THE SYSTEM OF RICE INTENSIFICATION: MEMOIRES OF AN INNOVATION

A question that philosophers have debated for millennia has some relevance here: If a tree in the forest falls and nobody hears it, did it fall? An innovation needs to be heard of and seen for it to have any meaningful existence. For an innovation to spread, various audiences need to hear and know about it. Whether people will accept it and use it depends on many considerations, such as accessibility, net benefits, and congruence with social relationships and norms. But certainly, communication is a precondition for any innovation to proceed.

​

Farmers are ultimately the most important recipients of knowledge about new and better ways to improve their production. But they number in the millions and are not easily accessed. There is a network, or perhaps better said a web, of connections and channels through which information flows (or does not) that can bring useful knowledge to farmers. In addition to farmers, this web includes scientists and researchers, a great variety of other professionals, the general public, business enterprises and investors, and policy-makers, to list some of the main groupings.

​

The channels for transmitting this information, apart from word of mouth, fall roughly into two categories that each influence the other, although not necessarily directly. There are first professional journals, often published by scientific societies or associations. These disseminate knowledge that has been vetted by these organizations to reflect what professional scientists with their respective disciplinary specializations consider valid knowledge: the state of the art, the most reliable knowledge that is available. This channel for communicating about ‘trees falling’ is the subject of this chapter. The second category includes a great variety of media, particularly the print and electronic media that reach the general public or certain segments thereof. This is the focus of the following chapter.

​

In recent decades there has been proliferation within both categories. This is most evident for the media as diverse electronic and digital forms of communication have claimed a place alongside the print media. This makes the two categories overlap to some extent as electronic publication has affected scientific publishing, with open-access journals being a positive development for scientific communication, and with ‘predatory publishing’ that has also sprung up being decidedly negative.[1] Electronic transmission of published material bridges the two domains.

​

A key consideration is the extent to which the time-honored process of ‘peer review’ is well-followed with the intent of providing knowledge to fellow scientists and to the public that is as correct and as precise and defensible as possible. Historically, scientific truth has been something approached asymptotically, getting closer and closer to a true understanding, but probably never arriving at truth in any absolute sense. The scientific enterprise involves the construction of understandings (paradigms) that serve us well until some better evidence and formulation of ideas improve upon them.

​

Reading a textbook in any discipline published 50 years ago, or even 20, will make this clear. From time to time, there are paradigm shifts that put the accumulation of knowledge and the direction of further accumulation on a somewhat different track. Paradigm change is generally much more difficult to get accepted than incremental changes in the body of knowledge. As suggested already, SRI experience and research have challenged some basic concepts and precepts for growing rice and other agricultural crops, representing a paradigm shift, calling more attention to the performance of roots and the contributions made by the life in the soil.

​

The story of SRI shines some light on the process by which scientific knowledge gets advanced and disseminated (or not), revealing some shortcomings in the idealized version of scientific endeavor. Of course, shortcomings permeate all human endeavors, not just science. In this chapter, there are examples of how imperfections in the peer-review process have disserved both the understanding and acceptance of SRI, and indeed the scientific enterprise.

​

Peer review has in principle meritorious effects, like not approving research findings that are poorly substantiated and delaying the publication of presentations until they are as coherent as they can be. The issue is not whether there should be peer review, but how it can serve better the purposes for which it is undertaken. In an era of rapid change and enormous challenges, having a scientific review process that is rigorous but not too rigid is more important than it was a century ago. Human faults cannot be eliminated, but there can and should be efforts to minimize the distortions that they can create. This is a concern that extends well beyond SRI.

 

PROFESSIONAL PUBLISHING

​

Long before the contemporary invention of ‘impact factors’ that are calculated to gauge the influence and significance of particular journals and articles, there was usually some consensus within a discipline on which journals were the most reputable and reliable: the ones that had the longest record of publishing cutting-edge articles, the journals that were considered most desirable to get one’s research published in, and which therefore attracted the best articles.

​

In recent years there has been a flood of new, mostly less-highly-regarded journals. In part this reflects the burgeoning of scientific research, not just a greater number of persons who for their own career advancement need to publish research. This kind of ranking makes it all the more important that the more prestigious journals uphold the idealized canons of science. In the SRI story, there is a cautionary tale to be told.

​

There is a need for gatekeeping of some kind, for some persons to be making judgments on behalf of the community of scholars about what research findings and which reports will get published in the most respected journals. These days with a proliferation of journals that range down to the purely mercenary, almost anything can get published somehow somewhere, and once published it becomes available through the internet. Still, because respect is as important as availability, there remains great competition for acceptance and publication of papers in top-tier journals.

