Adaptive Research with and without a Learning Alliance in Myanmar: Differences in learning process and agenda for participatory research

Highlights

•	Adaptive Research and Learning Alliances have complementary value.
•	Skill development and reconfiguration of practice are central in situated learning.
•	Stakeholder participation should be guided by the unfolding of experiments.
•	Functional linkages between practices are crucial in participatory innovation.

Abstract

The main challenge for researchers and project staff when implementing inclusive approaches in agricultural innovation is how learning and technology adaptation interact and how to reach jointly set targets. We provide a comparative analysis of the learning process induced by adaptive research (AR) in one case and a combined AR with Learning Alliance (LA) approach in another. The AR approach bridged farmers and researchers, but its implementation where researchers controlled experimentation, was not optimally conducive to experiential and discovery learning. The combined AR and LA approach expanded the number of stakeholders with whom farmers interact. This broadened the learning agenda beyond the initial objectives of the project. Although a LA provided added value in increasing the scope for learning with other stakeholders, limitations also emerged from the autonomy and informality of the learning process. Our analysis, based on the notion of situated learning, revealed practical concerns were a major driver in the participatory process. Incorporating insights and skills developed from experimentation to support the reconfiguration of practices, not only farming but also trading and milling, is needed. Lastly, the approaches have complementary value. Inclusion of more actors, as in a LA should not merely facilitate deliberations between actors but also support the reconfiguration of different practices and the functional linkages between these practices.

Keywords:

• Adaptive Research
• Learning Alliance
• Myanmar
• Learning
• Outcomes
• Technology

1 Introduction

Recent studies provide empirical evidence for the potential as well as the challenges of co-creation of knowledge in the spread of technologies (Rodela, 2014). Over the past decades, most agricultural research and innovation projects have developed approaches that are more inclusive towards end users (e.g. Horne and Stür, 2003;Palis et al., 2010; Adekunle and Fatunbi, 2012). These approaches open space for learning of various actors at different levels. A major overall aim is to boost learning, considering that agricultural innovations require learning beyond the farm level and coherent practices among a variety of stakeholders (Leeuwis, 2004). Although its importance has been recognized conceptually, empirical studies that target learning processes are still limited (Wals and Rodela, 2014). Triggered by the need to understand whether and how processes of learning can be facilitated, empirical examination is required of projects that use such approaches. The study presented here examined how learning is facilitated at two sites of a project in Myanmar, one based on Adaptive Research (AR) and another combining AR with a Learning Alliance (LA) approach. We investigated these sites to understand whether and how they support learning towards innovation in rice farming communities.

AR is an approach that characterizes the needs of farmers and then uses experiments in farmers’ fields to adapt a given technology to local conditions. The localized experiment thus becomes the learning activity to find out how introduced technological solutions can alleviate specific needs. The adapted technologies imply a learning effect for farmers and researchers (Krupnik et al., 2012; Palis et al., 2011; Dorward et al., 2007; Flor et al., 2016). There is also a learning effect for policy makers who can use the results for decisions on new investments for agricultural innovation (Horne and Stür, 2003).

The general principle of the AR approach is to improve the way science-based technologies are implemented in society. The underlying assumption, commonly held in the 1970s and 1980s when the approach developed, was that science produced knowledge and technologies independent of society. This is what Gibbons et al. (1994) termed Mode 1 science. Their Mode 2 science, which increasingly became visible in the later decades of the twentieth century, is knowledge production in the ‘context of application’. This is similar to what Funtowicz and Ravetz (2003) call post-normal science, raising questions about who, besides scientists, are involved in the identification of problems and formulation of research questions. One could argue that agricultural sciences have always depended on the context of application. Nevertheless, how to demarcate science from practice is a concern visible in organisations for agricultural research and extension and in debates over what distinguishes experiments from demonstrations (Maat, 2011; Maat and Glover, 2012).

The LA approach emerged against this backdrop and involves societal actors in the research process early on. The approach involves a network of various stakeholders, organised in linked stakeholder platforms, to identify, share and adapt innovative practices in specific contexts (Lundy, 2004; Lundy et al., 2005; Verhagen et al., 2008; Stelling et al., 2009). The addition of the LA as a new approach or the replacement of a tested one such as AR raises the question what added value a LA brings. We address this question from the perspective of situated learning, a notion that connects learning activities to specific practices. We examine the facilitation of innovation networks according to a LA approach and analyse if and how learning is effectuated differently from the more conventional AR approach.

The broader project in Myanmar was implemented to introduce improved cropping options to increase productivity in rice-based systems using the AR approach as its basic collaboration mode. In one of the sites, the LA approach was added to involve a broader network of stakeholders. In comparing the two cases, we examined how the AR approach influenced learning in the farming community in one site and how the involvement of a wider network of stakeholders (from the added LA approach) in the other site impacted the learning process.

