1Participatory crop improvement strategies in rice in the DFID Plant SciencesResearch Programme

JR WITCOMBE

Centre for Arid Zone Studies, University of Wales, Bangor, Gwynedd LL57 2UW, UK

Formal trials produce missing values indroughted conditions

There is a well-established system of multilocationaltesting of varieties on research stations to identifyvarieties that should be released and recommended.A common public-sector trial design is a randomisedcomplete block design with three replicates. Eventhough there are satisfactory statistical techniquesof dealing with missing values, it is difficult for theseto be accepted in a coordinated trial that involvesentries from several breeders and institutions.Moreover, once only a few entries actually producegrain (Figure 1), then a universal practice is todiscard the trial completely. Unfortunately, thismeans that the missing data are not considered, yetthey are not really missing. Instead, they providevaluable information that, under severe drought, thevarieties will fail.

PVS

Participatory research - in particular PVS - offers away out of this dilemma. If a variety fails in afarmer’s field and the local variety survives, thenthis is valuable information that can easily beaccommodated in assessing varietal performance.

Unfortunately, participatory research is notsimple, and the history of obtaining poor-qualitydata from 'minikits' is a deterrent. What is neededis reliable data from participatory trials. For

scientists in the formal system this means:

• Having a partnership with an NGO or extensionservice that has already built a relationship withfarmers.

• Staff dedicated to research within the partnerorganisation. (This avoids a conflict betweenpriorities for development activities and thosefor research).

• A systematic approach to participatoryevaluation such as a mother and baby trialssystem as pioneered by Snapp et al. 1999.

In PVS farmers are given a wide range of novelcultivars to test for themselves in their own fields.

A successful PVS programme has four stages:

1. Participatory rural appraisal (PRA) to identifyfarmers’ needs in a cultivar;

2. A search for suitable material to test withfarmers;

3. Experimentation on their acceptability infarmers' fields;

4. Wider dissemination of the cultivars farmersprefer.

The Department for International Development Plant Sciences Research Programme has fundedparticipatory crop improvement (PCI) projects over many years. PCI includes farmer participationin all methods of crop improvement, both genetic and agronomic, that place emphasis on farmerparticipation. Most of them have included rice as an important, sometimes the only, crop.

This paper describes the basics of two methods of PCI - participatory varietal selection (PVS)and participatory plant breeding (PPB). Initially, most of our PCI projects involved only PVS, inwhich farmers select among a choice of varieties that they can grow and evaluate on their ownfarms. These projects include ongoing programmes in Ghana (see Craufurd et al., this volume),Nepal, and India (in the Punjab and Gujarat). With the success of PVS in these projects, work beganon participatory plant breeding (PPB) in both Nepal and India. The results of these PPB programmesare described by Gyawali et al. (this volume) and Virk et al. (this volume). This paper reviews thereasons underlying a participatory approach, and outlines some of the methods that can be used.

Abstract

It is important tonote that this process isnot simply a relabellingof old techniques such asf r o n t - l i n edemonstrations orminikits. Traditionalapproaches do not startwith a PRA; they offerlittle choice in newvarieties but only the fewthat have been selectedafter years of formaltrials; they tend toinvolve only a fewfarmers; andmanagement isimproved by arecommended packageof practices that isbeyond the resources andrisk-taking capacity ofmost resource-poor farmers. Indeed, some of thepapers presented in this volume are initial attemptsat PVS and the reader can judge whether they aresimple relabelling. With time and experience thebenefits of more participatory approaches shouldbecome apparent.

In our mother and baby trials system all thevarieties are grown in mother trials in a one-field,one-replicate design. Typically, there will be aboutfive to six mother trials. Baby trials are morenumerous and each farmer grows a single entry andcompares it to his or her local variety. In the mothertrials quantitative estimates of yield are obtained,but in the baby trials we simply collect farmers'perceptions on yield. A full protocol is available onthe Plant Sciences Research Programme website(www.dfid-psp.org).

One of the great strengths of PVS is that it is anextension method as well as a research method. Forexample, PVS trials in upland rice in Ghana resultedin a dramatic spread of new varieties to new villagesover a single season (Craufurd et al., this volume).

PPB

The breeding strategy that has been followed in ourPPB programmes in Nepal and India involves:

• Making a careful choice of parents (often usingPVS to help identify them);

• Making only a few crosses;

• Using a large population size in each cross;

• Selecting in the target environment with theparticipation of farmers;

• Employing PVS to test the products derivedfrom the PPB programme.

Reason for the approachThe capacity of any breeding programme is limited,and as more crosses are made the populations derivedfrom them have to be smaller. However, theoreticalconsiderations provide strong arguments for usinglarge population sizes (Witcombe and Virk, 2001).

Hence, one possible breeding strategy is to selecta few crosses that are considered most likely to givedesirable segregants and produce large populationsfrom them. This strategy suits well the constraintsandadvantages of PPB:

• PVS aids the selection of parents. It is effectivein identifying locally adapted parental materialand in identifying breeding goals - for example,early maturity - that assists the selection ofcomplementary parents.

• Participatory breeding programmesconducted by NGOs will not have manyresources to devote to technical processes suchas making many crosses.

• Large population sizes are easy to deal withwhen these are grown by farmers. For example,in collaborative breeding a farmer can cost-effectively grow and select in a very largepopulation of rice. The farmer was, in any event,

Figure 1. One of the All India Coordinated RiceImprovement Project(AICRIP) Advanced Varietal Trials for very early, direct-sownupland rice. Photograph taken at Agricultural ResearchStation, Borwat, Rajasthan, in September 2001. Only a few entrieshave produced panicles and all the remaining entries will fail toproduce grain.

going to grow rice, and if the PPB material yieldsadequately, costs (or benefits) are only thedifference between the yield of the populationand the yield that the farmer would normallyhave obtained from his or her own variety.

There are two further important benefits fromfarmer participation:

• Selection is carried out in the target environment(minimising the untoward effects of genotype xenvironment interaction) and the selection is fortraits that farmers consider important. Whenbreeding for drought-prone environments inparticular, conventional multilocational trials aredifficult to analyse. The trials in the mostdrought-stressed environments produce manymissing values and are often excluded for thisreason, despite such trials being the most relevantfor farmers. Participatory trials do notsuffer from this disadvantage. Indeed, when avariety fails on a farmer’s field this givesvaluable information on the potentialacceptability of the variety and not a missingvalue.

• There is a very great reduction in the time neededto breed and test a new variety. Immediately anew product emerges from the PPB programmeit is included in PVS, and rapid feedback on thedesirability of the new material is gained.

Contrast to conventional breedingThe use of few crosses with large population sizes isnot a common strategy. Many breeding programmesuse many more crosses and hence restrict the size ofthe F2 populations that are evaluated. Depending oncircumstances, such a strategy may be correct. Theoptimum number of crosses will differ on howcompetitive the breeding is, how targeted it is to aspecific environment, the type of parental materialused, and whether the breeding can be considered asstrategic or adaptive.

Experimental evidence that the approach is effectiveWe now have empirical evidence that making fewcrosses in rice in PPB programmes is effective. InNepal only three crosses are at a sufficient stage tohave produced advanced lines. All three crosses haveproduced highly promising material, and farmershave already adopted several of them.

In India, three crosses are also at a stage whereadvanced lines have been produced, and again, allthree crosses have been successful. If a low-cross-number strategy were incorrect it would be unlikelythat of the five crosses attempted so far (one cross iscommon to both Nepal and India) that all would have

been successful.We are aware that some of the parents we have

used in our PPB programmes have been employedin many crosses in conventional breedingprogrammes, but have not produced advancedmaterial. We assume that the most likely explanationis that F2 and F3 population sizes have been too smallto recover and select desirable segregants.

However, PVS is limited to employing theexisting variation among cultivars, and sometimeswell-accepted cultivars cannot be found. PPB, inwhich farmers select from segregating material, isparticularly desirable when the PVS has failed toidentify new varieties with significantly greaterutility than existing ones. In our PPB programmeswe exploit the results of PVS by using identifiedcultivars as parents of crosses. Weaknesses incultivars are identified in the PVS programme andthey can be crossed with varieties that havecomplementary traits to eliminate those weaknesses,perhaps a high-yielding but low-grain-quality varietywith one with superior grain characteristics, asexemplified by the cross Radha 32 x Kalinga III(Gyawali et al. this volume).

What we have found is that PVS and PPB areused in combination. We start with PVS and thathelps to identify parents, then we carry out PPB. Assoon as there are products from this PPB, we testthem in PVS trials. This is a continuous processbecause new varieties, whether introduced fromclassical breeding programmes or from PPB, arealways becoming available and can be tested by PVS.

References

Craufurd et al. this volume.Pandey S, and Rajatasereekul S (1999). Economics of

plant breeding: The value of shorter breeding cyclesfor rice in Northeast Thailand. Field Crops Research64: 187-197.

Gyawali et al. this volume.Virk et al. this volume.Witcombe, JR and Virk, DS (2001). Number of crosses

and population size for participatory and classicalplant breeding. Euphytica 122:451-462.

We have adapted PPB methods to take advantageof the strengths of breeders and farmers. Thebreeders produce material that is geneticallyhomozygous but highly heterogeneous byadvancing the bulk populations from the F2 tothe F5 generations with minimal selection (Figure2). This means that we give bulks to farmers at aquite advanced generation when it is expectedthat there will be a good response to selectionbetween plants, and when segregation in the nextgeneration is no longer a major complicatingfactor.

A key element of PPB is the collaborativeparticipation of farmers who grow a bulk on theirown fields and select amongst it (see Gyawali etal. and Virk et al. this volume). Using thiscollaborative breeding, it is possible to replicateselection cost-effectively by giving seed of aparticular bulk to many farmers. The selection isthus replicated across physical environments(different farmers' fields) and across farmers(who may have different selection strategies andselect for different traits that best meet theirneeds).

One great advantage of PPB is that it is muchfaster than conventional breeding (Virk et al. thisvolume). The economic value of this reductionin time can be very large (Pandey andRajatasereekul, 1999).

FF22

FF55

FF44

FF33

FF66

FF77

FF11

ScientistScientist

FarmerFarmer

FF88 PVS/Breeder's trialsPVS/Breeder's trials

xxGenerationGenerationYearYear

33

33

22

44

44

55

22

11

11 PP1 x 1 x PP22

Figure 2. A schematic diagram of a PPBprogramme. Breeders control theprocess until the F4 generation,and then farmers collaborate fromthe F5 generation onward.Breeders include selected bulks informal trials from the F8generation. It is assumed that twocrops of rice are grown per year.

An example of PPB methods

2Participatory plant breeding (PPB) in rice in Eastern India - the success ofan NGO/GO partnership

DS VIRK1, DN SINGH2, R KUMAR2, SC PRASAD3, JS GANGWAR3, JR WITCOMBE1

1. Centre for Arid Zone Studies (CAZS), University of Wales, Bangor, UK2. Department of Plant Breeding and Genetics, Birsa Agricultural University (BAU), Ranchi, Jharkhand, India3. Indo-British Rainfed Farming Project, Gramin Vikas Trust (GVT), Kanke Road, Ranchi, Jharkhand, India

Introduction

Farmers in the Chotanagpur plateau of eastern Indiastill predominantly grow rice landraces. There areseveral reasons why the green revolution hasbypassed these areas. There has been less investmentin agricultural research for more marginal areas, andmodels of plant breeding that have worked for moreproductive, irrigated areas have been applied withoutany adaptation to harsher, lower yieldingenvironments. In addition, the needs of farmers innew varieties have been little considered. We usedmore participatory approaches to better meet theneeds of farmers, and they are outlined in this paper.

Release of rice varieties from PPB

The first ever superfine, high-yielding and early-maturing rice varieties have recently been releasedfor rainfed uplands in eastern India, following ahighly successful participatory plant breeding (PPB)programme. The programme was based on a uniquepartnership between an Indian non-governmentalorganisation (NGO), an Indian governmentalorganisation (GO), and a UK university funded bythe DFIF Plant Sciences Research Programme.

The PPB programme focused on the strategy ofusing carefully chosen parents, few crosses and largepopulations (Witcombe and Virk, 2001). Parents forcrosses were selected by participatory varietalselection (PVS) so that at least one parent was locallyadapted. The second parent was chosen to havecomplementary traits and be high yielding andgenetically distant in order to target a range of rice

ecosystems. We have used bulk-population,pedigree-bulk and modified-bulk populationbreeding.

Three crosses were made: Kalinga III/IR36,Kalinga III/IR64 and Kalinga III/Vandana. Of these,the Kalinga III/IR64 cross is the most advanced andis described below.

A cross between Kalinga III and IR64 was madeby the International Rice Research Institute (IRRI)at the request of the Centre for Arid Zone Studies(CAZS). The F2 seed of the cross obtained from IRRIwas grown in the rainy season of 1997 at the GVT-BAU research farm, Ranchi, Jharkhand, andadvanced in the off-season of 1997-98 at the CentralRice Research Institute, Cuttack.

In the subsequent F4 generation two types offarmer participation were employed:

• Collaborative participation: Farmersparticipated collaboratively using bulk-population breeding. Farmers were providedwith the entire F4 bulk from the cross to growin their own fields and make selections in themain season of 1998.

• Consultative participation: This involvedfarmers visiting the research station to evaluateF4 pedigree-bulk lines grown by scientists at theGVT-BAU research station in Ranchi, in themain (rainy) season of 1998. These farmers,both male and female, came from villagesserved by the GVT Rainfed Farming Project.

After advancing the material by one generation

We describe a participatory plant breeding programme (PPB) for upland rice targeted at the marginalareas where farmers appreciate very early maturing varieties that can escape terminal drought. ThePPB programme built on the results of participatory varietal selection (PVS) by using a variety identifiedin a PVS programme - Kalinga III - as a parent. Two types of farmer participation were employed -consultative and collaborative - and both were successful in producing varieties superior to thosecurrently available.

Abstract

importance of testing for quality before enteringvarieties in yield trials to save resources (Gyawaliet al., this volume).The varieties performed better in mother and babytrials (see Witcombe, this volume) particularly so inthe baby trials that were closest of all to farmers’conditions (Table 2).

As a result of their good performance in thetrials, BAU released Ashoka 200F (as Birsa GraminDhan 109) and Ashoka 228 (as Birsa Gramin Dhan110) in May 2001 for general cultivation inJharkhand in rainfed uplands. The breeding strategyof using a few crosses and large populations withboth consultative and collaborative participation hasrapidly produced improvements over Kalinga III, themost widely adopted upland rice variety in India.

PPB is also producing promising genotypes forsituations other than uplands. In F4 population bulksof two crosses we monitored farmers’ selection intheir fields in Jharkhand and West Bengal. Onefarmer, Saikya Mahoto, of Jhabra village in West

Table 2. Performance of Ashoka 200F and Ashoka 228 compared to the control varieties, BirsaGora 102 and Kalinga III, in the main season 2000 and 2001

1. BG 102 yield 1.08 t ha-1 2. Kalinga III yield 1.17 t ha-1

3. BG 102 yield 1.26 t ha-1 4. Kalinga III yield 1.33 t ha-1

during the dry season of 1998-99, the two most-selected pedigree bulks on the research station(Ashoka 228 and Ashoka 238), and one selected bulkpopulation produced by a farmer from an F4 bulkgrown in his field (Ashoka 200F), were promoted toformal research trials. These trials, conducted in themain season of 1999, were the state trials inJharkhand and the All-India Co-ordinated RiceImprovement Project trials (Initial Variety Trial -Very Early).

The performance of Ashoka 200F and Ashoka228 in the varietal trials at BAU, Ranchi wasoutstanding (Table 1). In these trials, both newvarieties matured in less than 95 days and were moreresistant to lodging than Kalinga III. They haveslender-grains and good cooking quality -particularly Ashoka 228. Both varieties are likedby farmers for the rainfed uplands.

Although Ashoka 238 performed about the sameas the other two varieties in yield trials it was droppedbecause of its poor grain quality. This illustrates the

Table 1. Performance of Ashoka 200F and Ashoka 228 compared to the control varieties, BirsaGora 102 and Kalinga III, in the main season of 1999, 2000 and 2001

Entry Yield (t ha-1)

Yield (% of BG 102)

Yield (% of Kalinga III)

Mother trials Ashoka 200F 1.15 23 18 Ashoka 228 1.19 56 23 Birsa Gora 102 0.97 - - Kalinga III 0.76 - - Baby trials Ashoka 200F v BG 102 1.52 411 - Ashoka 200F v Kalinga III 1.46 - 252 Ashoka 228 v BG 102 1.64 303 - Ashoka 228 v Kalinga III 1.57 - 184

Entry Yield (t ha-1)

Yield (% of BG 102)

Yield (% of Kalinga III)

Ashoka 200F 2.58 28 19 Ashoka 228 2.55 27 18 Birsa Gora 102 2.01 - - Kalinga III 2.16 - -

FF66 20002000

FF55 19991999

FF44 19981998

FF7 7 20012001

FF8 8 20022002

Bengal selected semi-dwarf segregants in theKalinga III/IR64 cross that were high yielding andearly maturing, to produce a promising variety forthe medium-upland rice ecosystem (Figure 1). Thefarmer used multiple-selection criteria such as plantheight (68 cm), early maturity (100-115 days),disease/pest resistance, no lodging, droughttolerance, more straw, high yield and superior grainquality.

It can be seen from Figure 1 that by the F7generation in 2001 the uniformity of the variety is,at the least, acceptable. Saikya Moto selected thisvariety from the F4 generation starting in 1998 andhas continued with it until at least 2002.

This is an example of the self-motivated interestthat farmers can take in collaborative PPB, and howcost-effective the process can be for both the farmer(who gains early access to superior material) andthe breeder (who can replicate selection acrossenvironments and farmers with limited input fromsalaried staff).

Another farmer in Jharkhand is continuing togrow selections from the Kalinga III/IR36 bulk forthe same ecosystem. He applied a mild selectionpressure, employing all criteria used by the WestBengal farmer. The resultant variety still retainsmuch variation. These farmer selections will beevaluated in on-station and on-farm trials in the mainseason of 2002.

References

Gyawali et al. this volume.Witcombe, this volume.

Figure 1. Saikya Mahoto of Jhabra village, West Bengal, in his selectedF7 bulk of the Kalinga III/IR64 cross, rainy season, 2001.

Witcombe, JR and Virk, DS (2001). Number ofcrosses and population size for participatory andclassical plant breeding. Euphytica 122:451-462.

3Participatory plant breeding in rice in low-altitude production systemsin Nepal

SANJAYA GYAWALI1, KRISHNA D. JOSHI2 AND JOHN R. WITCOMBE3

Introduction

We describe a participatory plant breeding (PPB)programme for low-altitude rice-growingenvironments in the inner terai of Nepal. Althoughthe target population of environments (TPE) wasinitially the rice fields of high potential productionsystems (HPPSs), less favourable rainfedenvironments or those with limited irrigation waterwere increasingly targeted.

The objectives of the project were to test theapplicability of PPB in more favourableenvironments, to develop varieties of a greater utilityto farmers through PPB, and to promote them morewidely in the target environments using participatorymethods.

Breeding strategy

The PPB programme used only a few crosses fromwhich large populations were produced (Witcombeand Virk, 2001). This was because it fits well withthe particular constraints and advantages of workingwith farmers, and because the approach is soundlybased in theory. Because few parents are employed,their choice is crucial. Participatory varietal selection(PVS) can help greatly in this process because it canidentify both parents and important target traits.Carefully chosen parents were crossed and a largepopulation size maintained in subsequentgenerations, all of which were grown in farmers'fields in different production systems, from uplandto lowland, in both the early (Chaite) and mainseasons.

We have used bulk, modified-bulk and pure-line-from-bulk breeding methods (Fig. 1). In the modifiedbulk method, a bulk population derived from a single

cross (Kalinga III/IR64) was divided at the F4generation into sub-bulks on the basis of traits thatwere known from participatory rural appraisals to beimportant to farmers - plant height, maturity and graintype (length x breadth ratio and length). At this stage,many of the sub-bulks were eliminated after farmerswere consulted on desirable grain shapes and targetmaturities.

The sub-bulks were still heterogeneous becauseseed was harvested from many selected F4 plants.Hence, farmers were given a heterogeneous bulk ofnearly homozygous plants for selection in their fieldsfor what we termed collaborative modified-bulkbreeding (Fig. 1).

The modified bulks were also used as sourcesfrom which pure lines were extracted for jointevaluation by farmers and plant breeders - consultativepure-line-from-bulk breeding in Figure 1.

The collaborative breeding programme allowedthe cost-effective replication of selection across a rangeof environments (farmers’ fields in several riceecosystems) and across farmers who employed varyingselection criteria and techniques.

Overall, the modified-bulk breeding programmehas produced a range of promising genotypes for allof the target rice ecosystems. Farmers are involved inthe evaluation of these entries using standard PVStechniques. We often refer to this as ‘PVS on PPBproducts’ to distinguish it from the initial process ofPVS that aided parental identification, or from PVSon introductions from classical breeding programmes.

Results

Collaborative breeding has produced at least onepromising varietyOne of the most promising entries for the Chaite

Abstract

Since 1998 we have undertaken a participatory plant breeding (PPB) programme in rice targeted atmore favourable rice-production environments in low-altitude areas of Nepal. We have employed fewcrosses from which large bulk populations were derived, and the bulks were selected using variousmethods and differing degrees of farmer involvement. Results showed that collaboration with farmerswas effective for improving the efficiency and cost-effectiveness of the breeding programme. Severalvarieties have been produced by the programme for contrasting niches that have performed better thanexisting check varieties in participatory trials.

season in 2001 participatory trials was Judi 141F, afarmer-selected bulk from sub-bulk ET (Early tall)of the Kalinga III/IR64 cross. It was preferred in thesetrials to Chaite variety CH 45, the most widely grownvariety in this season. It was greatly preferred to thebreeder-selected ET bulk because the breeders' bulkwas later maturing without having a higher yield.