​

So much gets published these days that nobody can keep up with the flood of articles that enter the public realm every month. Scientists and the public must rely on peer-review to select for publication the best, most important contributions to knowledge, or those which improve our knowledge by rebutting shaky or outmoded claims. Reviewers should certainly screen out unworthy and poorly substantiated reports.

​

The peer-review process does not always get implemented in its idealized form, however. There can be cronyism or bias among editors and reviewers. There can be defensiveness and self-protection by reviewers, or simple laziness. Most reviewers in my 50 years of experience with publication have been conscientious, and my own publications have certainly benefited greatly from their criticisms and suggestions. And most editors have been constructive, sometimes even overruling their own reviewers if they think the reviews were unfair or misdirected.

​

This chapter shares quite factually some of most important SRI experience with professional journals, much of it quite commendable. But some is hard to explain and defend. In the chapter that follows, we will consider roles that the various media have played in the acceptance of SRI, both positively and negatively.

 

INITIAL PUBLICATIONS ON SRI

​

The first published report on SRI was written by Fr. Laulanié in French and printed in the Belgian journal Tropicultura in 1993.[2] As far as we know, this article had little if any impact. The next report on SRI was published six years later by his Malagasy colleague Justin Rabenandrasana, the secretary of Association Tefy Saina, in the NGO journal ILEIA Newsletter as discussed in Chapter 24.[3] About the same time, I was able to publish my first paper on SRI together with other papers from the 1999 Bellagio conference on agroecological innovations in a new journal, Environment, Development and Sustainability.[4] However, none of these articles would be considered as putting SRI on the scientific map for persons who regard well-established peer-reviewed journals as the ‘gold standard’ for scientific communication.

​

Subsequently there were articles on SRI published in 2001 and 2002, respectively in ECHO Development Notes and further articles on SRI in the ILEIA Newsletter. These also would not be given much weight by researchers because the data reported were not extensive nor analyzed for statistical significance. Moreover, they were written more for practitioners than for scientists. Even so, they contributed to the start of experimentation with SRI and its evaluation in several countries.[5]

​

Most of the writing about SRI in the first several years after it began moving beyond Madagascar was published either as papers in conference proceedings or as invited chapters in books.[6] As the review process for these is not as formal or as rigorous as for professional journals, these reports could be read and studied for their inherent interest, but they could not be claimed, nor were they claimed, as being scientifically conclusive, even though extensive, consistent, and growing evidence was being reported. These articles did not give SRI much scientific traction because results employing standard experimental design and tests of statistical significance are preferred over on-farm data, even if the latter may be more relevant for making improvements in agriculture.

​

In the early 2000s, not much SRI data had yet been gathered that could be analyzed in ways which qualified for journal publication. Those who were learning about SRI faced a kind of chicken-and-egg situation: which comes first? On-station research or on-farm trials? There were not resources available for doing systematic research, beyond thesis research by students in Madagascar and then Cornell, in part because there was little yet published on SRI in the scientific literature. Indeed, some rice scientists discouraged funding agencies from investing in acquiring systematic knowledge about SRI.[7] So at first, book chapters and proceedings papers were the best and main channels available for reporting on SRI. We knew, however, that to gain scientific acceptance, journal articles published after peer-review vetting would be necessary. This was a slow and sometimes questionable process.

 

THE JOURNAL NATURE

​

This is one of the most respected scientific journals in the world, so it was probably unrealistic to expect that we could start with an article in Nature. In 2000 with data from Prof. Robert Randriamiharisoa and his student Jean de Dieu Rajaonarison, I submitted an article to Nature, ‘Evidence of synergy among cultivation practices that greatly increase rice yield.’ This was based mostly on Jean’s factorial trials (N=288), discussed in Chapter 7. The article cited also a number of other evaluations already done in Madagascar from government as well as NGO sources. But Nature’s deputy editor for biological sciences politely returned it without sending it out for review.[8]

​

In 2002, working with colleagues in China and Madagascar and at Cornell, I submitted another article on SRI to Nature, ‘Greatly increased yield from changing management practices.’ This was based on more data, but it was still not close enough to the prescribed format, and it got a quick pro-forma turndown. However, it did catch the eye of a senior biology editor for Nature, Chris Surridge, who telephoned me from England for an hour-long discussion of SRI. He became curious enough about SRI to write and publish a news article on SRI in March 2004. This reported on the ‘controversy’ surrounding SRI, rather than writing about SRI itself.[9]