1.1 Conceptual framework

Learning is a central notion used in many participatory approaches to agricultural innovation. In most approaches, social interaction is the basic principle for learning (Leeuwis, 2004). Differences appear when social interaction is further defined and located in a particular setting. In the AR approach, in-field experiments are the focal point of social interaction. The social interaction is mainly between the agricultural experts and farmers, discussing and reflecting upon the various field experiments. The AR learning model is basically a socialised version of Kolb’s (1984) experiential learning, a cyclical framework in which active use, through experimentation, follows reflection on earlier observations. Results of the experiment are then again observed, reflected upon and may lead to further experimentation (e.g. Krupnik et al., 2012; Palis et al., 2011; Dorward et al., 2007; Flor et al., 2016).

In the LA approach, learning is located at the level of the network or system. In principle, all actors that have some role in agricultural innovation can be included and typically a variety of issues is at stake (Lundy and Gottret, 2007). That also implies a wider variety of interests and views. Social interaction, therefore, may take the shape of negotiations. When managed well and leading to some sort of agreement, learning takes place in the entire network or system (Kilelu et al., 2014; van Mierlo et al., 2010; World Bank, 2006).

In both approaches, the ultimate effect is a change in knowledge and action of the actors involved. In AR, the learning process is principally geared towards the farmers. In a LA, all ‘system actors’ are potential learners. Of primary interest is how the learning process is affected by the issue at stake and the location of activities. These factors seem to influence the process ex-ante, as a pre-selection of the social actors that become involved in the learning process. Site-specific non-human conditions also are important. However, in both approaches these non-human conditions are not conceptually included in the learning process.

Stone (2007) conceptualises the learning process as a combination of social and environmental learning, the latter referring to evaluation of benefits in practice. He uses the distinction to explain why farmers in India were unable to learn about the use of Bt cotton, a genetically engineered cotton variety. The intense social interaction between farmers, traders offering different seed brands, and agricultural extension workers, obstructed farmers from doing meaningful experimenting in practice, therewith undermining the learning process. The reduced scope for environmental learning, Stone argues, results in de-skilling.

The notion of environmental learning resonates with the notion of situated learning, developed to understand location-specific conditions in the learning process. The theory is primarily developed by Lave and Wenger (1991), focusing on communities of practice (CoP) to describe the context where learning takes place. Situated learning conceptualises, more generally, the combined social and environmental learning process. The situated practice of farming consists of crops, animals, tools cultivation methods and so on. The learning is not only an effect of social interaction in terms of reflection and discussion or negotiation as a deliberative process, but also requires changes in practice as a performative process, resulting in improved skills and adjusted practices (Richards 1993; Stone 2007).

Skill development and adjustment of practices take time. In specialised practices, young people acquire skills by formal or informal apprenticeship arrangements (Jaarsma et al., 2011). In farm households, children develop their skills by observing and helping the elders. Introduced innovations can speed up skill development and change of practices. Rather than a one-off decision, adoption requires situated reconfiguration, a process of unpacking, testing and adjusting the new technologies. In the case of farming, this process typically takes several growing seasons (Glover et al., 2017).

The notion of situated learning is helpful to analyse the AR and LA approaches. Rather than asking who is involved, it raises the question how involvement of stakeholders is linked to practices. What new skills are gained and employed? What reconfigurations of practice become visible? The settings created by AR and LA approaches typically bring together people from different practices. Stakeholder platforms and other project activities suggest the creation of a new, shared practice. However, in most cases these are merely temporary shared practices, for example on-farm experiments ran by farmers and experts together. We also examine, in this paper, how practices of rice farmers, traders and millers change through the introduction of new rice varieties. Although a rice miller can also be a trader, socio-technical reconfigurations and skill formation are different for each of the activities. A small adjustment in skill, for example the visual recognition of a new rice variety, may lead to reconfiguration of the practice of trading, for example negotiation over price, and a different adjustment for milling, for example fine-tuning of the mill machinery.

2 Methodology

Situated learning requires an analysis of the learning effects from an AR approach and a combined AR and LA approach. We ask who is involved in the process, including who is considered to be the ‘learner’ and who is making decisions on what is to be learned? Secondly, what is the learning process? This directs examination on learning activities, the actors involved and roles they play in supporting or steering learning, and how that resulted in changed skills and reconfiguration of specific practices. The learning process also requires analysis over time. This relates to changes in skill and reconfiguration of practices, as well as the evolvement of linkages between these practices.

We compare two cases where the approaches were implemented through a research project in Myanmar. Case 1 is in Daik Oo Township (17°37′49.5“N and 96°34′29.1“E), where solely AR was used. Case 2 is in Maubin Township (17°37′49.5“N and 96°34′29.1“E), where the project used AR with LA approach.

A total of 119 farmers (landowners and non-landowners) were interviewed (Table 1). Interviews were done in two project villages each for Case 1 (Kyait Za Kaw and Pha Aung Kwe) and Case 2 (Nga Gyi Gayat and Nyaung Wine) (Table 1). Landowner respondents were purposively sampled striving for equal representation of farm locations that are flooded or non-flooded, and those near or far from trial sites. The co-operators, who were farmers implementing field experiments on a voluntary basis also were added. There were 14 co-operators in Case 1, and 8 co-operators in Case 2.

Table 1 Number of male and female interviewees^a by category from Case 1 (Adaptive Research) and Case 2 (Adaptive Research with Learning Alliance).