This confirms the findings in Chaite rice of Joshiet al. (2002) that breeders can be misled into selectingfor a maturity that matches the most popular varietythat farmers currently grow. In pure-line-from bulkbreeding (Fig. 1) breeders' and farmers' selectionswere compared in a nursery of several hundredadvanced Chaite rice lines. Lines selected by neitherof the two breeders but only by farmers were allearlier maturing than CH 45. In all cases, breedershad not selected them because they had a highincidence of disease.

Nonetheless, it underlines the higher emphasisthat farmers place on earliness compared to breeders.Farmers rarely selected entries that matured at thesame time as CH 45, let alone later entries, thusindicating their dissatisfaction with the long durationof this variety.

Early participatory testing of grain quality is cost-effectiveWe have also undertaken a breeding programme for

CollaborativeCollaborativebulkbulk

Collaborativemodified-bulk

Consultativepure-line-from bulk

FF22

FF33

FF44

M1M1 M2M2 M3M3FF55

����������������������FF66

FF77

FF88 PVS Trials PVS Trials PVS Trials

P1 P2x

FF11

Fig 1. Breeding schemes employed in PPB, indicating the role of farmers and breeders.Dark-shaded boxes represent generations grown by breeders and light-shaded boxesgenerations grown by farmers. Hatched boxes indicate consultative evaluation ofnurseries grown by breeders on farmers fields. The scheme is generalised and detailssuch as the generations at which activities commence with farmers will vary withcircumstances.

quality rice based on a very large irradiated populationof Pusa Basmati-1. This breeding programme,however, cannot be simply described as a mutationbreeding programme, since molecular marker studies(Katherine Steele, CAZS, pers. comm.) have shownthat the original population had many segregantsderived from natural outcrossing. We found thatparticipatory organoleptic testing cost-effectivelyeliminated many entries before the yield trial stage. Itis much cheaper to eliminate an entry for poor grainquality (grain shape, milling percentage, cookingquality, and taste) than to eliminate it for poor fieldperformance in either participatory or on-station trials.

Selecting for disease resistanceIn PPB it is important to select strongly for diseaseresistance and, during the initial generations advancedby breeders, resources are devoted to removing plantswith disease symptoms from the bulks. In collaborationwith the Nepalese Agricultural Research Council(NARC) the advanced PPB entries are tested in diseasenurseries (leaf and neck blast, and bacterial leaf blight)and none of the 15 lines tested so far has been rejectedfor disease susceptibility.

This result is a combination of exerting strongselection pressures and using crosses where bothparents were disease-resistant under terai conditions.We would not recommend using resistant x susceptible

e crosses for PPB unless there are adequate resourcesfor disease screening.

Conclusions

With limited resources and no research stationfacilities the PPB programme, conducted entirelyon farmers' fields, has produced 11 varieties for arange of ecosystems that have been preferred overfarmers’ current varieties in participatory trials(Table 1). These varieties are under seedmultiplication and there acceptability is being testedin further trials.

In collaboration with the Department ofAgriculture and several NGOs, seed is beingsupplied to the informal seed supply system in 18of the 21 districts in the Terai of Nepal.

References

Witcombe, JR and Virk, DS. 2002. Number ofcrosses and population size for participatory andclassical plant breeding. Euphytica 122: 589-597.

Joshi, KD, Sthapit, BR, Subedi, M and Witcombe,JR. 2002. Participatory plant breeding in ricein Nepal. in: DA Cleveland and D Soleri (Eds).

Farmers, scientists and plant breeding, intergratingknowledge and practice. CABI, Wallingford. pp239-267.

†U = upland; MU = medium upland; ML = medium lowland; L = lowland. §Also suitable for upland main-seasonconditions. ‡Pusa Basmati-1

Table 1. Promising genotypes in participatory trials, main season 2001 and Chaite season 2002,from the PPB programme and the ecosystem and seasons to which they are adapted.

Variety Pedigree First tested in PVS

Ecosystem† Grain type

Chaite Judi 141F§ Kalinga III/IR64 2002 U Medium coarse Judi 503 Kalinga III/IR64 2002 MU, ML Medium coarse Judi 580 Radha 32/Kalinga III 2002 ML, L Coarse long Judi 582§ Radha 32/Kalinga III 2002 ML, L Coarse long

Main season Sugandha 1 Irradiated PB-1‡ 2001 U, MU Medium coarse, aromatic Barkhe 1027 Kalinga III/IR64 2001 U Medium coarse Barkhe 2001 Irradiated PB-1‡ 2001 MU, ML Medium fine Barkhe 2026 Irradiated PB-1‡ 2001 MU, ML Medium coarse Barkhe 2027 Irradiated PB-1‡ 2001 MU,ML Medium coarse Barkhe 3004 Kalinga III/IR64 2001 ML Medium fine Barkhe 3010 Kalinga III/IR64 2001 ML, L Medium

4The role of participatory crop improvement for upland rice in Ghana

W DOGBE2, R BAM3, PQ CRAUFURD1, K MARFO3, P DORWARD1, A OPOKU-APAU3, IBIMPONG2, D DJAGBETY2, K GYASI2, E ASIEDU2

1. Plant Environment Laboratory, The University of Reading, Reading, UK RG2 9AD2. Savanna Agricultural Research Institute (SARI), PO Box 52, Nyankpala, Ghana3. Crops Research Institute (CRI), Kumasi, Ghana

Introduction

Rice is an important staple in Ghana and per capitaconsumption has increased steadily. However, mostof the increase in consumption is met from imports.Increasing rice production in Ghana is a priority.About half the rice area is in the upland andhydromorhpic ecosystem, divided equally betweenthe northern and southern parts of the country. Todate, few improved cultivars of rice adapted toupland or hydromorphic ecosystems have beenreleased and their seed is not readily available. Thispaper reports data and experience from a PVSprogramme initiated in Ghana in 1997.

Upland and hydromorphic rice productionenvironments and practices

In northern Ghana (Northern, Upper East and Upper

West Regions), which is in the savannaagroecological zone, rice is grown underpredominately hydromorphic conditions (i.e. somestanding water for part of the season). Mean annualrainfall is about 1000 mm, and falls between Mayand October. Drought is a major constraint. Rice isusually sown on ridges prepared by animal tractionor tractor. Weeding and harvesting are donemanually. Fertilizer is used, but there are no otherinputs. All rice is par-boiled during post-harvestprocessing. Both male and female farmers grow rice.A wide range of cultivars is grown in this region -up to 11 in one village, ranging in age from five to>50 years old, and none of them has been formallyreleased.

In southern Ghana (Volta Region in the east, andWestern Region), which is in the humid forest zone,rice is grown in inland valleys (rainfed-lowland) andon the slopes (uplands). Mean annual rainfall is

To date, few improved cultivars of rice adapted to upland or hydromorphic conditions have been releasedin Ghana, and seed of improved cultivars is not readily available. A formal Seed Release Committeeexists in Ghana but there is no multilocational testing system in rice to identify new varieties. Mostupland rice farmers continue to grow traditional O. glaberrima or old O. sativa varieties or landracesobtained from informal sources. High yield and weed competitiveness are important characteristics wantedby all farmers, male and female. However, female farmers also rank harvest and post-harvest characteristicshighly.

Since 1997, a PVS program has been implemented in one or two communities in the forest zone(Volta and Western Regions) and savanna zone (Northern, Upper East, Upper West Regions) of Ghana.Rice in the forest zone is pure upland while in the savanna zone it is hydromorphic. These programmesvary in whether they are researcher, NGO and extension or community managed. In Volta Region, wherethe PVS process started first, several varieties have been adopted and given local names (e.g. 'You'll notbe tired' to reflect the ease of threshing). Seed has spread rapidly, initially kin to kin both within andbetween neighbouring villages, and seed has also been sold on the open market at a premium.

In fora where farmers, extension officers and plant breeders have discussed these participatoryactivities, there seems to be a consensus that participatory methods are a good way to promote newvarieties and are useful to breeders because farmer preferences can be determined and farmer practicesbetter understood. However, there is no agreement on how to combine these approaches with a conventionalmulti-location testing and release system. Neither is there agreement on how to test and release varietiesusing participatory means in the absence of a conventional multilocational testing system. Experiencefrom upland rice in Ghana and elsewhere in West Africa has shown the potential of participatory cropimprovement approaches, but these approaches are yet to be incorporated into the formal testing andrelease process.

Abstract

about 1500 mm and the rainfall pattern is weaklybimodal; rains start in March and end in December,with a break in August. There are therefore twoseasons; a major season from March to July and aminor season from September to December.

Upland rice is grown as part of a slash-and-burnsystem and is usually broadcast or dibbled followingclearing and burning. All operations are done by handand few inputs are used. Both male and femalefarmers grow rice. Cultivar diversity is low in theforest zone, and in Volta Region most farmerscontinue to grow cultivars of O. glaberrima, such asViono or Kawumo. These are valued for thepreparation of local dishes.

In contrast, in Sayerano in the Western Region,95% of the area is sown to a single variety, AgyaAmoah. It is an indica type introduced from Coted'Ivoire in 1983 by Mr Agya Amoah.

Production constraints

Male and female farmers identify a number ofdifferent production constraints. In the forest zonemale farmers cite weeds as the biggest constraint,followed by bird damage, drought and seedlingdamage by rodents. Female farmers regard birds asthe main constraint, followed by a lack of fencing(to keep out rodents).

In the savanna zone, weeds are also a majorconstraint, along with low fertility, access to seed,and drought. Female farmers also cite lack of accessto animal traction for land preparation as a constraint.

Characteristics farmers would like to see in newvarieties

Male and female farmers in the savanna and forestzones identified and ranked characteristics theyconsidered important (Table 1). This was done byconsidering the good and bad characteristics of

traditional cultivars and by showing farmers plantand grain samples of new varieties.

Male farmers in both zones wanted varieties thatwere high yielding (and also chose characteristicsassociated with yield) and suppressed weeds. Diseaseresistance was wanted in the forest zone and droughttolerance (early maturity) in the savanna zone.

Female farmers also regarded high yield anddrought tolerance and early maturity that allowsescape from terminal drought as important. Incontrast to male farmers, female farmers alsoidentified several post-harvest characteristics asimportant. These included taste, aroma, ease ofthreshing, good expansion ability and good millingquality without par-boiling.

PVS in Ghana

PVS programmes have been conducted in fourlocations in Ghana; Hohoe in Volta Region, Sayerano& Aferi in Western Region, Nyankpala & Tolon inNorthern region and Tambalug & Nyorigu in UpperEast Region. These PVS programmes have all usednurseries (i.e. a mother trial) in the first year andthen seed distribution to individual farmers in thesecond or subsequent years (i.e. baby trials).

However, the way in which these nurseries andseed distribution have been managed has varied fromthe whole process being managed by researchers,through to NGO and extension facilitation, to thembeing wholly community managed.

The first PVS in Ghana started in 1997 at Hohoe.Approximately 100 varieties were selected for thePVS from throughout the West . African region alongwith new interspecific (O. sativa x O. glaberrima)progenies. These were grown in 10 m2 plots on afavourable upland site under either improved (cleanweeding, added fertiliser) or farmer (one weeding,no fertilizer) management.

Approximately 30 male and female farmers

Source: unpublished PRA in Hohoe, Sayerano, Tambalug and Nyorigu

Table 1. Characteristics desired in new upland rice cvs by male and female farmers in twoagroecological zones in Ghana.

Forest zone Savanna zone Men Women Men Women High yield High yield Suppress weeds High yield Suppress weeds Taste Drought tolerant Drought tolerant Disease resistant Suppress weeds High market value Easy to thresh High tillering Early maturity High yield Grow well at low fertility Will not lodge Aroma High tillering Good grain expansion Large grains Large grains Early maturity Mill without par boiling

evaluated the varieties several times during thegrowing season. Post-harvest cooking and tasteevaluations were also conducted with farmers andwith rice traders in two major markets. This processwas repeated in 1998 with about 50 varieties selectedfrom the original 100.

In Hohoe, variety IDSA 85 was chosen by allmale and female farmers in the flowering and harvestevaluations (Figure 1). It was chosen because of itslong slender grains and henceits high market price (i.e. it hadsimilar grain to that ofimported rice). IDAS 85 is alsohigh yielding and easy tothresh. Subsequent on-farmcomparisons confirmed itsacceptability, along withseveral other varieties chosenby farmers during the PVS.These have all been givenlocal names (Table 2) and arespreading rapidly from farmerto farmer.

Seed disseminationpathways

Following the PVS in Hohoe,seed of eight varieties wasdistributed to farmers aroundHohoe in 2000 through anumber of different 'channels'.

Figure 1. Frequency of selection of cvs by male (open bars) and female (solid bars) farmers atharvest at Hohoe in 1998. Source: Opoku-Apau (2001).

0

5

10

15

20

25

30

35

40

IDSA

85

WA

B16

0-24

-H-

WA

B22

4-16

-HB

WA

B45

0-I-

B-P

-

WA

B45

0-24

-3-2

-

WA

B34

0-B

-B-9

-

WA

B12

6-15

-HB

IGU

APE

TOX

3377

-34-

3-3-

2

TOX

4004

-8-1

-2-3

TOX

3100

-37-

3-3-

TOX

3440

-171

-1-1

-

Freq

uenc

y Male

Female

These included: individual farmers who participatedin the PVS evaluations; a seed producer group;through the chief farmer or lead farmer; through anextension officer, by wealth categories; and througha mobilisation officer/local politician. Seed givento six villages in 2000 had spread up to 40 km by2001 (Figure 2). In general, seed was distributed firstto kin or sold outside the village, though in somecommunities farmers did not sell or give any seed

Source: CRI, unpublished

Cultivar Local name Translation IDSA 85 Idana You’ll not be tired

(threshes easily)

Iguape Cateto Leyawor I will not perish

WAB 340-B-B-9-13-4-B Kaeme Remember me

WAB 209-5-H-HB Elebode Is good to eat

WAB 160-24-H-HB Adime Saviour

WAB 224-16-HB Eleme It’s good for me

WAB 450-24-3-2-P18-HB Malamanyo I’ll take care and see

WAB 56-50 Eteyale It will be well

Table 2. Local names give by farmers in Todzi, near Hohoe, VoltaRegion to cvs selected from a PVS in 1997 and 1998

In discussions with breeders and those involvedin seed release, most felt that PVS and participatoryon-farm testing processes are a useful andappropriate means for disseminating releasedcultivars. However, these approaches were notdeemed appropriate for release. There was a cleardifference between those who took the view that allthat mattered was whether farmers grew a newvariety or not, and that therefore the priority shouldbe to get seed to farmers to test on their own farms,and those who thought the decision should rest withthe research institutes. One of the major objectionsto PVS and on-farm testing processes for release wasthat the quantitative data needed for the releaseprocess could not, or would not, be collected. To alarge extent, the 'mother and baby' trial systemadvocated by many experienced PVS practitionersmeets these objections.

Conclusions and recommendations

Based on the experience gained with PVS in Ghanasince 1997, we recommend the following:

• Any proposed PVS programme should bepreceded by a seed multiplication and a CultivarNeeds Assessment (see below). The success of

away, preferring to multiply it for themselves. Wordabout the new varieties spread rapidly through kinrelations, and demand for seed was very high. Whereseed was sold, it fetched a premium of 20 to 30%above the price for local seed.

The most successful dissemination mechanismwas that of the mobilization officer, who set up avillage committee to run a seed fund on the basis ofone kg borrowed and two kgs returned. After thefirst year, when the seed fund had grown, the originalcommittee set up similar committees in neighbouringvillages.

PVS and varietal release in Ghana

Ghana has a formal variety or seed release systemand a publicly funded foundation seed productionsystem. For release the Seed Release Committeerequires evidence of yield and stability, evidence oforganoleptic quality and consumer acceptance, andevidence of on-farm performance and farmeracceptance. However, no varieties have beenformally released for upland or hydromorphicecosystems. One reason for this is that there is nomultilocational testing system for rice. Given theabove, can a PVS programme contribute to therelease and seed dissemination mechanisms?

Figure 2. Spread of seed in Hohoe District between 2000 and 2001.

Source: CRI, unpublished

?

Todzi

4

Alavanyo

?

26

Fodome 26km

24

2001, distance

2000, no. farmers

L. Ashambi

24

Mempeasem

Odomi

2km

15km

15

Kukurantani

Korforidua

Abrani Todome

10km

13km 4km

L. Kumasi

40km

L. Bakua

Mate 5km

Tetema

Abotiase

Bala 7km

D. Papese

?

?

S-Benua 10

Agbozome

14km

Sokpo

S. Benua

Adorkor 38km

36km

8km

References

Opoku-Apau, A, (2001) Participatory Approaches inRice Varietal Improvement in Ghana: Lessons andImplications for Future Research. PhD. Thesis,Department of Agriculture, The University ofReading.

the PVS process is highly related to theavailability of an adequate quantity of seed andthis should be produced before starting theprogramme. Seed multiplication can also beused for evaluation by farmers.

• A Cultivar Needs Assessment - a focusedparticipatory rural appraisal (PRA) - to discussdesired varietal characteristics is very valuable.We propose that this should be done in thegrowing season when plants and grain samplescan be shown. This can be combined with, orreplaced by, a visit by some members of thecommunity to another PVS nursery. The processis also facilitated by researchers being able tovisit farms and see the local cultivars. This needsassessment should include traders andconsumers, as well as farmers. If possible,cooking and taste evaluations of candidatevarieties for PVS should be included at thisstage, as these are such important characteristics.Time spent on post-harvest traits at the start ofthe process will save time and effort later, butthey do require sufficient seed.

• In the first year a nursery or mother trial(s)is necessary, and this should probably beresearcher or facilitator managed. However, thecommunity should be actively involved anddecide how many there should be and where theyshould be sited. Researchers can collectmeaningful quantitative data from these trialsfor release purposes.

• In the second and subsequent years, a motherand baby trial system or, if data for release arenot needed, a baby system (i.e. on-farm, pairedcomparisons) is preferred.

• Data collection. In our experience communitieshave been more than capable of keeping goodrecords when they are motivated.

• Community managed PVS. We have had mixedresults with this, at least in the first year of a PVSprocess. Initially, many farmers lack motivationand are unsure of the benefits and sometimessuspicious of the motives of others. It may bebetter, therefore, to initially target lead orrespected farmers if the trial is not going to beresearcher managed.

5.1Combining molecular marker technology and participatory techniques: Acase study for drought-tolerant rice in eastern India. I: Molecular breedingstrategy

KA STEELE, G EDWARDS AND JR WITCOMBE

Centre for Arid Zone Studies, University of Wales, Bangor, Gwynedd, LL57 2UW, UK.

Participatory plant breeding (PPB), like conventional breeding, is predicted to give a low response toselection for traits with low heritability. These include many traits that farmers consider desirable, suchas drought resistance. The technology of marker-assisted selection (MAS) can be used to make selectionfor traits with low heritability more effective. We describe how MAS can be used to assist in the breedingof drought tolerance in upland rice.

Abstract

Introduction

We are carrying out MAS for drought-tolerance-related traits. We have done this on large populationsof young rice plants grown in the glasshouse atBangor, or on rice plants grown in India from whichDNA has been sampled. We are following threedifferent strategies for combining PPB withmolecular marker techniques. These are:

1. Backcross-MAS to pyramid root QTLs(quantitative trait loci) that should imparttolerance to drought, and a QTL controllingaroma in the grain.

2. Modified single large-scale MAS, to create bulkpopulations for PPB.

3. Marker-evaluated selection in which changes inallele frequency are assessed in a range of PPBproducts to enable the design and creation ofideotype varieties.

Strategy 1. MAS for root QTL and aroma

Drought is a serious abiotic constraint for rice farmersof Eastern India (Pandey et al., 2000). Their land isoften drought-prone because of sloping fields andshallow soils, and rainfall is unpredictable. Thesefarmers predominantly grow landraces or varietiesthat are selections from landraces. Modern, improvedvarieties often have poorer drought tolerance thanlandraces. This can, in part, be due to poorlydeveloped root systems, so farmers prefer to growlocal landraces with deep root systems (Ekanayakeet al ., 1985; Lafitte et al., 2001). PPB andparticipatory varietal selection (PVS) in eastern Indiahave been successful (Virk et al., this volume)indicating that farmers appreciate being given a

greater choice of varieties and are willing to adoptthem if they are superior.

The popular variety Kalinga III is susceptible tolodging and has a poor root system, making it proneto early-season drought. Several mapping studieshave identified QTL for root traits in segregatingpopulations and have identified QTLs in the varietyAzucena that increase root length, thickness andpenetration ability (Price and Tomos, 1997; Price etal., 2000; Price et al., 2002; Yadav et al., 1997).

We targeted four QTLs controlling root traits(on chromosomes 2, 7, 9 and 11) that had large effectsand were stable across experiments. These regionshave all been shown to contain QLTs for drought-related traits in several crosses evaluated by otherworkers (for review see Zhang et al., 1999). Weselected for these four target regions in a backcrossmarker-assisted selection (BC-MAS) programmeusing Kalinga III as the recurrent parent. A fifth,highly heritable QTL for aroma on Chromosome 8was also selected. The flanking markers used forselection were restriction fragment lengthpolymorphisms (RFLP) and simple sequence repeats(SSR or microsatellites). Selection was made for thetarget regions in BC1, BC2 and BC3. At the BC3generation, lines having at least two of the targetregions were also screened with 30 backgroundmarkers at non-target regions; two lines with higherfrequencies of Kalinga III alleles at backgroundregions were selected. These lines were advancedto fix the target Azucena alleles. Crosses betweenlines containing different target regions were madein order to pyramid all five target QTLs in theKalinga III genetic background (Figure 1a).

Root screening of partial pyramid lines(containing 1, 2 or 3 target root QTLs) has beencarried out in soil-filled pipes under field conditions

in Bangalore, India, (HE Shashidar, pers. comm.)and in root boxes under glasshouse conditions inAberdeen, UK (A Price, pers. comm.). Somecombinations of Azucena alleles at root QTL wereshown to significantly increase root length, thicknessand volume compared with Kalinga III.

A locus in the target root QTL on Chromosome7 was found to considerably delay flowering in theglasshouse at Bangor, and experiments are plannedto test if this is due to linkage drag or pleiotropy.