​

Comments by several skeptics who were interviewed by Surridge and then quoted in the article cast enough doubt about SRI so that, although the article may have prompted some curiosity, it elicited, as far as we could tell, no evident interest sufficient for getting scientific evaluations done. The article’s title, “Rice cultivation: Feast or famine?” had as its leader: “Proponents call it a miracle. Detractors call it smoke and mirrors. Will the System of Rice Intensification feed the hungry world, or needlessly divert farmers from tried and true practices?” In fact, no proponents of SRI whom we knew called it “a miracle,” but rather regarded it as something real and explainable, and replicable. Yet the article’s eye-catching presentation made SRI less interesting to researchers who on the whole prefer to avoid contentious subjects.

​

The journal did, however, publish soon thereafter a letter from the director of extension at Andhra Pradesh’s state agricultural university in India, who wrote in response to the article. Based on his own evaluations across all the districts of that state (described in Chapter 7), Alapati Satyanarayana confirmed the reports on SRI yield increases.[10] But the Nature article put SRI in the realm of ‘controversy’ rather than regarding it as something meriting scientific investigation.

​

A dozen years later, in December 2017, Nature’s on-line journal Nature Plants ran an editorial titled “Strategies for survival” which began with this statement: “The United Nation’s Sustainable Development Goals contain a commitment to abolish world hunger. Sounds like a job for a plant scientist!”[11] The editorial then proposed that the world can move toward abolishing hunger without necessarily relying on the development of new varieties:

“Much can be achieved in these areas without changing the plants we use. Adjustments to farming practices can reduce the inefficiencies found at particular locations, ranging from improved irrigation, transport and storage infrastructures, to more extensive changes in approach such as no-till or organic systems.

​

The system of rice intensification (SRI), for example, has been employed by over ten million farmers in Asia, Africa and the Middle East. It was developed in Madagascar in the 1980s by the French Jesuit Father Henri de Laulanié and adapted for use in other countries during the 1990s and 2000s, in part by the work of Norman Uphoff of Cornell University, New York.

​

The approach uses less water than a traditional paddy, keeping soil moist rather than inundated, and involves transplanting young seedlings into grids of greater spacing. In the North Indian province of Bihar, where many hundreds of thousands of hectares of rice are under SRI cultivation, the Bihari government reports 40% higher yields than for more traditional farming. But SRI is always adapted to local conditions, making direct comparisons difficult to perform.”

​

This reporting was positive and encouraging, but it remains to be seen how much influence it will have in eliciting scientific or general interest in understanding and utilizing SRI methods.[12] The journal Nature could have made a large contribution to getting scientists to take SRI seriously enough to evaluate it, but SRI and/or our presentations of it did not mesh with the conceptions and criteria of its editors during the first decade of SRI’s emergence on the international stage. This was probably more our fault than theirs as with more time and effort we might have been able to prepare a manuscript that could pass through the eyes of their editorial and peer-review needles.

 

AGRICULTURAL SYSTEMS

​

This is one of the best-known and most-respected agricultural journals. In 2002, it accepted and published an article that introduced SRI to agricultural scientists,[13] even though the journal’s reviewers were not very supportive of its publication.[14] The article reported on what was known at the time about SRI methods and results and proposed that SRI methodology warranted further examination and validation. It offered some explanations for the SRI results reported, but did not assert them as established facts.

​

The editor accepted this article without the revisions called for by the journal’s reviewers because he did not think that the claims being made in the article went beyond the evidence reported, and probably because he considered SRI to be something deserving of investigation. While journal editors are usually reluctant to second-guess their reviewers, in this case the editor was supportive of exposing the new ideas of SRI for others’ consideration and testing. This was immensely helpful for getting SRI laid before the scientific community.

​

The next year there followed in this journal two articles that cast SRI in a negative light, the first in mostly economic terms, and the second from an agronomic perspective. Both articles contributed to the visibility of SRI among researchers, but they probably slowed its acceptance.[15]

​

The first article, by Moser and Barrett, discussed in Chapter 7, was well-researched and clearly written. But unless readers read the article carefully, their take-away message was likely to that SRI is inherently labor-intensive and too labor-intensive to be of much benefit to poor households, so it would not be relevant for reducing poverty. There was growing evidence at the time from several countries that for most farmers SRI either was or could become labor-saving, which was an opposite assessment to the article’s. However, its characterization of SRI as too labor-demanding for poor households contributed to a stereotype that for a decade or more was mostly impervious to evidence to the contrary.