	Case 1 (Daik Oo):		Case 2 (Maubin):
	AR		AR with LA
Respondent	Male	Female	Male	Female
In the village
Farmers (landowners)	51	3	43	6
Farmers (non-landowners)	6	3	6	1
Service providers (equipment, labour coordinators)	1	3	2
Outside the village
Finance group	1		1
Input sellers (seeds, pesticides, fertilizers)	1	3	7	1
Research and extension staff	2	2	1	2
Manufacturers, artisans	2		2
Millers	4	1	5	1
Traders, trader federation	2	1	2	1
Local leaders	1		1
Seed producer			1

a All had connections to the village but not all were involved in the learning activities.

The questions addressed farming conditions, cropping practices, involvement in the learning trials, observations from these trials, and what they planned to do differently. All farmers were interviewed twice (at the end of two cropping seasons). An exception is farmers from Nyaung Wine who sold their land and were either no longer farming or farming only for household consumption.

Other stakeholders (Table 1) were interviewed regarding their interactions with farmers. Aside from community-level stakeholders, those involved in AR or LA outside Daik Oo and Maubin (e.g. researchers) were also asked about their involvement, activities, the process of implementation, and initial outcomes.

Data from on-farm conditions and practices were summarized through IBM-SPSS 22. Qualitative and observational data were coded through Atlas.ti and analyzed along themes of involvement, observations from learning activities, interactions, changes implemented or planned, and initial outcomes.

3 Results: dynamics of learning in interactions

3.1 Defining involvement and agenda

In 2012, project activities started similarly for both cases with parallel partners and activities using the AR approach. The roles of the four actor groups, farmers, researchers from IRRI and Department of Agricultural Research (DAR), and extension staff from Department of Agriculture (DOA) involved in AR were defined beforehand, around providing varieties, technical advice, funding, and capacity-building support, coordinating outreach activities with farmers, and management of trials in villages. Some farmers volunteered to do the trials on their farms; they are termed co-operators of the project. Other farmers were considered ‘learners’ of the activities.

3.1.1 Shaping the learning agenda: translating diverging views

At the start of the project, different actors highlighted several problems as important, creating similarities between the two case-study areas (Table 2). While there was some overlap and congruence in views about the problem (e.g. farm-level issues affecting productivity, and issues related to harvesting), there were also clear differences in emphasis.

Table 2 Actors in Adaptive Research and their characterization of main challenges in rice production in 2012.

Actor	Characterization of the problem	Key interest	Details
DOA	Low yields, lack suitable technologies; adjustments in cropping plan; labor scarcity	New crop management technologies/best management options; extension; capacity-building; research	“Our main role is to transfer good agricultural practices to farmers”; “The DG is very interested in opportunities to train staff”; "…we follow (recommend to farmers) a cropping plan. "; “farmers lose profits because of transporting cut crop in flooded harvest time”
DAR	Farmers lack suitable varieties	Assessment of new varieties; seed production	“DAR is mandated to test and produce new rice varieties for Myanmar”
IRRI	stacking leads to losses and poor quality; lack technology to shorten growing period or hasten postharvest	Research on rice technologies AR trials; Dissemination of technologies from research	"…the main constraint in rice production is poor road access. Because farmers do not have good road access to millers, they stack their rice after harvesting and leave on the levee for 15 days… They do not have dryers in both Townships”
Farmers	flooding and drought, financing (high input costs, credit, fluctuating selling prices), pest and diseases, harvesting during rainy period, poor soil conditions	Improve yield and profits; location-specific constraints like water, pests or labor; link with sources of information/government	Constraints discussed in needs assessment: lack of irrigation (at flowering period in monsoon and in dry season); financing (high costs); harvesting during rainy period; pest and disease (leaf streak, bacterial blight, army worm); flooding/heavy rains delay crop establishment; problems with credit and fluctuating rice prices; soil conditions

Sources: Interviews with DOA and DAR staff (2014), Trip reports (IRRI 2012, 2013), Needs Assessment Report (IRRI, 2012).

Researchers from IRRI and national organizations made decisions on the key questions investigated. These decisions considered their perspectives of the problem, and those identified by farmers. Constraints brought out by farmers that researchers deemed beyond the scope of the project (e.g. credit, better drainage after flooding, fixed prices for rice and pulses, land consolidation) were reframed as a post-harvest problem: the lack of time to process the rice crop immediately after harvest resulting in poor quality rice, high losses, and low selling price. This opened ways for actors to agree in general terms and at the same time address various sub-problems.

When researchers and DOA selected village sites, these were characterized to have specific cropping patterns, rice-rice and rice-pulse. These were however, general characteristics only, and not all farmers in the village follow the same cropping pattern. Only one village, Nga Gyi Gayat had a homogenous cropping pattern for farmers (Table 3). This shows that the options tried may not be relevant for all.

Table 3 Characterization of village sites and mean area in ha (range) cultivated per farmer, by cropping pattern in the two cases from Lower Myanmar.