Strategy 2. SLS-MAS

The approach of single large-scale MAS (SLS-MAS) has been described (Ribaut and Betrán, 1999).This interactively combines the use of DNA markersand conventional breeding. It involves screening alarge population of F2 or F3 plants for favourableloci while still maintaining, as much as possible,the allelic segregation in the rest of the genome. Weused a population of more than 500 lines, derivedfrom the previously selected BC2, and screenedthem for markers at the four target root QTLs (Figure 1b).

In contrast to the conventional SLS-MAS

Figure 1. Backcross-marker-assisted selection programme used to transfer five target regions fromAzucena to Kalinga III, indicating the number of plants genotyped at each generation.a) Selection and advancement of BC3 lines, to pyramid QTL in Kalinga III geneticbackground. b) Modified single large-scale MAS used to derive five bulks, each containinga different target QTL, and one control bulk.

method, we only attempted to identify lines withone root QTL. We detected, on average, 25 linesthat were fixed (homozygous) for each Azucenatarget locus. The selected lines with at least one rootQTL were then screened with a marker linked toaroma to identify a subset of lines containing aroma.Of these, four were homozygous and seven wereheterozygous for the aroma locus. The lines havingQTLs identified by the markers were used to makebulks for PPB (Table 1). All the bulks weresegregating for Azucena alleles in the non-targetgenomic regions.

Strategy 3. Marker-evaluated selection

In PPB we have used material from Kalinga IIIcrossed to:

• IR64 and IR36 - both are elite modernvarieties for irrigated conditions;

• Vandana, an Indian upland variety;

• Radha 32, a popular Nepalese variety(Gyawali et al. this volume).

These populations generated successful

Kalinga III / Azucena

F 1 BC1F 1 7 / Kalinga III

/ Kalinga III

22MAS at 5 target regions

BC2 1

BC3 1

BC3 2

4 / Kalinga III 59

31

40

MAS at 5 targets (and 30 non-target chromosomes for Kalinga III)

Individual lines crossed to pyramid all targets in Kalinga III genetic background

a

BC2 3

BC2 2 565

6 bulks for PPB

and selected MAS at 4 targets for root QTL, lines tested for aroma (RM223).

b

2 selfed

Table 1. Bulks selected via modified SLS-MASfor PPB, indicating target regions from Azucenapresent.

genotypes in Eastern India (Kumar et al., 2000) andNepal. Some of the lines and bulks performed wellacross both countries, and in different seasons andecosystems. These lines and bulks represent asignificant genetic resource for testing the usefulnessof marker-evaluated selection. The most promisinglines selected by farmers may hold the key to thebest choice of future PPB crosses. We are currentlyusing molecular markers to evaluate 62 populationsfrom these crosses (bulks and pure lines), whichwere selected in PPB.

Farmer-preferred traits are those havingadaptive or commercial value. Markers linked tothese farmer-preferred traits can theoretically beidentified using marker-evaluated selection, and weaim to test this hypothesis and develop the technique.

DNA fingerprint linkage blocks (DFLBs, Zhuet al., 1999) are being used to determine farmer-preferred genomic regions. If marker allelefrequency in the PPB-selected population differssignificantly from the allele frequency in a non-selected population derived from the same cross, itis assumed to be linked to a locus that contributesto a farmer-preferred trait. This is a one-tailedselective genotyping (OTSG) approach, similar tothe trait-based analysis for detection of marker-QTLlinkage of Lebowitz et al. (1987).

QTLs explaining 30% of the genetic variancemay be detected in samples of 15 or more selectedlines, but linkages involving QTL of smaller effectsare unlikely to be detected (G Atlin, pers. comm.).We expect that loci controlling plant height andflowering time will definitely be detected, and thatother farmer-preferred traits might also be identified.The unselected population used as a control willprevent detection of false DFLBs, which differ in

allele frequency due to segregation distortion.Segregation distortion in rice mapping populationsis often found in regions containing gametophyticgene loci and/or sterility loci (He et al., 2001).

By exploring the graphical genotypes of farmer-selected lines, 'ideotype' plants can be designed forspecific environments and crosses will be made toselect for those idetypes in the progeny using MAS.The optimum combinations of genomic regions forfarmer-preferred traits and root QTL could be readilytransferred, through MAS, to any farmer-preferredvariety.

References

Ekanayake IJ, O'Toole JC, Garrity DP and Masajo TM(1985). Inheritance of root characteristics and theirrelations to drought resistance in rice. Crop Science25: 927-933.

Gyawali S et al. (2002). Extending PPB in rice tofavourable environments - the LI-BIRD experiencein Nepal. This volume.

He P, Li JZ, Zheng XW, Shen LS, Lu CF, Chen Y, andZhu LH (2001). Comparison of Molecular LinkageMaps and Agronomic Trait Loci between DH andRIL Populations Derived from the Same Rice Cross.Crop Science 41: 1240-1246.

Kumar R, Singh DN, Prasad SC, Gangwar JS, VirkDS, and Witcombe JR (2000). Participatory plantbreeding in rice in eastern India. Paper presented inthe international symposium Participatory plantbreeding and participatory plant genetic resourceenhancement: An exchange of experiences fromSouth and South East Asia, held at Pokhara, Nepal,1-5 May, 2000. In press.

Lafitte HR, Champoux MC, McLaren and O'Toole JC(2001). Rice root morphological traits are related toisozyme group and adaptation. Field Crops Research71, 57-70.

Lebowitz RJ, Soller M, Beckmann JS (1987). Trait-basedanalyses for the detection of linkage between markerloci and quantitative trait loci in crosses betweeninbred lines. Theoretical and Applied Genetics 73:556-562.

Pandey S, Behura DD, Villano R, and Naik D (2000).Economic Cost of Drought and Farmers' CopingMechanisms: A Study of Rainfed Rice Systems inEastern India. IRRI Discussion Paper Series No. 39.Makati City, (Philippines): International RiceResearch Institute. 35pp.

Price AH and Tomas AD (1997). Genetic dissection ofroot growth in rice (Oryza sativa L.) II: mappingquantitative trait loci using of root growth in rice(Oryza sativaof root growth in rice (Oryzasativamolecular markers. Theoretical and AppliedGenetics 95: 143-152.

Price AH, Steele KA, Moore BJ, BarracloughPB, andClarke LJ (2000). A combined RFLP and AFLPlinkage map of upland rice (Oryza sativa L.) used to

Bulk Target regions present I None (control)

II Root QTL (Chromosome 7)

III Root QTL (Chromosome 9)

IV Root QTL (Chromosome 11)

V Root QTL (any one target) and

aroma (Chromosome 8)

VI Root QTL (Chromosome 2)

identify QTLs for root penetration ability. Theoreticaland Applied Genetics 100: 49-56.Price AH, Steele KA, Moore BJ and Wyn-Jones G (2002).

Upland rice grown in soil-filled chambers and exposedto contrasting water-deficit regimes: II. Mapping QTLfor root morphology and distribution. Field CropsResearch 76: 25-43.

Ribaut J-M and Betrán J (1999). Single large-scale marker-assisted selection SLS-MAS. Molecular Breeding 5:531-541.

Virk DS, Singh DN, Kumar R, Prasad SC, Gangwar JSand Witcombe JR (2002). Participatory plant breeding(PPB) in rice in Eastern India - the success of an NGO/GO partnership. This volume.

Yadav R, Courtois B, Huang N, and McLaren,G (1997).Mapping genes controlling root morphology and rootdistribution on a double-haploid population of rice.Theoretical and Applied Genetics 94: 619-632.

Zhang J, Zheng HG, Ali ML, Tripathy JN, Aarti A, PathanMS, Sarial AK, Robin S, Nguyen TT, Babu RC, NguyenBD, Sarkarung S, Blum A and Nguyen HT (1999).Progress on the molecular mapping of osmoticadjustment and root traits in rice. In: Ito O, O'Toole Jand Hardy B (eds.) 1999. Genetic improvement of ricefor water-limited environments. Proceedings of theworkshop Genetic Improvement of Rice for Water-limited Environments, 1-3 December 1998, Los Banos,Philippines. Manila; International Rice ResearchInstitute. pp 307-317.

Zhu JH, Stephenson P, Laurie DA, Li W, TangD, and GaleMD (1999). Towards rice genome scanning by map-based AFLP fingerprinting. Molecular and GeneralGenetics 261: 184-195.

.

5.2Combining molecular marker technology and participatory techniques: Acase study for drought-tolerant rice in Eastern India. II: Farmer evaluationof SLS-MAS bulks in participatory plant breeding

KA STEELE1, DN SINGH2, R KUMAR2, SC PRASAD3, DS VIRK1, JS GANGWAR3, JRWITCOMBE1

1. Centre for Arid Zone Studies, University of Wales, Bangor, Gwynedd, LL57 2UW, UK.2. Birsa Agricultural University, Ranchi, Jharkhand, India.3. Gramin Vikas Trust, 280 Kanke Road, Ranchi, Jharkhand, India.

Background

Modern biotechnological technologies have thecapacity to create rapidly varieties with newcombinations of genes, which are unlikely to bederived through conventional breeding methods.MAS is a potentially valuable strategy for breedingresistance to abiotic stresses but its value can onlybe proven by testing the products of MAS in marginalenvironments by the end-users; small-scale farmers.By involving farmers directly in the MAS breedingprogramme through PPB methods, the twotechnologies address the needs of farmers. However,the PPB methods that can be used in combinationwith MAS can differ greatly according to the degreesof farmer participation (Witcombe et al., 1996) withthe extremes being on-station selection by farmers(consultative) and on-farm, farmer-led PPB(collaborative).

Results from collaborative PPB

In collaborative PPB farmers are given materialto select amongst which they grow on their ownfarms (Witcombe, this volume; Virk et al., thisvolume). Three farmers, one in each of the threestates, Orissa, Jharkhand and West Bengal, were

given the six bulks (at the BC2F3 generation) to growin plots alongside Kalinga III in the rainy season2000. Harvested seed was returned to breeders fromthe farmer in Orissa only.

In the rainy season 2001 10 farmers were giventhe bulks, but a full set of yield data was onlyreceived from seven farmers. In both years farmersused their usual practices and were asked to makeselections within each bulk. The farmers willcontinue the PPB programme with the selectedmaterial from all six MAS bulks in rainy season2002, and the material will then be at the BC2F6generation (Figure 1).

There was a great deal of variation within allthe bulks in both years for height, tiller number,spreading or compact plant type, maturity, awnpigmentation, stem thickness and panicle length. Toa lesser extent, grain shape varied, although all bulkshad fine, slender grains. The bulks all had increasedpanicle length compared to Kalinga III, and wereless prone to lodging. All bulks containing Azucenagenes had greater biomass than Kalinga III and werenot significantly earlier.

In the rainy season 2001 the farmers were askedto rank the plots (six bulks and check variety KalingaIII) in their order of preference. A score of 1 wasgiven to the least preferred plot and 7 to the most

We wish to test if participatory approaches and modern molecular marker techniques can be effectivelycombined. To do this we have integrated participatory plant breeding (PPB) with marker-assisted selection(MAS). The PPB has been carried out on six bulks selected with molecular markers via modified singlelarge-scale MAS (SLS-MAS). The bulks were derived from the second backcross generation betweenKalinga III and Azucena where Kalinga III was the recurrent parent (See Part I: Molecular BreedingStrategy, this volume). Four of the bulks had been selected to contain a quantitative trait locus (QTL) forroot growth. One bulk was a selected subset of those with root QTLs which had also inherited a markerfor aroma. A control bulk was selected from the same generation, which contained none of the targetsand was used to test the effect of the target root QTL. Bulks were given to farmers in three states ofEastern India for evaluation over two seasons of collaborative PPB. The same bulks were also tested on-station through consultative PPB, with off-season seed multiplication.

Abstract

preferred plot, for each criteria. Thecriteria used for ranking were:maturity, height, yield, straw, lodging,pest attack, grain colour and overallpreference. Four farmers respondedand they ranked all six of the MASbulks higher than Kalinga III foroverall performance, including thecontrol bulk with no QTL (Table 1).Farmers' rankings for other criteriawere confounded by the high levelsof intra-bulk variation.

In making selections the farmersgenerally, although not exclusively,selected early lines in all bulks (Figure2). They stated that earliness was themost important selection criteria, withplant height (tall plants with morestraw for fodder) the second. Farmersdiffered in their opinion about grainshape, and while most farmers said thatpossession of fine grains was the thirdmost important selection criteria, onefarmer said that bold grains were moreimportant.

Many farmers ranked the aromaticbulk highly and were excited to smellthe aroma in the field during flowering.Due to limited seed availability thecooking quality still remains to beevaluated. This is the first time thataromatic upland rice has beenavailable in Eastern India, and farmers'reactions were generally positive.Aromatic lines have greater marketvalue than non-aromatic varieties andthis could benefit poor farmerseconomically.

However, the introduction ofaroma may have both environmentaland economic consequences. Theincidence of insect pests on thearomatic bulk was greater than on non-aromatic bulks. This, combined withthe greater market value, could lead toincreased use of insecticides.

Results from consultative PPB

Consultative PPB was carried out atthe GVT/BAU upland farm and off-season multiplication carried out atCRRI, Cuttack, Orissa. In the rainyseason 2001 all six bulks plus twocheck varieties, Kalinga III and BirsaGora 102, were grown in a randomised

Figure 1. Schemes for collaborative and consultative PPBusing SLS-MAS bulks. Each generation consistedof only selected material from the previousgeneration, and all six bulks were advancedsimultaneously.

BC2F3

Collaborative Consultative

Off-season 1999-2000

Rainy season 2000

Off-season 2001-2002

Rainy season 2001

BC2F4 BC2F4 3 farmers GVT-BAU farm

BC2F5

10 farmers GVT-BAU farm

BC2F6

Rainy season 2002 BC2F7 BC2F6

Farmers continue

GVT-BAU

CRRI, Cuttack

Collaborative PPB

Drought trials

CRRI, Cuttack

AICRIP trials

1From Azucena present on chromosome.2From a maximum of 24 given by four farmers for overall rank.

Bulk QTL1 Total score2

I None (control)

19

II 7 (root length and mass)

22

III 9 (root length)

15

IV 11 (root length and penetration)

14

V 8 (aroma) plus one of either 2, 7, 9 or 11 root QTL

12

VI 2 (root length, thickness and penetration)

18

Kalinga III none (check variety)

7

Table 1. Target QTL present in six bulks selected throughmodified SLS-MAS and the overall rank scoresgiven by farmers in two states duringcollaborative PPB evaluation alongside KalingaIII check in the rainy season 2001.

block experiment. Within the all bulks and checksthere was no significant difference for plant heightor flowering time. For yield, three bulks weresignificantly different from Kalinga III; bulks IV andV yielded less, while bulk VI had a greater yield.

Twenty-one farmers (men and women) visitedthe plot 81 days after sowing. Selection and taggingof plants was carried out for four differentenvironments: upland, medium upland, mediumlowland and lowland. These selected bulks wereadvanced in the off-season and the BC2F7generation will be given to farmers for pairedcomparisons with Kalinga III in the rainy season2002 (Figure 1). In the off-season, plots of theselected bulks II, III and III were uniform after twogenerations of consultative selection. They havebeen recommended by BAU for entry into the All-India Coordinated Rice Improvement (AICRIP)trials and in state trials.

All six bulks at the BC2F7 generation will betested in replicated field trials in both Eastern andWestern India for drought resistance in the rainyseason 2002. They will be grown under uplandconditions with no irrigation to test the hypothesisthat bulks containing root QTL are more adapted towithstand drought than the control bulk.

An additional 16 individual lines selected fortarget QTLs by pure MAS (from BC3F2 andpyramid generations) were also grown at the BAUfarm in the rainy season 2001. Eleven of these werearomatic, four bulks were partial pyramids for twoor three target root QTL and one was an advancedgeneration control line with no target QTL. Visitingfarmers made selections within these plots andtagged selected plants according to four targetenvironments.

Conclusion

MAS appears, from these preliminary results, to besuccessful in producing promising material.However, whether this is the result of introducing

Figure 2. Upland farmers in Udalidistrict, Orissa, September2000, making selections insegregating bulks selectedvia modified SLS-MAS forroot trait QTLs and aromafrom Azucena.

(Photo AK Paria, CommunityOrganiser, GVT East).

target QTL, or more generally the result ofintroducing small genomic contributions of Azucenainto Kalinga III is not yet known. So far, the bulkwith contributions from Azucena but without thetarget QTLs also appears superior to Kalinga III.Moreover, much of the assessments have, so far, beenmade on the heterogeneous bulks rather than inselections from them. We can conclude that there isno barrier to combining 'high-technology' breedingapproaches with participatory ones.

Indeed, it provides an acid test for the real valueof target QTLs. Only if MAS products are preferredby farmers can MAS be deemed successful.

References

Virk DS, Singh DN, Kumar R, Prasad SC, Gangwar JSand Witcombe JR (2002). Participatory plant breeding(PPB) in rice in Eastern India - the success of an NGO/GO partnership. This volume.

Witcombe JR, Joshi A, Joshi KD and Sthapit BR (1996).Farmer participatory crop improvement. I Varietalselection and breeding methods and their impact onbiodiversity. Experimental Agriculture 32: 445-460.

Witcombe JR (2002) Participatory crop improvementstrategies in rice in the DFID Plant Sciences ResearchProgramme. This volume.

Introduction

Demand for rice in West and Central Africa(WCA) has been growing at the rate of 6% per annumsince 1973, and now amounts to over 8 milliontonnes per annum. Increased consumption is due bothto population growth (average of 2.9% per annum)and to the increased share of rice in the diet of theAfrican population.

Over the same period, production of rice hasbeen expanding at the rate of 5.1% per annum, with70% of the growth due to an increased area ofcultivated rice, but only 30% due to higher yields.Much of the expansion has been in the rainfedsystems, particularly the two major ecosystems thatmake up 78% of rice land in WCA: namely, theupland and rainfed lowland systems.

To cover the increasing gap between ricedemand and supply, rice is being imported at anincreasing rate. Shipments of rice to Africa rose by11% to slightly more than 5 million tonnes in1999/2000; Nigeria being the major importer at 0.9million tons, closely followed by Côte d’Ivoire. Thecost of this importation is around $1 billion of foreignexchange each year.

With population growth rate exceeding that ofregional food production, and limited foreignexchange for increased levels of imports, the futurefor Africa’s poor appears grim. For food security,WCA needs to increase production capacity for whencheap rice can no longer be imported.

The birth of a new rice for Africa

The Asian rice Oryza sativa was brought toAfrica over 500 years ago. Within a short period,because of its high yield potential, it replaced most ofthe indigenous African rice, Oryza glaberrima,which was pushed to inhospitable, scraggy andinfertile upland soils or flood-prone inland valleysand deltas. In Africa, however, O. sativa sufferedattack by pests and diseases known only to theAfrican continent. Toxic soils with limited nutrientsalso reduced its yield potential. The hardy Africanrice, O. glaberrima, with a history of 3,500 years ofcultivation, withstands these problems better than O.sativa. To combine the best traits of the Asian andAfrican rice species, the challenge is to transfer intothe heavy yielding but stress-prone O. sativa thedesirable genes in O. glaberrima while shedding itsundesirable traits of lodging, fewer grains and grainshattering.

Conventional breeding efforts to developinterspecific hybrids had failed due to a high level ofsterility of the hybrids. In 1991, WARDA launched anew effort to combine the genetic potential of the tworice species, using conventional and anther culturetechniques to overcome sterility and to hasten thebreeding process. Crosses were made and embryorescue was used to remove fertilized embryos andgrow them in artificial media. Anther culture allowedrapid fixation and helped to retain interspecific linescombining desirable features of the two rice species.

Development of New Rice for Africa (NERICA) and participatory varietalselection

HE GRIDLEY, MP JONES, M WOPEREIS-PURA

West Africa Rice Development Association (WARDA), Bouaké, Côte d’Ivoire

AbstractOne of the major research thrusts of the West African Rice Development Association (WARDA)

to meet the ever-increasing demand for rice in West and Central Africa (WCA) has been thedevelopment of higher yielding varieties. By 1996 WARDA had developed a range of new interspecificvarieties derived from crosses between the Asian rice, Oryza sativa and the African rice O. glaberrima.These varieties, later termed NERICAs (New Rice for Africa), have shown a stable and high yield andtolerance to major biophysical production constraints in a range of upland environments. Their rapiddissemination to small rice farmers in WCA has been achieved through participatory varietal selection(PVS), an applied and adaptive research mechanism in which farmers play an active role in varietalselection, development and spread. PVS lasts for three years during which farmers select and evaluatevarieties on their own farms. PVS was initiated in Côte d’Ivoire in 1996 and by 2000 all 17 WARDAmember countries had initiated PVS involving 4000 farmers at 105 sites. Particular progress has beenmade in Guinea where from 1997 to 2000 the number of farmers participating in PVS rose from 116 to20,000 the area sown to NERICAs from 50 to 8,000 hectares.

The generous support of donors in Japan and theRockerfeller Foundation in the USA to the JointInterspecific Hybridization Project allowedconsiderable progress to be made, and by the late1990s interspecific lines of a radically different planttype, tailored for dryland rice farmers of West andCentral Africa, had been developed and were beingtested and evaluated in a range of environments.

Agronomic characteristics of interspecific lines

Since the mid-1990s many interspecific lines havebeen generated, evaluated and characterised for arange of agronomic traits and reaction to importantdiseases and pests (WARDA 1999, 2000, and 2001).In 2000, the interspecific lines were dubbed NERICA(New Rice for Africa).

NERICAs have been identified with one ormore of the following characteristics:• Wide and droopy leaves that help to smother

leaves in early growth.

• Strong stems that can support heavy heads of grain.

• More tillers with longer grain-bearing panicles than either parent and non-shattering grains.

• Stems with secondary branches on theirpanicles that can carry up to 400 grains.

• Early maturity, 30-50 days earlier thancurrently grown cultivars.

• A good height that allows easy harvest ofpanicles.

• Good tolerance to drought.

• Resistance or tolerance to Africa’s mostserious pests and diseases - African rice gallmidge, rice yellow mottle virus and blast(Magnaporthe grisea).

• Tolerance to acidic soils but responsive tolimited organic and inorganic fertilizers.

