firms can “shape markets, shape technology and innovation agendas, and shape policy and governance frameworks” [S43]. - The Ninth Circuit’s dicamba ruling documents a near-monopoly dynamic in trait markets. Glyphosate-tolerant seed passed 90% of US soybeans by 2008, and dicamba-tolerant seed reached about 50% by 2018, partly through defensive adoption [S7]. - Policy responses to IP concerns. - The European Patent Office’s Enlarged Board (G 3/19, 14 May 2020) held that plants exclusively obtained by essentially biological processes (conventional breeding) are not patentable for patents filed from 1 July 2017 [S44]. - The EU’s 2026 NGT Regulation requires applicants to disclose patents and optional licensing declarations. It sets up a public database, provides for a Union code of conduct on patent licensing, and orders a Commission assessment of patenting’s impact on breeding. The recitals cite “concerns that patents relating to NGT plants might limit the access of breeders” [S12]. - Where IP is weak. - In Argentina, soybean seed saving is widespread. “The difficulties in controlling intellectual property” have led multinational seed firms to limit their exposure or exit [S2]. - In Pakistan, lack of IP protection and regulatory uncertainty have produced “technology stagnation” [S2]. - These cases confirm the chapter’s premise that IP and seed saving pull against each other. They also show the other side of the trade-off: where seed saving prevails, private trait investment withdraws. - Conventional breeding. I did not retrieve a post-2013 primary dataset on conventional-variety breeding effort in GM-dominated crops, so the POST claim is unchecked here. US adoption of GM varieties exceeds 90% for every GM crop except alfalfa [S2], so the market for conventional seed of those crops remains small.

Complications - The concentration was approved, with remedies, by competition authorities who explicitly recognised innovation harms [S40, S41]. The feedback loop the chapter describes therefore operated through lawful, reviewed transactions. Divestitures preserved the number of competitors in some segments but did not restore the pre-merger count of integrated R&D players. - Some public-sector counter-currents emerged: the EPO ruling [S44] and the NGT patent-transparency provisions [S12].

Verdict: strengthened. Concentration went further than the chapter could document in 2013, and a major competition authority, applying its own tests, confirmed the mechanism of reduced innovation competition. Seed saving of patented seed was definitively ruled out in the US. The sub-claim that firms lose the incentive to breed conventional varieties remains plausible but was not re-verified.

Implications for weight. The chapter’s concentration lesson can carry substantial weight, particularly the specific mechanism that consolidation among a few integrated R&D players reduces innovation competition upstream (pp. 467, 477). The IP–seed-saving tension is real in both directions. The chapter only presents the side where IP suppresses farmer practice.

Claim 6: Genetically engineered drought tolerance will lag classical breeding and marker-assisted selection because drought tolerance involves many genes; top-down genotype fixes will only “shift the problem in time or space” (pp. 474–475)#

Original claim. Progress on drought tolerance “is being made in some crops through classical breeding … especially augmented through marker-assisted selection. Yet similar genotypic approaches using genetic engineering have not been as successful. As drought tolerance depends on the action of multiple genes …” (pp. 474–475). Top-down solutions “will only shift the problem in time or space, addicting us to finding and producing even more extreme genotypes” (p. 475).

Subsequent developments - GE drought traits. - Monsanto’s MON 87460 (cold-shock protein B; DroughtGard) increased grain yield by 6% on average under water-limited conditions in 2009–2011 field trials, with no consistent difference under well-watered conditions [S45]. - In eastern and southern Africa, the WEMA/TELA partnership tested MON 87460 in more than 120 conventional drought-tolerant hybrids over 3–6 years. Significant yield gains appeared in only a subset: 36–62% in three hybrids at one South African site, and 7–13% in five of 34 adapted hybrids [S46]. (This study has partnership-affiliated authors.) - TELA maize (MON 87460 × MON 89034) was authorised in Nigeria (2024) and Ethiopia (2024). Commercial release in Kenya is on hold pending litigation [S2]. - HB4 wheat (sunflower HaHB4 transcription factor), in 37 field experiments, gave 6% higher yield on average, 16% under stress and 3% without stress [S47]. It is grown on about 55,000 ha in Argentina and approved but not commercialised in Brazil [S2]. - Conventional drought-tolerant maize in Africa. - The CIMMYT/IITA Drought Tolerant Maize for Africa project (2006–2015) reports that 60 drought-tolerant hybrids and 57 open-pollinated varieties were marketed or made available in 2007–2012 alone, across 13 countries [S48]. - CIMMYT’s 2021 review describes deployment of “an array of elite stress-tolerant maize cultivars across sub-Saharan Africa, Asia, and Latin America” through public breeding and public–private partnerships [S49]. - Household impact studies: - Uganda: +15% yield and a 30% lower probability of crop failure [S50]; - Zambia: +15% yield, 38% lower yield variance and 36% lower downside risk [S50]; - Tanzania: about +1 t/ha and reduced food insecurity, with larger benefits for poorer households [S50]; - on-farm trials in eight countries: yield advantages of 4–19% over commercial checks [S50]. - The WEMA project’s commercial “DroughtTEGO” hybrids are themselves conventionally bred. The GE trait was layered onto them [S46].

Complications - GE drought traits do now exist and show modest benefits under stress [S45–S47]. The chapter’s implication that GE drought tolerance might not be achievable is too strong. - The “shift the problem in time or space” and water-table depletion claims (p. 475) were not tested by anything I found.

Verdict: held up. Conventionally bred drought-tolerant maize reached far more smallholders, sooner, with documented yield-stability and food-security gains. GE drought traits arrived later, delivered modest and inconsistent gains (roughly 6% on average), cover small areas, and in Africa depend on conventionally bred backgrounds.

Implications for weight. The underlying mechanism can carry weight: complex, multigenic traits favour approaches that work on the whole genome and the environment over single-gene insertion. It should not be over-extended to “GE cannot deliver”. The better-supported lesson is about delivery channels: publicly funded breeding with local testing networks reached smallholders at scale.