Case and village	Characterization by the project	Rice-Rice	N	Rice-Pulse	N	Rice-Fallow	N	Rice – Rice/Pulse/Fallow	N
Case 1 (AR):
Kyait Za Kaw	Rice-Rice	5.3 (2–22)	12	7.3	1	9.1 (4–13)	11
Pha Aung Kwe	Rice-Pulse	3.7 (1–5)	8	3.5 (1–12)	17	2.2 (2–3)	2	5.5 (3–8)	2
Case 2 (AR + LA):
Nyaung Wine	Rice-Rice	3.2 (0.4–9)	10	1.9 (0.8–3)	7			2.8 (1–4)	8
Nga Gyi Gayat	Rice-Pulse			6.1 (0.8–40)	23				s

Technical options benchmarked in all sites selected were alternate wetting and drying (AWD), fertilizer management, four different herbicides, and direct seeding using a drum seeder. The results of benchmark trials fed into decisions for AR activities in 2013–2014.

3.1.2 Starting the learning alliance: expansion of the network in Case 2

With the addition of LA for Case 2, stakeholders such as landless farmers or labour providers, the Agricultural Bank, and seed producers were invited. Private sector actors such as millers, traders, service providers and manufacturers of threshers and dryers also were involved. One was Pioneer Postharvest Development Group (PPHDG), a company based in Yangon, dealing with rice milling, manufacturing of agricultural equipment and selling farm inputs, seeking ways to implement their corporate social responsibility policy (interview with MAK, 2015). In this wider network, roles and activities were less-clearly demarcated. Involvement was on voluntary basis, depending on interest in the discussions and possible activities.

In the first learning activity, various stakeholders shared their perspectives of the underlying causes on why farmers get low profits from rice. They then discussed possible solutions and topics to explore. The involvement of various stakeholders brought out the importance of linkages between practices. Farmers stated that ‘millers control prices no matter what the quality’ while millers emphasized ‘farmers sell low quality paddy because they use mixed varieties in the first place’ (workshop report 2013). The selected topic focused on improvement of the connection between farming and milling: improving quality through good seeds and varieties, and improving postharvest management (e.g. better threshers). In addition, they explored how traders and millers would assess the quality rice and pay a premium price. With the addition of LA in Case 2, research and extension stakeholders remained influential in decisions on the agenda. They made sure that activities developed under AR were linked as well.

3.2 Experimentation, interaction and reflection

In both settings, the AR component implied similar experiments in 2013. These included seedbed management, new varieties, improved fertilizer management, and assessment of storage losses. Participatory variety trials (PVS) introduced 12 varieties with potential for higher yields, drought tolerance and salinity tolerance. Researchers implemented trial protocols that included support in the management of labour, fertilizer, and pest control. The farmers were instructed about field monitoring and irrigation intervals. At the end of the season, the harvested products were given to the farmer after data collection.

In the first season, field trials under AR for both cases were managed completely by the local IRRI staff and a DOA officer. Although farmers jointly selected plots, researchers decided on details such as plot sizes, varieties, and replicates. Their objective was to compare the performance of different new technologies under farm conditions. Despite initial stakeholder meetings, many farmers expressed they ‘did not know what was done in their fields’, and were unable to explain details of the research trial (interview notes, 2014). Although there were field days where the staff met with farmers to view and review results the learning directly related to doing the experiment seems mainly to imply increased skills for local staff and DOA officers on how to implement field trials and collect research data.

In both cases, the involvement of farmers increased in the following season. The co-operator farmers joined in setting up experiments, using treatment plots and control plots, where they grow the crop as they normally would. Experiments were guided by protocols, instructing the farmers for example about irrigation schedules or pest control. As a co-operator explained: “the staff provided fertilizers in bags weighed according to the application protocol and left a list of when it should be applied” (interview notes 2014). The farmer applied these on the plots upon phone confirmation from the staff. Local staff also said ‘there were cases when the staff or DOA officer did the application’, and the farmer was not present or aware when it happened. Harvesting, threshing and recording data were done by the IRRI staff with hired labour. A few farmers helped manage labourers during the harvest and recording of data.

Farmers and researchers agreed that involvement of farmers in field activities and data collection was limited because ‘farmers were busy’ and ‘data collection entailed meticulous work over small plots’. Activities such as monitoring water depth every 2 days, harvesting 1 × 1 m plots separately, counting panicles per m², or the number of filled and unfilled grains for 10 panicles was considered too much work by farmers (interview notes, 2014). Such meticulous data collection was required mainly for variety trials.

The low level of involvement by farmers as well as diverging interests on data or observed metrics, resulted in delayed reflection on the trial results. Meetings to discuss the findings and to plan activities for the next season generally occurred at least three weeks after the harvest was completed because farmers were busy with postharvest activities. The activity with higher farmer participation was the participatory variety selection (PVS), which included observing crop performance, sensory evaluation of cooked rice, and voting on preferred varieties. Researchers consolidated and presented results to farmers.

3.2.1 Learning Alliance activities, Case 2

The addition of a LA approach to Case 2 implied that besides the AR activities, which included varieties and seed-trials of interest to the LA, farmers were involved in testing new types of threshers. This failure; LA members therefore embarked on other activities.