NERICAs have shown stable yields under bothlow- and high-input conditions and are expected toreduce risk and increase productivity in farmers’fields, thereby reducing the need to clear new land.Reduced risk will also give farmers incentives to usemore inputs, intensify land use, and graduallyabandon shifting cultivation, thus improving systemsustainability. The rapid introduction of NERICAs to

small farmers in WAC was, therefore, considered avitally important first step towards the sustainableintensification of Africa’s fragile uplands. TheNERICAs were not intended as a total replacementfor local varieties for integrating into the existingvarietal portfolio of farmers.

PVS methodology

In conventional breeding schemes selection andtesting procedures involve a series of multi-locationtrials over eight to 12 years of a diminishing numberof varieties (Figure 1). This favours the selection of afew, widely adapted lines for release, often with littleregard to farmer and consumer preferences or toaddressing the myriad of farm levelmicroenvironments.

At WARDA, this conventional approach totransfer of new varieties has given way to an appliedand adaptive research mechanism, termedparticipatory varietal selection (PVS) that favoursfarmers playing an active role in varietal selection,development and spread (Figure 1).

The goal of PVS is to efficiently transferimproved rice varieties to farmers in order to:• Reduce the time required to move varieties

onto farmers’ fields;

• Determine the varieties that farmers want togrow;

• Learn the traits that farmers value in varietiesto assist breeding and selection;

• Determine if there are gender differences in varietal selection criteria.

WARDA’s approach to PVS research (PVS-R)is a three-year programme (Figure 1). In the firstyear, breeders identify centralised fields near villagesand plant a ‘rice garden’ trial of up to 60 uplandvarieties. The varieties range from traditional andpopular O. sativas to NERICAs, African O.glaberrimas and local checks. Men and womenfarmers are invited to visit informally the plot asoften as possible but the farmers are brought asgroups for formal evaluation of the varieties at threekey stages: maximum tillering, maturity and post-harvest. For the first two, farmers compareagronomic traits, including weed competitiveness,growth rate, height, panicle type and growth cycle,whilst the third focuses on grain quality attributessuch as size, shape, shattering, ease of threshing andhusking and palatability. Each farmer’s varietalselection and the criteria for selection are recorded

and later analyzed.In the second year, each farmer receives as

many as six of the varieties he or she has selected inthe first year to grow on his or her farm. Thus geneticdiversity enters the communities. PVS observers,who may compromise breeders and/or techniciansfrom NGOs and Extension Services, visitparticipating farmers’ fields to record performanceand farmer appreciation of the selected varieties. Atthe end of the year, farmers evaluate threshability andpalatability to provide an overall view of thestrengths and weaknesses of the selected varieties.For the third year, farmers are asked to pay for seedsof the varieties they select providing evidence of thevalue they place on them. Thus, in three years, PVS-R allows the farmers to select varieties with specificadaptation and preferred plant type and grain qualitycharacters. These, in turn, can be integrated into thebreeding programmes totailor varieties forfarmers. In theconventional scheme atleast 12 years is necessaryto reach this point and,even then, farmers andconsumers may notappreciate the varietiesselected.

A PVS extension

(PVS-E) phase has recently been introduced tocompliment PVS-R and accelerate dissemination andofficial release. Four to six of the more commonlyselected varieties in the second year of PVS-R in anecoregion are disseminated widely to farmers withinthe region for evaluation in the third year. After twoyears of PVS-E, the more preferred of these varietiesare enrolled in multi-location trials to generate datafor official release. Simultaneously, these varietiesenter community-based seed systems (CBSS) formultiplication to ensure adequate seed supplies forrapid dissemination of the varieties officiallyapproved for release.

PVS activities in West Africa

PVS-R was initiated in 1966 through a smallproject in Boundiali in Côte d’Ivoire, where farmers

1 PVS-R and -E: PVS Research and Extension, respectively. 2 CBSS: Community Based Seed System.

Year

1 2 3 4 5

PVS-R1

30-60

varieties PVS-E1 4-6 varieties

evaluated by farmers eco-

regionally Official trials for release

Multiplication and diffusion

by CBSS2 of released

varieties

PVS 5 years

Crossing and selection Evaluation of lines in:

! Observational Nursery ! Preliminary Yield Trial

No varieties

Conventional 8-12 years

Advanced yield

Elite variety

Regional Elite

On-farm

2-3

2-3

2-3

2-3

24

24

14

4

Years in trial Trial type

Figure 1 . Diagrammatic representation of relative time scales for conventional and PVS to deliver new varieties to farmers.

Number of: Year Countries with PVS-R activities Countries Sites Farmers 1996 Côte d’Ivoire 1 1 55 1997 Benin, Ghana, Guinea, Togo 5 17 570 1998 Burkina Faso, The Gambia, Guinea-

Bissau, Nigeria, Sierra Leone 7 38 1293 1999 Cameroon, Chad, Liberia, Mali,

Mauritania, Niger, Senegal 17 64 2491 2000 17 WARDA member countries 17 105 >4000

Table 1. Evolution of PVS-R activities by year in West Africa

were provided with a rice garden of 57 varietiesamongst which they were able to select those that mettheir own needs. The results were so encouraging thatWARDA decided to extend the participatory

approach. By 1999 all 17 WARDA member countrieshad instigated PVS-R at 64 sites with 2,491 farmersparticipating; by 2000 over 4,000 farmers at 105 siteswere involved (Table 1) (WARDA 1999).

1S: O. sativa; I: interspecific (O. sativa x O. glaberrima).

Table 2. The 58 varieties in advanced PVS testing by country in 2000

Type1

Variety B

enin

B

urki

na F

aso

Cam

eroo

n C

ôte

d’Iv

oire

G

hana

G

uim

ea B

issau

G

uine

a M

ali

Mau

rita

nia

Nig

er

Sene

gal

Sier

ra L

eone

Tc

had

The

Gam

bia

Togo

To

tal

S Bouaké 189 " 1 S BW 293-2 " 1 S BW 348-1 " 1 S Chinese " 1 S CICA 8 IR 2042-178-1 " 1 S FKR 21 " 1 S FKR 243 " 1 S FKR 33 " 1 S FKR 41 " 1 S IDSA 10 " 1 S IDSA 85 " 1 S IR 12979-24-1 " 1 S IR 28228-45-3-3-2042-178-1 " 1 S IR 46 " 1 S IR 47701-6-3-1 " 1 S IRAT 144 " 1 S IRAT 216 " 1 S ITA 150 " 1 S ITA 222 " 1 S ITA 398 " 1 S Niono 2 " 1 S RP 1045-25-2-1 " 1 S RP 17846-111 " 1 S SK 51-5-2 " 1 S Suphanbouri " 1 S TGR 75 " 1 S TOX 3093-35-2-3-3-1 " 1 S TOX 3098 " 1 S TOX 3100 " 1 S WAB 181-18 " " " 3 S WAB 365-B-1-H1-HB " 1 S WAB 50-HB " 1 S WAB 515-B-16A2.2 " " 2 S WAB 515-B-16H2-2 " 1 S WAB 56-104 " 1 S WAB 56-50 " " " 3 S WAB 570-10-B-1A2.6 " " 2 I WAB 450-11-1-1-P31-HB " " " 3 I WAB 450-11-11-P50-HB " 1 I WAB 450-11-1-2-P41-HB " 1 I WAB 450-24-2-3-P33-HB " 1 I WAB 450-24-2-3-P38-HB " 1

As early as 1997, significant genderdifferences in varietal selection were detected atDanane in Côte d’Ivoire but further analysis of thelarge body of data collected from other countries isneeded to determine if this is common. Markedspecificity in varietal selection between countries,however, was evident (Table 2), reflecting acombination of differing varietal adaptation to thewide range of on-farm (micro-) environments anddiverging consumer preference. Of the 58 varieties inadvanced PVS-R testing in 15 countries in 2000, 46were selected in only one country. These accountedfor 79% of the varieties, with the rest accounting for1% or less - namely, six in two countries, five in threecountries and four in one country (Table 2).

The most frequently selected varieties werederived from WARDA crosses designated by thecode ‘WAB’, and amongst these the NERICAs wereas popular as the O. sativa varieties. The NERICAswere particularly popular in Côte d’Ivoire,accounting for all varieties selected, and in Guinea,where NERICAS were all but one of the varietiesselected. Five NERICAs have now been released inCôte d’Ivoire and two in Guinea. On average, five

varieties were selected in each country, ranging froma minimum of three for Mauritania to a maximum ofseven for Guinea and Tchad (Table 2).

Selection criteria applied by farmers amongst17 countries were more consistent (Table 3). Themost frequently applied criteria across countrieswere:• Yield (14 countries).

• Height (13 countries).

• Short growth cycle (11 countries).

• High tillering (11 countries).

• Grain size (9 countries).

• Large grain (9 countries).

Thus across countries, heavy yielding, earlyvarieties with profuse tillering and large grain are themost desired, but breeders must be alert to variationbetween countries, especially taking account of theless-frequently noted characters.

Country Selection criteria

Yie

ld

Hei

ght

Shor

t gro

wth

cyc

le

Hig

h til

lerin

g

Gra

in si

ze

Gra

in-la

rge

Wee

d co

mpe

titiv

enes

s

Gra

in c

olou

r

Gra

in-b

old

Goo

d re

spon

se to

ferti

lizer

Lodg

ing

resi

stan

ce

Pani

cle

size

Tast

e

Non

-stic

ky g

rain

Dro

ught

tole

ranc

e

Med

ium

gro

wth

cyc

le

Brid

dam

age

resi

stan

ce

Ada

ptab

ility

Emer

genc

e ra

te

Aro

ma

Dis

ease

resi

stan

ce

Togo " " " " The Gambia

" " " " " "

Tchad " " " " " " Sierra Leone

" " " " " " " "

Senegal " " " " " Niger " " " " Nigeria " " " " Mauritania " " " " " " Mali " " " " " " " " " " Liberia " " " Guinea " Bissau " " " " " " " " " " " Guinea " " " " " " " Ghana " " " " " Cote d’Ivoire

" " " " " " " " " "

Cameroon " " " " " Burkina Faso

" " " " " " " " "

Benin " " " " Total 14 13 11 11 9 9 4 4 4 4 4 3 3 3 3 2 2 1 1 1 1

Table 3. Farmers' selection criteria applied PVS-R in 1999 in 17 countries.

PVS in Guinea

PVS-R and NERICAs have made a major impact inGuinea. The first phase of PVS-R began in 1997 withsupport from the World Bank that enabled TheInstitut de Recherche Agronomique de Guinea(IRAG) and Service National de la Promotion Ruraleet de la Vulgarisation (SNPRV) to collaborate withWARDA to identify varieties that could be rapidlytransferred to farmers. A second phase was started in1999.

The number of farmers participating rose from116 in 1997 to 20,000 in 2000, with the area ofNERICAs sown over this period increasing from 50to 8,000 hectares (Table 4) (WARDA 2001). In 2000,the gain in production and in income from NERICAsis estimated at 15,000 tonnes of paddy and US$2.5million, respectively, with additional incomeestimated to rise to US$300,000 in 2002 (Table 3).

References

WARDA (West Africa Rice Development Association),1999. Rice Interspecific Hybridization Project:Research Highlights 1999. WARDA, Bouaké, Côted’Ivoire. 34 pp.

WARDA (2000). Rice Interspecific Hybridization Project: Research Highlights 2000. WARDA, Bouaké, Côted’Ivoire. 34 pp.

WARDA (2001). Bintu and Her New Rice for Africa. WARDA, Bouaké, Côte d’Ivoire. 32pp.

WARDA (2001). Annual Report 2000. WARDA, Bouaké,Côte d’Ivoire. 84pp.

WARDA (2000). Annual Report 2000. WARDA, Bouaké, Côte d’Ivoire. 74pp.

Year Number of farmers

Area (hectares) sown to NERICAs

Estimated gain in production and income

1997 116 50 1998 380 150 1999 1,680 250 2000 20,000 8,000 15,000 tonnes; US$2.5

million 2002 300,000 tonnes

Table 4 . PVS in Guinea: Number of farmers participating,area sown to NERICAs and estimated gain inproduction and income from NERICAs, 1997-2002.

7Enhancing on-farm conservation of traditional rice varieties in situ throughparticipatory plant breeding in three contrasting sites from Nepal

B STHAPIT1, J BAJRACHARYA2, A SUBEDI3, K JOSHI4, R RANA3, S KHATIWADA2, S GYAWALI3,P CHAUDHARY3, P TIWARI3, D RIJAL3, K SHRESTHA2, B BANIYA2, A MUDWARI2, M UPADHAYA2,D GAUCHAN3 AND D JARVIS1

1. International Plant Genetic Resources Institute (IPGRI), APO, Serdang, Malaysia (based inNepal: 3/10 Dharmashila Buddha Marg, Nadipur Patan, Pokhara, Nepal).

2. Nepal Agricultural Research Council (NARC), Khumaltar, Lalitpur, Nepal.3. Local Initiatives for Biodiversity, Research and Development (LIBIRD), PO Box 324,

Mahendrapul, Pokhara.4. DFID Plant Sciences Research Programme, CIMMYT, Kathmandu, Nepal.

Introduction

The study was initiated in 1998 and is part of theIPGRI global project, Strengthening the ScientificBasis of in situ Conservation of AgriculturalBiodiversity in Nepal. It was jointly implementedby the Nepal Agricultural Research Council(NARC), an NGO (Local Initiatives for BiodiversityResearch and Development, LI-BIRD) and localinstitutions (community-based organisations[CBOs], a network of nodal farmers and farmingcommunities).

Understanding rice diversity

We found that communities maintained a richvarietal diversity ranging from 21 cultivars (100%of which were landraces) in the high-mountain Jumlasite to 69 cultivars (91% local, farmer-namedcultivars) in the middle-hill Kaski site. Farmers atthe highly accessible Bara village, where extensionservices had made many interventions, close to thenational rice research centre and Indian bordermaintained 33 cultivars (62% of them landraces).More landraces were maintained at both the

Participatory plant breeding (PPB) is one of the strategies employed for on-farm management oftraditional rice varieties in three contrasting in situ conservation villages of Nepal. PPB programmes inNepal were designed to investigate whether:

1. Traditional farmers' cultivars can be conserved per se.

2. PPB contributes to the enhancement of biodiversity by broadening the genetic base that providesbenefits to community.

3. The PPB process encourages farmers to maintain the evolutionary processes that allow the crop toevolve over time, while remaining adapted to diverse, local production niches.

4. Landraces can be conserved by improving them for traits that farmers consider important to makethem more competitive against alternative varieties.

This paper describes the processes of PPB that encourage farmers to locate diverse rice varietiesand their custodians; understand the amount of rice diversity (the number of local and modern varietiesat both the household and community level) and its distribution (the frequency of households growingeach variety); analyse the preferred traits of rare, endangered and locally-common farmers' varieties;and develop options for adding benefits to cultivating local landraces. It describes preliminary resultson understanding genetic diversity in terms of its value to local communities and the participatorymethods used to select landrace parents for PPB. Participatory methods, such as biodiversity fairs,diversity blocks, diversity kits and community biodiversity registers, were integrated to sensitise thelocal community to varietal diversity issues, to understand the value of local crop diversity, and tostrengthen the roles of farmers and the informal sector in the local crop development process.

Abstract

household and community level than moderncultivars in all three sites. The area covered by ricelandraces in Bara was 17%, followed by 73% inKaski and 100% in Jumla.

Landrace enhancement

We found that resource-poor farmers, compared toresource-endowed households, were more dependenton landraces for their food security. This wasparticularly true in marginal rather than in better-offenvironments (Rana et al., 2002). We also found thatsome landraces were very competitive to moderncultivars in certain niches and such landraces couldbe promoted more widely for similar agroecologicalniches without further improvement. Results alsoindicated that a few culturally important cultivars,such as Anadi, could be conserved through marketlinks and consumer awareness.

Genetic diversity of landraces

We also assessed the genetic diversity of a few locallycommon and well-known landrace populations oflandrace Jetho Budho in Kaski district, and landracesKariya Kamod and Lalka Basmati in Bara district.The landraces from Jumla had a low level of diversity(Jaccard similarity index of 0.87 and 12.8% ofpolymorphic loci) (Bajracharya et al., 2001). Thishigh degree of similarity indicates a possibility thatthese landraces, which have unique traits fortolerance to cold water chilling injury but aresusceptible to neck blast, could have originally comefrom the same source. Compared to the low diversityamong Jumla landraces, landraces from Kaski andBara had a considerable level of genetic diversity(Jaccard similarity index = 0.48-0.50), reflecting thehigh ethno-botanical variability indicated by manydifferent farmer-given names and diverse seedrelated phenotypes (Bajracharya et al., 2001).

Though intra-population genetic variation ofJetho Budho and Basmati was generally low,variation for quality traits has been found and is apotential opportunity for landrace enhancement.Through consumer surveys we have established thequality traits of Jetho Budho, which include in order:softness of cooked rice, aroma, elongation ability,and taste. The project plans to help farmers marketlocally produced, quality-assured rice seed and grainusing local brand names as a way of improving rurallivelihoods and conserving genetic diversity.

Social seed network

We also found that 96-100% of farming householdsare dependent upon an informal seed source

(Baniya et al., 2002). The seed flow occurs basicallythrough farmers' social networks. The communitymanaged a rich rice diversity through this socialsystem, such as exchange on a barter basis (60-63%),gift (20-25%), borrowing of either seed or seedlings(10-12%) and purchase (<5%) (Subedi et al., 2002).Nodal farmers who search, select and maintain ricediversity have been found to play an important rolein seed flow in the informal seed system and wereselected as collaborators in the PPB programme.

Performance of PPB bulks

The 13 segregating materials from crosses betweenlocal landraces, and local landraces with exoticvarieties, were distributed to interested and nodalfarmers for growing an F2 or F3 bulk of their choice,and a farm walk was carried out by researchers andfarmers to assess the field performance of differentpopulations.

In Begnas village F3 segregants of Mansara(locally adapted to marginal drought-prone rainfedand low-input conditions) and Khumal-4 (a good-quality modern cultivar having the local varietyPokhereli masino as one of its parents) showedpromising results in upper Begnas areas (1000-1300m) (Figure 1). The breeding goal is to incorporatethe good 'eating' quality and yield potential ofKhumal-4 into Mansara without losing the adaptivetraits of Mansara. The cross between Pusa Basmati-1 and Jetho Budho (a local high-quality rice with ahigh market demand) was also doing well inKholakochew environments (650-690m). In Baradistrict, six large segregating populations wereevaluated in farmers' fields, among which farmerspreferred the populations of the cross Lajhi withIR62161-22-1-2-1-1 and selected among them forlodging resistance and post-harvest traits. Theperformance of these segregants was very good.

Figure 1. Promising plants in the bulk of thecross Kumal 4 x Mansara

Future plans

Selected populations of Jetho Budho landraces willbe evaluated for quality traits and taste with diverseconsumers - hoteliers, millers, housewives, farmersand researchers.

The PPB bulks of the Mansara x Khumal-4 havebeen divided into three bulks based upon grain colorand phenotypes. These bulks have been distributedto all nodal farmers and interested farmers for on-farm testing and selection. Researchers will monitorthe spread of these PPB bulks in 2003 through farmwalks (Figure 2) and household interviews.

Figure 2. Participants in a farm walk, Begnas,2002.

References

Bajracharya, J, Steele KA, Witcombe JR, Sthapit BR andJarvis DI (2001). A study of genetic relationships inrice landraces of Jumla, Kaski and Bara ecosites usingmicrosatellite DNA markers. Paper presented at theNational Workshop of in situ Conservation ofAgrobiodiversity On-farm, held at Lumle, Pokhara,Nepal, 23-25 April, 2001.

Baniya, BK, Subedi A, Rana RB, Tiwari RK, ChaudharyP, Shrestha S, Tiwari P, Yadav RB, Gauchan D andSthapit BR (2002). What are the processes used tomaintain the genetic diversity on-farm? In: GauchanD, and Sthapit BR, (eds.) Nepal's Contribution toAgrobiodiversity Conservation in-situ: A ScientificBasis for Policy Recommendation. IPGRI/NARC/LI-BIRD.

Rana, RB, D Gauchan, DK Rijal, A Subedi, MPUpadhaya, BR Sthapit and DI Jarvis (2002). Factorsinfluencing farmers’ decision on management of localdiversity on-farm and their policy implications. In: DGauchan, and BR Sthapit, (eds.) Nepal's Contributionto Agrobiodiversity Conservation in-situ: A ScientificBasis for Policy Recommendation. IPGRI/NARC/LI-BIRD.

Subedi, A, Chaudhary P, Baniya B, Rana RB, Tiwari RK,Rijal DK, Jarvis D and Sthapit BR (2002). Who

maintains genetic diversity and how? Policyimplications for agrobiodiversity management. In: DGauchan, and BR Sthapit, (eds.) Nepal's Contributionto Agrobiodiversity Conservation in-situ: A ScientificBasis for Policy Recommendation. IPGRI/NARC/LI-BIRD.

8Farmer participatory breeding and participatory varietal selection in easternIndia: Lessons learned

TR PARIS1, RK SINGH2, G ATLIN1, S SARKARUNG3, G MCLAREN51, B COURTOIS4, KMcALLISTER1, C PIGGIN5, S PANDEY1, A SINGH10, HN SINGH10, ON SINGH6, S SINGH6,RK SINGH7, NP MANDAL7, K PRASAD7, RK SAHU8, VN SAHU8, ML SHARMA8, RKPSINGH9, R THAKUR9, NK SINGH9, D CHAUDHARY9, S RAM10

1. IRRI, DAPO Box 7777, Metro Manila, Philippines2. IRRI, 1st Floor, CG Block, NASC Complex, Dev Prakash, Shanstri Marg, Pusa, New Delhi-

110012, India3. Rice Research Institute, Department of Agriculture, Chatuchak, Bangkok, 109004. CIRAD Montpeiller, France5. ACIAR, Canberra, New South Wales, Australia6. Central Rainfed Upland Rice Research Station, P.O. Box 48, Hazaribagh, Bihar, India7. Narendra Deva University of Agriculture and Technology Kumarganj, Faizabad 224 229

Uttar Pradesh, India8. Indira Gandhi Agricultural University F/2 Krishak Nagar, Rapir, 492 012 Madhya Pradesh,

India9. Rajendra Agricultural University, Pusa, Samastipur, 848125, Bihar, India.10. Central Rice Research Institute, Cuttack, 753 006, Orissa, India.