Claim 7: Agroecological and organic systems can match or exceed industrial yields, with mature conversions out-producing industrial farms (per the chapter’s reading of Badgley 2007); African projects roughly doubled yields (+116% across 114 cases; 2.13-fold across 40 projects); the potential to scale up is “immense” (pp. 473–474)#

Original claim. - “The world’s largest meta analysis … found that the latter could match the former in the common metric of yield (Badgley, 2007)”. Agroecology “achieved the same or superior yields”. “Mature agroecological conversions (those in excess of five years old) consistently out produced industrial operations” (pp. 473–474). - UNEP-UNCTAD’s 114 African cases (+116%), Pretty’s 286 projects (+79%) and the UK Foresight 40 projects (2.13-fold). “The results speak for themselves” (p. 474). - “The potential scale-up for agroecological based bottom-up approaches appear to be immense” (p. 474).

Subsequent developments - What Badgley 2007 actually reported. The paper’s own abstract gives the average organic-to-non-organic yield ratio as “slightly <1.0 for studies in the developed world and >1.0 for studies in the developing world”. The comparison was against “conventional or low-intensive” production. The abstract also notes that review raised issues about crop rotations and “the reliability of grey-literature sources” [S51]. - The abstract does not support “same or superior yields” against industrial systems generally. I could not verify the chapter’s “mature conversions … consistently out produced” claim from the sources I could access. - Contemporary critiques (Connor 2008; Connor 2013) disputed the analysis [S52, titles only]. - In 2026 the original authors wrote that later studies “found yield ratios (yield gaps) similar to or somewhat larger than ours”, while maintaining that organic methods could contribute substantially to food supply [S51]. - Later yield-gap meta-analyses. - Seufert et al. (April 2012, before the chapter but not cited): organic yields about 25% lower on average [S53; background]. - Ponisio et al. (2015; 115 studies, over 1,000 observations): organic yields 19.2% (±3.7%) lower. Multi-cropping and crop rotation cut the gap to 9% and 8% [S54]. - Kniss et al. (2016; USDA data for 2014 from over 10,000 organic farms): organic yields averaged 67% of conventional, with wide variation and parity for some crops [S55]. - Meemken & Qaim (2018): organic is “lower yielding on average”, and the gap may widen if more farmers convert [S56]. - Aizen et al. (2026; 199 studies): the gap narrows with pollinator dependence and becomes negligible for moderately to highly pollinator-dependent crops [S60]. - Ben-Ari & Makowski (2026 preprint): differences in yield stability are largely explained by differences in mean yield [S60, preprint]. - Ecosystem services and food security. - Tamburini et al. (2020; second-order meta-analysis of 98 meta-analyses, 5,160 studies): diversification “enhances biodiversity, pollination, pest control, nutrient cycling, soil fertility, and water regulation without compromising crop yields”, with context-dependent trade-offs [S57]. - Bezner Kerr et al. (2021; 55 cases, 1998–2019): 78% of studies found positive food-security or nutrition outcomes from agroecological practices in low- and middle-income countries [S58]. - Pretty et al. (2018): 163 million farms (29% of all) on 453 million ha had “crossed a redesign threshold” for sustainable intensification [S59]. This rests on a broad definition covering IPM, conservation agriculture and irrigation management, and on initiative-reported figures. - Scaling experience. - HLPE (2019) summarised the evidence and highlighted significant knowledge gaps “on: relative yields and performance of agroecological practices compared to other alternatives across contexts”. It concluded that, given underinvestment, “it remains unclear how representative the cases so far documented are” [S23]. - Andhra Pradesh Community-managed Natural Farming, launched 2016, is the largest government-led programme. A cluster-randomised evaluation (BLOOM) was registered in 2021 and its protocol published in 2023 [S62]. I found no published results. - Sri Lanka’s 2021 ban on agrochemical fertiliser imports led to “a notable decrease in production” [S61, from a modelling paper’s background]. The ban was not a participatory agroecological transition of the kind the chapter advocates. But it shows that rapid, top-down input withdrawal is not a proxy for agroecology’s potential.

Complications - The chapter relied on uncontrolled before-and-after project syntheses (UNEP-UNCTAD, Pretty, Foresight). Those gains are measured largely against low-input baselines, and do not compare like with like against industrial systems. HLPE’s caveat applies directly [S23]. - The chapter’s own asymmetric standard (see digest) is thrown into relief. The same authors demand rigorous, independent, adequately powered evidence for GM safety (pp. 468–469), but accept such syntheses as “speak[ing] for themselves” (p. 474).

Verdict: weakened. - Weakened: the headline yield-parity and “mature conversion” claims. Every major later meta-analysis finds an average gap, and the original study’s own authors now accept gaps similar to or larger than theirs. - Held up in a more modest form: - diversification narrows the gap; - diversification raises ecosystem services without yield penalty; - agroecological practices often improve smallholder food security relative to low-input baselines. - Untested: the “immense” scaling potential.

Implications for weight. Give low weight to the chapter’s quantitative yield claims, and moderate weight to its claims about ecosystem services and smallholder food security. The episode is itself a lesson about evidential symmetry. An argument that turns precautionary rigour on one pathway and not on the rival weakens its own credibility.

Claim 8: Top-down innovation “will fail in the long run to produce food security”, “will always serve last” smallholders on plots under 2 ha, and “increases food insecurity”; bottom-up models are “the responsible innovation models” (pp. 470, 476–477)#

Original claim. - The top-down framework “will fail in the long run to produce food security because it does not have the necessary incentives to create resilient and sustainable production systems” (p. 470). - Top-down innovation “undermines the stated national and international goals of poverty reduction, sustainability, and increases food insecurity” (p. 476). - “Top-down providers are invariably attracted to the largest markets … They therefore will always serve last … the presently small and subsistence farmer on < 2 hectare plots” (p. 476). - “Bottom-up innovation models are the responsible innovation models” (p. 476). - Top-down approaches “will most likely fail to deliver on the large promises of food security and alleviation of poverty” (p. 477).

