The history and future of agricultural experiments

Abstract

An agricultural experiment is usually associated with a scientific method for testing certain agricultural phenomena. A central point in the work of Paul Richards is that experimentation is at the heart of agricultural practice. The reason why agricultural experiments are something different for farmers and agronomists is not their capacity to experiment as such but the embedding of experiments in a specific ecological, material and institutional environment. Using a historical perspective, changes are examined in the organization of agricultural experiments focusing on the Netherlands and colonial Indonesia during the first half of the 20th century and the international agricultural research institutes for the period thereafter. The results show a gradual shift in the role of experiments in the connection between science and practice. Initially, the link was considered to be established through various forms of experiments, rooted in an integrated social and technical understanding of agronomy. Gradually, this turned into a connection primarily established through various forms of communication. Recent work of Richards incorporates ideas that address key issues emerging from the history of agricultural experiments, dealing with an integrated social and technical understanding of agriculture.

Keywords:

• Agricultural science
• Farming practice
• Experimentation
• Research policy
• Scientific disciplines

1 Introduction

What is an agricultural experiment? Within the agricultural sciences the answer to this question will vary among disciplines. The common features are a treatment, a hypothesized process or causal mechanisms to be tested. Living creatures or parts thereof are usually the object of an experiment. Today, each branch of the agricultural sciences will have its manual or guidelines for experimentation, depending on the object of the experiment, the place where the experiment is done, the treatment or process that is tested and the methods used. The connection between agricultural experiments and agricultural science seems obvious. However, at the beginning of the 20th century, agricultural scientists were very much in doubt about the validity of the commonly used experimental approach. In recent years, anthropologists like Richards and others claim that many of the basic agricultural activities carried out by farmers are experimental in nature as well. Based on anthropological fieldwork among rice farmers in Sierra Leone, Richards in particular emphasized how farmers deal with the agro-ecological conditions as a performance. In farming practice, experimentation is a crucial act to improve farming results in subsequent seasons [1,2]. At first glance, the experiments by farmers look completely different from the scientific experiments performed on controlled experimental plots, often in climate-conditioned greenhouses. Experiments by (African) farmers and scientists seem as distinct as (Western) scientific knowledge and (non-Western) indigenous knowledge [3]. However, Richards has pointed out that the principles of farmer experiments are basically the same as the principles of scientific experiments [2,4]. For him, claims about fundamental distinctions between the cognitive processes underlying experiments of African farmers and knowledge production in (Western) science carry an “implicit notion of intellectual apartheid”. What makes agricultural experiments something different for farmers and agronomists is therefore not the capacity to experiment as such but the embedding of experiments in a specific ecological, material and institutional environment.

This paper puts different environments of experimentation and the linkages between them in a historical perspective. Experts dealing with agricultural experiments were most of the time worried about both the scientific validity of their experiments and the connection with farming practice. Throughout the years the growth of research activities as well as the changes in experimental capacity transformed the nature of the connection between scientific experiments and on-farm experimentation. In later work, Richards offered concrete suggestions for new ways of establishing a connection between farmer experiments and scientific experiments. What is argued here is that his ideas about reconnecting farmer experiments with scientific experiments fit with the historical trajectory of agricultural experiments and therefore have a broader relevance than the focus on West African rice cultivation might suggest. The historical trajectory of agricultural experiments is characterized by a growing distance between farmer experiments and scientific experiments. In other words, establishing the connection between science-based experimentation and agriculture became more complex in terms of methodology and organization. This resulted in a weakening of the link between the scientific experiment and the experimental nature of farming. Agricultural advisors or extension officers are crucial actors in bringing science and practice together. The changes in agricultural science also affected the relation between research and extension. The historical development of agricultural extension, as presented here, suggests that extension work moved from an agronomic orientation towards a more sociological and psychological orientation. This may explain why in recent decades attempts to bridge the gap between scientific experiments and farming practice are usually framed in behavioural and communicative terms.

The material for this paper is a selection of key events, decisions and circumstances in the history of agricultural science that resulted in the establishment of various forms of agricultural experiments, its connections and disconnections. Like in most of Richards’ work, examples and case material relate to rice cultivation. Rather than looking at West Africa, the historical developments in agricultural science presented here primarily relate to the case of the Netherlands. The involvement of Dutch scientists in agricultural experiments, applied to a variety of crops including rice, started in the early 20th century when besides the booming agricultural sector in the Netherlands, the Dutch built up much of their agronomic expertise in the Netherlands Indies. In the colonies, cash crops formed the dominant economic part of agriculture, although much effort went into stabilizing the food situation for which agricultural science was considered a necessary input. After the independence of Indonesia in the 1940s, Dutch agronomists continued to be active in rice cultivation, partly in the remaining Dutch South American colony Suriname and partly in the emerging international research institutes and expert networks spreading out over the globe. The continuing involvement of Dutch agronomists in research on tropical crops like rice illustrates how agricultural science has become an activity relatively independent of immediate linkage to constituent farmers. This is not to say that agricultural science has entirely lost its connection with practice but to make clear how the commonalities between scientific experimentation and farmer experimentation have become obscured and received less attention. Combining the history of agricultural experiments with the work of Richards provides some interesting perspectives for the future of agricultural science and the role of experimentation in creating effective linkages between science and practice.