Introduction

Poverty in Asia is most severe in rainfed areas wherepeople depend on rice for subsistence. To these ruralpoor, rice is not only the staple food but has manylivelihood uses. Sustaining their livelihood andalleviating poverty require a major increase inagricultural productivity. This means, among otherfactors, developing higher yielding well-adaptedrice varieties.

Classical breeding approaches have beensuccessful in developing improved varieties of ricefor favourable rice environments. However, theseapproaches have been less successful in rainfedenvironments because they fail to account for thehigh levels of social and agro-ecological diversityin these areas. Witcombe et al. (1998), reportweaknesses in the formal testing system in Indiathat have reduced the chances that varieties releasedfor marginal areas would meet farmers’ needs. Thefailure of the system is evidenced by, for example,

the rejection of many varieties by farmers, and the rapidand high adoption by farmers of non-released varietiesthat had been rejected in the formal testing, such asMashuri rice.

The goal of this project was to increase foodsecurity by providing suitable varieties for higher andstable yields. The specific objectives were:

1. To test the hypothesis that farmer participationin rainfed rice breeding can help developsuitable varieties more efficiently.

2. Identify the stages along a breeding programwhere farmer participation is most necessary.

3. Enhance the capacities of the nationalagricultural research systems (NARS) in farmerparticipatory research and gender analysis forrice plant breeding and varietal selection(Courtois et al., 2001).

This paper describes the objectives and methods used in a participatory breeding (PPB) project inseveral rice ecosystems in eastern India. Farmers were interested in a large range and combinations oftraits. The selection criteria of farmers were determined by hydrological conditions (water depth) andland types (toposequence) followed by the adaptation of the variety to different user needs (food,livestock fodder, thatching, cash). Other factors determining preference were: the compatibility of thevariety in the cropping systems, socioeconomic status, gender and culture. Several lessons learnedfrom the three years’ experience led to the refinements in the participatory methods in the next phase ofthe project.

Abstract

Methodologies and major findings

Selection of the research sitesThis project was conducted in eastern India, (easternUttar Pradesh, eastern Madhya Pradesh Assam,Bihar, West Bengal and Orissa). Eastern India is thelargest rice-growing region in the country andaccounts for about 63% of India's rice area (26.8million ha).

However, about 80% of the rice farming in thisregion is rainfed. The project involves two riceecosystems and six sites: one in the upland ecosystemand five in the rainfed lowland ecosystemrepresenting different agro-ecologies (drought prone,submergence prone or both). In each site, two to threevillages were chosen on the basis of their access tomarket, agro-ecological diversity and extent ofadoption of improved varieties. A team of socialscientists (mostly agricultural economists) and plantbreeders from each center conducted this research.

Finding out farmers’ selection criteria based ontheir agro-ecological and socioeconomicenvironmentFarm household surveys and participatory ruralappraisal (PRA) methods were conducted tocharacterize farmers' agro-ecological andsocioeconomic environment, farming systems, ricediversity, farmers’ crop management practices,gender roles, and seed management practices.Farmers were consulted with regard to theirpreferred traits as well as the positive and negativetraits of local and improved varieties. Throughgraphic illustration of traits, male and female farmersvalued the importance of traits of varieties accordingto land types (lowland, midland and upland), bysocioeconomic groups and by gender. The criteriafor varietal choice were:

• Adaptation of the variety to the hydrologicalconditions (water depth) and landtype (topsequence location) of their fields. For thelowlands and submergence-prone ecosystems,long-duration, photoperiod-sensitive, semi-talland tall varieties were well adapted. For theuplands and shallow-depth medium lands, earlyto medium varieties were preferred to escape theterminal drought. (Courtois et al., 2001; IRRI2001).

• Adaptation to different user needs such as food,livestock fodder, thatching and cash. Farmersgrew more than one variety not only due toheterogeneous rice fields but also for differentend-uses.Different varieties fulfilled different

livelihood functions (food, livestockfodder, thatching, and cash). For example,farmers preferred varieties with long, finearomatic grain, because these are used as giftsfor special occasions (marriage) and for religiousceremonies.

Poor farmers were more interested in thequality of leftover rice, which should remaintender and soft - characteristics found intraditional varieties. Similarly, traditionalvarieties were perceived to be better forpreparing puffed rice and other rice products(Paris et al., 2001). In the uplands, farmerspreferred tall varieties such as Brown Gora,Vandana, RR 151-3 and Kalinga III because theyneed the straw for animal fodder (IRRI, 2001).Other farmers grew traditional varieties with apurple-pigmented base in drought-prone areasin Bihar, Madhya Pradesh, and Cuttack. Thistrait helped farmers distinguish rice from weeds,especially in direct-seeded rice where weeds area major problem (Sahu et al., 2001)

• Compatibility with existing cropping systems.Farmers preferred medium-height varieties inthe medium lands of early duration (90-110 days)that allowed the timely sowing of a followingcrop of pulses, wheat, or vegetables. InBhubaneswar, Orissa, farmers rejected certainrice varieties that had dense root growth becausethese inhibited the establishment of a relay cropof pulses (black gram) broadcast into thestanding rice crop in rice (IRRI 2001).

• Socioeconomic status of farmers. Farmerspreferred different grain types according tosocioeconomic status or degree of marketintegration. For example, in Faizabad, easternUttar Pradesh, farmers and field workersof lower castes with small landholdingspreferred varieties with coarse grains, which givethem a feeling of fullness due to their slowdigestibility. the higher caste farmers with largelandholdings who sell rice to the marketpreferred fine slender grains, which command ahigher price.

In general, smallholder farmers from thelower castes used rice mainly for consumption,while farmers from the upper caste with moreland sold their surplus (Paris et al., 2001).Farmers in the uplands of Hazaribagh preferredvarieties that do not require high inputs. Farmerswho depend on family labour preferred varietieswith a range of maturity dates so that harvestscan be staggered (IRRI 2001).

• Gender-specific roles in rice production, postharvest, consumption, and livestock care. At allthe sites, there were gender-specific tasks in riceproduction. Women from poor farminghouseholds provided most of the labor in riceproduction (pulling of seedlings, transplanting,weeding), post harvest (winnowing, handthreshing, seed drying), and seed management(selection, storage). Men were mainlyresponsible for land preparation, application ofchemicals, and transporting inputs and products.Male and female farmers in Faizabad agreed thatgrain yield and duration were the most importanttraits when choosing varieties for upland andlowland areas. However, women gave moreimportance to traits such as competitiveness toweeds and post-harvest qualities such as ease ofdehusking, ease of threshing, and high millingrecovery or suitability for different foodpreparations (puffed rice). In Raipur, MadhyaPradesh, women consistently identified strawquantity and quality as important, whereas themen never mentioned these as important.Women’s criteria for varietal selection werelikely related to their roles and responsibilities(Paris et al., 2001; Sahu et al., 2001).

Selecting new varieties suitable to farmers'preferences and agro-ecological conditions

Participatory plant breeding (PPB). Farmers andbreeders selected individual plants from segregatingpopulations (F5) from different crosses, using thepedigree selection method. Trials were held on-farmand on-station. Plants selected by breeders andfarmers were advanced separately through severalgenerations until fixed. Farmers and breedersevaluated these genotypes at maturity on the basisof panicle and grain characters, and susceptibility tostem borer. Breeder-selected and farmer-selectedmaterials were then compared. Promising lines wereselected and these genotypes were multiplied andsupplied to farmers for evaluation.

Participatory varietal selection (PVS). In PVS,similar sets of fixed varieties (13-25 advanced linesand a local check) suited for the specific hydrologicalconditions in the area, were tested on-station and onfarmers' fields. The advanced lines were from theIRRI Shuttle Breeding Program and from breedingprograms of project partners in eastern India. Twoto three farmers per villages conducted the on-farm

trials under their level of management. At vegetative(pre-flowering) and reproductive stage (maturity) ofrice, farmers and breeders visually ranked the ricelines grown on-station and on-farm trials. TheKendall coefficient of concordance and theSpearman rank correlation coefficient were used toanalyze the agreement of ranking of the genotypesamong farmers, among breeders and betweenfarmers and breeders (Courtois et al., 2001). Theresults were as follows:

• There was strong agreement among farmers'visual rankings, but not always among those ofplant breeders.

• Agreement of breeders with farmers was fairlygood on plant traits but many quality traits wereoverlooked.

• Agreement between visual ranking of traits bymen and women during pre-harvest stage wasgood. However, they differed during post-harvest assessment.

• Yield and duration were important traitsconsidered by farmers. However, rankingpreferences were not always correlated,indicating that these traits are not the only factorstaken into farmers when selecting rice varieties.

• Sensory evaluation showed that the mode of ricepreparation (parboiled or not parboiled)influences farmer selection. Sensory ranking didnot correlate with results of classical physico-chemical analysis (Singh et al., 2001).

Lessons learned

Several lessons were learned from the three years ofexperience in developing and testing themethodologies for farmer participation. Theselessons are related to the concerns:

• Choice of representative sites. During the firstyear, a few of the sites selected for on-farm trialswere not representative of the environmentstargeted in the breeding work. They were chosenfor convenience reasons (e.g. close to thestation). Thus new trial sites, which betterrepresented farmers' environments, replacedsome of the sites.

• Number of villages to represent a specificagro-ecology. Due to limited resources and staff,only two to three trials and farmers wereincluded in each village. Thus the risks of losinginformation due severe drought, poormanagement of trials, etc. were high. The‘mother-baby’ trial model may provide an

alternative in reliably testing a large number ofcultivars under farmer management (Atlin et al.,2001; Witcombe this volume).

• Choice of the varieties to be included in theexperiments. The material chosen for the PVStrials was not always ideal. The lack of clear-cut differences between some of the varietiesincluded in the PVS made it more difficult forthe farmers to rank them. The number ofvarieties tested should be balanced betweenwhat is useful for the breeders (many lines),acceptable by the farmers (fewer lines) and whatis possible on a reasonable plot size.

• Number of varieties to rank. Farmers haddifficulty in visually ranking too many (13-25)rice lines from 1 (best liked) up to n (least liked).Farmers were willing to test a maximum of fivevarieties on their own field. A rating system forexample, 1-3 (bad, average, good) or 1-5numerical scale, for traits is a simpler method(Atlin et al., 2001).

• Constraints in post-harvest operations of toomany lines. The need to harvest, thresh, andweigh the different varieties in small quantitieswas too cumbersome. Moreover, this workincreased the burden of the women cooperatorswho did the post-harvest work. The number oflines for testing and selection should be reduced.A local field technician is required to assist inpost-harvest work and to ensure that the varietiesare not mixed.

• Number of varieties to include in sensoryevaluation. Men and women found it difficultto evaluate the cooking and eating quality oftoo many lines. Orgonoleptic tests should bemodified so that fewer varieties are tested at anyone time.

• Access to new seeds. During the first two yearsof the project, the availability of seeds was quitelimited and not all farmers who wanted toparticipate in the selection of new rice lines ontheir own farms could be included.

• Institutional constraints. There was a lack of plantbreeders at centers and NARS with experienceon participatory approaches. In some centers, itwas difficult for the breeders to change theirpractices and incorporate the participatoryapproach into their formal breeding program.There were fears that farmer participatorybreeding will replace, rather than complement,

conventional breeding. Moreover, the skills indoing this kind of work, which involves multi-institutional participation, diverse socio-culturalsettings and many stakeholders, were not welldeveloped. In many of the centers, there wereno female social scientists or female plantbreeders included in the team. Thus, it wasdifficult at the beginning for the male plantbreeders to increase the number of womencooperators in the PVS on-farm trials. Thus thereis a need to develop partnerships with NGOsand extension research institutions.

Conclusions

PVS is essential in unfavourable rainfedenvironments and diverse socioeconomic groups thatuse rice for their livelihood. Farmer participationimproved the selection of suitable varieties forcomplex rainfed environments in eastern Indiabecause: a) farmers screened new varieties on theirown farms under their own levels of management,and b) breeders better understood farmers' qualityrequirements.

The close collaboration between scientists andfarmers helped ensure that farmers had access touseful germplasm that was adapted to theircircumstances and met the requirements of thecommunity for quality and livelihood use. Providingfarmers with diverse materials also help enhancegenetic diversity. Continuous feedback mechanismsbetween breeders and farmers should be establishedto ensure appropriate materials are disseminated tofarmers.

There was little evidence of major differencesin the way farmers and breeders visually ratedvarieties in the field. Hence, simply having farmerstake over the early generation visual selection isunlikely to result in significant improvements in theacceptability of cultivars. On the other hand,extensive farmer-managed trials appear to be apromising approach to reducing the effects of randomvariability and increasing gains from selection.

Although it is too early to show the impact ofPVS through the spread of the materials generatedfrom that PVS, there are promising signs that farmer-selected varieties are easily spreading throughfarmer-to-farmer exchange. Plant breeders arebeginning to change their mindsets and have realizedthe necessity of interacting with farmers, as well aswith socio-economists. However, the challenge liesin institutionalizing farmer participation as integralin the formal breeding programs and in scaling upPVS as well in dissemination of quality new seeds.

References

Atlin G, Paris T and Courtois B (2001). Sources ofvariation in rainfed rice PVS trials: implicationsfor the design of mother-baby trial networks.Paper presented at the workshop QuantitativeAnalysis of Data from Participatory Methods inPlant Breeding, 23-25 August 2001, Giessen,Germany.

Courtois, B, Bartolome B, Chaudhary D, McLarenG, Misra CH, Mandal NP, Pandey S, Paris T,Piggin C, Prasad K, Roy AT, Sahu RK, SahuVN, Sarkarung S, Sharma SK, Singh A, SinghHN, Singh ON, Singh NK, Singh RK, Singh RK,Singh S, Sinha PK, Sisodia BVS, Takhur R(2001). Comparing farmers and breedersrankings in varietal selection for low-inputenvironments: A case study of rainfed rice ineastern India. Euphytica 122: 537-550.

Courtois B, Singh RK, Pandey S, Paris T, SarkarungS, Baghel SS, Sahu RK, Sahu VN, Sharma SK,Singh S, Singh HN, Singh A, Singh ON, SisodiaBVS, Misra CH, Roy JK, Chaudary D, PrasadK, Singh RK, Sinha PK, Mandal NP (2000).Breeding better rainfed rice varieties throughfarmer participation: some early lessons fromeastern India. In: Lilja N, Ashby J and SperlingL, (eds.), Assessing the Impact of ParticipatoryResearch and Gender Analysis. ParticipatoryResearch and Gender Analysis Program,International Center for Tropical Agriculture,Cali, Columbia. 208-223.

IRRI (2001). Farmers and scientists: building apartnership for improving rainfed rice in easternIndia. Final narrative report and technical reportsubmitted to the IDRC, Canada, September2001. IRRI, Metro Manila, Philippines.

Paris TR, Singh A, Luis JS, Hossain M, Singh HN,Singh SS, Singh ON (2001). Incorporatinggender concerns in participatory rice plantbreeding and varietal selection: preliminaryresults from eastern India. In: Lilja N, Ashby Jand Sperling L (eds.), Assessing the Impact ofParticipatory Research and Gender Analysis.Participatory Research and Gender AnalysisProgram, International Center for TropicalAgriculture Cali, Columbia. 109-121 pp.

Paris TR, Singh A, and Luis J (2001) Listening tomale and female farmers in rice varietalselection: a case in eastern India. In: CIAT, AnExchange of Experiences from South andSoutheast Asia. Proceedings of the InternationalSymposium on Participatory Plant Breeding andParticipatory Plant Genetic ResourceEnhancement, Pokhara, Nepal, 1-5 May 2000.Centro Internacional de Agricultura, Cali,Columbia.

Sahu RK, Sahu VN, Sharma ML, Paris TR,McAllister K, Singh RK, and Sarkarung S(2001). Understanding farmers' selection criteriafor rice varieties: a case in Madhya Pradesh,

eastern India: a case study in eastern UttarPradesh, India.

Singh RK, Prasad K, Mandal NP, SinghRK, Courtouis B, Singh VP (2001). Sensoryevaluation of rice varieties with farmers: anexperience in eastern India. In: CIAT, AnExchange of Experiences from South andSoutheast Asia. Proceedings of the InternationalSymposium on Participatory Plant Breeding andParticipatory Plant Genetic ResourceEnhancement, Pokhara, Nepal, 1-5 May 2000.Centro Internacional de Agricultural, Cali,Columbia.

Witcombe JR, Virk DS, and Farrington J (eds.)(1998). Seeds of Choice. Making the Most ofNew Varieties for Small Farmers. Oxford/IBHNew Delhi, and Intermediate TechnologyPublications, London.

10Participatory scaling up of participatory varietal selection

KD JOSHI1, S GYAWALI2 AND JR WITCOMBE3

1. DFID Plant Sciences Research Programme (PSP), CIMMYT South Asia Office, Kathmandu, Nepal.2. Local Initiatives for Biodiversity, Research and Development (LI-BIRD), Nepal.3. Centre for Arid Zone Studies (CAZS), University of Wales, Bangor, UK

Background

For over five years in Nepal, LI-BIRD, incollaboration with CAZS, has been developing andpromoting participatory crop improvement (PCI)and participatory plant breeding (PPB) approaches.These approaches have been found to be simple,rapid and effective in identifying and disseminatingfarmer-preferred technologies (Joshi et al. 2001;Joshi and Witcombe, 2002; Witcombe, 2001).

The next important step was to find better waysof scaling up to improve cost-effectiveness. The firstassumption was that the active involvement offarmers in this process - 'participatory scaling up' -would greatly improve cost-effectiveness.Participatory scaling up involves farmers as seedproducers and disseminators. When there is a highdegree of farmer involvement it is easier to benefitresource-poor farmers in remote regions, but it alsoworks well for more favourable environments.Another important advantage is that technologyverification and adoption takes placesimultaneously.

We begin by briefly reviewing the role offarmers before discussing the different stakeholdersin this process and the partnerships that are evolvingin Nepal.

The role of farmers' participation

Participatory scaling up is designed to maximise thecontribution of farmers in the process. For example,farmers are identified that are key individuals in theinformal seed supply system in the village. Suchfarmers are contracted to produce seed for theproject for dissemination by informal research and

development methods (Annexe 1) or they are givengood quality source seed and asked to disseminatethe harvest through their usual informal network.During the participatory scaling up processindividual initiatives, the actions of farmers' groups,and the results of informal farmers networks havebeen documented to illustrate the role of farmers inthe process.

One example of the role of an individual isMr. Daya Ram Khandka, Chairman of the DhadwarVillage Development Community (VDC), BardiyaDistrict. He is also a farmer, and a member of theKulanath Krishak Samuh (a farmers group). Mr.Khandka took on the responsibility of distributingseeds, periodically monitored activities, andmaintaining records of the yield and yieldcomponents of all the varieties on offer from a LI-BIRD and District Agricultural Development Office(DADO) Bardiya collaboration. The farmers' groupsaved the entire seeds of all the varieties for plantingnext year. They have their own norms for exchangingseeds; any farmer wishing to get seed of new varietyshortly after harvest within the month of Mangsir(Nov 16 to Dec 15) could do so just by exchangingseeds for rice grains. However latecomers will haveto buy the seed paying cash at a rate of two rupeesless than the standard price for seed fixed byAgricultural Inputs Corporation (AIC). Thiscommunity initiative for varietal evaluation and seedmanagement indicates a sustainable way ofachieving impact from PVS.

One of the greatest displays of interest fortesting and adopting new varieties from a single-season's exposure was at Beladevipur village, KailaliDistrict. Most of the people in the village whom themonitoring team could see could tell the names of

AbstractFarmers are the ultimate beneficiaries of new agricultural technologies, and the best indicator for thesuccess of a technology is its acceptance and uptake by them. We discuss the benefits of the activeinvolvement of farmers in scaling up the results of participatory varietal selection (PVS) and participatoryplant breeding (PPB). We discuss the stakeholders and partnerships for participatory scaling up ofvarieties. In Nepal, we describe the evolving partnerships between government organisations (GOs)non-government organisations (NGOs), and community-based organisations (CBOs) that is designedto combine the strengths of the stakeholders whilst sharing responsibilities and resources.

all the varieties tested in that site, the relativeperformance of all the varieties, and which particularvariety they would grow next year. Such interest isonly possible when an external agency has been thecatalyst of providing seeds and information so thatfarmers have interacted intensively with their fellowfarmers on the merits of the new varieties.

There was also good evidence of farmer-to-farmer seed spread in Kailali District. Nearly 1500kg rice seeds of the new varieties would be plantedthis season in all the sites in Kailali from an initial300 kg that was distributed (Ram AyoudhayaMahato, DADO, Kailali pers. com.).

The spread of variety BG 1442 in Chitwan andMakwanpur districts provides one of the mostconvincing examples of the effectiveness of farmernetworking. The variety was introduced in farmers'field trials during the early nineties as a Chaite ricevariety but it is yet to be formally released by thenational research system. The variety is promisingfor the Chaite season; for partially irrigated well-drained conditions as a main season rice variety andis equally suitable for Bhadiaya system (rainfed landin the terai where early- maturing rice varieties aregrown). The crop is planted in May and is calledBhadaiya because it is harvested in the Nepali monthof Bhadra (late Aug-Sept).

The variety is spreading very quickly fromfarmer-to-farmer. LI-BIRD introduced a few kilosof seeds of this variety in participatory varietalselection (PVS) trials in 1998 1999 and subsequentlyin informal research and development (IRD) from2000 to 2002. The variety increased rapidlyin area from 2001 to 2002 to cover nearly1200 ha, about one third of the Chaite areain Chitwan district. The spread wasencouraged by the breakdown of thepredominant variety CH 45 to blast diseases.The variety is also spreading in villages inthe neighbouring district of Makwanpur andcovered nearly 50 ha area in the main seasonof 2002. Most of this is from farm-savedseeds spreading through the farmer-to-farmer network, facilitated by distributionof source seeds by DADO, Chitwan, LI-BIRD and some farmer cooperatives ineastern Chitwan.