Subsequent developments - Where the tendency held. - Commercial GM portfolios remain concentrated on large-scale commodity crops in the Americas. Central and South America plus North America account for about 88% of GM area [S2]. - Outside South Africa, total GM area in Africa was about 0.12 million ha in 2025 [S2]. - NASEM warned that seed cost and credit constraints may limit adoption by resource-poor smallholders, and that GE crops “are not able by themselves” to address smallholders’ problems [S3]. - Where smallholder GM reached farmers, largely through public channels. - Bangladesh. Bt brinjal was bred into local varieties by the public Bangladesh Agricultural Research Institute. A cluster-randomised trial found [S21]: - yields up 51%; - pesticide costs down 37.5%; - net revenue up 128%; - pesticide toxicity down up to 76%; - fewer pesticide-poisoning symptoms among farmers with pre-existing conditions. But adoption fell from a 2021 peak of 11.9% to 3.4% of brinjal area by 2025 [S2]. - Nigeria and Ghana. Bt cowpea (Nigeria 2021, Ghana 2025) reportedly cut sprays from six or seven to two per season, but seed multiplication limits uptake [S2]. - China and India. In China, more than 77% of GM cotton uses varieties from the public Chinese Academy of Agricultural Sciences. India’s Bt cotton is grown overwhelmingly by smallholders [S2]. - Where outcomes were poor or reversed. - Burkina Faso’s exit (lint quality) [S20]. - South Africa’s Makhathini Flats smallholders: adoption shrank to 10% of initial levels (HLPE citing Schnurr 2012) [S23]. - India’s pest-resistance and rising-pesticide-spend trajectory (contested) [S15, S16, S18]. - Hunger trends (SOFI 2026) [S70]: - hunger affected an estimated 645 million people (7.8%) in 2025, down from 8.6% in 2022; - compared with 2015 the prevalence is slightly lower (8.0% then), but the number is about 50 million higher; - Africa now hosts the largest number of hungry people (309 million, 20.0%), and its long upward trend has only just halted; - between 510 and 520 million people are projected to face hunger in 2030.

Complications - Global hunger trends cannot test the chapter’s causal claim. They are driven by conflict, economic shocks, climate extremes, the pandemic and food prices, and no source I found attributes them to GM adoption or its absence. - The strongest evidence of smallholder benefit from a GM crop came from a publicly developed product delivered through public institutions [S21]. That supports the chapter’s view that incentives and institutions, not the genotype-based tool itself, decide distribution (p. 466). It undercuts the categorical claim that top-down technologies “will always” fail smallholders. - Many African and Asian smallholder GM products were slowed or halted by litigation, regulation and seed-system bottlenecks, not only by provider incentives [S2].

Verdict: partly held up. - Held up: the distributional tendency. Commercial providers still target the largest, most uniform markets, and smallholder-oriented GM remains tiny after thirteen years. - Weakened: the categorical claims (“will always serve last”, “increases food insecurity”). A randomised trial and several public-sector cases show real smallholder gains. - Untestable against the evidence available: the long-run food-security prediction.

Implications for weight. Carry forward the mechanism: providers are attracted to large, uniform, paying markets, so smallholder needs are served late unless public or philanthropic channels exist. Do not carry forward the absolute predictions. The better-supported lesson is that the governance and delivery channel matters as much as the technology’s origin.

Claim 9 (recommendation): Put the IAASTD and EU SCAR 2012 findings into practice; give low-input, high-output agroecological research “the highest priority for funding”; create a public research sector “free from political incentives for top-down innovation” (p. 475)#

Original claim (p. 475). Rebalancing innovation “means operationalising the outcomes from the IAASTD (2009a) and SCAR (2012) reports”. The chapter quotes SCAR: “Approaches that promise building blocks towards low-input high-output systems … should receive the highest priority for funding”. “To achieve this, a public sector free from political incentives for top-down innovation is an essential capacity.”

Subsequent developments - FAO and the UN food-security bodies. - FAO’s Committee on Agriculture supported the “10 Elements of Agroecology” in October 2018. The FAO Council (197 members) approved them in December 2019 “to guide FAO’s vision on agroecology” [S63]. - The HLPE’s 2019 report gave agroecology high-level legitimacy. It also documented how little money follows [S23]: - public investment in agroecological approaches is “estimated at between 1 percent and 1.5 percent of total agricultural and aid budgets”; - UK aid for agroecology is under 5% of agricultural aid; - US research on diversified systems is under 2% of public agricultural research funding; - about 8% of FAO’s 2018–2019 work contributes to agroecological transitions. - Donors. The NGO-produced Money Flows study (IPES-Food/Biovision, 2020) found that 85% of Gates Foundation projects and over 70% of Kenyan research-institute projects targeted industrial agriculture or productivity-and-efficiency approaches. Only 3% of Gates projects included agroecological redesign [S69]. - CGIAR. CGIAR ran an Agroecology Initiative in its 2022–2024 portfolio. The page now redirects to a “Multifunctional Landscapes” science programme using “Agroecology+ solution bundles” for 2025–2030 [S68]. (I could not retrieve budget figures.) - EU. - Farm to Fork (20 May 2020) promised a “dedicated partnership on agro-ecology living laboratories”. It also made agroecology eligible for CAP eco-schemes, and set targets of at least 25% organic land and a 50% cut in the use and risk of chemical pesticides by 2030 [S64]. - The Horizon Europe AGROECOLOGY Partnership launched its network of living labs and research infrastructures in June 2024 (79 members by 2026, planned to run to 2033), with co-funded calls in 2024, 2025 and 2026 [S67]. - The proposed Sustainable Use Regulation for pesticides was rejected in the Parliament plenary on 22 November 2023 and withdrawn by the Commission on 6 May 2024 [S65]. - The Commission’s Vision for Agriculture and Food (19 February 2025) [S66]: - does not use the word “agroecology”; - promotes “new genomic techniques to produce more climate resilient crops”; - says the Commission “will carefully consider any further ban of pesticides if alternatives are not yet available”; - says “continuous support for organic farming remains essential”. - The NGT Regulation (2026) creates a lighter regime for most gene-edited plants [S12].

Complications - The chapter’s recommendation was framed as an either/or rebalancing. The institutions that took it up (FAO, HLPE, the EU partnership) generally adopted a both/and framing: agroecology alongside new breeding technologies. - Whether a public research sector “free from political incentives for top-down innovation” is achievable was not tested. The EU’s 2024–2026 shift shows that research programmes and political strategy can diverge.

Verdict: partly held up. The recommendation gained real institutional footholds: FAO’s endorsement (2019), an HLPE report (2019), an EU research partnership (2024) and a CGIAR programme. But funding stayed marginal on every measure I found, and the EU’s political commitments were partly reversed in 2023–2025. As a diagnosis of chronic underinvestment, the chapter’s premise has been reinforced by HLPE’s own figures.

Implications for weight. Treat the recommendation as a durable agenda that institutions adopt rhetorically faster than they fund it, and which is vulnerable to political cycles. The lesson that “alternatives whose proceeds flow to adopters rather than providers lack political champions” (p. 476) gains support from how the EU trajectory unfolded.