​

It was certainly true that SRI is more labor-intensive than it is capital-intensive, which is the basis for economists to classify a technology as being either one or the other. However, the relevant question for farmers is whether a new technology requires more or less labor from them than their present practices, and whether it returns more rice per hour or per day of labor than at present. With 90% of the world’s production of rice occurring in Asia where growing rice is usually pretty labor-intensive, most rice farmers were finding that taking up SRI was reducing their labor requirements per hectare, at least once they had learned and mastered its methods. And certainly with SRI practices farmers were producing more rice per hour or per day of labor input, as Moser and Barrett acknowledged in their 2003 article. But this matter of fact got little attention among researchers, donor agencies and policy-makers once the label ‘labor-intensive’ had been attached to SRI.[16]

​

The other article on SRI published in Agricultural Systems in 2003 was written by Achim Dobermann, who was soon to become IRRI’s deputy director for research.[17] This article was based on a priori reasoning from the published literature rather than on experimentation or any empirical evaluation of SRI. As discussed in a box in Chapter 10, the article relied mostly on results from an exercise in crop modeling that Dobermann had carried out using coefficients taken from the literature, all derived from rice plants that had been grown under very different conditions from SRI, in other words, with very different phenotypes.

​

Considering only variables such as mean daily temperature and day length, Dobermann calculated that under Madagascar conditions, rice plants could not accomplish enough photosynthesis daily to support a yield of more than 18 tonnes per hectare. Therefore, according to his calculations, a reported SRI yield of 21 tonnes per hectare (Chapter 10) was beyond the biologically-determined limits on yield. Any yield reported as more than 18 tonnes per hectare should therefore be attributed to errors in measurement (or to naïve misrepresentation).

​

The feedback that I had given to Dobermann before his article was published was ignored. I had pointed out that rice plants under SRI management having much greater root growth than conventionally-grown rice phenotypes would have greater productive capability than he was assuming. Dobermann, however, considered the growth and functioning of root systems to be irrelevant to his estimations of the process of photosynthesis.[18]

​

In his article, Dobermann acknowledged that SRI practices indeed produce deeper rice root systems, but he asserted nevertheless that modern-variety rice plants in favorable environments with good water and nitrogen management would outperform SRI-grown plants. He accepted that SRI practices might be beneficial for resource-limited farmers who were cultivating on poor soils and who could not afford to purchase fertilizers. But in his conclusion, Dobermann maintained that SRI was not relevant for improving rice production in general and would be, at best, only ‘a niche innovation.’ Such a prediction was already being contradicted by accumulating experimental results and field experience, as reported throughout this book.

​

After publishing these three articles, Agricultural Systems did not remain the forum for publication and controversy on SRI. Instead, the journal Field Crops Research moved into that role, as discussed below. Agricultural Systems published nothing more on SRI for more than a decade, except for a follow-up rebuttal in 2005.[19] In 2016, Agricultural Systems published an article on SRI by Oxford University and Indian researchers which summarized their rather comprehensive evaluation of the impacts of SRI use in Andhra Pradesh state of India.[20]

​

The main negative feature of SRI that these researchers identified was that households which adopted SRI methods had less need for labor in their rice production, so they reduced their wage payments to agricultural laborers, mostly women, by 50%. This was an unfortunate and undesirable social impact. But this this single negative effect was offset by many substantial benefits that the researchers documented: substantial increases in farmers’ yield (by 60%) and in income (>200%), with reductions in net greenhouse gas emissions (40%), in water requirements (60%), and in fossil fuel consumption (70%). This study refuted the conclusion that Moser and Barrett had drawn 14 years before, that SRI is inherently and necessarily more labor-intensive.

​

This 2016 article, one of the most thorough examinations of SRI impacts published, received much less attention than the earlier articles published in this journal, however. That same year Amod Thakur and I submitted another article to Agricultural Systems, on SRI as ‘climate-smart agriculture.’ It was rejected by the editors without being sent it out for review, with the explanation that the journal was only publishing articles that had a ‘systems’ perspective (apparently the System of Rice Intensification did not qualify). Fortunately, this article was fairly quickly accepted and published by another, equally reputable journal.[21]

 

INTERNATIONAL JOURNAL OF AGRICULTURAL SUSTAINABILITY

​

This journal was launched in 2003 under the editorship of Jules Pretty, who had learned about SRI from participating in the Bellagio conference in 1999 (Chapter 35). Given this new journal’s focus and mission, advancing knowledge and practice for more sustainable agriculture, SRI was a subject evidently relevant to editors and readers.[22]

​

The first issue of IJAS contained an article on SRI which became one of its most-often cited articles.[23] A number of other SRI papers were subsequently accepted and published in IJAS, helping to bring SRI knowledge and experience to a readership oriented to considering this kind of new knowledge.[24] Although IJAS was not considered a major mainstream agronomy journal, it became one of the best channels for getting SRI known within research circles and among persons in the agricultural sector who were looking for ways to improve current practice.