Although farmer-to-farmer spread ofseeds is very important and has been themajor source of seed in most developingcountries, including Nepal, it is not thefastest system. Farmers consume seed on-farm and sell it as grain to the markets soonly a proportion can be used as seed forsowing. Our experience with most of thefarmer-to-farmer spread has been that a three

to four fold increase is often found, but a sixty foldincrease can easily be achieved with an organisedseed increase and distribution programme. Thisefficient seed source of officially multiplied seed canbe used most effectively when it is distributed widelyand to many farmers.

Stakeholders and modes of partnership

Once the central role of farmers was established, afirst step in scaling up is the identification ofstakeholders and devising the most appropriate modeof partnership (Table 1). Lack of lasting effects ofsome of the research and development projectsdesigned and run in top-down approach in the pastwas due to lack of farmers' participation, and ignoringthe central role that farmers can play. There was alsoa lack of identification of all stakeholders and theirroles and responsibilities.

Scaling up is more complex than simpletechnology generation or verification because itinvolves more stakeholders. Technology generationoften ignores farmers and multi-partner approaches(Table 1). A multi-partnership is the most suitableone for scaling up (Table 1) in which farmers play acentral role.

The benefits of this approach are particularlyimportant when, as is common, linkages betweenresearch and extension are poor and governmentextension services are restricted to serving moreaccessible areas by limited resources.

Donor CGIAR NARS Ignores farmer, extension and NGO sector

Donor CGIAR NAES Ignores farmer, research

and NGO sector Donor CGIAR NGOs Ignores farmer and GOs Donor NGOs Ignores farmers, GOs and

CGIAR Farmer NARS NAES Multi-partner

collaboration of all stakeholders. Farmers’ first

NGOs CGIAR Donor –

Table 1. Stakeholders for participatory scaling up(CGIAR: Consultative Group on InternationalAgricultural Research, NARS: NationalAgricultural Research System, NAES: NationalAgricultural Extension System, NGO: non-governmental organisation, GO: governmentalorganisation).

NGO and Community partnershipsAn important form of collaboration that is emergingfor participatory scaling up is collaboration betweenNGOs and Community-based organization (CBO).The CBOs are usually farmers' groups that werealready established for cooperative enterprises orfor specific users' interest, e.g. drinking water. ThreeCBOs were as follows:• Women’s group (Narayani Mahila Milan Club,

Amaltari, Kumarti-9);• Mixed farmers group (Gramin Swablamban

Bikas Kendra, Bishaltar, Devchuli-1);

Box 2. Letter of Agreement between DADO Chitwan and LI-BIRD for participatory scaling up.

Roles of DADO·• Full responsibility for implementing scaling-up activities through their networks.• Preparing a joint annual plan and obtaining the approval of the Ministry through the regular planning

process.• Periodical monitoring - collecting field information and preparing progress and technical reports.·• Providing financial and other institutional support to seed production of the new, unreleased varieties,

as applicable to released varieties.• Farmer and staff skill development on participatory approaches through orientation, training and

exposure visits.• Organizing periodical coordination and review meetings with LI-BIRD.

Roles of LI-BIRD·• Arranging source seeds of the farmer-preferred varieties for the scaling up.• Assisting in preparing the annual plan.• Imparting skills to DADO staff through joint activities and through orientation and training

programmes.• Sharing in data analysis and report writing.• Organizing seasonal monitoring tours involving DADO, DoA, NARC & IAAS staff.• Providing agreed financial and institutional support.

The process·• Participatory rural appraisals (PRAs) to identify community-based organizations suitable for participatory

scaling up, and the analysis of secondary information through agricultural service centres.• Documentation of profile of short-listed farmers' groups by PRA.• Selection of established farmers' groups.• LoA with three farmers' groups (FGs), two mixed groups and one women farmers' group• Implementation of programme through the groups.

Role of farmers' groups·• Responsible for carrying out all the field activities: farmer identification, (focus on poor and medium-poor

farmers), distribution of seeds, record keeping, including the monitoring of varietal adoption.• Seed production and farmer-to-farmer distribution.• Group decision making.

Role of LI-BIRD·• Imparting necessary skills for scaling up through orientation, training and exposure visits.• Providing technical backstopping.• Assisting in monitoring and evaluation.• Providing financial and institutional support for scaling up.

Box 1. Scaling up through farmers' groups in Nawalparasi District

• Water users' group (Naya Belhani KhanipaniUpabhokta Samiti, Naya Belhani-2).

LI-BIRD signed LoA with three CBOs inNawalparasi, one of them was formed by DADOoffice while two got organized on their own.

In this type of partnership, both organisationshave a high degree of institutional flexibility. Itssuccess will depend on both the type of partnersrespecting each other's roles and responsibilities(Box 1) and, in the case of LI-BIRD and three CBOs(Narayani Mahila Milan Club, Amaltari, Kumarti-9, Gramin Swablamban Bikas Kendra, Bishaltar,Devchuli-1 and Naya Belhani Khanipani UpabhoktaSamiti, Naya Belhani-2), there were three separate

Partnership modes for participatory scaling up

LoAs to make these clear. However, the scale ofoperation of NGO and CBO partnerships will usuallybe limited as most farmers groups are localorganisations.

NGO and GO partnerships

LI-BIRD and DADO Chitwan. One mode of workingis collaboration between an NGO and a GO. Anexample in Nepal is the partnership between LI-BIRD and the Government line agency the DADOin Chitwan district. One unique feature of thispartnership was the signing of a letter of agreement(LoA) for three years, which agreed the roles, andresponsibilities of the two partners (Box 2). However,the sustainability of this approach could be improvedif funding was from the recurrent budget of thegovernment rather than a special project funded bythe Plant Sciences Research Programme.

This agreement between an NGO and a GO maybe one of the first examples of its kind in Nepal (LoAsigned by district level authority of the Departmentof Agriculture rather than at the Department level)although there are other examples of GO-NGOcollaboration for specific development objectives atthe Department level, such as collaboration betweenCARE Nepal and the Departments of Soil and WaterConservation and the Department of Health Services.

The initial interest for the formal collaborationcame from the GO. However, informal collaborationstarted before that; when LI-BIRD initially set upthe research project it consulted DADO Chitwan onidentification of project villages and DADO staffregularly participated in the monitoring of theparticipatory varietal trials. DADO then startedscaling up of the unreleased farmer-preferred cropvarieties from their own funds for over a year.

Expanding the collaboration to more districtoffices. A stakeholders meeting was held inKathmandu in February 2001 to discuss building onthe existing partnership between LI-BIRD andDADO, Chitwan. The meeting agreed on theadvantages of the participatory scaling up and set upa joint Working Group of the Department ofAgriculture (DoA), Nepal Agricultural ResearchCouncil (NARC) and LI-BIRD. The Workingdeveloped a proposal to scale up the activities inChitwan to four more districts: Dhanusha, Sarlahi,Bardiya and Kailali. This was subsequently approvedby the authorities from DoA, NARC & LI-BIRD.The World Bank-funded Agricultural Research andExtension Project (AREP) agreed to provide a smallgrant for this work. Farmer-preferred rice varietieswere produced and disseminated in 2001 in the majorrice growing domains in each of four districts.

Farmers tested the new rice varieties in long-standingwater, irrigated, partially irrigated and rainfedconditions, and for the Bhadiaya system.

Consolidation and further expansion to moreorganisations.

A workshop was organized in January 2002 jointlyby Department of Agriculture (DoA) and LI-BIRDto share the findings from the five collaboratingdistrict offices. Representatives from seven DADOsparticipated and developed seven project conceptnotes (PCNs) for participatory scaling up of PVS-identified and PPB-produced varieties.

The possibility was discussed of incorporatingthe work plans of the PCNs into the regular planningprocess of the DoA to attract regular funding. Someof the DADOs have already done so, indicating agreater acceptance and uptake of the approach bythe DoA. Although the extent of activities differsgreatly, currently activities have been extended to17 terai and two hill districts. In these initiatives,LI-BIRD is also collaborating with other NGOs suchas FORWARD, REGARDS, CARE and PLANInternational, in addition to working closely withthe DoA.

Opportunities

Currently there is a helpful environment forparticipatory research and scaling up in Nepal. TheMinistry of Agriculture and Cooperatives haveadopted partnership as one of the main approachesto agricultural research and development. HisMajesty's Government of Nepal has set up acompetitive funding system for action research andscaling up called the National Agricultural Researchand Development Fund (NARDF). Agriculturalextension will very soon be devolved to the localgovernment, which necessarily involvesdecentralization and a greater emphasis on people'sparticipation. DADOs are also considering thecontracting out of some of their service deliveryfunction to improve the efficiency.

References

Joshi, KD, Sthapit, BR, and Witcombe, JR, 2001. Hownarrowly adapted are the products of decentralisedbreeding? The spread of rice varieties fromparticipatory plant breeding programme in Nepal.Euphytica 122:589-597

Joshi, KD and JR Witcombe 2002. Participatory varietalselection in rice in Nepal - a comparison of twomethods by farmers' selection and varietal adoption.Euphytica. In press.

Witcombe, JR, 2001. The impact of decentralised andparticipatory plant breeding on the genetic base ofcrops. In: HD Cooper, C Spillane & T Hodgkin (Eds.)Broadening the genetic bases of crops, pp. 407-417.CAB International, UK.

Introduction

Most upland rice farmers, particularly in easternIndia, are resource poor and are not able to usepurchased inputs. Improved varieties developed forrainfed uplands were also not fully adopted byfarmers. The local cultivars have the desirablecharacters to withstand unfavourable growingconditions but they yield less than a tonne perhectare. Traditional breeding approaches fordeveloping improved rice varieties have not beenvery effective. By the time on-farm trials areconducted, the majority of the genotypes haveusually been discarded which might have appearedpromising from the farmers' point of view. Farmersprefer varieties that will perform well under multiplecropping and integrated farming situations. They arealso looking for varieties that could provide productsand by-products for household needs.

Past experience clearly shows how breedershave failed to adequately address the problems ofupland ecosystems. Variety adoption has beenmeagre and spread has been slow and restricted.Impact of modern varieties was poor mainly due totheir shorter stature, lack of early vigour and poorresponsiveness to low input management. Farmersare not in favour of taking risks in unpredictableenvironments. On-farm research with farmers’participation aims at involving farmers to select andtest new varieties along with existing traditional

types under their own management practices. Aparticipatory breeding project was started in 1997with this objective, under the ICAR-IRRIcollaborative programme. The project has two majorcomponents - plant breeding and social science. Thebreeding component consisted of two distinct parts:

• Participatory varietal selection (PVS), where existing cultivars or fixed breeding lines wereexposed to the farmers for selection(Witcombe et al., 1996).

• Participatory plant breeding (PPB), in which farmers were involved in selecting lines fromsegregating populations.

Materials and Methods

Site selection Three villages in Hazaribag district were selected forthis project on the basis of representativeness of theenvironments targeted in the breeding work, diversityand range of ecological conditions, involvement ofwomen in farm activities, availability of previoussurvey data and easy access to the site.Approximately 20% of the total cultivated areacomes under upland rice in these three villages.Farmers grow Gora rice in upland though yield ofGora rice is very low (usually less than 1 tonne ha-1).

Farmers’ participatory breeding for upland rice in eastern India

NP MANDAL1, PK SINHA1, RK SINGH1, M VARIAR1 RK SINGH2 AND GN ATLIN2

1. Central Rainfed Upland Rice Research Station (Central Rice Research Institute),PO Box-48, Hazaribag- 825 301, Jharkhand, India

2. International Rice Research Institute, DAPO Box 7777, Metro Manila, Philippines

AbstractLack of access to new, high-yielding rice varieties, and to information about new varieties, are majorconstraints for resource-poor farmers in eastern India. There is no effective local seed system, and newvarieties have not been adopted because of their lack of early vigour, short stature and poor response tolow-impact management. Although local cultivars can survive the poor environmental conditions, theydo not have high yields. Traditional breeding approaches have been ineffective in producing modern,high yielding varieties that can withstand growing conditions under local management practices.A participatory breeding project begun in 1997 aimed to get farmers involved in testing new varieties ontheir own fields along with traditional cultivars. Results were outstanding. Farmers consistently selectedmore successful varieties than breeders, and were able to select superior progenies from the segregatinggeneration. The project showed that participatory varietal selection (PVS) and participatory plantbreeding (PPB) are excellent ways of helping breeders to understand the problems facing farmers, andallow the farmers to take an active part in driving their farming practices forward.

Farmer selectionTen farmers from each village were selected for thisproject. Most of the farmers selected in the first yearcontinued to participate up to third year except in afew cases. Farmers were selected based on theirprimary occupation being rice farming, especiallyupland rice, ownership of the operational landholding and some degree of literacy.

PVSThe PVS trials were conducted for three consecutiveyears during 1997-1999, with 15-16 upland ricegenotypes (including the local check, Brown Gora).In the first year, the trial was conducted infarmers’fields in three villages, one set managed byfarmers and another set managed by breeders inadjacent blocks. Breeders also conducted the sametrial at the research station. From the second year thetrial was conducted in all three villages under farmermanagement, except for the seeding, and at theresearch station by breeders. Most of the test entriesof the PVS trial were continued for three years. A fewentries ranked as very poor by farmers and breederswere replaced with new ones in the second and thirdyear of testing. The trials were unreplicated CRD infarmers’ fields and replicated RCBD at stations. Theparticipating farmers of a given village and thebreeders evaluated all the trials conducted in thevillage and also at the research station. The genotypeswere ranked at two phenological stages (vegetativeand reproductive) as most liked to least liked, on thebasis of their selection criteria. In addition, breedersalso recorded duration, height, yield and yieldcomponents, and reaction to disease and pests in allthe trials. To compare the ranks given by farmers andbreeders, the Kendall coefficient of concordance (W)was used as described in Siegel (1956). Farmers’ andbreeders’ rankings were compared following theSpearman Rank coefficient of correlation. Detailsabout the methodology used can be found in Courtoiset al. (2001).

PPBPPB started one year after PVS, after farmers weregiven some basic training on the objectives andmethodology of single plant/line selection in thesegregating population. One hundred segregatinglines from 12 crosses in the F5 generation weregrown in one on-farm site (Khorahar) and at theresearch station. Lines were scored by farmers andbreeders at both sites at different stages of the crop.At maturity, single plant selections were madeseparately by farmers and breeders at both sites. Theplants selected by farmers and breeders on -stationand at Khorahar were grown in separate blocks in the

same field in respective sites in the 1999 wet season.The farmers and breeders continued their selection onthe respective materials and places and, during thewet season 2000, four sets of final (bulk) selectionswere made - i.e. breeders’ and farmers’ selection on-station, and breeders’ and farmers’ selection on-farm.These four sets were pooled, along with three checks(Brown Gora, Kalinga III and Vandana) andevaluated both on-farm and on-station during the2001 wet season in an alpha-lattice design with tworeplications. Analysis of the PPB trial data wasconducted using a mixed model, with selector andselection environments taken as fixed, and lineswithin selection environment x selector combinationsconsidered random. The analysis was conducted withthe REML algorithm of SAS PROC MIXED.

Results

PVS trialsThe coefficient of concordance among farmers washighly significant in all the trials conducted in threeyears. This indicated that farmers’ rankings were notrandomly attributed and there was good agreementamong them. The concordance among breeders’rankings was also high but often not significantbecause of the smaller number of breeders. The rankcorrelation between farmers’ and breeders’ averagerankings was highly significant except in a few cases(1997 on-farm results and at station in 1999). Thisindicated a good agreement between farmers andbreeders in varietal choice. The farmers’ or breeders’rankings at maturity were not always highlycorrelated with yield, especially with farmers’ranking in low-yielding trials. This may be becausefarmers consider other criteria than grain yield whileselecting varieties for their farms. It may also bebecause ranking was undertaken before harvest.

There was no correlation between farmers’ andbreeders’ rankings and duration or plant height, witha few exceptions. The variance component analysisfor the on-farm trials showed that there were largeand significant differences among cultivars in grainyield under farmer management. The grain yield dataaveraged over the four on-farm sites and three yearsshowed that one elite line (RR 348-5) significantlyoutyielded Vandana, and produced about 100% moreyield than Brown Gora (Table 1).

No variety x location or variety x yearinteraction was detected (Table 2). The three-wayinteraction, variety x location x year, was large, butcannot be separated from the within-trial error in thisanalysis, as the trials were unreplicated. This resultdoes not support the hypothesis that varieties exhibit

specific adaptation to particular sub-environmentswithin the target region of the research station. Therepeatability or broad-sense heritability of differenton-farm and on-station trials were estimated to judge

the precision of the trial. It was found thatrepeatability of on-station trial was poor, compared toon-farm trials. This may be because of the low yieldof the on-station trials, because upland rice is beinggrown continuously, resulting in poor soil fertility.

The regression of on-farm performance on on-station performance was non-significant. This meansthat on-station testing is not able to predict the on-farm performance of the genotype, and there is a needfor integration of on-farm testing at an early stage ofthe breeding programme. At the end of three years ofPVS trials farmers selected five varieties - RR 354-1,RR 347-166, RR 151-3, RR 166-645 and RR 51-1 -using their own selection criteria. The highestyielding line, RR 348-5, was not selected by thefarmers. Farmers do not consider yield as the onlytrait of importance.

RR 166-645 was preferred for its long, slendergrains. RR 51-1 appealed to the farmers for darkgreen leaves and high tillering ability. Farmerspreferred tall varieties as they need fodder for theircattle. Farmers’ perception on cooking quality wasthat RR 354-1 and RR 347-166 were better than theirlocal variety or even Vandana.

During the 2000 wet season, seeds of thesevarieties were given to farmers in two villages togrow in their field along with the check (BrownGora). In spite of severe drought, RR 347-166, RR151-3 and RR 354-1 performed better than thefarmers’ variety in most of the farmers’ fields.However, RR 51-1 failed as it was susceptible todrought. Drought tolerance was not taken intoaccount by farmers, as the last three years werealmost favourable. Now we need to improve thisotherwise good genotype for drought tolerance.

Mother-Baby trials (see Annexe 1) werestarted during the 2001 wet season. Theanalysis of mother trial data at three on-farmsites showed that there were large andsignificant differences among cultivars ingrain yield under farmer management (Table3). The on-station data are not includedbecause there were large differences in on-station and on-farm performance. There wasno significant cultivar x location interaction(Table 4). This implies that there was no rankchange in varietal performance within thissub-environment.

Table 1. Mean grain yield (t ha-1) of upland cultivars and breeding lines evaluated under farmer management in four villages near Hazaribag, Jharkand, over four wet seasons.

Cultivar No. of trials Mean1 Brown Gora 12 1.12 Vandana 11 1.74 RR 347-166 12 1.92 RR 348-5 12 2.25 RR 151-4 9 1.11 RR 203-16 9 1.47 RR354-1 12 2.12 RR 51-1 12 1.78 RR 151-3 12 1.68 RR 166-645 12 1.49 RR 50-5 9 1.63 RR 139-1 12 1.25 LSD.05 for means over nine trials

0.47

Table 2. Variance components for cultivarsevaluated under farmer management in four villages near Hazaribag, Jharkand, over four wet seasons.

Source Variance component Cultivar 0.077 Cultivar x location 0 Cultivar x year 0 Cultivar x location x year 0.246

Cultivar Chichi Khorahar Peto Mean1 CRURRS2

RR 354-1 1.32 1.47 1.70 1.49 2.67 RR 347-1 1.46 1.53 1.30 1.43 2.92 RR 347-166 1.61 1.55 1.10 1.42 3.25 RR 348-5 0.74 1.65 1.50 1.30 2.78 RR 345-2 1.47 1.38 1.00 1.29 2.30 Kalinga III 1.13 1.30 0.80 1.08 1.83 RR 361-1 1.03 1.28 0.90 1.07 2.35 CR 876-6 1.29 1.20 0.70 1.06 1.95 RR 51-1 1.44 0.87 0.80 1.04 2.67 Brown gora 1.26 1.22 0.60 1.03 1.85 RR 151-3 1.09 1.10 0.50 0.90 2.10 Vandana 1.01 0.95 0.70 0.89 2.43 RR 166-645 1.24 0.87 0.40 0.84 2.25 RR 363-737 0.90 0.92 0.60 0.81 2.38 RR 139-1 0.55 1.32 0.35 0.74 1.87 RR 361-783 0.51 0.90 0.33 0.58 1.15 LSD.05 0.70 0.70 0.70 0.42 0.76 1Over villages. 2On-station

1Over years and locations

Table 3. Mean grain yield (t ha-1) of upland cultivars and breeding lines evaluated under farmer management in three villages and on-station in Hazaribag, Jharkhand, 2001 WS.

Averaged over the three sites, RR 347-1, RR354-1, and RR 347-166 significantly outyieldedVandana (Table 3). These lines also performed wellduring 1999-2000. They yielded about 40% morethan Brown Gora and about 30% more than KalingaIII in 2001, and have consistently outperformed theselines in previous years. They should be evaluated inlarge-scale PVS trials. Variance component analysiswas conducted to estimate the precision of on-farmtrials through estimation of repeatability or broad-sense heritability. The broad-sense heritability ofgrain yield measured in the on-farm trials wasrelatively high, as in previous years. The resultsshowed that evaluation of genotypes at three sites,with three replicates per site, would give adequateprecision for the detection of cultivar differences(Table 5).

In Baby trials, three varieties were given toeach farmer to grow along with the local variety.Farmers’ perceptions for different varieties wererecorded. At Chichi, two farmers preferred RR 354-1, and one farmer each preferred RR 347-1, RR 363-737, Vandana and Kalinga III. In most of the trialssevere weed infestation was the reason given byfarmers for failure of the trials. In a few cases droughtdamaged the crop at key phenological stages.

PPBFarmers concentrated on fewer crosses compared to

breeders. They rejected crosses that did not produceplants that fit their criteria. In general, farmerspreferred breeding lines with tall stature, hightillering and long panicles. Plants generated fromcertain crosses (VHC 1253 x Sathi 34-36, N 22 x RR20-5, Annada x RR 151-3 and RR 139-1 x IR 57893-08) were preferred by both farmers and breeders.