As presented in Table 4, representatives from different groups took the initiative to discuss relevant topics, and start new activities. The AR trials on varieties were used to present the top three varieties selected by farmers. These were discussed with millers and government members including possible options to produce export-quality rice. A variety called Sin Thwe Latt was considered more marketable because the ‘millers know about it’, or ‘the government has opened possibility for export of this variety’ (LA meeting notes, 2014). One miller set-up a trial in collaboration with farmers, wherein he bought fresh grains, then processed it using improved postharvest practices. He then milled the grains to assess the grain quality and profitability for millers. This activity emerged from discussions between farmers and millers about the assumptions on quality of grains sold by farmers.

Table 4 List of Learning Alliance agenda from 2013 to early 2014 with actors involved in the activity in Maubin, Myanmar (Case 2).

Activity	Actors involved
Sharing insights from variety trial (linked with AR)	AR trial co-operators, local staff, farmers
Discussion on quality standards, varieties, price	millers from Maubin, DOA, farmers
Meeting between farmers, millers and traders	farmers, millers and traders
Reflection activity on varieties and markets
Exploring standards in wholesale market (Yangon)
Miller trial (buy from farmers, good pH practices)*	PPHDG
Introduce lightweight threshers	IRRI, thresher manufacturer (Bogalay)
Work with local manufacturer in MMR
Import units into MMR

Aside from *, all activities were coordinated by LA facilitators from IRRI.

WS = wet season, AR = Adaptive Research, MMR = Myanmar, PPHDG = Pioneer Post-harvest Development Group, DOA = Department of Agriculture, IRRI = International Rice Research Institute.

Prior to harvesting in early 2014, DOA officers and researchers met with farmers during organized field days. The visit was to explain the AR trials and allow farmers to observe the standing crop. Ideally, farmers shared their comments, and decided what they would implement next. This did not happen at the initial stages of the learning activities. In fact, only 17% in Case 1 (N = 54), and 41% in Case 2 (N = 49) had heard about the AR trials in early 2014.

Overall, with the LA approach more farmers were involved in discussing what was happening in the experiment plots. Of farmers interviewed in late 2014, 54% (n = 54) from Case 1 and 69% (n = 49) from Case 2 were aware of the ongoing trials and project activities. The number of land-owning farmers who joined activities was 27 in Case 1 and 28 in Case 2. There were 3 tenant farmers who joined activities in Case 2.

Farmers had a general idea of trials, recognizing for example a fertilizer trial, but could not reproduce details about treatments and comparisons. They felt they had no influence on what to try next. They said the IRRI and DOA extension staff suggested the trials and they just agreed. The extension staff also said they were not familiar with the exact protocols for some trials (interview notes, 2014).

Increased involvement was most notable in variety trials. After two seasons in 2014, farmers noted performance of the varieties, differences in yield, new techniques, and implementation of the trial (Table 5). Farmers also identified related aspects that they would like to know more about, or indicated why they would not practice the new technique.

Table 5 Comparison of observations by farmers from learning activities in 2014 after the first wet and dry season for rice production in two townships in Lower Myanmar; AR = Adaptive research; LA = Learning Alliance.

Category	Case 1: AR	Case 2: AR + LA
Varieties	PVS varieties were good but normal, not different from own field	Sin Thwe Latt has good eating quality, many plants per hill, tillers; panicles look good
	Liked HmawBi	Liked IR7754-90-111
	Liked Site Pyo Yay 1 (pulse variety)	Liked Sin Thwe Latt and IR7754-501
	Liked Yezin 11, Yezin 9, Site Pyo Yay 1 (pulse)	Saltol Sinn Htwe Latt was good: tall plants, good panicles; similar to own varieties
	Liked HmawBi 2	Liked Saltol Sin Thwe Latt: suitable duration and fit with own area
	Saw varieties for rice and pulse, don’t know much about it	Sin Thwe Latt is suitable for the area; IR7754-90-111 has good tillers and panicles, and high yield
		Liked the second most-preferred variety (don't remember the name)
		Saw two promising varieties
		Liked one variety but Nan Gar is still best for own field
		Sin Thwe Latt and Sin Thukha performance was good
		Wants to try Manaw Thukha 2 variety for monsoon
Yield	In comparison with monsoon, summer trials performed better	High yield in trial plots
Techniques	Interested in grain purity and grain cleaner	Some varieties would work in flooded area with tidal effects
	Regarding new variety: how much fertilizer and when? (wants to know; information is missing)	Knows seeds can be reduced by 1.5 baskets but still uses 3 − 5 basket; maybe it only works in small trial plots, too risky
		First time to use drum seeder (different from broadcast seeding)
		Good crop establishment: many plants despite low seed rate
Implementation	Do not know trial, not invited because don't grow summer rice	Only son and daughter joined, has not seen the trials
	PVS from monsoon different from summer PVS	Crossed trial plots but did not notice what was in it
	Did not see anything different or new	The meetings are informative: now farmers know they need pure seeds, and have machine options they did not know before
	Far from trial, only join with DOA activities	Wants to join the preference analysis next time
		Liked the trial because of scientific method

PVS = Participatory variety selection.