There are few studies that examine agricultural experiments as performed by agronomists or other agricultural scientists. The social science literature on scientific experiments more generally is much larger and this paper therefore first addresses some of the central features emerging from that literature and how this applies to experiments in agricultural science. In the following sections historical information is mobilized to show what developments resulted in the displacement of agricultural experiments from the farmer's field to various other environments. The case of rice is of particular interest because initiatives to set up experiments for rice improvement were taken by administrators of the colonial government who were concerned about the food situation on Java, Indonesia. During the late 1880s and 1890s, district officers located in different parts of this island were instructed to set up experiments with various cultivation methods to demonstrate to the local farmers how to grow rice more efficiently. These administrators had no training in agriculture, no experience with rice cultivation and, with few exceptions, never took it very seriously. Initially, when agricultural experts entered the scene there was little commitment to engage in rice farming. However, once agricultural advisors were appointed with a mandate to perform on-farm experiments things started moving. Prompted by the advance in statistical inference calculation, the design and validity of the experiments became a controversial issue. It will be shown how a particular solution established by the late 1920s, resulted in a hierarchy of experiments held together by the bureaucracy of the agricultural research organization. In the decades that followed, a variety of factors resulted in an increasing differentiation of agricultural experiments. This differentiation had an impact on most agricultural research and extension services across the world and is still the dominant mode of operation today. The last section discusses some of the shortcomings of the current mode of agricultural experimentation. It is shown how recent work of Richards offers some suggestions for alternative ways of setting up experiments and how agricultural experimentation might be organized differently.

2 Agricultural experiments in the social science literature

Many forms of experiments can be classified as an agricultural experiment. Rather than making a list of all the appearances of agricultural experiments, the social science literature, in particular the history and sociology of science, is used to highlight some of the common features and processes related to experimentation. In several studies the theory and practice of scientific experiments are examined. The overall message is that historically and socially determined factors play an important role in establishing what counts as a scientific experiment. The common association between experiments and laboratories, for example, is a feature of present-day science that is very different from the situation in the past. A common feature of all forms of experiment, in past and present, is demonstration. More specifically, there is a close connection between what experiments try to demonstrate and the public they want to convince with the demonstration.

Examining the activities of Royal Society Fellows in 17th century England, the historian and sociologist of science Steven Shapin [5] showed that most experiments were conducted in private houses. Other possible locations were a coffeehouse or the royal palace. More than a geographical space and material setting, these locations were demarcated by social regulations. “[A]ccess to most experimental venues (and especially those located in private residences) was obtained in a highly informal manner, through the tacit system of recognitions, rights, and expectations that operated in the wider society of gentlemen.” [5: 389]. The location where experiments were done varied with the audience called in to be convinced as witness. A similar point emerges from Bruno Latour's study of the discovery of an anthrax vaccine by the microbiologist Louis Pasteur. Not the discovery as such but the process of convincing veterinarians and livestock farmers, Latour argues, is what made Pasteur a great scientist. To accomplish this, Pasteur organized ‘staged demonstrations’ at the countryside in which he managed to replicate what he did in the laboratory [6].

Besides management of the audience, the objects and findings resulting from scientific experiments require alignment with the material environment outside the scientific experimental setting. The social studies on the role of laboratories in science make clear how relocation of a laboratory experiment in a real situation requires both physical and social adjustment in order to make clear that what works in the laboratory also works in society. Success in science implies optimized mobility of experimental results between the protected environment of a laboratory and the messy world outside [7,8]. In particular for scientific fields that work on technical applications in a certain domain of society, careful adjustments of what works in a scientific experiment and what works in society is required. In other words, for experiments on new technical devices or modifications it is accuracy and reliability of the technical procedure or system that is tested [9].

Is there something special about experiments in the domain of the agricultural sciences? Although experiments in agricultural science have specific features, there is no reason to give it a special status or to assume that scientific agricultural experiments operate principally different from experiments in other scientific fields. Experimental practices might even become hard to recognize as ‘agricultural’ when research objects are taken out of their agricultural environments. An experimental setting of a molecular biology laboratory at an agricultural institute will look very much the same as the scenery of a molecular laboratory in medical research. Under laboratory conditions, experimental objects are often hard to recognize as organisms or simply invisible for the naked eye. It is interesting to explore how in agricultural science the linkages between experimental objects, organism and practice are made. The main features of scientific experiments emerging from social studies of science and technology are key elements in agricultural science as well [10,11]. A point that can be added is the specific role of on-farm field trials in agricultural science. This type of experiment tries to establish a connection between experimental work at a research institute and the activities of farmers. However, on-farm field experiments not only take place at a specific place, they also have dynamics of their own. Field experiments therefore are not just a communication channel between science and practice but require particular knowledge and skills of the experimenter about setting up the experiments, materials to use and the involvement of farmers, technicians and other actors [12]. The importance of these intermediary forms of experimentation amends the strong researcher-controlled portrayal of experimentation emerging from the laboratory studies by Latour and others. This is taken even one step further by Richards’ observations about the experimental nature of farming itself. Agricultural experiments take place at various locations, not necessarily in a scientific environment. Consequently, scientific researchers are key persons in experiments but not the only actors nor always the ones in charge of experiments.