The results indicate that, on average, farmerselections significantly outperformed breeders’selections, and that selection on-station was superiorto selection on-farm (Table 6). Farmers’ selectionsyielded almost twice as much as breeders’ selections(Table 7). Of the five highest-yielding lines, fourwere selected by farmers on the research station(Table 8). They were also found superior to checkvarieties.

Discussion

Farmer agreement on ranking varieties was highly

Table 4. Combined analysis of variance for cultivars evaluated under farmer management in three villages near Hazaribag, Jharkand. Wet season 2001.

Source MS F Pr>F Variance component

Cultivar 0.64 3.43 0.002 0.05

Cultivar x location

0.19 1.49 0.077 0.02

Residual 0.13 0.13

Table 5. Predicted broad-sense heritability (H) of grain yield for cultivars evaluated under farmer management in 3 villages near Hazaribag, Jharkand.

No. of sites No. of replicates H

1 1 0.23 1 3 0.44 3 1 0.50 3 3 0.70

Source F Pr. >F

Selector (farmer vs breeder) 16.6 0.0002 Selection environment 20.4 0.0001 (station vs farm) Selector x selection 5.5 0.0246 environment

Table 7. Effect1 of selection by farmers versus breeders and selection on-farm versus on-station on the grain yield of selected lines evaluated under farmer management at Khorahar, Jharkand, 2001.

Selection Selector environment Farmer Breeder Mean g/plot On-farm 243 162 202 On-station 566 264 415 Mean 405 213

1F-tests of main effects of selector and selectionenvironment were significant (Pr>F 0.0002 and 0.0001),respectively.

Table 6. F-tests for effects of selection by farmers versus breeders and selection on-farm versus on-station on the grain yield of selected lines evaluated under farmer management at Khorahar, Jharkand, 2001

significant, although there were differences inopinion. The relatively high agreement contradictedour initial assumption that farmers’ preference wouldvary because of the diversity in their socioeconomicbackground. This may be due to the limited numberof farmers involved in the project, or to the lowdiversity in wealth, caste, ethnicity etc. The highagreement among breeders indicated similarity inselection. The agreement between farmers andbreeders was good in most of the cases. Participationbrings little improvement when there is closeagreement between farmers and breeders.

The degree of agreement is highly influencedby the materials used for the selection. For example,in uplands of eastern India, the variety has to be tall,maturing within 100 days. Less variability within thetested genotypes perhaps influenced the farmers’ andbreeders’ preference for variety or traits. Therankings of farmers and breeders were correlatedwith yield only in few cases, as farmers also considerother factors in their decision. Though duration andstature had little association with yield, these areimportant characters in varietal choice. Again,because of the low variability in the tested material,these were not identified as important. Therefore,ranking of varieties must be combined with surveydata about farmers’ criteria of selection to get theactual information (Courtois et al., 2001). The results of three years of on-farm testing indicatedthat the varieties tended to rank similarly acrossfarms. Therefore, the genotype x environmentinteraction was not a significant factor influencingvarietal performance. This indicated that an on-station breeding programme can serve a purpose inthe targeted region. But the precision of on-farmtrials was as high as that of on-station trials, and threeyears of on-station trials failed to predict the on-farmperformance. This clearly indicates that PVS shouldbe integrated at an early stage of testing for betterprediction of varietal performance.

Farmers’ intentions of voluntarily testing the

new varieties in their own fields indicated thataccess to the new varieties or information is themajor constraint, as the local seed system is notvery effective. Out of five varieties they testedduring the 2000 wet season two did not performwell and were rejected by farmers. At the earlystage farmers may adopt more varieties, butsome will later be dropped, as the farmers testthem over a number of years. Once the varietyis adopted by farmers it can be spread fromfarmer to farmer, as the farmers from theneighbouring villages also took seeds from theadopted village (Joshi et al., 2001). In PPB, where both farmers and breeders made

selections from the segregating population,farmers selected progenies from fewer crosses.Farmers immediately rejected the crosses that did notproduce progenies according to their criteria.Knowing the background of the crosses, perhapsbreeders were expecting desirable segregants in thelater generation. The significant superiorityof farmers’ selection over breeders’ selection wasquite unexpected. It has established that farmers arealso able to identify high-yielding entries (Ceccarelliet al., 2001). Better performance of on-stationselection over on-farm selection indicated thatselection at the research station can be useful in thetarget region. This may be due to the low, butconsistent, yield of the research station. Of course,we need to be confirm this result on a larger scale inthe next season.

Conclusion

The PVS programme has given the breeder asystemic way to approach the farmers. Theinteraction with farmers and social scientistsinvolved in the project helped breeders and farmersdevelop a better understanding of the complexity ofthe problem. As the local seed system is non-functioning, access to new varieties or informationabout new technology seemed to be major constraint.Proper allocation of resources for varietal testing isneeded, as on-station evaluation alone cannot predictthe performance of the cultivar in the target region.

The PPB work has given more interestingresults. Although it is very early in the testing stage,it appears that farmers can manage a segregatingpopulation and were able to select superior progeniesfrom the segregating generation. This indicates it ispossible to expose the breeding materials at an earlystage to farmers. References

Courtois, B, Bartholome B, Chaudhary D, McLaren G,

Table 8. Means and selection history of the checks and five highest-yielding lines

Line Selection history Grain (g/plot)

RR356-77 Farmer-selected on station 831 RR356-72 Farmer-selected on station 825 RR356-74 Farmer-selected on station 769 RR356-71 Farmer-selected on station 569 RR356-51 Breeder-selected on station 519 Kalinga III 413 Vandana 413 Brown Gora 256 LSD.05 412

Misra CH, Mandal NP, Pandey S, Paris T, Piggin C,Prasad K, 2001. Comparing farmers’ and breeders’rankings in varietal selection for low-inputenvironments: A case study of rainfed rice ineastern India. Euphytica 122: 3, 537-550.

Witcombe JR. Joshi A, Sthapit BR,1996. Farmer participatory crop improvement. I. Varietalselection and breeding methods and their impact onbiodiversity. Experimental Agriculture 32: 4, 445-460.

Siegel S, 1956. Non-parametric Statistics for theBehavioral Sciences. Macgraw Hill, New York,312pp.

Ceccarelli S, Grando S, Bailey E, Amri A, El-Felah M, Nassif F, Rezgui S, Yahyaoul A, 2001. Farmerparticipation in barley breeding in Syria, Moroccoand Tunisia. Euphytica 122: 3, 521-536.

Joshi KD, Sthapit BR, Witcombe JR, 2001. How narrowly adapted are the products of decentralized breeding?The spread of rice varieties from a participatoryplant breeding programme in Nepal. Euphytica122: 3, 589-597.

Introduction

Between 1992 and 2000 more than 13,000 traditionalrice samples (representing over 3000 differentvarieties) were collected from Laos. More than 7,300of these samples were upland rice varieties. Usingthis collection, a system was set up to identifysuperior varieties for wide release. While germplasmfrom breeding programmes are also evaluated, thefocus is on the traditional varieties. The rationalebehind this is that the Lao prefer glutinous rice andthere are relatively few improved glutinous varieties.

The process of evaluation involves four steps.The Huay Khot Research Station in Luang Prabangprovince annually receives a new set of 300accessions from the Lao genebank, which areevaluated on-station in an observational nursery(OBN-1). From this, 20-30% of the varieties areselected for a second observation nursery (OBN-2).A limited number of varieties (8-16) are selectedfrom the OBN-2 for multi-location yield trials

(MLYT). Variety evaluation in these first three stagesis based primarily on yield and duration (preferencefor early maturing varieties). Varieties selected fromthese are then evaluated on-farm. Between 1992 and2000, over 2000 varieties were evaluated through thissystem and 27 varieties were selected for on-farmevaluation. For on-farm evaluation, varieties(typically 60 g/variety/farmer) were sent to districtofficers, who identified cooperating farmers. Thesuccess rate in this final stage was poor. Between1994 and 2000, 235 farmers were involved in on-farm evaluation. However, data were only collectedfrom 15 farmers (6%). As of 2001, there were novarieties that can be recommended for the uplands,due mainly to limited data on farmer preference andperformance under farmer management.

In initiating a PVS programme there were twobroad objectives. First, there was a need to acceleratethe evaluation of upland material. Continuing at thecurrent rate (about 300 varieties entering theevaluation process each year), it will take over 20

Participatory varietal selection: Lessons learned from the Lao uplandprogramme

K SONGYIKHANGSUTHOR1, GN ATLIN2, S PHENGCHANH1 AND B LINQUIST3

1 Northern Agriculture and Forestry Research Station, Luang Prabang, Lao PDR. 2 IRRI, DAPO Box 7777, Makati, Philippines.3 IRRI-Lao Project, PO Box 600, Luang Prabang, Lao PDR.

Abstract

Several thousand traditional upland rice varieties have been collected in Laos. Between 1992 and 2000,approximately 2000 of these were evaluated for grain yield and other agronomic characteristics inresearcher-managed trials in northern Laos, in an effort to identify high-yielding traditional varieties fordissemination to upland farmers. A few of the tested varieties were given to farmers to evaluate incollaboration with district extension officers before 2001, but data were returned in only 6% of the on-farm trials. In 2001, a new participatory varietal selection (PVS) programme was designed andimplemented to obtain information about farmer preference and on-farm performance of the uplandvarieties, and to increase the efficiency of the screening effort. In the first stage, on-farm PVS wasconducted with 32 farmers in five provinces. Farmers evaluated eight early or eight mid-durationvarieties selected by researchers in the earlier agronomic testing programme. Preference data weresuccessfully obtained from 84% of the farms and yield data from 63%, a 10-fold increase in the on-farmtrial success rate over the previous programme. Farmers strongly preferred early-maturing, large-seededvarieties with large panicles and strong stems, and disliked varieties with few tillers or tillers that ripenednon-uniformly. These criteria will be incorporated into the initial stages of varietal screening in thefuture. The correlation between variety means over farms for grain yield and preference rating was 0.82in the medium maturity trial and 0.54 in the early trial. Occasionally, high-yielding cultivars were notpreferred, but the lowest-yielding cultivars were never preferred, indicating that agronomic selection forgrain yield helps to select varieties farmers prefer. A farmer-preferred, early-maturing variety (Nok) thatsignificantly outyielded the local checks was identified. This and five other lines are being evaluated ina scaled-up PVS programme in 2002.

years to evaluate all the upland varieties. Identifyingand using farmer-preferred characteristics, it ispossible to rapidly screen varieties using passportdata (passport data are available for all rice samplesheld in the genebank) and in the OBN-1. Second, itwas recognized that changes needed to take place inthe final stage of evaluation (on-farm testing). Whilethe programme was able to collect agronomic datafrom the first three stages of the evaluation process,the final stage has produced no information on farmerpreference.

To achieve these objectives, activities in thisfirst year of PVS involved:• Training of researchers and district officers in

participatory research methods;• Evaluating varieties on-farm using

participatory research methods;• Identifying characteristics farmers prefer in

their rice varieties.

Materials and Methods

On-farm PVS was conducted with 32 farmers in fiveprovinces: Luang Prabang (10 farmers), Oudomsay(six farmers), Luang Namtha (four farmers),Sayaboury (six farmers) and Xieng Khouang (sixfarmers). Seed availability limited the number offarmers able to evaluate the 16 glutinous upland ricevarieties included in the trials. Twelve varieties wereselected from the 2001 multi-location yield trials andfour from previous trials (Vieng, Hom, Dam andMakthoua). These varieties were divided in twogroups depending on their duration (Table 1)

For the 2001 PVS trials, participating farmerswere given one set (early or medium) of eightvarieties (110 g per variety). This was enough to plantan area of 15 m2 of each variety. Farmers were givenassistance with plot layout and planting but managedthe plots according to their normal practices.

Between flowering and maturity, researchers

visited each farmer and conducted a PreferenceAnalysis (PA). In the PA the farmers were asked torank each variety using seeds, placing more seeds forvarieties they liked.They were then asked to indicatethe positive and negative characteristics of thevarieties. Finally, they were asked which varietiesthey would continue to evaluate next year on theirown. At harvest, the researchers and farmersharvested each plot for grain yield.

Analysis

Of the 32 farmers evaluating varieties, 27participated in the PA and we were able to collectyield data from 20. Yield data from four of the siteswhere the data was collected were not used in theanalysis because of poor plot layout.

Individual farmer variety rankings werestandardized across farmers by expressing the ratings(seed placed by each farmer on the board) as aproportion of the total number of seed applied to theboard by each farmer? The ratings were furtherstandardized to a mean of 0 and SD of 1 for eachfarmer. Least-squares means were estimated usingPROC MIXED METHOD=REML in SAS, withvarieties considered fixed, and farms random effects.Correlations between measured grain yield andfarmer preference ratings were estimated for thosefarms for which both types of data were collected.

Results and Discussion

General

Compared to previous years, the 2001 on-farm trialswere very successful. PA data were collected from84% of the farmers and yield data from 63%. Goodsuccess can be at least partially attributed to timelytraining (particularly in participatory researchmethodology) and continuous follow-up during the

course of the season. In some cases, the PA could

not be conducted due to early cropdamage by livestock. Yield datacould not be collected from somefarmers due to damage bylivestock, wild pigs or rats. Theseare factors that the researchers cando little about. However, at foursites, PA and yield data werecollected, but due to poor plotlayout, the data could not be usedin the analysis. Related to this,some of the trials were in fieldswhere the local variety (used as

Early Maturing Medium Maturing Accession no. Variety name Accession no. Variety name

LG-7347 Taa roon LG-7051 Do hom LG-6911 Deng said LG-6432 Mak hine

LG-7771 Kou yongke LG-7023 Da cheung LG-6499 Nok LG-7084 Paneur LG- Vieng LG-6655 Lebmeu LG- Hom LG-1724 Dam LG-6905 Sang LG-6593 Sangon LG-6501 Deng LG-2387 Makthoua

Table 1. Varieties and Accession numbers evaluated during the 2001 PVS trials

the control) had longer growth duration than thosebeing evaluated. This resulted in worse than normalpest damage. In future, there will need to be trainingin these areas, and simpler designs involving fewervarieties will be used

Preference Analysis: preferred plant characteristics

During the PA, farmers were asked to identifyreasons why they liked or did not like particularvarieties. These reasons were summarized from allthe PAs and are given in Figure 1. Large panicles, bigseed and strong stems were the most commonlyreferred to positive characteristics. In contrast,farmers did not like varieties that produced fewtillers, had short panicles or had panicles that maturednon-uniformly. All the varieties evaluated were ofearly and medium duration, therefore, preference forearly duration did not stand out prominently in thisPA. However, early maturaton was mentioned as apositive character and late maturation as a negativecharacter. Earlier research had identified farmerpreference for early- and medium-maturing varieties,and therefore, only such varieties were included inthe on-farm evaluation.

Using these criteria, varieties can now beeliminated early in the evaluation process, savingtime and limited resources. For example, passportdata (including data for some of these characteristics)exist on most varieties in the collection. Using thisdatabase some varieties could easily be eliminatedfrom the evaluation process. Also, in the OBN-1,

variety scoring should include these characteristics,to ensure that varieties selected for further testingwill not be rejected by farmers for reasons that arealready obvious.

Grain yields and farmer variety ranking

In the early trial the highest grain yield (GY) was2.14 t ha-1 (Nok). This was significantly higher thanthe average of the local early-maturing varietieswhich yielded 1.5 t ha-1. In the medium trial thehighest yield was 1.94 t ha-1 (Makthoua) which wassimilar to the average yield of the local varieties (1.82t ha-1).

An analysis of data for only those farmers fromwhom both yield and preference data were obtainedshowed significant differences in yield andpreference in the medium maturity group, but not inthe early group (yield, P=0.06). Yields and farmerpreference ratings (PR) (standardized) wereassociated. The correlation between variety meansover farms for GY and PR was 0.82 in the mediummaturity trial and 0.54 in the early trial. Occasionally,high-yielding cultivars were not preferred, but thelowest-yielding cultivars were never preferred. Thisindicates that agronomic selection for grain yieldhelps to select varieties farmers prefer.

A test of cultivar by province interaction wasconducted, using the within-province cultivar xfarmer interaction as an error term. There was nodetectable cultivar x province interaction for GY. A

Figure 1. Results from the preference analysis (PA) summarizing positive and negative plant characteristics most mentioned by farmers. Twenty-seven farmers participated in the PA. The number of times a characteristic is mentioned can be more than 27 because farmers were evaluating eight varieties.

0 10 20 30 40

Low t illering

Short panicle

Non-uniform panicle

Small grain

Small plant

Sterile spikelets

Poor emergence

Grows in some soil

Non-uniform ripening

Susept ible to pests

Few grain

Late durat ion

M any leaves

Non-uniform plant

Itchy seed

Number of t imes ment ioned by farmers

0 20 40 60 80

Large panicle

Large grain

Strong stem

M any seeds

High t illering

Early durat ion

Tall plant

Looks good

Uniform panicle

Grows indif ferent soils

M any panicles

Rapidemergence

Resistant tolodging

Weedcompetit iveness

Few leaves

Number of t imes ment ioned by farmers

significant interaction was observed for PR in theearly cultivar set; the cultivar Kouyonke was stronglypreferred in Luang Prabang, but had a very lowpreference score in Sayabouli. The reason for thisrank reversal is not clear. Certainly, one may expectinteractions between cultivars and different farmers,ethnic groups, or environments. However a largernumber of farmers must be surveyed for such ananalysis to be reliable.

What next?

Based on farmers’ preferences and grain yield, Taaroon, Nok, Do Hom, Makhine soung, Dam andMakthoua were selected for further evaluation in2002. Seed of these varieties are being multipliedduring the 2002 dry season. A Mother-Baby design(Annexe 1) will be employed in the 2002 analysis.

One Mother trial will be placed in each village,which will permit many farmers to evaluate varietiesside by side. Baby trials will consist of one farmerreceiving one variety to compare with their variety. Ineach village, we anticipate three or more farmersevaluating each variety (a total of 18 or more farmersper village). The number of villages depends on thesuccess of the seed multiplication. Yield data will becollected from the Mother trial. In the Baby trials,farmers will be asked to compare the two varieties, aswell as providing information collected on growingconditions, soil type, etc. This information can begathered during and after (provide some comparisonof grain quality) the growing season.

Farmers need to be involved earlier in theselection process. We have already indicated thatresults from the PA could be used to help screen out

varieties early in the evaluation process. However,farmers could also be involved in the MLYT. Thiswas done in 2001 but results are not yet available.Farmer involvement at this stage requires a differentmethodology.

Lessons learned

• Compared to earlier on-farm work the successof on-farm evaluation using participatorymethods was very good.

• In the future, more training emphasis needs toplaced on good plot layout if a number ofvarieties are to be evaluated in a singlelocation.

• Better care needs to be taken in working withfarmers, so that the duration of the varietiesbeing tested matches the duration of the localcheck variety.

• Too many varieties were evaluated with eachfarmer. This resulted in several problems: (1)farmers could not remember variety names, (2)the PA took too long and (3) it was easy forfarmers and researchers to get varieties mixedup. Based on our experiences, no more thanfive varieties should be evaluated with any onefarmer. When evaluating many varieties asimple question could be, "What varieties willyou plan to further evaluate next year?"Farmers were always quick to identify suchvarieties. These would naturally be the one thefarmer is most interested in.

• In future evaluations it would be good to gather additional information from the farmer related to the environment (such as, soil type, soil colour, drought occurrence [if so, early middle or late season]). This would allow us to test interactions between cultivar and environment.

• In future evaluationswe need to ask farmers to evaluate post-harvest

Table 2. Average grain yields and ranking in 2001. Varieties in boldthose selected for evaluation in 2002.

Early varieties Medium varieties Variety

Grain yield (t ha-1)

Standardized farmer preference*

Variety Grain yield (t ha-1)

Standardized farmer preference*

Nok

2.14 0.53 Makthoua

1.94 a

0.46

Hom 2.08 -0.44 Mak hine 1.75 ab 0.62 Taa roon 1.98 0.60 Do hom 1.72 ab 0.49 Kouyongke 1.88 0.02 Dam 1.55 bc 0.39 Vieng 1.86 -0.16 Peek 1.48 bc -0.70 Deng said 1.67 0.04 Da cheung 1.33 c -0.11 Deng 1.62 0.06 Sangon 1.31 c -0.41 Sang 1.34 -0.65 Lebmeu 1.19 c -0.76 LSD ns ns 0.39 0.83

* Farmer rank is standardized so that the mean is 0.0

characteristics of the rice, such as threshability, milling quality and taste. This was not done this year due to the limited amount of seed each farmer received.

• More seed (at least 0.5 kg and preferably 1 kg)needs to be provided to each farmer forevaluation. This will require that a system ofseed multiplication is in place. Seedmultiplication during the dry season iscurrently being evaluated.

Conclusion

Based on this year’s results, we identified how ourtwo broad objectives of the variety improvementprogram can be achieved. First, the evaluationprocess could be accelerated by taking into accountfarmer preferences early in the evaluation process(either by using passport data or in screening theOBN-1). Second, a system for on-farm evaluationhas been identified, which produced encouragingresults in this first year.

14Upland rice varietal access, test and multiplication (ATM)

RFD HONDRADE AND EG HONDRADE

University of Southern Mindanao, Kabacan, Cotabato, Philippines

Introduction

The Rice Varietal ATM Project is a research/extension strategy that involves seed testing,evaluation, selection and multiplication by farmersof upland rice varieties from farmer-managed trials.

Methodology

The Rice Varietal ATM project (Figure 1) followedthe Public Education Campaign (PEC) model. It wasguided by the following concepts: facilitate accessto information and technology through training,farmers test (farmers try the technologies on theirfields), and multiply (produce in larger scale ordisseminate to other farmers for wide use oftechnology).

Breeding network trialPromising breeding lines of upland rice, collectedfrom several countries, were tested in the breedingnetwork trial. From this trial, promising lines wereidentified based on the following criteria:

• Adaptation to local conditions• High yield• Resistance to drought and blast.