 

FIELD CROPS RESEARCH

​

Balancing the receptivity of IJAS toward SRI knowledge was the less open posture of this well-established journal. Its attitude toward SRI became evident from the way that it reviewed and dismissed an article on SRI in 2006 that had been submitted by senior researchers of the China National Rice Research Institute in Hangzhou. Their manuscript was summarily rejected without being sent out for peer review, and with no suggestion of how the article might be revised to be acceptable for publication. Giving constructive feedback to authors is a common editorial practice.

​

This article, discussed in Chapter 9, showed with replicated on-station trials conducted over two seasons with two hybrid rice varieties that rice yield was higher on plots that had (a) much-reduced plant populations, (b) with alternate wetting and drying rather than flooding, and (c) with more organic and less inorganic fertilization. These findings contradicted the standard rice management practices in China and elsewhere.

The lead author of the paper Lin Xianqjing was dismayed by the negative response of the FCR editors in 2006 and wrote to me at Cornell, asking if I could help them revise their manuscript to get it published. The rejection surprised him because just one year earlier, the editors of FCR had accepted and published a very similar research article from him and CNRRI colleagues.[25]

​

This new article, for which the director-general of the Institute was also a co-author, had used the very same research methods that they had used in the article that FCR had published the preceding year. Indeed, their trials for this new article had been conducted on the very same research station. But the earlier article had been evaluating the effects of modifications in irrigation rather than SRI methods. It had been quickly accepted and published by FCR.[26]

​

The new article that Lin sent me was very substantial and easy to reorganize and make more readable. The research design was state-of-the-art, and the data base was extensive. I hoped that the article would be re-submitted to Field Crops Research, to see how its editors would assess a more polished presentation. But Lin and his colleagues wanted to get the article published quickly and submitted it to a new journal in China, the Journal of Agricultural Biotechnology and Sustainable Development.[27] So we cannot know whether FCR editors would have been willing to at least review a revised and better version of the article.

​

We have some idea of how the editors of FCR looked upon SRI because they had accepted for publication with unusual speed two articles that have been frequently cited as conclusive refutations of SRI claims.[28] These articles were accepted, respectively, within 40 days and 10 days after being submitted to the journal.[29] It takes most articles months rather than weeks or days to be reviewed and get acceptance. The 2004 article by John Sheehy and colleagues was not only made the lead-off article in its issue, but a short discussion piece with a pejorative title commending the critical article and endorsing its conclusions, was published along with the article itself.[30]

​

There were quite evidently a large number of errors and misrepresentations in both the 2004 and 2006 articles that should have kept them from being approved by any peer-reviewers who were objective and knowledgeable on the subject. Conversely, when responses to these articles were written and submitted to FCR, it took many months for these responses to get accepted by the editors and published.[31] And in both cases the editors solicited and printed rebuttals to the SRI responses without giving the authors of the responses any opportunity to challenge and correct errors in the rebuttals.[32] Nor did the editors raise any objection to the emotional and derogatory language that was used in the rejections of SRI. Ironically, the critics who were hostile to SRI did not write in the kind of language that is expected for scientific discourse which attempts to establish what is the truth of the matter based on evidence, not argumentation.[33]

​

The hastiness of FCR’s review process and the lack of review by persons who had thorough or direct knowledge of SRI resulted in the journal publishing articles that should not normally have passed through peer review. Below is an identification of factual problems identifiable in the papers, foreshadowed by their pejorative subtitles, that serious reviewers should have caught.

​

Fact or Fallacy?

​

The conclusion of this article by Sheehy et al. -- that “SRI has no major role [to play] in improving rice production generally” -- was based (a) on a priori reasoning rather than on empirical evidence, (b) on crop modeling by Dobermann, a co-author, and (c) on the results from three small on-station trials conducted in three provinces of China. It ignored a large body of evidence that had already been assembled by dozens of rice researchers in China.

​

The modeling, as noted already, took no account of the large and measurable phenotypical differences between rice plants grown with SRI vs. conventional methods.  It assumed without justification that SRI plants’ having larger, more vigorous root systems would not have an impact on the plants’ capacity for photosynthesis in their leaves or on their grain production.