A final point to be made is about the historical framework of this paper. Agricultural science has its roots in science disciplines, most notably biology and chemistry, which have a long historical record. At specialized research and training institutes the history of agricultural science emerged halfway the 19th century [13]. A crucial role in this process is that of national governments trying to stimulate the agrarian economy with the help of science. The history of agricultural science picked up when the colonial government started to realize that agricultural experiments could benefit agriculture, i.e., in the second half of the 19th century. This paper highlights particular events and developments in the history of agricultural science that implied significant changes for the way agricultural experiments were set up and organized.

3 District officers and agricultural experts

One of the earliest accounts of experiments with rice cultivation on Java is from government sources in the mid 1870s. The Dutch tea planter Karel Frederik Holle had developed an interest in rice farming and in 1874 wrote an advice to the government on the improvement of rice cultivation. The report was based on several trials he and some of his friends in the colonial service had carried out in different parts of Java. Their conclusion was that yield and productivity could increase significantly. The list of recommendations primarily focused on the planting stage, arguing that reduction of seed use by sowing more thinly would result in firmer plants that recovered more quickly after transplanting and produced more tillers. This allowed for wider spacing, which in combination with planting in rows made weeding easier [14]. The colonial government circulated instructions to all district officers to start field demonstrations based on Holle's advice. The main recommendation, thin seeding and row planting with wider spacing, was not further specified and so left ample room for district officers to apply it with flexibility, which is indeed the picture that emerges from the various reports, mostly summarized and published by Holle himself [15]. These reports make clear that some district officers with an interest in rice cultivation carried out serious trials and that results were considered positive, meaning a measured yield increase and saving on seed. There is no evidence of a wide uptake of similar experiments by Javanese rice farmers.

The experiments set up by Holle were controversial but not because of doubts about his methods or lack of results. What was considered off the mark was the fact that trials were set up by colonial administrators. The rice experiments by the district officers were part of a wider transformation of the colonial administration. From the 1850s onwards there was growing opposition against the role of the government in the forced production of cash crops for international markets, such as coffee, sugar, indigo and tobacco, known as the Cultuurstelsel. This was fed by a process of national political reforms in the Netherlands leading to increased control over government activities by the parliament. One of the consequences was the dismantling of the entanglement of state interest and the private interests of the Dutch royal family in colonial cash crop production. Holle was among the progressive colonials who considered the coercive regime as a direct cause of economic stagnation and inefficient farming practices of the local population. Labour employed in cash crops implied that farmers neglected rice cultivation.¹ After the forced labour regime had been abolished, Holle's on-farm experiments were to re-educate rice farmers in taking better care of their rice fields.

A handful of colonial administrators took the field demonstrations seriously and sent detailed reports about what they experimented with and the results obtained. Most administrators, however, ignored the instructions or simply ordered local village elders to implement Holle's suggestions [16]. The colonial government looked for other ways to improve food crop production and in 1899 the Botanical Garden in Buitenzorg was put in charge of the field experiments and demonstrations. In other words, responsibility over the experiments moved from administrators to experts with a background in biology or agronomy. The director of the Botanical Garden, Melchior Treub, was to transform the garden into a Department of Agriculture, realized in 1905. Under Treub's guidance, more intensive research on the rice plant and rice cultivation practices was set up [11]. The approach taken by Treub put scientific research central. In his view a substantial set of on-farm experiments, as proposed by Holle, was not necessary. Instead, he prioritized on-station trials and the creation of some model rice farms to demonstrate the latest findings. The colonial government, however, pushed for more experimental activities in the rice fields. On-farm experiments were resisted by Treub until his retirement in 1909. His place was taken over by Herman Lovink who had just reorganized the agricultural research and extension system in the Netherlands. Lovink had no university degree, reason for Treub and other biologists to qualify his appointment as a director of the Department of Agriculture as a betrayal to pure science [17,18]. Lovink, however, did not reduce any of the research activities but added agricultural teachers to the department. This extension service for indigenous agriculture started in 1910 with nine Dutch and four Indonesian agricultural experts, numbers going up in four years to 26 and 32, respectively. Although the initial increase of extension staff levelled off in ensuing years, the new branch in the organization of the colonial Department of Agriculture was a fact.

What were the implications for agricultural experiments? A great deal of the extension activities concerned agricultural education, organized in village schools that were run by the Indonesian staff members. Many of these schools operated as demonstration plots. The annual report of 1913, for example, makes clear that a major part of the teaching was done through demonstration in the school garden [19]. The same report mentions all sorts of trials in various parts of the archipelago, primarily focusing on rice varieties and fertilization. The Dutch staff of the extension service had occasional meetings with the agronomists of the Department of Agriculture to discuss the type of experiments to be done and how to perform them. Thus, over the two decades before and after the turn of the century there were several shifts, both in responsibility over agricultural experiments and in the audiences to which experiments were addressed. By attaching demonstration trials to agricultural schools, Lovink particularly aimed at new generations of farmers. Moreover, parallel to on-station experiments, field trials were considered equally important in developing scientific insight.² In following decades, the organization, planning, reporting and analysis of the different types of experiments became the central issue on both the research and the policy agenda in Dutch agricultural science.