Claim 10: GM adoption is patchy because high-income consumers reject GM, because rich-country subsidies concentrate GM in subsidised commodity crops (OECD support costing developing countries about USD 17 billion a year), and because seed and input costs exclude poor farmers (p. 464)#

Original claim (p. 464). “First, significant markets of high-income consumers have rejected GM”. Second, rich-country subsidies undermine developing-country markets (“about USD 17 billion per year”). Third, “the high rent of GM seeds and associated management inputs … confines these tools … to countries that redistribute wealth to farming for export”.

Subsequent developments - EU cultivation and opt-outs. - Directive (EU) 2015/412 let member states restrict or ban cultivation on their territory [S72]. - Nineteen member states (some for part of their territory) asked for MON 810 exclusions in 2015. The Commission’s Decision (EU) 2016/321 (3 March 2016) adjusted the authorisation accordingly [S71]. - Europe’s GM area fell to 70,653 ha in 2025, almost all Bt maize in Spain [S2]. - Imports and politics. The EU continued to authorise GM food and feed imports. Parliament passed dozens of objections, and member states repeatedly failed to reach a qualified majority [S38] (Claim 4). - Consumer attitudes. - In the 2019 Special Eurobarometer, 27% of EU respondents who had heard of at least one food-safety topic named “genetically modified ingredients in food or drinks” among their main concerns. It ranked eighth, behind antibiotic/hormone residues (44%), pesticide residues (39%), environmental pollutants (37%) and additives (36%) [S73]. - The report cautions that its 2010 predecessor (66% “very or fairly worried” about GMOs) is not directly comparable [S73]. - Seed and input costs. NASEM: “The cost of GE seed may limit the adoption of GE crops by resource-poor smallholders.” The seed-price difference is usually a small fraction of total costs, but credit constraints and up-front risk matter [S3]. High seed prices also appear in HLPE’s account of Burkina Faso [S23]. - Rival explanations. Evidence since 2013 points to additional drivers: - long regulatory timelines (industry-reported 16.5 years to market) [S4]; - litigation and court orders (Philippines, Kenya) [S2]; - weak IP enforcement deterring trait introduction (Pakistan, Argentina) [S2]; - quality mismatches (Burkina Faso) [S20]. - In India, illegal cultivation of unapproved GM crops (Bt brinjal, HT soybean) is reported [S2]. Demand exists in places where approval does not. - Subsidies. I could not retrieve OECD’s current support estimates (the site refused automated access), so the USD 17 billion figure is not re-checked. [Background, not re-verified: Brazil and Argentina, now holding about 44% of world GM area [S2], have low or negative producer support in OECD estimates. That is hard to square with subsidies as a main driver of where GM is grown.]

Complications - The chapter treats high-income consumer rejection as the first reason for patchy adoption. EU cultivation did collapse (opt-outs, Spain-only cultivation), but the EU imports large volumes of GM feed. The rejection operates mainly through cultivation politics and food labelling, not through feed markets.

Verdict: partly held up. Consumer and political rejection in Europe did shape cultivation, and seed and credit costs remain a recognised barrier for poor farmers. But the explanation is incomplete. Regulatory and legal friction, IP enforcement conditions, product-quality mismatches and seed-system capacity have all visibly shaped adoption since 2013. The subsidy mechanism is not borne out by the geography of the largest post-2013 growth (Brazil, Argentina, China).

Implications for weight. Use the chapter’s list as a partial account. The lesson for a lens is that adoption patterns rarely have a single cause. Regulatory friction, legal contestation, market and trade structure, and the provider’s own targeting all interact, and a one-factor story, whether “consumers rejected it” or “regulation blocked it”, should be treated with suspicion.

Summary of verdicts#

Technology-neutral lessons this check supports (for later use as a lens)#

Each is tied to the section’s pages and to the later evidence above.

1. Single-tactic control of an adaptive system produces a treadmill, and stacking tactics buys time rather than escape. Resistance to each successive layer (glyphosate, then auxins, then glufosinate) appeared within a few years of its large-scale deployment (p. 462; [S5–S7]).
2. Externalities can compel adoption. When one user’s use of a product damages non-users, non-users may adopt defensively. This produces lock-in and near-monopoly that market-choice framings miss, and which regulators may fail to acknowledge until courts intervene (pp. 462, 477; [S7]). This mechanism goes beyond what the chapter described.
3. Broad early promises narrow to what can be packaged and sold in the largest markets, and the narrowing persists for decades. Headline scale depends heavily on the chosen denominator (p. 463; [S1–S3]).
4. Consolidation among a few integrated R&D players reduces innovation competition upstream. Competition authorities can recognise this and still approve with remedies, so the feedback loop can operate through lawful, reviewed transactions (pp. 467, 477; [S40–S41]).
5. Developer-controlled evidence carries a measurable funding effect, and integrity failures can surface decades later. Institutional fixes arrived about six to eight years after the warning: study registration, proactive disclosure, verification studies and judicial limits on deference to applicant data (pp. 468–469; [S29–S31, S36]).
6. Whether the evidence process is trustworthy and whether the product is harmful are separable questions. Independent long-term replication can settle the second while process reforms address the first. Early “indications” from single contested studies are a weak basis for a late-lesson claim (pp. 468–470; [S24–S27]).
7. Benefits measured early can erode as target organisms adapt and input systems shift. Time horizon matters as much as sign when appraising claimed benefits (pp. 463, 468; [S9, S15, S17, S18]).
8. Who develops and delivers a technology shapes its distribution as much as what it is. Publicly developed products delivered through public institutions reached smallholders with measurable gains where commercial offerings did not (pp. 466, 476; [S2, S21]).
9. Complex, multi-factor performance goals favour whole-system and whole-genome approaches over single-component fixes. Those approaches often reach users sooner through public breeding and testing networks (pp. 474–475; [S45–S50]).
10. Commitments to alternatives whose benefits flow to adopters rather than providers are adopted rhetorically faster than they are funded, and are politically reversible (pp. 475–476; [S23, S64–S69]).
11. Asymmetric evidential standards undermine an argument’s credibility. Claims for the preferred alternative made on uncontrolled syntheses did not survive later meta-analysis. The precautionary rigour demanded of one pathway should be applied to its rival (pp. 468–469 vs 473–474; [S51, S53–S56]).
12. Adoption patterns rarely have a single cause. Consumer politics, regulatory time, litigation, IP enforcement, product-quality fit and seed-system capacity all interact (p. 464; [S2, S4, S20, S71–S73]).