​

A major limitation on the article’s conclusion was the way in which the three trials in China were conducted, on small plots (40 m²) with four replications. One of the basic rules for scientific evaluation is that trials be conducted according to prescribed, agreed-upon protocols. In these trials there was:

• No soil aeration: One of the most important elements of SRI methodology is to use mechanical weeders that stimulate the growth of roots and the soil biota. In these trials, the SRI plots had only manual weeding, so no active soil aeration.

• Excessive synthetic fertilizer: Another key element of SRI methodology is the enhancement of soil organic matter to promote the abundance and diversity of the soil biota, which is inhibited or even suppressed by an excess of inorganic nitrogen.

     The Sheehy article said that all N applications on all of the test plots (SRI and control) “followed recommended practices in each location.” Yet all of the plots, both SRI and BMP, received 180-240 kg of N per hectare, with SRI plots also getting 1.5 tonnes of rapeseed cake per hectare to enhance organic material. In fact, there has never been SRI recommendation for applying anything like this amount of inorganic N fertilizer because it will adversely affect the life in the soil, unbalancing it with some species expanding rapidly at other species’ expense.

     Inexplicably, at the same time that these trials were being done at the three locations in China, five of Sheehy’s seven co-authors were writing a paper for IRRI that was based on their extensive fertilizer trials conducted across China. Their paper concluded that in China, the optimum N fertilizer application for rice is just 60 to 120 kg per hectare.[34] So this means that the SRI trial plots in these comparison trials were given than most of Sheehy’s co-authors were recommending at the time based on their concurrent research for IRRI.

     As a consequence of this excessive application of inorganic N fertilizer, the SRI rice plants in the trials in Hunan Province, one of the three evaluations, lodged, reducing these SRI plots’ yield by 32%, according to the article. We saw in Chapter 12 that SRI rice plants seldom lodge, even under the stresses of strong wind and rain. So, the lodging of the Hunan trials was evidently due to the excessive application of N fertilizer, which is contrary to SRI recommendations. This reduced rather than enhanced the SRI yield.

• Longer crop cycle: A clear sign that SRI methods were not used appropriately in these trials was the article’s report that the SRI rice plants matured 2 weeks later than those in the control plots. As seen in Chapter 11, when SRI methods are used as recommended, rice plants usually mature about 2 weeks sooner, not maturing later. This disparity indicates to anyone who knows SRI that its methods were not used as recommended.

 

Ironically, even with these significant deviations from an SRI protocol, the average yield from the three SRI trials was slightly higher than from the control plots, by 100 kg per hectare. This increase was not large enough to be statistically significant, however, given the small number of data points.

​

Everyone can agree that in these trials, SRI did not perform significantly better than best management practices (BMP). But what was the conclusion of Sheehy and his co-authors? Not that SRI and BMP were equally good, but that somehow BMP remained preferable to SRI.

​

If the plots in Hunan had not been over-fertilized and had not lodged, average SRI results for the three sets of comparisons would in all probability have been significantly greater than the BMP yield. Indeed, all of the SRI plots would have performed better with active soil aeration and with more organic (and less inorganic) fertilization.

​

Note also that the comparison between SRI and BMP was made only in terms of grain yield. No consideration was given to SRI’s lower costs of production or its reduced requirements for external inputs, things that are important to farmers if not to researchers. FCR reviewers did not give any weight to such factors, however. The reviewers and the journal’s editors approved the Sheehy article for publication in less than 6 weeks.

​

An evident shortcoming of this article was that its authors based their conclusion on just three small on-station trials, not considering the large body of evidence that was already available from Chinese rice scientists, such as was reported at the Sanya conference in 2002, attended by one of the article’s co-authors, Shaobing Peng.

​

Moreover, the authors ignored the favorable conclusion about SRI that had been reached by Prof. Yuan Longping (Chapter 21), probably the most respected rice scientist in China.[35] A basic canon of scientific research is that one’s findings should always be considered within the context of other scientists’ previous work. Since a majority of the article’s co-authors were Chinese, they could certainly have reviewed the literature in Chinese language that reported the results of other scientists’ research.

​

In fact, the article’s conclusion was completely contradicted by an evaluation of Chinese researchers’ studies of SRI. Ten years later, a meta-analysis was done of all the published Chinese research that had compared SRI yields with those of BMP. Quantitative analysis showed that good, not even full, use of SRI methods gave SRI a 20% yield advantage over BMP.[36] In addition to this conclusion about yield, there are cost-saving and water-saving for farmers to be considered, plus soil health and water quality benefits for the environment.