4 Convergence of research and field experiments

Parallel to organizational changes in agricultural experiments, researchers concerned with the validity of their trials developed new experimental methods for agriculture. The earliest accounts are from the mid-19th century. Chemists like J.F.W. Johnston became aware of the problematic nature of testing the effects of agricultural inputs.³ In such tests experts try to reveal the influence of one or more variables on the performance of a crop. But a field with a crop, like any array of organisms, always contains a considerable degree of natural variation. For example, if a fertilizer can increase yields by 10%, how can an experiment confirm the yield increase when ‘natural’ yield variation between fields is already larger than 10%? To solve this problem, inference calculation or statistical significance testing strongly affected the organization of agricultural experiments.

By the time the colonial Department of Agriculture was established, similar structures were already operational in the Netherlands for several years. An overview of government activities in agriculture in 1905 reports 592 field experiments conducted in agriculture and 217 in horticulture. The majority of these experiments were on fertilizer effects and performance of crop varieties [23]. The value of these experiments was openly questioned by Joost Hudig, a chemist working at one of the agricultural research stations. Based on a review of work done in Germany, Denmark and Britain, he argued that the Dutch agricultural advisors, just like most of their colleagues in other countries, lacked a proper understanding of the mathematical complexity of experimentation [24]. His main worry was the habit of the advisors to combine experiments from different locations and draw conclusions based on the averages of the results of these experiments. Using the example of a fertilizer experiment with potato, he argued that such experiments provided little information on the effect of the applied fertilizer. “When we consider that the report says nothing about the type of sandy soil, gives no details about the kind of fertilizer, explains nothing about previous fertilization, nothing about rainfall, and still calculates the average, everyone must agree with me that the result of these calculations will not inspire much confidence” [25]. In addition to questioning the experiments as such, Hudig warned that chemical companies used such shaky figures to recommend their products.

The agricultural advisors, however, had other concerns than experimental methodology. The advisor who performed the experiments Hudig had used as an example wondered in a response “does the practising farmer have to wait for science to solve this problem before he can apply Chile saltpetre or ammonium sulphate?” [26: 19]. As an advisor he felt ‘morally obliged’ to inform farmers about results of experiments, even if the experiments were not carried out according to the latest insights. Hudig rebuked by pointing out that no one forced advisors to draw vague conclusions. “Practice does not benefit from wrong calculations; it benefits more from the plain acknowledgement that the data obtained were inadequate.” [24]. Hudig's objectives were broader than changing the experimental practice of agricultural advisors. The underlying problem, he argued, was the organization of agricultural experiments. There was no mechanism that provided advisors with fixed methods for experimental design and analysis, nor was there a central authority that streamlined their experiments. It took until the 1930s before such an organization was established.

Meanwhile, the Agricultural College in Wageningen responded to the concerns by appointing a mathematics professor. In 1913, M.J. van Uven, a mathematician and one of the Dutch pioneers in inference calculation, was appointed a full professor. He was to teach statistics to the agricultural researchers. “Wherever an analysis is made of the laws governing a large set of living organisms, for example entire fields of crops, entire generations of animals and plants, where problems like heredity and breeding have to be solved, it is statistics that − by combining the results of observation − will contribute to finding causal connections. It is therefore advisable, if not to say imperative, that particularly agricultural education be supplemented with a special course in probability calculation and statistics.” [27]. Van Uven's lectures – he taught in Wageningen until 1950 – gave future agricultural advisors and researchers a firm basis in inferential statistics.

Getting the mathematics right was not enough. During the 1920s and early 1930s the organization of agricultural experiments was worked on by the government. A first initiative came at an Agricultural Congress in 1922. A paper from J.D. Koeslag, an agricultural advisor working for the Plant Breeding station in Wageningen, gave an overview of current experimental practices. His conclusion was that most experiments had the character of demonstrations. Producing results was considered more important than properly analysing these results and advisors had no incentive to do otherwise. “Until now the agricultural advisor operated completely autonomously in his area” [28]. Field experimentation, Koeslag argued, needed to be centralized and experiments should be carried out according to ‘the new scientific methods of experimenting’ making demonstrations redundant. In response to his presentation several agricultural advisors emphasized the value of demonstration fields. Demonstrations, they argued, helped farmers to get an impression of the effects of certain treatments and are more effective than scientific reports they would never read. In other words, the advisors were convinced that their demonstrations convinced the audience better than the field trials and other experiments done by researchers. A committee was installed, chaired by Koeslag, that proposed a new set-up for the agricultural experiments as the connecting element between extension and research. The advice of the committee resulted in a manual for field experiments, issued under ministerial responsibility. The manual distinguished five different types of field experiments: (1) demonstrations of clearly observable differences, (2) observation experiments for disease resistance, (3) exploratory experiments to determine which factors needed more precise testing, (4) experiments where yield was the decisive element, and (5) ‘institute experiments’ to be conducted at research stations and not suitable for the field [29].