Sources#

All retrieved 25–26 September 2026 unless noted. EU legal texts were read in their Official Journal form through the Publications Office Cellar (https://publications.europa.eu/resource/celex/<CELEX>); ELI/EUR-Lex addresses are given for readers.

GM crop area and traits - [S1] ISAAA. Brief 55: Global Status of Commercialized Biotech/GM Crops in 2019 (executive summary). ISAAA, 2020 (industry-funded). https://www.isaaa.org/resources/publications/briefs/55/executivesummary/default.asp - [S2] AgbioInvestor. Global GM Crop Area Review 2025. May 2026 (45 pp.; downloaded from the GM Monitor downloads page). https://gm.agbioinvestor.com/downloads ; summary at https://gm.agbioinvestor.com/ - [S3] National Academies of Sciences, Engineering, and Medicine. Genetically Engineered Crops: Experiences and Prospects. Report in Brief, May 2016. https://nap.nationalacademies.org/resource/23395/GE-crops-report-brief.pdf ; full report https://nap.nationalacademies.org/catalog/23395 - [S4] AgbioInvestor. GM Trait Study (April 2022 survey of BASF, Bayer, Corteva and Syngenta on the cost and time to commercialise a new GM trait; industry-sourced). https://gm.agbioinvestor.com/gm-trait-study

Herbicide resistance, drift and herbicide use - [S5] International Herbicide-Resistant Weed Database. “Weeds resistant to Inhibition of Enolpyruvyl Shikimate Phosphate Synthase, HRAC Group 9 (Legacy G)” (glyphosate). Accessed 25 September 2026; counts are mine from the downloaded table. https://www.weedscience.org/Pages/MOA.aspx?MOAID=12 - [S6] International Herbicide-Resistant Weed Database. Auxin mimics, HRAC Group 4: https://www.weedscience.org/Pages/MOA.aspx?MOAID=24 ; glutamine synthetase inhibitors (glufosinate), HRAC Group 10: https://www.weedscience.org/Pages/MOA.aspx?MOAID=13 (accessed 25 September 2026). - [S7] National Family Farm Coalition v. U.S. EPA, No. 19-70115 (9th Cir. 3 June 2020). https://cdn.ca9.uscourts.gov/datastore/opinions/2020/06/03/19-70115.pdf - [S8] US EPA. “Registration of Dicamba for Use on Dicamba-Tolerant Crops” (approval for 2026 and 2027 seasons; last updated 6 February 2026). https://www.epa.gov/ingredients-used-pesticide-products/registration-dicamba-use-dicamba-tolerant-crops - [S9] Perry, E.D., Ciliberto, F., Hennessy, D.A., Moschini, G. “Genetically engineered crops and pesticide use in U.S. maize and soybeans.” Science Advances 2(8), e1600850 (August 2016) (abstract). https://doi.org/10.1126/sciadv.1600850 - [S10] Benbrook, C.M. “Trends in glyphosate herbicide use in the United States and globally.” Environmental Sciences Europe 28, 3 (2016) (abstract). https://doi.org/10.1186/s12302-016-0070-0 - [S11] Schulz, R. et al. “Applied pesticide toxicity shifts toward plants and invertebrates, even in GM crops.” Science 372, 81–84 (April 2021) (abstract). https://doi.org/10.1126/science.abe1148 - [S12] Regulation (EU) 2026/1388 of the European Parliament and of the Council of 17 June 2026 on plants obtained by certain new genomic techniques and their products. OJ L, 2026/1388, 26.6.2026 (recitals 19, 20 on herbicide tolerance, the patent recitals and the patent-information articles; applies from 17 July 2028). http://data.europa.eu/eli/reg/2026/1388/oj (via CELEX 32026R1388); procedure file https://oeil.europarl.europa.eu/oeil/en/procedure-file?reference=2023/0226(COD)

Agronomic and economic outcomes - [S13] Klümper, W., Qaim, M. “A meta-analysis of the impacts of genetically modified crops.” PLOS ONE 9(11), e111629 (November 2014) (abstract). https://doi.org/10.1371/journal.pone.0111629 - [S14] Pellegrino, E., Bedini, S., Nuti, M., Ercoli, L. “Impact of genetically engineered maize on agronomic, environmental and toxicological traits: a meta-analysis of 21 years of field data.” Scientific Reports 8, 3113 (February 2018) (abstract). https://doi.org/10.1038/s41598-018-21284-2 - [S15] Kranthi, K.R., Stone, G.D. “Long-term impacts of Bt cotton in India.” Nature Plants 6, 188–196 (March 2020) (Perspective; abstract). https://doi.org/10.1038/s41477-020-0615-5 - [S16] Qaim, M. “Bt cotton, yields and farmers’ benefits.” Nature Plants 6 (November 2020), https://doi.org/10.1038/s41477-020-00788-8 ; and Kranthi, K.R., Stone, G.D. “Reply to: Bt cotton, yields and farmers’ benefits.” Nature Plants 6 (November 2020), https://doi.org/10.1038/s41477-020-00790-0 (metadata only). - [S17] Tabashnik, B.E., Fabrick, J.A., Carrière, Y. “Global patterns of insect resistance to transgenic Bt crops: the first 25 years.” Journal of Economic Entomology 116(2), 297–309 (April 2023) (abstract). https://doi.org/10.1093/jee/toac183 - [S18] Pink bollworm resistance in India: Madhu, T.N. et al. “Insecticide resistance in pink bollworm … across cotton growing regions of India.” Journal of Applied Toxicology (May 2026), https://doi.org/10.1002/jat.70241 ; Srilekha, K. et al. “Status of insecticide resistance and detoxification enzymes activity in pink bollworm … populations from India.” Ecotoxicology (August 2026), https://doi.org/10.1007/s10646-026-03138-0 (abstracts). - [S19] Pakistan: Ahmad, S. et al. “Resistance status of Helicoverpa armigera against Bt cotton in Pakistan.” Transgenic Research 28, 199–212 (April 2019), https://doi.org/10.1007/s11248-019-00114-9 ; Jamil, S. et al. “The level of Cry1Ac endotoxin and its efficacy against H. armigera in Bt cotton at large scale in Pakistan.” GM Crops & Food 12(1) (2021), https://doi.org/10.1080/21645698.2020.1799644 (abstracts). - [S20] Dowd-Uribe, B., Schnurr, M.A. “Briefing: Burkina Faso’s reversal on genetically modified cotton and the implications for Africa.” African Affairs 115(458), 161–172 (January 2016) (abstract via OpenAlex). https://doi.org/10.1093/afraf/adv063 - [S21] Ahmed, A.U., Hoddinott, J., Abedin, N., Hossain, N. “The impacts of GM foods: results from a randomized controlled trial of Bt eggplant in Bangladesh.” American Journal of Agricultural Economics (online 13 November 2020) (abstract via Crossref). https://doi.org/10.1111/ajae.12162 - [S22] Heinemann, J.A., Massaro, M., Coray, D.S., Agapito-Tenfen, S.Z., Wen, J.D. “Sustainability and innovation in staple crop production in the US Midwest.” International Journal of Agricultural Sustainability 12(1), 71–88 (online 14 June 2013; issue 2014) (chapter author’s own later work; abstract). https://doi.org/10.1080/14735903.2013.806408 - [S23] HLPE. Agroecological and other innovative approaches for sustainable agriculture and food systems that enhance food security and nutrition. Report 14 of the High Level Panel of Experts on Food Security and Nutrition of the Committee on World Food Security. Rome, July 2019 (summary para. 10; section 1.4.3 on yields; section 1.4.5 on knowledge gaps and funding; Box 14 on Bt cotton). https://www.fao.org/3/ca5602en/ca5602en.pdf