​

But the 2004 article by Sheehy et al. was published in a prestigious journal after going through what was expected to be a rigorous and objective peer review. Based on this assumption, the article’s evaluation was quite influential in justifying opposition to or neglect of SRI. At a minimum, persons who had in-depth knowledge about SRI should have been included among the reviewers for such an article. But FCR editors apparently chose to involve in their review process only proponents of BMP and skeptics about SRI.

​

A Synopsis of the Empirical Record?

​

This 2006 article shaped opinion against SRI at least in part because it was written by three Cornell faculty or staff. Unfortunately, there were evident flaws in both the methodology and the data base that any serious reviewer should have picked up, although some of the deficiencies and biases in the data base would have been obvious only to someone like myself who was thoroughly acquainted with the data that was being analyzed.

​

Actually, I had provided the authors with almost half of the data sets that they analyzed for the article, not having been informed about how they would be used. SRI data are made available to anyone who asks for them, so sharing the data was done as a matter of course.[37] But if the authors had conferred with me about the data, which they could have done easily since our offices were not far from each other, I could and would have helped them avoid some serious errors in constructing their data base. Most of these errors would not have been realized, however, by a reviewer who came to the article without much knowledge of the subject.

• Selection of ‘SRI’ cases: The first and most important flaw in the study was discernable simply from the authors’ presentation of their methodology, from their own words. They wrote that their data base included only comparisons in which the SRI trials “closely approximated” the principles of SRI, which they had listed correctly in the article. This sounded like a very scientific approach, but its operationalization contradicted the justification. The authors wrote in their Methods section: “Experiments with treatments that did not closely approximate the principles of SRI (e.g., included only one or two SRI elements) … were excluded from the database.”

     This sounds reasonable but it means that when the authors constructed their data base, they included as ‘SRI cases’ comparisons where as few as 3 of the 6 recommended SRI practices had been used. Thus, the ‘SRI’ data could be from trials where old seedlings were transplanted, or where there was no water control, or no organic soil amendments, or no active soil aeration. How using 50% of a protocol could represent “close approximation” was never explained; the term “close approximation” usually means something close to 100%. This disjunction should have been noticed and challenged by any reviewer who read the article’s Methods section carefully and objectively.

• Selection of ‘BMP’ cases: The inclusion of data from many sets of practices that were classified as ‘best management practices’ was also questionable since the authors never specified in the article what constituted ‘best practices,’ as they had done (and then ignored) for SRI. In fact, one-quarter of the reported BMP yields were below the world average yield, with one reported BMP yield as low as 1.1 tonnes per hectare! This made the concept of ‘BMP’ rather spurious, but FCR reviewers did not object to this.

     ​In fact, most of the data sets did not have enough accompanying information so that the authors could know what practices had produced the respective average yields – age of seedling, spacing, irrigation schedule, etc. The comparisons analyzed thus were not between ‘apples and oranges’ so much as between different kinds of fruits in both categories.[38] There were also other flaws in the data set that was analyzed to arrive at the authors’ conclusion that ‘BMP’ yields on average surpassed ‘SRI’ yields by 11%.

• Exclusion of Madagascar cases: Five of the 40 comparisons in the article’s data set were from this country, and they showed SRI yields to be on average triple those of BMP. These data sets were excluded from the authors’ analysis, however, with the simple explanation, for which no evidence was provided, that the soils in Madagascar are “distinctive.” In fact, the Madagascar comparisons were some of the few in their data base that had been made to test the article’s hypothesis explicitly, employing replicated trials with random block design. It appears that these results favorable to SRI were excluded from the statistical analysis to get the results desired, but only the authors can clarify this. Presenting the Madagascar data as part of the whole data set made the article’s data base look larger and more diverse than it actually was when the data were ‘crunched.’ This a reviewer should have noted.

• Cases from China: The data base included the three comparisons from China reported in the 2004 article by Sheehy et al. Why these were not valid representations of differences between SRI and BMP has been discussed above.