In the colonies, similar developments took place. In fact, a centralized organization of field experiments was already functioning in the late 1920s. In 1931 a Wageningen PhD study on field trials in rice cultivation on Java explains the overall procedure. First, a general plan of trials is made on the basis of which budget is allocated to each district. The district agricultural advisor makes specific proposals for each field trial. A copy of the plan is sent to the central office, another copy sent for advice to the research station. Once the trial is centrally registered, the forms and the advice from the research station are returned to the district extension officer who then starts with the experiment. Outcomes of the trials are sent back to the office, results calculated and administered [30]. The description of experimental practice by Ossewaarde shows a rather efficient government service running a large number of experiments. With respect to rice, the experiments mainly focused on variety tests and fertilization [31]. What is striking in the procedures is the paper work this required. Co-ordinating a network of agricultural advisors doing different experiments in different areas and connecting this to on-station research resulted in an experimental bureaucracy that had to facilitate the information flows between the field sites and research institutes. How effective was this system?

One indicator of the efficacy of the experimental bureaucracy is the response by the private sector. The private agricultural sector in the Netherlands Indies, mainly consisting of plantations producing cash crops for the international market, by and large copied the experimental system developed by the colonial Department of Agriculture. Although the planters often scorned the bureaucracy of the government services, they had adopted the same centralized co-ordination of field experiments and even argued for integration with the government system during the economic crisis of the 1930s. In the Netherlands, the private agricultural sector mainly consisted of commercial firms producing inputs such as seeds and fertilizers. The largest fertilizer company, Dutch State Mines, for example, had experts posted at the major public agricultural research stations, not only to share expertise but also to make use of the effective linkages with extension officers doing field experiments all over the country [32]. Similar public–private connections existed extensively in the colonies but primarily for European-run plantation agriculture. The common element of the two situations was a relatively small number of farmers who overall were well informed about what research and extension had on offer. Conversely, the centralized experimental bureaucracy was quite effective in reaching its audience, given its size and education level. That was not the case for the colonial food crop sector where farmers were many times more numerous, less educated and spoke different languages. Although a substantial number of Dutch and Javanese extension officers were active, demonstrating the advances of science for such a huge farmer audience was hard to organize. Consequently, extension officers questioned whether distribution of, for example, improved rice varieties with often only marginal benefits was worth the effort [11]. In 1941, the head of the rice breeding station, Van der Meulen, estimated that the area covered with varieties from the station was about 9%. Creating a balance between different forms of experiments on the one hand and creating accuracy and effectiveness on a large scale on the other was nearly impossible for the centralized agricultural research and extension services.

5 Research and extension as separate disciplines

The organization for field experiments in the Netherlands Indies was dismantled with the Japanese occupation and subsequently with independence of Indonesia. In the Netherlands, the organization continued operation and expanded after World War II due to increasing investments in the agrarian sector in response to the immediate after-war food shortages. The organizational structure implied an effective linkage between the work of researchers working in the agricultural institutes and the advisors of the agricultural extension service, working with farmers in all agricultural areas of the country. Field experiments, however, were not the only activity for research and advisors. For the researchers in agricultural institutes, field experiments had to fit in their wider set of research activities. For the advisors of the agricultural extension service, experiments were just one among many activities. The newly established Central Institute for Agricultural Research (CILO), the nerve centre of agricultural experiments, therefore, was a sort of umbrella organization that co-ordinated activities of researchers and advisors that were formally employed by other institutes. Abolished in 1957, this central body lasted for less than two decades. Co-ordination of field experiments was handed over to the Research Station for Arable Crops and Pastures [33]. This organizational modification reflects a broader pattern in which field experiments were increasingly a concern for researchers, and less for agricultural advisors. Concerns about the dominant role of research were also expressed in the colonies.

During the first half of the 20th century the training of Dutch agricultural experts at the Agricultural College in Wageningen implied a broad agronomic basis with a variety of specialization options in later years. By and large, graduates who were employed in the agricultural extension service, either in the Netherlands or the colonies, were agronomists with a specialization in social science disciplines, primarily agricultural economics [11]. Establishing a basic understanding of agricultural activities in the region where advisors were stationed formed a core element of the extension activities. The advisors in the Netherlands Indies, most of them stationed on Java and a smaller number placed on other major islands in the archipelago, faced the challenge to set up extension activities for a type of agriculture they were hardly familiar with. Collecting data of local agricultural activities formed a substantial part of the activities of these advisors. Several of the colonial advisors turned their data analysis into a thesis, defended in Wageningen. Although varying in topic and focus, these publications all contained elements of social geography, ethnography, agricultural economics and agronomy.