Health evidence and risk-assessment institutions - [S24] Séralini, G.-E. et al. “Long term toxicity of a Roundup herbicide and a Roundup-tolerant genetically modified maize.” Food and Chemical Toxicology 50, 4221–4231 (2012), https://doi.org/10.1016/j.fct.2012.08.005 ; Retraction notice, Food and Chemical Toxicology 63, 244 (January 2014; announced 28 November 2013), https://doi.org/10.1016/j.fct.2013.11.047 (metadata only; the notice’s reasons are as reported in secondary sources); republished as “Republished study: long-term toxicity of a Roundup herbicide and a Roundup-tolerant genetically modified maize.” Environmental Sciences Europe 26, 14 (June 2014), https://doi.org/10.1186/s12302-014-0014-5 - [S25] Portier, C.J., Goldman, L.R., Goldstein, B.D. “Inconclusive findings: now you see them, now you don’t!” Environmental Health Perspectives 122(2), A36 (1 February 2014) (metadata only). https://doi.org/10.1289/ehp.1408106 - [S26] Steinberg, P. et al. “Lack of adverse effects in subchronic and chronic toxicity/carcinogenicity studies on the glyphosate-resistant genetically modified maize NK603 in Wistar Han RCC rats.” Archives of Toxicology 93, 1095–1139 (April 2019) (G-TwYST; abstract). https://doi.org/10.1007/s00204-019-02400-1 - [S27] Coumoul, X. et al. “The GMO90+ project: absence of evidence for biologically meaningful effects of genetically modified maize-based diets on Wistar rats after 6-months feeding comparative trial.” Toxicological Sciences 168(2), 315–338 (April 2019) (abstract). https://doi.org/10.1093/toxsci/kfy298 - [S28] Commission Implementing Regulation (EU) No 503/2013 of 3 April 2013 on applications for authorisation of genetically modified food and feed. OJ L 157, 8.6.2013, p. 1 (recitals 10–12; Art. 12 review clause; Annex II section 1.4.4.1). Publications Office catalogue lists only the original consolidated version (02013R0503-20130608) and a corrigendum. http://data.europa.eu/eli/reg_impl/2013/503/oj (via CELEX 32013R0503) - [S29] Regulation (EU) 2019/1381 of the European Parliament and of the Council of 20 June 2019 on the transparency and sustainability of the EU risk assessment in the food chain. OJ L 231, 6.9.2019, p. 1 (recitals 66–68, 74; Arts. 32a–32d; applies from 27 March 2021). http://data.europa.eu/eli/reg/2019/1381/oj (via CELEX 32019R1381) - [S30] Court of Justice (Grand Chamber), Case C-616/17, Criminal proceedings against Mathieu Blaise and Others, judgment of 1 October 2019, ECLI:EU:C:2019:800. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:62017CJ0616 (via Cellar) - [S31] Guillemaud, T., Lombaert, E., Bourguet, D. “Conflicts of interest in GM Bt crop efficacy and durability studies.” PLOS ONE 11(12), e0167777 (December 2016) (abstract). https://doi.org/10.1371/journal.pone.0167777 - [S32] Hilbeck, A., Binimelis, R., Defarge, N., Steinbrecher, R., Székács, A., Wickson, F., Antoniou, M., Bereano, P.L., Clark, E.A., Hansen, M., Novotny, E., Heinemann, J., Meyer, H., Shiva, V., Wynne, B. “No scientific consensus on GMO safety.” Environmental Sciences Europe 27, 4 (24 January 2015) (metadata only; a chapter author is a co-signatory). https://doi.org/10.1186/s12302-014-0034-1