• Cases from Laos: The IRRI representative in Laos had gotten several organizations working there to do SRI trials in the rainy season of 2002, as discussed in Chapter 7. In his message transmitting the data from these trials to IRRI and to Cornell, this representative stated that neither IRRI/Laos nor the organizations doing the trials thought that the trials had been conducted properly. So, these results should not be considered as a valid evaluation of SRI vs. other methods, they said.[39] Yet these comparisons from Laos included in the article’s data base constituted 10% of that data base.[40]

• Cases from Nepal: When asked by Peter Hobbs for data comparing SRI with BMP, I provided him with results from two large sets of comparison trials from the Sunsari-Morang irrigation system in Nepal, where a Dutch consulting team (NEDECO) was implementing an irrigation improvement project funded by British aid (DFID) (Chapter 8). The team had gotten farmer field schools at different locations within the irrigation system to do in-field evaluations of SRI vs. BMP vs. their own rice cultivation methods. Replicated trials were done at 12 farmer field school locations in 2002 and at 13 in 2003. These trials showed SRI yields averaging 40% more than BMP (8.28 vs. 6.01 tonnes per hectare) and almost double the yield from farmers’ practices (4.29 tonnes per hectare).[41]

     These data from Sunsari-Morang, which met the authors’ criteria for inclusion in the analysis, were not included in the data base, however, for no apparent reason. Instead, the only comparative data set reported from Nepal was from an agricultural experiment station at Bhairahawa in 2002, which showed an on-station SRI yield of 5.4 tonnes per hectare vs. 5.7 tonnes for BMP. No consideration was given to the station’s reporting that in that same season, the researchers had also conducted some off-station (on-farm) trials; the SRI yields were 8.8 tonnes per hectare at 20x20 cm spacing and 7.6 tonnes at 30x30 cm spacing.[42] Further, the next year, 2003, the same researchers’ on-station comparison trials showed SRI yield to be 33% higher than their BMP yield, a reversal that the Nepali researchers themselves attributed to their having gained experience with the new methods. However, in their analysis published in 2006, the authors included only the 2002 on-station, less-favorable comparison. A synopsis should presumably consider all available data, not selected sets. But reviewers would not have known of these other data.

• Cases from Bangladesh: Twenty percent of the article’s data base came from evaluations of SRI done in this country, with 3 of the comparisons taken from a single article published previously in Field Crops Research.[43] One of these comparisons was from an on-station trial, while the other two comparisons were from small on-farm trials (each N=10). The authors chose not to include the results from a much larger set of on-farm comparison trials (N=1,278) that I had given them, from a two-year study, 2002-2004, funded through IRRI and conducted by three NGOs (BRAC, SAFE and POSD) plus Syngenta Bangladesh Ltd., discussed in Chapter 7.

     The small study by Latif et al. concluded that BRRI’s BMP recommendations gave both higher yield and greater profitability than SRI, while the much larger study that was ignored supported the opposite conclusion.[44] When I asked why the data from the larger study by BRAC, SAFE, POSD and Syngenta, carried out under IRRI auspices with DFID funding, had not been considered in the meta-analysis, the answer was that this report was not in the published literature, whereas the smaller study by Latif et al. was.[45] However, this justification made little sense because the article’s data base included three data sets from Bangladesh that were listed just as ‘personal communications,’ i.e., not published.

​

These specifics on the article by McDonald et al. are not so much a rebuttal as a demonstration of how superficial and mistaken the peer-review process can be, even for a respected journal like Field Crops Research. The conclusion that the authors drew -- that (excluding the Madagascar trials) SRI methods gave an average yield 11% lower than BMP -- has been refuted by a more conclusive meta-analysis of Chinese researchers’ findings noted above with reference to the 2004 article.[46]

​

This experience speaks to the questions that I raised with one of McDonald’s co-authors, Susan Riha, reported in Chapter 23: Is everything that is true in the peer-reviewed literature? And is everything in the peer-reviewed literature true? The answer to these questions is unfortunately ‘no.’

 

OTHER JOURNALS

​

Experience with other journals was mixed, but in general more constructive. The journal Plant and Soil, similarly prestigious as Field Crops Research, published the meta-analysis of Chinese data on SRI vs. BMP note above, as well as an important article on SRI plants’ nitrogen use efficiency by Amod Thakur and colleagues.[47] So it was open to research that considered new directions.

​

The respected journal Experimental Agriculture published by Cambridge University Press in the UK was one of the first major journals to publish articles on SRI, and it published several analyses that illuminated different aspects of SRI research and experience.[48] Another British journal, CAB Reviews, put out by the Commonwealth Agricultural Bureau International, accepted and published the first article on SRI in Indonesia, and later on SRI’s application to the production of teff (STI) in Ethiopia.[49] Two agronomic journals started in Germany, Archives of Agronomy and Soil Science and Journal of Agronomy and Crop Science, were also among the first journals to accept and publish research articles on SRI.[50]