One of these publications was the PhD thesis of W.J. Timmer, defended at the agricultural college the Dutch had established in Bogor (Buitenzorg) just before the war broke out. Timmer's dissertation was basically an attempt to synthesize the work of the colonial extension service and turn it into a methodology for agricultural extension, for which he coined the term ‘social agronomy’ [34]. The basis of social agronomy was a detailed appraisal of the rural area, something Timmer considered as the most important element of agricultural extension. “Because once again, the principle ‘know your district’ is for every social-agronomist a crucial requirement and if this is not met, any form of appropriate extension is out of the question” [34: 176]. Timmer worked out a step-wise procedure for extension. Following collection and analysis of information, prioritizing topics to work on was followed by testing and disseminating adjusted cultivation practices, treatments or techniques. Despite the emphasis on field studies, Timmer realized that advisors were primarily occupied with the last phase, testing and promotion. The main reason, he explained, was pressure put by research institutes on the extension service to implement their findings. Advisors thus hardly had time for an elaborate exploratory phase. As a result, Timmer remarked with some acrimony, introduction of any change of cultivation methods or input was based on sheer coincidence and the hope for “a good guess” [34: 182].

For Timmer and his fellow advisors who had published on Indonesian agriculture, extension made sense if based on a profound understanding of the local patterns and variation of agriculture, including the social and economic conditions of farming. By this, they argued for an additional element to agricultural experiments. Structuring on-farm experimentation and demonstrations on the basis of what research stations offer is putting the cart before the horse. A thorough assessment of the agronomic and social-economic situation of a certain area is what should drive the experimental agenda of agricultural science. For Timmer, the heart of agricultural extension was empirically driven field research, integrating agronomic and social science perspectives. Timmer's main concern was the effectiveness of agricultural science, not by intensifying the downstream flow of scientific findings to the field but by creating an interchange between information from the field and information from science. Rather than questioning how various forms of experimentation and demonstration should be organized, Timmer raised a more fundamental issue about the type of information needed to make experiments and demonstrations meaningful and effective. This could only be based, he argued, on extensive field studies combining analysis of agronomic and social-economic facts. After the 1940s this line of thinking was continued primarily in anthropological studies. Extension, as a branch of agricultural science, moved in a different direction.

Over the 1950s and 1960s the Agricultural College in Wageningen saw a rapid expansion of the number of professors appointed, often in line with a disciplinary separation of study programmes [11]. The social science disciplines, starting with rural sociology and agricultural economics expanded as well. In March 1965, a chair for ‘extension studies’ was created and the person appointed was A.W. van den Ban. van den Ban had a background in rural sociology and in his inaugural lecture he emphasized that the discipline of extension studies should focus on farmers’ behaviour. Agricultural extension, he argued, “provides information to farmers on which they decide to change their behaviour” [35: 4 – emphasis in original]. How behaviour and behavioural change worked was considered the major academic challenge. Therefore, the main “supporting disciplines” for extension studies were, according to van den Ban, sociology, psychology and cultural anthropology [35]. His inaugural lecture further sketched examples of how decision-making worked, of the organizational challenges of extension services and the difficulties advisors had to deal with in convincing farmers to adopt certain techniques.

The focus on adoption behaviour and communication, as proposed by van den Ban, followed an international trend in agricultural extension. By the late 1950s and early 1960s, agricultural extension in most countries had detached itself from its roots in agronomy. Extension work focused on promoting and distributing the output provided by research institutes. This type of extension work was given theoretical grounding by the work of van den Ban in the Netherlands and many other extension scholars worldwide of which undoubtedly Everett Rogers’ work on the diffusion of innovation is most illustrative [36]. Rather than developing integration between agronomy and social science as suggested by Timmer, extension as an academic discipline specialized on behaviour and communication. Where Timmer argued for ‘the field’ in all its aspects as the starting point for advisory work, the unit of analysis for extension became the farmer as an individual or as a group. Consequently, extension was reduced to studying and working with the notion that farmers behaved in certain ways and had a voice to which agricultural science might want to listen. What farmers had to say about their fields, crops and how this could inform agricultural experimentation became a blind spot in agricultural extension.

6 Experiments in international research

In the 1960s, the emerging international system of research institutes played a prominent role in further emphasizing the communication element of extension. The international campaign to fight poverty and hunger through dissemination of nitrogen-responsive varieties of wheat and rice, known as the Green Revolution, implied a world-wide investment in extension. Next to distributing high-yielding crop varieties, fertilizer and credit, extension officers instructed farmers how the high-yield potential could be attained. The international organizations, most prominently the World Bank, developed an extension package called the Training and Visit System. As the name suggests, the method was based on regular visits to farmers in order to instruct them in improved farming methods. The method was primarily set up to increase adoption rates of the Green Revolution varieties [37,38]. This further increased the dominance of on-station and in-field experimentation in control of researchers. Agricultural advisors in developing countries primarily organized the stage and drummed up an audience to watch the scientific demonstrations. The many studies on the impact of the Green Revolution suggest that only in certain areas the audience was easily convinced while in many places they probably left halfway, head shaking.