Glyphosate - [S33] Guyton, K.Z. et al. (IARC Monograph Working Group). “Carcinogenicity of tetrachlorvinphos, parathion, malathion, diazinon, and glyphosate.” Lancet Oncology 16(5), 490–491 (May 2015; online 20 March 2015) (metadata). https://doi.org/10.1016/S1470-2045(15)70134-8 ; IARC Monographs Vol. 112 (not retrieved), https://publications.iarc.who.int/549 - [S34] Tarazona, J.V. et al. “Glyphosate toxicity and carcinogenicity: a review of the scientific basis of the European Union assessment and its differences with IARC.” Archives of Toxicology 91, 2723–2743 (August 2017), https://doi.org/10.1007/s00204-017-1962-5 ; Portier, C.J. et al. “Differences in the carcinogenic evaluation of glyphosate between IARC and EFSA.” Journal of Epidemiology and Community Health 70, 741–745 (2016), https://doi.org/10.1136/jech-2015-207005 (abstract/metadata). - [S35] Commission Implementing Regulation (EU) 2023/2660 of 28 November 2023 renewing the approval of the active substance glyphosate. OJ L, 2023/2660, 29.11.2023 (recitals on the 6 July 2023 EFSA conclusion, the 30 May 2022 ECHA RAC opinion, indirect biodiversity effects and the committee votes). http://data.europa.eu/eli/reg_impl/2023/2660/oj (via CELEX 32023R2660) - [S36] Retraction of Williams, G.M., Kroes, R., Munro, I.C. “Safety evaluation and risk assessment of the herbicide Roundup and its active ingredient, glyphosate, for humans.” Regulatory Toxicology and Pharmacology 31, 117–165 (2000), https://doi.org/10.1006/rtph.1999.1371 ; retraction notice https://doi.org/10.1016/j.yrtph.2025.106006 (February 2026 issue; Retraction Watch database date 5 December 2025; metadata only). Reasons as reported by Retraction Watch, “Glyphosate safety article retracted eight years after Monsanto ghostwriting revealed in court” (4 December 2025), https://retractionwatch.com/2025/12/04/glyphosate-safety-article-retracted-elsevier-monsanto-ghostwriting/ ; and “Publisher investigating two more papers on glyphosate safety over ghostwriting claims” (15 June 2026), https://retractionwatch.com/2026/06/15/glyphosate-roundup-safety-ghostwriting-claims-critical-reviews-toxicology/ (journalism; secondary). - [S37] Supreme Court of the United States. Docket No. 24-1068, Monsanto Co. v. Durnell (certiorari granted 16 January 2026; argued 27 April 2026; judgment reversed and remanded 25 June 2026, Kavanaugh J. for the Court; Jackson J. dissenting, joined by Gorsuch J.) (docket only; opinion not read). https://www.supremecourt.gov/docket/docketfiles/html/public/24-1068.html - [S38] European Parliament resolutions on draft Commission decisions authorising GM food and feed: 63 resolutions 2015–2026 listed in the Publications Office catalogue (SPARQL query, https://publications.europa.eu/webapi/rdf/sparql). Example: European Parliament resolution of 16 December 2025 on genetically modified sugar beet KWS20-1, P10_TA(2025)0328, OJ C/2026/2151, 6.5.2026 (CELEX 52025IP0328). http://data.europa.eu/eli/C/2026/2151/oj

Seed industry and intellectual property - [S39] Bowman v. Monsanto Co., 569 U.S. 278 (13 May 2013) (syllabus). https://www.law.cornell.edu/supremecourt/text/11-796 - [S40] Summary of Commission Decision of 27 March 2017 declaring a concentration compatible with the internal market (Case M.7932 — Dow/DuPont), C(2017) 1946 (paras 52–58 on innovation competition). CELEX 52017M7932(02). https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52017M7932(02) - [S41] Summary of Commission Decision of 21 March 2018 declaring a concentration compatible with the internal market (Case M.8084 — Bayer/Monsanto), C(2018) 1709 (competition concerns; BASF divestment package; vegetable seeds divestment). CELEX 52018M8084(02). https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52018M8084(02) - [S42] Summary of Commission decision of 5 April 2017 (Case M.7962 — ChemChina/Syngenta), C(2017) 2167. CELEX 52017XC0610(01) (title and date only). https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52017XC0610(01) - [S43] Clapp, J. “The problem with growing corporate concentration and power in the global food system.” Nature Food 2, 404–408 (June 2021) (abstract). https://doi.org/10.1038/s43016-021-00297-7 - [S44] European Patent Office, Enlarged Board of Appeal. Opinion G 3/19 (Pepper), 14 May 2020. https://www.epo.org/en/boards-of-appeal/decisions/g190003ex1

Drought tolerance - [S45] Nemali, K.S. et al. “Physiological responses related to increased grain yield under drought in the first biotechnology-derived drought-tolerant maize.” Plant, Cell & Environment 38(9), 1866–1880 (September 2015) (abstract). https://doi.org/10.1111/pce.12446 - [S46] Obunyali, C.O. et al. “Efficacy of Event MON 87460 in drought-tolerant maize hybrids under optimal and managed drought-stress in eastern and southern Africa.” Journal of Genetic Engineering and Biotechnology 22, 100352 (March 2024) (abstract; authors include TELA/WEMA partnership staff). https://doi.org/10.1016/j.jgeb.2024.100352 - [S47] González, F.G. et al. “Field-grown transgenic wheat expressing the sunflower gene HaHB4 significantly outyields the wild type.” Journal of Experimental Botany 70(5), 1669–1681 (March 2019) (abstract). https://doi.org/10.1093/jxb/erz037 - [S48] CIMMYT. “Drought Tolerant Maize for Africa (DTMA)” project page (2006–2015). https://www.cimmyt.org/projects/drought-tolerant-maize-for-africa-dtma/ - [S49] Prasanna, B.M. et al. “Beat the stress: breeding for climate resilience in maize for the tropical rainfed environments.” Theoretical and Applied Genetics 134, 1729–1752 (June 2021) (abstract). https://doi.org/10.1007/s00122-021-03773-7 - [S50] Drought-tolerant maize adoption and impact (abstracts): Simtowe, F. et al. Land Use Policy 88, 104091 (November 2019), https://doi.org/10.1016/j.landusepol.2019.104091 ; Amondo, E. et al. International Journal of Climate Change Strategies and Management 11(4) (2019), https://doi.org/10.1108/IJCCSM-03-2018-0024 ; Gebre, G.G. et al. Food and Energy Security 10(4), e313 (November 2021), https://doi.org/10.1002/fes3.313 ; Setimela, P.S. et al. Agronomy Journal 109(2) (March 2017), https://doi.org/10.2134/agronj2015.0540 ; Lunduka, R.W. et al. Climate and Development 11(1) (2019), https://doi.org/10.1080/17565529.2017.1372269