Local staff and extension workers also reflected on the design of the trials. They observed, for example, that ‘the positioning of the plots did not work for AWD trials’, ‘the weedy experimental side make my adjacent (own) field also look weedy (even if it is not)’, ‘farmers are not convinced with observations from small plots’ (interview notes, 2014).

Overall, the difference of the LA approach in case 2 was the interaction between farmers and millers. One miller explained to the farmers what grain quality of rice implied for their milling practice. A major concern for millers is mixed white and red grains, which can be ‘corrected’ by the millers with an additional milling operation. The miller showed farmers the calculation of additional costs and effects on the price. The example shows how the LA approach invoked a discussion about adjustment in related practices.

3.3 Comparing cases

3.3.1 Integration of learning topics

By the end of 2014, there were AR trials that were completely farmer-managed in Case 2, but not in Case 1. The farmer-managed trials were specifically on varieties. Aside from seeds, inputs or protocols were not provided for farmer-managed trials. Researchers only collected data on crop management and output.

Activities under the LA approach clearly increased the social learning in terms of exchange of information between a larger number of actors from diverse practices. The mentioned exchanges between farmers and millers was extended to traders and seed producers. Farmers interacted with millers and traders at the wholesale market in Yangon and visited a nearby seed farm. Moreover, the farmers of case two became more active in the on-farm experiments. Farmers implemented trials and activities on their own, with outcomes noted by the farmer and shared at LA meetings. They also interacted with farmers from another province, Shwebo, who were more experienced with marketing good quality Paw San variety, which is priced higher in markets. These interactions created incentives to further experiment on their farms. Some farmers in Case 2 decided to produce Sin Thwe Latt variety, hoping this variety would have a higher farm-gate price.

Farmers in Case 1 primarily interacted with the researchers. Researchers discontinued experiments with drum-seeders, for example, because trials did not yield positive results and farmers showed little interest. The experiments focused mainly on varieties, weed management, and fertilizer management. Although the differences between the two cases seem substantial in terms of social learning, differences in acquired knowledge do not automatically translate into differences in acquired skills or reconfigured practices.

3.3.2 Emergent adjustments in practice

An examination of what co-operators said they would do differently in 2015 shows how farmers in both cases were primarily focusing on further variety trials (Table 6). Although some co-operators indicated they had no additional plans, it seems that the experiments introduced under the AR and LA approaches induced further experimenting among farmers. As visible in Table 6, the addition of the LA approach implied a broadening of the items to experiment with. Adjustments in technical practices of farmers regarding seed rate, use of raised seed bed or separate nursery, fertilizer use, varieties, use of threshers, as well as market practices were put on the agenda. There were however, no significant adjustments measured at farm level due to the introduction of the technologies. Traders and millers had indicated interest for the varieties and buying in bulk from the farmers in the sites, but these were still being discussed and experimented as of 2015.

Table 6 Comparison between Case 1 and 2 of what co-operators will do differently after WS 2014: Data from co-operator farmers in Lower Myanmar.

Case 1: AR	N	Case 2: AR with LA	N
Own trial for 2 varieties (2.5 acres)	1	Grow local variety in half the farm to test if seed is adaptable	1
Own trial Yezin 11 (1.5 acres), and Yezin 2 (1 acre)	1	No plan for monsoon; for summer crop plans to use drum seeder for 5–6 acres to manage weed problem	1
Own trial for Yezin 11 and Yezin 9	1	Test Yezin 2 (<1 acre) to compare with own variety	1
Nothing because yield is same as local variety	2	Own trial for some varieties; noticed that having fertilizers is better than no fertilizers (not Foliar only) for pulse (rainfall at seedling stage still a problem)	1
Nothing	7	Use good seeds	1
		Own trial for pulse	1
		Nothing	1
		Plans to store to wait for higher price	1

Results from the interviews reveal an important factor that obstructs farmers from growing different varieties. Several farmers, not only co-operators, who showed an interest in trying new varieties, mentioned that acquiring seeds was a major concern. Most farmers (76%) used their own seeds for about three years. Eight farmers from Case 2 used seeds that they sourced from seed farms linked through the project. None of the farmers from Case 1 did this although there were seed farms nearby. The expressed concerns of the farmers may reveal that seed production, as an off-farm activity, implies another practice that requires adjustment when farmers change the varieties they grow. Interviews with seed producers and sellers show choices about varieties offered as seed are based on large demand of a few varieties.

The concerns expressed by farmers, presented in Table 7, show a number of other specialised activities, such as transport and credit, which may be part of the practices of traders or may constitute specific practices. The concerns also show the awareness among farmers that innovation implies changes in the required skills to make an introduced technology operational. Farmers may not have the time to develop the various skills related to a change in variety all at once; and certain skills can only be developed with support from outside. For example, the mentioned concern of leaf rot disease, implies a need to develop the skill to control this disease, which may need chemicals to be retrieved from agro-shops, a training on effective control measures, or both. In case the scope for such skill development is limited for any reason, this may imply that despite all the things farmers know about the new variety and its benefits, they will not reconfigure their farming practice around the new variety.