By the 1970s, extension was established as a special field within academic circles and in national and international agricultural research organizations. Agricultural experiments were not seen as part of the expertise and skills of the agricultural advisors. The importance of a connection between research and extension was acknowledged and establishing such linkages was considered part of the reforms of extension services induced by the Training and Visit System. This would require “regular training sessions and workshops for extension and research personnel, joint participation in planning each season's extension and adaptive and applied research activities, shared responsibility for farm trials, joint field trips to review specific crop problems and to obtain a better idea of actual production conditions, and visits by extension staff to research stations” [37: 39]. Although adequate as a description of an ideal situation, it has only rhetorical value. As the described colonial system for the Netherlands Indies shows, co-operation between different branches of the agricultural services not only required an efficient organization and a sophisticated experimental methodology, to be effective it also has to meet the challenge of reaching large numbers of farmers. As the Dutch experienced in the colonies, this required an immense organizational capacity. Regardless of whether it would at all be possible to realize this, most developing countries lacked the means even to set up smaller organizations for agricultural experimentation. In most cases, priority was given to promotional activities.

The fervour with which Green Revolution varieties were introduced further emphasized promotion and distribution activities of extension work. Moreover, the international research institutes seemed little interested in creating a linkage between on-station experiments and experiments in the farmers’ fields. To take the International Rice Research Institute (IRRI) in the Philippines as an example, attempts to connect the research of the institute with on-farm experiments were not very much supported. In 1963, IRRI created an Office of Communication, headed by Frank Byrnes, that set up a methodology to organize feedback from the farmer level to IRRI. The proposed method was not put into action and the office was closed in 1968. According to Byrnes this was due partly to the successful release of the rice cultivar IR8, partly because of fears of rice scientists “to be contaminated with activities of lower-status extension workers” [39: 126]. Besides status differences, direct interaction between the international centres, national extension services and farmers was considered politically sensitive [40]. So the international research institutes relied on the various national research centres and related services to communicate research results to the field as well as getting feed-back from the various countries. What were the implications for experimentation?

Given the wide scale on which tropical crops were grown, even if only focusing on rice, the organization of international agricultural research implied a growing gap between experimentation in agricultural research and experimentation at farm level. This gap also widened in terms of experimental methodology. Because systematic linkages between information from the field and research activities were problematic, experiments in agricultural science had to rely more on a theoretical understanding of agriculture. An example is the use of crop physiological models. Following developments in systems ecology where cybernetic principles were applied to biology, agricultural crops were represented as a set of linear equations, acting as machines [41]. In international agricultural research the work of the Australian agronomist Colin Donald became influential. In 1968 he coined the term ‘ideotype’, a blueprint of an optimally shaped plant based on detailed calculation of the relevant physiological processes. “It is the familiar approach in aircraft production, building construction and instrument design, and its validity for these physical purposes is generally accepted. Can this principle be applied to biological needs?” [42: 387]. An immediate answer to the question could not be given but for Donald this was merely a matter of time. Donald referred to work of the IRRI breeders as an example of how a focus on key physiological functions (in particular photosynthesis and plant competition) resulted in an optimized plant architecture (erect foliage). The notion of plant design through theoretical ‘calculation’ of high-yielding varieties became an important element in the development of the IRRI varieties. Mid 1980s a research programme, called SARP (Simulation and Systems Analysis for Rice Production) was set up to develop a crop physiological model for rice (named ORYZA) for which data from the experimental fields of IRRI were used [43]. The emergence of crop physiological models implied a new dimension for agricultural experiments. Besides checking plant and crop performance in the field, the functioning of new varieties or variation in water, sunlight or nutrient uptake could be tested using computer simulation. The idea of designing a plant based on modelled optimal conditions (referred to as ideotype breeding) became a key focus in the work of international research institutes like IRRI [44].

The development of plant physiology and the use of crop models did not make field experiments redundant. Models need verification with real situations (known as calibration) for which the collection of field data is an important element. In fact, the data sets needed to make a model reliable are substantial and require a lot of work. Most models are based on data collected from research stations and thus have value for experimental work under similar (optimal) conditions. In situations where various stress factors constrain plant growth, the accuracy of most physiological models is limited and requires substantial cross-reference with field data [45]. The development of crop physiological models added another dimension to advanced experimental work in agricultural research. Moving back and forth between the linear computer models and the stochastic models of field experiments inserted an extra interface between different forms of experimentation. Testing accuracy and reliability of the models requires substantial effort of the researcher. Moreover, these procedures have to be iterated for new situations in order to make the crop physiological models effective in developing solutions or advices to farmers. Crop physiological models can increase the efficacy of scientific experiments but further complicate the connection between different forms of experimentation, in particular for making linkages with on-farm experiments in highly variable circumstances.