Agroecology and organic yields - [S51] Badgley, C. et al. “Organic agriculture and the global food supply.” Renewable Agriculture and Food Systems 22(2), 86–108 (June 2007) (abstract), https://doi.org/10.1017/S1742170507001640 ; Badgley, C., Perfecto, I., Moghtader, J. “Revisiting organic agriculture and the global food supply, twenty years on.” Renewable Agriculture and Food Systems 41 (2026) (abstract), https://doi.org/10.1017/S1742170526100416 - [S52] Connor, D.J. “Organic agriculture cannot feed the world.” Field Crops Research 106(2), 187–190 (March 2008), https://doi.org/10.1016/j.fcr.2007.11.010 ; Connor, D.J. “Organically grown crops do not a cropping system make and nor can organic agriculture nearly feed the world.” Field Crops Research 144, 145–147 (March 2013), https://doi.org/10.1016/j.fcr.2012.12.013 (titles and metadata only). - [S53] Seufert, V., Ramankutty, N., Foley, J.A. “Comparing the yields of organic and conventional agriculture.” Nature 485, 229–232 (online 25 April 2012) (pre-publication background; the 25% figure is from background knowledge, not retrieved). https://doi.org/10.1038/nature11069 - [S54] Ponisio, L.C. et al. “Diversification practices reduce organic to conventional yield gap.” Proceedings of the Royal Society B 282, 20141396 (January 2015) (abstract). https://doi.org/10.1098/rspb.2014.1396 - [S55] Kniss, A.R., Savage, S.D., Jabbour, R. “Commercial crop yields reveal strengths and weaknesses for organic agriculture in the United States.” PLOS ONE 11(8), e0161673 (2016; corrected) (abstract). https://doi.org/10.1371/journal.pone.0161673 - [S56] Meemken, E.-M., Qaim, M. “Organic agriculture, food security, and the environment.” Annual Review of Resource Economics 10, 39–63 (October 2018) (abstract). https://doi.org/10.1146/annurev-resource-100517-023252 - [S57] Tamburini, G. et al. “Agricultural diversification promotes multiple ecosystem services without compromising yield.” Science Advances 6(45), eaba1715 (November 2020) (abstract). https://doi.org/10.1126/sciadv.aba1715 - [S58] Bezner Kerr, R. et al. “Can agroecology improve food security and nutrition? A review.” Global Food Security 29, 100540 (June 2021) (abstract via Semantic Scholar). https://doi.org/10.1016/j.gfs.2021.100540 - [S59] Pretty, J. et al. “Global assessment of agricultural system redesign for sustainable intensification.” Nature Sustainability 1, 441–446 (14 August 2018) (accepted manuscript, White Rose Research Online, https://eprints.whiterose.ac.uk/133144/). https://doi.org/10.1038/s41893-018-0114-0 - [S60] Aizen, N., Sáez, A., Morales, C.L., Aizen, M.A. “The conventional-to-organic yield gap diminishes with increasing crop pollinator dependence.” Proceedings of the Royal Society B 293 (28 January 2026), https://doi.org/10.1098/rspb.2025.2553 ; Ben-Ari, T., Makowski, D. “Mean yield gap explains differences in yield stability between organic and conventional systems.” Preprint, 8 June 2026, https://doi.org/10.64898/2026.06.04.730123 (abstracts; the second is not peer-reviewed). - [S61] Rathnayake, H., Mizunoya, T. “Assessing the global warming potential impact of organic fertilizer strategies in rice cultivation in Sri Lanka.” Environmental Science and Pollution Research (July 2024) (abstract; used only for its background statement on the 2021 import ban). https://doi.org/10.1007/s11356-024-34348-w - [S62] Jaacks, L.M. et al. “Co-Benefits of Largescale Organic farming On huMan health (BLOOM): protocol for a cluster-randomised controlled evaluation of the Andhra Pradesh Community-managed Natural Farming programme in India.” PLOS ONE 18(3), e0281677 (2023) (abstract). https://doi.org/10.1371/journal.pone.0281677

Policy uptake of agroecology - [S63] FAO. “The 10 Elements of Agroecology” (COAG support October 2018; FAO Council approval 2–6 December 2019). http://www.fao.org/agroecology/overview/the-10-elements-of-agroecology/en - [S64] European Commission. A Farm to Fork Strategy for a fair, healthy and environmentally-friendly food system. COM(2020) 381 final, 20 May 2020. Via https://publications.europa.eu/resource/celex/52020DC0381 ; https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52020DC0381 - [S65] European Parliament Legislative Observatory. Procedure file 2022/0196(COD), Sustainable use of plant protection products (plenary vote 22 November 2023; proposal withdrawn by the Commission 6 May 2024; “Procedure lapsed or withdrawn”). https://oeil.europarl.europa.eu/oeil/en/procedure-file?reference=2022/0196(COD) - [S66] European Commission. A Vision for Agriculture and Food: Shaping together an attractive farming and agri-food sector for future generations. COM(2025) 75 final, 19 February 2025 (full text searched; no occurrence of “agroecology”). Via https://publications.europa.eu/resource/celex/52025DC0075 - [S67] European Partnership AGROECOLOGY (Horizon Europe). “About the Network” (network launched June 2024; 79 members after the 2026 wave; planned to 2033) and call pages (2024, 2025, 2026). https://agroecologypartnership.eu/en/about-the-network ; https://www.agroecologypartnership.eu/ - [S68] CGIAR. Initiative: Agroecology (page now presenting the Multifunctional Landscapes Science Program, 2025–2030). https://www.cgiar.org/initiative/agroecology/ - [S69] IPES-Food and Biovision. Money Flows: What is holding back investment in agroecological research for Africa? June 2020 (NGO/advocacy). https://ipes-food.org/wp-content/uploads/2024/03/Money-Flows_Full-report.pdf

Hunger - [S70] FAO, IFAD, UNICEF, WFP and WHO. The State of Food Security and Nutrition in the World 2026 – Understanding and addressing the high cost of a healthy diet. Rome, 2026 (core messages; section 2.1). https://doi.org/10.4060/cd8306en ; https://www.fao.org/3/cd8306en/online/state-food-security-and-nutrition-2026/ending-hunger-food-security.html

EU cultivation and public opinion - [S71] Commission Implementing Decision (EU) 2016/321 of 3 March 2016 adjusting the geographical scope of the authorisation for cultivation of genetically modified maize MON 810. OJ L 60, 5.3.2016, p. 90 (nineteen member states’ demands, 3 July–2 October 2015). http://data.europa.eu/eli/dec_impl/2016/321/oj (via CELEX 32016D0321) - [S72] Directive (EU) 2015/412 of the European Parliament and of the Council of 11 March 2015 amending Directive 2001/18/EC as regards the possibility for the Member States to restrict or prohibit the cultivation of GMOs in their territory. OJ L 68, 13.3.2015, p. 1 (as cited in [S71]). http://data.europa.eu/eli/dir/2015/412/oj - [S73] European Commission / EFSA. Special Eurobarometer Wave EB91.3: Food safety in the EU. Report, April 2019 (QD4T concern rankings; footnote 19 on non-comparability with 2010). https://www.efsa.europa.eu/sites/default/files/corporate_publications/files/Eurobarometer2019_Food-safety-in-the-EU_Full-report.pdf