Table 7 Activities planned and concerns discussed by learning alliance in 2014 (Case 2), from farmers in Maubin, Myanmar.

Activity to try in 2014	Concern/s
Produce pure seeds	Available seeds for farmers, market price for new variety
Use the tools (e.g. drum seeder)	What are the tools/machines suitable for the technique? Money to buy equipment
	Needs skilful labor
Market	Safe credit support (risk with fertilizers bought on credit)
	No fixed market for the harvested crop
Plant Sin Thwe Latt variety	How to control leaf rot disease (in times of heavy rain?)
Trial of varieties suitable for area	Different soil types (different farming systems)
Establishment methods	How does it work for different land level and soil type
Which fertilizer should be used	How to do this? Need to learn the techniques
Herbicide/pesticide trial	Need to learn (knowledge to care for new variety)
Transportation from field to home	Poor transportation; still a problem
Thresher (fast, can finish threshing crop from 5 to 10 ac/day)	Not enough time for threshing in short period and for sowing pulses

4 Discussion

The main purpose of AR is to facilitate interactions between researchers and farmers. This entails a process of first defining a technical agenda based on needs assessments of farmers, benchmark data and researchers’ knowledge. The on-farm experiments trigger a learning process that is different for farmers and experts. Our results show that local research and extension staff express a learning effect on operational details of the experiment. Even though the experts frequently interacted with the farmers and shared results, the involvement of the farmers in the experiments was fragmented and limited. This suggests that the concerns of doing the experiment correctly overruled the attempts to include farmers in the experimental activities. A study by Henke (2000) shows that field technicians arranging on-farm experiments strike a balance between making the experiment ‘lab-like’ through standardizing data collection and making them ‘field-like’ so that meaningful results are obtained. However, such experiments are primarily focused on collecting meaningful data for researchers. Sharing results with farmers and formulating advice on how to manage their farms requires additional activities. Our results suggest that the AR experiments in Myanmar may have been enough field-like but perhaps not enough farm-like. In other words, the experimental design included the agro-ecological conditions in farmers’ fields but the operational procedures hardly mirrored the way farmers operate in their practice. Consequently, experiential learning was limited for farmers. The LA approach in Case 2 encouraged farmers to take the lead in subsequent experiments. Our results show that most farmer-led experiments continued with the variety tests, as initiated by the AR. Although the learning effects for farmers are likely to be higher in terms of acquired knowledge about different varieties, factors obstructing the uptake of these varieties in the farmers’ fields impeded farmers from using that knowledge for situated learning.

The addition of the LA approach brought in more stakeholder groups in the interaction and created visible differences in the learning process. The varied stakeholders with differing interests represent, in terms of situated learning, different communities of practice (CoP). What binds these CoPs is the processing of rice. However, each of these practices involves different rice-related activities. In each practice different properties of the rice crop become salient and, vice versa, different features of the crop enable and constrain different activities for the practice (Glover et al., 2017). The level of homogeneity of seeds produced at seed farms, affects the capacity of farmers to produce a crop that meets the demands of millers. The involvement of a wider set of actors, resulting from the combined AR and LA approach, increased the number of interactions. As van Mierlo et al. (2010) argue, a certain level of agreement and coordination between stakeholders is required in order to realise learning effects through the network or system. The notion of situated learning helps to explain that such agreements also require operational coordination. In other words, the coordination implies adjustments of functional linkages between various practices, through mutual reconfigurations and development of skills.

Over two years, there were important differences in the learning agenda between the two cases in Lower Myanmar (Fig. 1). The LA broadened the learning topics in Case 2. It not only tackled a greater variety of technical issues and topics, but also expanded experimentation into institutional topics such as standards, markets and collective action. Studies have noted that assembling a network of actors may not necessarily benefit the learning process (Beers et al., 2014). Our findings show AR could engage farmers and researchers for learning about on-farm agronomic solutions, but engaging wider stakeholders can further the experimentation into concerns that support innovation on these technical solutions.

Fig. 1 Comparison of learning agenda between Case 1 (Adaptive Research) and Case 2 (Adaptive Research with Learning Alliance) from 2013 to 2015 in Lower Myanmar.

The yellow/shaded boxes refer to the learning topics of the LA. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Findings on the learning process point to important differences in network formation between the two cases. There were observed advantages in facilitating interaction outside the immediate CoP of farming. Primarily, it was to allow more opportunities for other actors to steer learning towards enabling conditions for farmers to use technologies of interest. The interactions brought out tensions in the form of diverging perspectives and views, which are important for learning collectively. Learning about the performance of a variety on-farm, producing quality grains from it through better postharvest management, and observing how this would sell in the market was one integrated agenda for example (Fig. 1). In contrast, where only researchers and farmers interact (Case 1), the focus of learning was limited to on-farm technical concerns–topics of interest to influential CoPs involved. This suggests the LA approach can widen the learning process, as several studies show (Hounkonnou et al., 2006; Lundy, 2004; Stür et al., 2009).