7 The future of agricultural experiments

The development of various forms of scientific experiments for crop improvement has led to considerable advancement in finding the optimal agricultural conditions that, in combination with modification of the genetic composition of crop varieties, has resulted in a considerable yield increase. However, under most farming conditions, in particular in developing countries, such optimal conditions are difficult to attain. The difference between potential yield and actual yields attained in the field, known as the yield gap, has become a growing concern for international agricultural development agencies. This paper has shown that changes in how the connection between scientific experiments and the (experimental) activities of farmers has developed over the years have resulted in an ‘experimental gap’. Although many factors reduce yields in farmers’ fields and vary from place to place, dealing with the experimental gap is an effort to be made in order to effectively tackle the yield gap. Brief explorations of two present-day examples with respect to rice farming may clarify how narrowing the experimental gap might be achieved. The first example is the scientific debate over the System of Rice Intensification. The second example brings us back to the work of Paul Richards on rice farming in West Africa.

Emerging from the work of the Jesuit priest de Laulanie on Madagascar, the System of Rice Intensification (SRI) has developed into a widely promoted method for rice cultivation [46; in this issue]. Questioned by various rice scientists in the international research network, a controversy emerged that appeared in various agronomic journals, which was serious enough to get coverage on the editorial pages of Nature [47]. Although the debate focused mainly on the yield potential of SRI, an issue that was stressed in particular in contributions from the soil scientist Willem Stoop, was the nature of experimental work required to adequately test the various components of SRI. Given that various forms of SRI are applied by farmers under different conditions, considerable experimental work is needed, ranging from relative simple on-farm experiments to more complicated multi-factorial analysis [48,49]. Having worked in international research centres for many years, Stoop observes that the experimental work at these centres is too remote from the type of field experiments needed to test SRI in an adequate manner. Given that SRI is promoted widely in various countries by government and non-government organizations without any form of scientific test, this resembles the situation in the Netherlands Indies at the end of the 19th century. In fact, the practices promoted by Holle, i.e., reduction of seed and wider plant spacing, are among the key principles of SRI.

Where the SRI case illustrates some of the limits of the national and international research institutes in connecting to the field level, the second example illustrates the capacity of farmer-based experimentation. Through collection of rice varieties at several locations in West Africa, a team of African and European researchers demonstrated the existence of interspecific hybrids (between Oryza glabberima and O. sativa) in farmers’ fields. Resulting from intercropping of the two species in the same or nearby fields, farmers consciously selected for and maintained these hybrids [50]. The farmer hybrids were used independently and long before another interspecific hybrid, called NERICA (New Rice for Africa), was introduced to the region in 2002 by the African Rice Centre (WARDA). NERICA was exalted by scientists and policy makers as a showcase of international agricultural research. Researchers knew about the existence of interspecific hybrids in farmer's fields. However, no initiative was taken by the African Rice Centre to connect to and learn from rice farming practice with respect to interspecific hybrids. Although the importance of farmer management of rice varieties is known for many years to both local and international rice breeders, there is no mechanism in place that enables the integration of field information, on-farm experimentation and on-station and laboratory experiments by researchers. More than coincidence and the hope for a good guess, recalling the concerns expressed by Timmer in the 1940s, the example shows that the current research system is predominantly focused on scientific solutions, excluding available information obtained from field studies.

What, then, is needed to close or at least narrow the gap between scientific experiments and the experimental activities of farmers? As pointed out above, the notion of farming as a performative, experimental activity as worked out by Richards suggests that there is no fundamental obstacle to overcome for creating such a connection. From his involvement in several shared experiments between farmers and scientists, Richards recently elaborated a suggestion for a different organizational principle of agricultural experiments [51]. Taking up a distinction drawn in artificial intelligence research, he points out that most conventional research is organized as supervised learning, being the instruction of fixed solutions within a given set of conditions. By contrast, unsupervised learning works on the basis of feedback mechanisms that, within a network, lead to solutions that emerge as the combined outcome of the various feedback mechanisms. Comparing Richards’ arguments with the history of agricultural experimentation, the unsupervised learning structures provide greater space for farmer experimentation and do not require an extensive experimental bureaucracy to organize a pre-structured connection between various forms of experimentation. This also implies a different role for extension and agricultural advisors. Rather than setting up staged demonstrations and make farmers behave as a compliant audience for the miracles of modern science, advisors have to rediscover their agronomic roots, assess different forms of information and collectively design creative experiments that serve a mixed audience of farmers and scientists. Moreover, integration of agronomic and social-economic factors requires experimental forms that allow for integration of both types of data.⁴ Both the farmer-based seed selection activities, exemplified by the West African interspecific hybrids and the System of Rice Intensification seem to offer interesting cases to start working on such an approach.

Notes

1 “From experience we learned that nowadays the small farmer is doing much better and, even without being chased, cultivates his field even better than before, if only his time is not taken by services for cash crops, colonial or village elders.” [14: 8; translation HM].

2 Lovink openly criticized Treub's approach. “The question is not what maximum possible amount of rice can grow on a certain area, but how it will be possible, once acquainted with rice cultivation as conducted by the Javanese, to increase together with the Javanese farmer his rice yields economically, taken into account his development, workforce and his capital.” [20: 387; translation HM].

3 Johnston was a (critical) follower of the founder of agricultural chemistry Liebig and an influential figure in Dutch agricultural science [21,22].

4 In January 2009 Richards organized a workshop in Wageningen that focused on experiments in social science.