Increased involvement of stakeholders and engaging a broad network however, may not always be the most efficient practice. When the network process started, topics aligned with the technology preferences from AR. The addition of a LA approach implied a selection of stakeholders based on an ex-ante assessment of practical relevance. The notion of situated learning implies a closer examination of the linkages between experiments and stakeholder meetings. Stakeholder meetings thus require a continuous evaluation and asking which members of which CoPs to include at what moment. This may also allow adjustment of participation over time. For example, involvement of bankers or pesticide sellers may become useful once results from experiments suggest farmers will opt for different varieties, requiring different pest management and credit schemes.

Timing is also important with respect to the technologies experimented with. Our results show that the stakeholder meetings in Case 2 resulted in a broader array of topics brought up in the meetings. Researchers had a strong voice in reframing issues and creating focus in topics for experiments. Concerns of farmers and other stakeholders were nevertheless important. Our results suggest that practical concerns played an equally important role. Not only does available time and resources limit what can be experimented with, cropping calendars and growing seasons also determine what can and what cannot be done at certain moments. In more general terms, the scope for experimentation and practical complications of experimenting are major factors in making decisions. This is underlined by the argument made by Richards (2004) that practical performance is as importance for the participation process as deliberation. Consequently, the added value of an LA approach would lie in an increased focus on the experimental process. Although this necessarily implies enough scope for reflection and discussion, deliberations should be guided by the experiments rather than the other way round.

These insights, as well as the nature of on-farm experiments, discussed above, also raise questions about how involvement in experimenting is arranged. The way the AR was set up in both cases did not create much space for the experience and knowledge of farmers. The experiments were mainly run by local staff and DOA officers and, consequently, the learning effects for farmers mainly emerged through the brokering activities of the experts rather than through self-discovery. Limited participation was not the only issue since attention to the quality of participation was also important (van de Fliert et al., 2010). In another case of AR from Indonesia, implementers were more aware of participatory engagement with farmers, and engaged farmers better during implementation of the trials (Flor et al., 2016). Thus, where farmers were more involved to make their own observations, they built upon their knowledge to make modifications on the technologies. Clearly, the introduction of participatory research approaches in an environment where people are used to top-down approaches required a learning and adaptation process of its own. Emphasis on experimentation and the insights out of these beyond research data concerns are also relevant, pointing to the levels of control of experiments by researchers, and hence the need for researchers themselves to adapt to different cultural conditions.

There are indications that in Case 2, where the LA was added, the degree of involvement and ownership of stakeholders increased more rapidly. Some members of the alliance started more quickly to define their own activities, trials, and agenda based on the reflections from previous learning cycles. It resulted in an agenda where specific technologies were integrated with other learning topics.

4.1 Limitations

Our findings are based on a comparison of only two cases, which implies that we need to be very cautious in drawing conclusions about the added value of Learning Alliances. While the projects in the case-study areas were carried out by the same organisation, by the same international researchers and under very similar ecological conditions, there could well be differences in the historical context, infrastructure, and/or local people involved that have shaped the differences that we have signalled above. Nevertheless, the kinds of differences we find are largely congruent with the processes and objectives strived for by proponents of Learning Alliances and the shortcomings of participatory research activities that address only locally-specific technical issues (Hounkonnou et al., 2006).

5 Conclusion

In this article, we examined the dynamics that emerged in the context of a project using approaches to enhance learning and adaptation of rice-farming technologies in Myanmar. We compared a case of only Adaptive Research with another case combining AR with a Learning Alliance approach. Our findings suggest that the implementation of AR could miss opportunities to facilitate learning if there is extensive control on experimentation. Although AR has potential to support agronomic learning, the way in which AR was implemented was not optimally conducive to experiential and discovery learning. There were indications that technologies trialled did not always match with farmer’s context and interest.

The inclusion of LA in Case 2 expanded the number of stakeholders with whom farmers interact. With this broader network, the learning agenda also expanded out of the initial concerns or interests targeted by the project. Although having a LA had added value in including a wider variety of stakeholders and increasing the scope for farmers to engage with other stakeholders, this also created limitations. The greater autonomy and informality of the learning process through LA implied that lessons were not always explicated and shared. This may well have led to a learning agenda that fitted only a specific segment of farmers.

Our analysis, based on the notion of situated learning, reveals that practical concerns were a major driver in the participatory process. Practical concerns imply the technicalities and required skills to be able to make experimentation meaningful and, consequently, to incorporate the lessons of experiments through the reconfiguration of the agro-ecological setting and, therewith, the development of skills. This not only implies adjustments in the practice of growing rice but also in related practices such as buying rice and milling.

We conclude that the inclusion of more actors in LA or similar participatory approaches should not merely facilitate deliberations between actors but also support the reconfiguration of practices other than farming and in establishing linkages between practices. Nonetheless, reducing control from the research side and allowing a wider set of stakeholders to engage and guide the agenda and flow of learning is likely to be conducive for aligning interdependent practices.

Acknowledgements

Funding for this research was provided by the Australian Centre for International Agricultural Research (ACIAR) through the MyRice Project, the United Nations Office for Project Services (UNOPS) through the Livelihoods and Food Security Trust Fund Project, and the Swiss Agency for Development and Cooperation (SDC) through CORIGAP.