The presence of the breeding network in theirlocal municipality allowed farmers to observe thevarieties or lines. Some of these maintained stablehigh yields, which were 50% higher than their own Figure 1. Rice Varietal ATM methodology

Varietal assessment of farmers

Breeding Network (researcher-managed)

Advanced Yield Trial (AYT) (researcher-managed)

Farmers’ evaluations on rice lines in the AYT

I. Seed Testing Phase (small area varietal

evaluation)

II. Seed Multiplication Phase

(wide area varietal evaluation)

Upland rice farmers usually conduct their own on-farm evaluations when presented with a test materialof a new technology, such as seeds of a new upland rice variety. This strategy was conceived as the RiceVarietal ATM (Access, Test and Multiplication) Project. The project was designed to provide both thehardware and software of selected upland rice production technologies to resource-poor rice farmers.Farmers were provided with small quantities of new, high-yielding upland rice varieties, as well asother researcher-generated rice production technologies. If the farmers found the material promising, itwas multiplied for wider use. All cooperators underwent a one-day training prior to the start of afarmer-managed trial.

Results revealed that all farmer-cooperators who grew the new rice lines in upland areas succeededin their seed testing phase. Usually, successful Rice Varietal ATM trials were conducted by farmerswho were prepared to plant upland rice during the cropping season, who had the required land available,and who were willing to fund and manage the Rice Varietal ATM trial.

Most importantly, the new seed materials yielded better than the farmers' existing or local varieties.

Abstract

varieties, and had resistance to pests, diseases anddrought.

Advanced yield trialThe advanced yield trial (AYT) was composed ofpromising, high-yielding lines from the breedingnetwork trial. The advanced yield trial further testedthe performance of the promising upland rice linesfrom the breeding network trial. Yield, productive

tillers per linear meter, days to maturity, plant height,and pest and disease occurrence were all measuredin the AYT.

Farmers' evaluations on rice lines in the AYTFarmers were usually asked to evaluate the uplandrice lines in the AYT planted in the trial site duringfield days. The holding of field days apparentlybecame a necessary component of the trials by theUpland Rice Research Consortium (URRC)1 inArakan and allowed farmers to observe closely andevaluate of entries in the AYT using their owncriteria. Farmers were asked to write down the ratingof their most-preferred varieties and their least-preferred ones.They were also asked to identify thecriteria they commonly used in judging a preferredvariety. Results of farmers' evaluations were usuallyconsidered in succeeding trials or activities. Entriesthat performed well in the AYT that were highlyfavored by farmers were given to them for furthertesting in their own fields.

Farmer-managed trialsSeed Testing Phase: small area varietal evaluation.The varietal trials were carried out by farmers infarmer-managed trials. Each participating farmer wasassigned at least two promising varieties (5 kg ofseed each) and asked to grow them alongside thelocal variety in the same field under the samemanagement. The farmers were trained to use a two-page monitoring instrument with a calendar, withwhich they recorded their farm practices and generalobservations on the performance of the varieties.

The monitoring instrument was based on anearlier survey conducted on the criteria used byupland rice farmers in evaluating a variety. Farmer-to-farmer sharing of information on promising linesoccurred in this phase. Eight cultivars were testedon 22 farmers' fields. Among the promising andrecommended varieties tested were: B6144,PR23813-2-53, IR55423-01, PR23706-26, C-22,UPL Ri-5, PSB Rc-1 and PSB Rc-5.

Seed Multiplication Phase: wide area varietalevaluation. If farmer-cooperators are satisfied withthe results in the seed testing phase, they will likelymultiply the seeds for wide use. Usually, the producefrom the seed testing phase is used as material forseed multiplication in a wider area.

It is hoped that farmer-to-farmer informationdissemination will be strongest at this phase.

Rice Farming Practices at Pre-ATM Phase

Fertilization Practices. Most (68%) of the uplandrice farmer-cooperators used complete fertilizer ontheir farms. The rate and time of application rangedfrom 1 to 3 bags ha-1 at tillering stage (47%). Otherfarmers used urea (36%) and ammonium sulfate(23%) at the rate of 1 to 3 bags -1 at tillering andpanicle initiation stage.

Pest management practices. Most (91%) of themused pesticides to control rice pests and diseases.Very few (9%) claimed using biological controlagents such as plants with pesticidal properties. Theupland rice farmers employed manual weeding incontrolling weeds, usually by the family members.

Yield from using their own varieties. Besidestraditional varieties and promising upland selections(all the test materials introduced to the farmers fortesting, such as B6144, UPL Ri-5, PR 23706-26, etc.)the upland rice farmer-cooperators also plantedlowland rice varieties such as MASIPAG lines andIR36. In fact, the highest average yield obtained byrice farmers was from IR36, a lowland rice variety.The average yield of IR36 was 1.25 t ha-1 followedby Dinorado (1.00 t ha-1). The MASIPAG linesyielded an average of 0.78 t ha-1.

Seed testing phase (small area varietal evaluation)

Number of successful farmer-cooperators in the seedtesting phase. Only six of the 22 farmer-cooperatorswere able to plant the seed kits given them on theirupland farms (Table 1). Each farmer tested at leasttwo upland rice varieties/promising selections. Therest reported that they planted them in the lowlandareas.

1The URRC is a collaborative programme of IRRI andthe member country's research agency involved in uplandrice (in this case the Philippine Rice Research Institute)and the university, which co-implements the rice trials.

1 Each farmer received seed kits containing at least twovarieties2 Successful under upland conditions3 Planted in lowlands

Table 1. Number of successful trials pervariety per location.

Variety No. of farmers1

Upland2

Lowland3

B6144 10 5 5 UPL Ri-5 2 2 - PR23706-26 1 1 - PR23813-2-53 7 2 5 IR55423-01 6 2 4 C22 2 2 - PSB Rc-5 1 1 - PSB Rc-1 1 1 -

pre-released varieties, B6144 and IR55423-01, wereobserved to be susceptible to lodging. B6144 lodgedat milk grain stage. All the other entries showed ahigh degree of resistance to lodging.

The farmer-cooperators who tested B6144claimed that this line was high tillering but wassusceptible to narrow brown spot, leaf-folder andbacterial leaf blight.

IR55423-01 and PR23813-2-53 were observedto be susceptible to bacterial leaf blight and narrowbrown spot.

Conclusions

The primary objective of the project was to find outwhether selected researcher-generated technologies(e.g. new promising upland rice varieties) performwell on farmers' fields under their own management.The Rice Varietal ATM project has demonstrated itsvalue as a way by which farmers can evaluate theconventional technology package (e.g. variety) undertheir own field conditions.

This project also facilitates the provision offeedback from the farm to the experiment station.This information could be used in setting researchpriorities and goals to serve the varietal needs of thefarmer-clientele.

Results of the farmer-managed trials indicatedthat farmers can better make objective evaluationswhen experiments are done by them, since the projectprovides a forum for rice farmers to integrate theirindigenous technologies with introduced research-generated technologies.

Although one of the released varieties (UPL Ri-5) performed better than any of the pre-releasedvarieties, this is not yet conclusive since it is onlytested for only one cropping season.

Yield performance of some upland rice varieties/selections in farmer-cooperators' fields. UPL Ri-5gave the highest average yield (Table 3). This wasfollowed by PR23813-2-53 and IR55423-01. B6144PSB Rc-1, PSB Rc-5 and C-22, yielded less.

Farmers' preferences. Generally, most farmer-cooperators preferred the new seed materials to theirexisting variety. A good number of farmers preferredPR23813-2-53 because of the limited number ofrespondents for the other rice varieties.

Positive and negative qualities of new upland ricevarieties as observed by farmers. The releasedvariety UPL Ri-5 is high yielding but is susceptibleto bacterial leaf blight, leaf-folder and stemborer, asobserved by the farmer-cooperators (Table 4). Two

Table 2. Varieties/promising rice lines identified and tested by the Rice Varietal ATM Project.

Table 3. Yield (t ha-1) of upland rice varieties/promising rice lines and farmers' preferences.

Varieties/Promising rice lines identified by the RiceVarietal ATM Project and performance ofrecommended cultivars. Four varieties (C-22, UPLRi-5, PSB Rc-1 and PSB Rc-5) and four promisingrice varieties (B6144, PR23813-2-53, IR55423-01,and PR23706-26) were identified by the Rice VarietalATM project. (Table 2).

Cultivars Cultivars tested Year Released Pre-released B6144 PR23813-2-53 IR55423-01 PR23706-26 C-22 UPL Ri-5 PSB Rc-1 PSB Rc-5

! ! ! !

! ! ! !

- - - -

1972 1980 1990 1997

Farmers’ preference Variety tested

Farmers’ variety

Variety Number of cooperators

Average yield (t ha-1)

Yield range (t ha-1)

f % f %

C-22 2 2.28 2.14-2.42 1 50 1 50 PSB Rc-1 1 1.78 - - - 1 100 PSB Rc-5 1 1.63 - 1 100 - - UPL Ri-5 2 3.22 2.85-3.60 2 100 - - B6144 10 2.44 1.80-3.38 6 60 4 40 IR55423-01 6 2.86 2.63-3.10 5 83.3 1 16.7 PR23813-2-53 7 2.91 2.60-3.23 7 100 - - PR23706-26 1 2.23 - 1 100 - -

Table 4. Positive and negative qualities of upland rice varieties as observed by farmers.

Variety Positive qualities Negative qualities

UPL Ri-5

High yielding

Susceptible to bacterial leaf blight, leaf-folder and stemborer

PSB Rc-1 -

Susceptible to narrow brown spot

PSB Rc-5 - - C-22

-

Susceptible to sheath blight

B6144

High tillering

Lodges at milk grain stage Susceptible to narrow brown spot, leaf-folder and bacterial leaf blight

IR55423-01

High tillering

Susceptible to lodging Susceptible to bacterial leaf blight

PR23813-2-53

Susceptible to narrow brown spot and bacterial leaf blight

PR23706-26 Early maturing -

Introduction

In Uttar Pradesh rice is grown in 5.5 million ha, ofwhich 2.2. million ha (40%) of rice lands are rainfedlowland and 0.7 million ha are upland (13%). Theimpact of the Green Revolution is hardly visible insuch areas due to several factors:• Poor and inconsistent performance of on-

station developed varieties when grown underfarmers’ management;

• An inability to meet farmers’ preferences forquality;

• A lack of information about new varieties orpoor access to seeds of them, or bothinformation and seed are lacking.

The major production constraints in rainfed riceinclude an unpredictable combination of droughtsand floods, build-up of pests and diseases, poorphysical soil conditions and socioeconomiclimitations. Farmers are resource poor, which limitstheir risk-management strategies in rice cultivation.Therefore, the development of rice varieties suitablefor these fragile ecologies will make a positiveimpact on the millions of poor people whoselivelihoods depend on them.

Narendra Deva University of Agriculture andTechnology (NDUAT) in Faizabad, eastern India, incollaboration with IRRI, initiated a farmer

participatory breeding (FPB) programme in 1997. Inthis approach, the needs and priorities of farmerswere taken into account during the early evaluationof rice genotypes, both on the station and on farmers’fields.

This paper discusses the FPB approach and itsimpact on on-farm diversity conservation by theparticipating farmers.

Methodology

Three typical rainfed sites were chosen - Sariyawanand Mungeshpur in Faizabad, and Basalatpur inSiddharthnagar districts. Basalatpur represented afavourable lowland environment, while Mungeshpurand Sariyawan represented shallow butsubmergence- and drought-prone areas.

Three approaches were followed:

Participatory plant breeding (PPB)This included F4 generations of six different crossesinvolving diverse parents. These segregatingpopulations were jointly assessed by farmers andbreeders on-station as well as on farmers’ fields. Twobreeders and five farmers selected separately from F4to F6 during the 1997, 1998 and 1999 croppingseasons. Altogether, 50 populations were developedand tested in three replications on-station and in tworeplications on-farm. From these, six lines forFaizabad and five lines for Siddharthnagar conditions- identified by breeders, farmers or both - were

Addressing genetic improvement and on-farm diversity through farmerparticipatory breeding: A case study of rainfed rice in the Faizabad andSiddharthnagar districts, Eastern Uttar Pradesh, India

ON SINGH1, S SINGH1, VN SINGH1, JL DWIVEDI1, A SINGH1, HN SINGH1, RK SINGH2, TR PARIS3, GN ATLIN3 AND VP SINGH3

1. NDUAT, Kumarganj, Faizabad, India.2. IRRI-India Office, NASC Complex, Pusa, New Delhi, India.3. IRRI, DAPO, Box 7777, Metro Manila, Philippines.

AbstractResource-poor farmers in Uttar Pradesh, India, face a number of production constraints to rainfed ricegrowing, ranging from poor performance of traditional varieties and a lack of access to new ones, tounpredictable weather. A farmer participatory breeding programme begun in 1997 sought to addressthose problems by developing new rice genotypes.

During the four years of the study, several important points were established: Farmers couldvisually rate for yield with high accuracy, and their visual yield ratings were the best predictor of farmerpreference. Some cultivars produced very high yields under low-input management, and were highlypreferred by farmers. However, some high yielding varieties were not preferred, while some farmerspreferred varieties that did not give a high yield.

selected and compared in field trials. Observationswere recorded on traits such as yield.

Participatory varietal selection (PVS) This was initiated during the rainy season of 1997and continued in the 1998 rainy season. Fifteen to 20entries were tested on-station and on farmers’ fields.The entries included advanced breeding lines fromthe IRRI-led Shuttle Breeding project, releasedvarieties, and the most common local varieties. Twofarmers each from the village of Mungeshpur(drought prone), Sariyawan (drought andsubmergence prone), Faizabad and Basalatpur(submergence prone), and Siddharthnagar werechosen for PVS experiments. Final PVS trials wereconducted during 2001, both on-station and on-farmin Faizabad district.

Mother-Baby PVS programmeThis was initiated during the wet season of 2001, andconsisted of replicated researcher-managed mothertrials with 15 genotypes at Masodha rice researchstation, Faizabad. In these experiments recommendedfertilizer doses were used.

Nine farmers in Siddharthanagar districtconducted the baby trials in three villages. Each babytrial included five genotypes and were entirelymanaged by farmers. The alfa-lattice incomplete-block design was used for the layout and analysis ofthe baby trials. Simple rating scales (1-5) for differentcharacteristics were used for statistical analysis. Asimple form was used to elicit additional informationfrom the farmers on the traits and to get suggestionson how to improve the trials.

Important findings of PPB and PVS

PPB With few exceptions, the genotypes flowered earlieron farmers' fields than on-station. However, yields onfarmers’ fields were generally lower, as thegenotypes suffered prolonged waterlogging,particularly at Siddharthnagar. While a coarse graintype (IR 700803-43-NDR-2-9335) selected by thefarmers performed best both on-farm and on-stationat Faizabad, the farmers in Siddharthnagar preferredan early type (123 day duration) that showed amaximum yield of 5.1 t ha-1 on station fields and 3.2t ha-1 on farmers’ fields.

PVSTen farmers (five women and five men) visitedindividual plots and ranked the genotypes grown on

different farmers’ fields. Farmers’ selection criteriawere based on ecological needs, livelihood uses,gender and social and economic background.Mungeshpur and Sariyawan farmers preferredgenotypes with drought tolerance and short-mediummaturity. In contrast, Basalatpur farmers selectedslightly late-maturing genotypes with submergencetolerance and long, slender grains. In general, ricegrain yields observed on farms were lower than on-station. However, some of the cultivars producedvery high yields (> 4 t ha-1 under low-inputmanagement) and were highly preferred by farmers.

Farmers did select more than one variety (eightin this case), indicating that PVS/PPB can helpmaintain varietal diversity on-farm. Traits such asthreshability, submergence tolerance, and grain shapewere also preferred in this set of cultivars. Strawtraits, and pest and disease resistance ratings were notselected by farmers. Data for grain yield collectedfrom the on-farm trials were highly repeatable.

Mother-Baby PVS trialsSignificant differences were observed for grain yieldamong the genotypes tested in the mother trials.However, there was little difference in yield and itsattributes between the two treatments. Farmersperceived significant differences between thecultivars for grain yield, submergence tolerance,threshability (particularly women farmers), grainshape and overall acceptability.

Differences in measured grain yield werehighly significant. The reliability of yieldmeasurements was much higher than expected foron-farm trials. Two of the highest-yielding genotypesgave over 4 t ha-1 on-farm, and were highly preferredby farmers over their own varieties. However, onehigh-yielding genotype was not preferred, but noclear reason for this anomaly could be determinedfrom the questionnaire results.

On a single-plot basis, farmer visual yieldratings were highly correlated with measured yield (r= 0.84) (Table 1). Farmer yield rating and farmers'preference rating were also highly correlated (0.75).Breeder preference rating was highly correlated withmeasured yield (-0.81) and farmers’ yield rating(0.94).

On a genotype mean basis, farmers’ yieldrating and measured yields were highly correlated(0.90). Farmers had a strong preference for cultivarsthey rated highly for yield (r = 0.81) (Table 2).

On-farm diffusion and varietal diversity

While PVS trials continued, farmers picked up a few

lines that they found promising and grew them onlarger plots. For example, three genotypes, PVS1, 7and 10, were grown in two villages of Faizabaddistrict on 23 farmers’ fields in 2000, and 36 fields in2001. Similarly, six genotypes, PVS 1, 2, 3, 7, 10 and13 were cultivated on 45 farmers’ fields in 2000, and48 fields in 2001, covering seven villages inSiddharthnagar district (Table 3). This clearlyindicates that FPB not only enriches the existingdiversity of rice varieties grown under rainfedconditions, but also allows farmers to grow improvedvarieties they have chosen themselves.

Conclusions

• Farmers could visually rate for yield with highaccuracy.

• Farmer visual yield ratings were the bestpredictor of farmer preference.

• Some preferred varieties were not high yielding (e.g. NDRSB9830102, which gave2.5-3.6 t ha-1), and some high yielders, such asNDR 40032 (3.2-4.1 t ha-1) and NDRSB9730020 (4.2-5.8 t ha-1) were not preferred.

• Some cultivars produced very high yields (> 4 t ha-1) under low-input management, and werehighly preferred by farmers.

• Over the years of the study, farmers did not continue all the lines that they had selected inprevious years. This was because they tookinto account additional information from

Grain yield

Farmer preference

Breeder preference

Farmer yield

Farmer submergence tolerance

Farmer threshability

Grain yield -0.59 -0.81 -0.84 -0.53 -0.69 -0.46 Farmer preference

0.75 0.74 0.48 0.61 0.59

Breeder preference

0.94 0.49 0.69 0.6

Farmer yield 0.49 0.74 0.62 Farmer submergence tolerance

0.34 0.45

Farmer threshability

0.33

Table 1. Correlations among plot values for measurements and ratings taken by farmers andbreeders in baby trials in three villages in Siddarthnagar (rainy season, 2001).

Table 2. Correlations among cultivar means for measurements and ratings taken by farmers andbreeders in baby trials in three villages in Siddharthnagar (Kharif 2001).

Grain yield

Farmer preference

Breeder preference

Farmer yield

Farmer submergence tolerance

Farmer threshability

Grain yield -0.67 -0.87 -0.90 -0.61 -0.89 -0.57 Farmer preference

0.80 0.81 0.69 0.77 0.75

Breeder preference

0.96 0.74 0.83 0.74

Farmer yield 0.70 0.84 0.75 Farmer submergence tolerance

0.70 0.75

Farmer threshability

0.53

various trials and later observations (such assensory evaluation, milling recovery, anddrought incidence), in order to finally decidewhether to select or reject a variety.

Years and location

Designation

No. of farmers

Total area (ha) Yield range (t ha-1)

2000 Faizabad

NDR-40032 NDRSB-9830102 NDRSB-9730020

2

18 3

0.16 2.17 1.50

3.5-5.5 2.8-3.7 3.5-4.8

Siddharthnagar NDRSB-9730004 NDRSB-96005 NDRSB-40032 NDRSB-9730020* NDRSB-96012 Kamini Sugandha

9 6 8

12 3 6 1

1.00 1.15 0.80 2.85 0.38 0.85 0.40

1.8-2.5 2.4-4.2 3.5-5.2 3.7-5.6 2.5-3.6 1.6-2.5 1.5-2.2

2001 Faizabad

NDR-40032 NDRSB-9830102 NDRSB-9730020

4

23 9

0.56 3.10 1.18

3.2-4.1 2.5-3.6 4.2-5.8

Siddharthanagar NDRSB-96005 NDRSB-9730020 NDR-40032 Kamini NDRSB-9830102

5 19 8

11 5

0.90 4.60 1.20 1.70 0.95

3.4-4.7 3.9-5.6 3.6-4.8 1.5-2.5 2.6-3.5

*Spread in Nepal by relatives of cooperator farmer but not monitored

Table 3. On-farm diffusion of PVS lines, 1998-2001

17Participatory selection on upland rice in SumatraSUWARNO1, B KUSTIANTO1, WS ARJASA2, AND G ATLIN3

1. Research Institute for Rice, Sukamandi, Subang, Indonesia.2. Tamanbogo Experiment Station for Food Crops, Tamanbogo, Central Lampung, Indonesia.3. International Rice Research Institute, IRRI DAPO 7777, Makati, Philippines.

Participatory varietal selection is a method to select breeding material by taking into account farmers’preferences. A mother-baby trial was conducted in the upland rice cultivation area in Central Lampung,Indonesia, to evaluate promising lines and improved varieties. The experiment was planted late andthere was heavy rainfall during the growing season so there was heavy infestation by blast. Limboto, anewly released variety, was the highest yielder and was rated highly by almost all of the participatingfarmers. Most of the tested material rated low, in comparison with the local variety, were blast-susceptiblebreeding lines. The promising lines rated highly by most farmers and that produced a relatively highyield were TB47H-MR-10, B9071F-TB-7, and BIO511B-TB-61-2-4. Continuous deployment of breedinglines through mother-baby trials and the adoption by farmers of ones they prefer will increase thegenetic variability deployed in farmers’ fields and might improve blast resistance. It would help maintainthe current diversity - more than 10 different traditional varieties were observed to be grown in the area- which might result in a high genetic diversity for blast resistance.

Abstract

Introduction

Indonesia has about 1.5 million ha of upland rice,about 12 % of the country’s total rice growing area.Generally, upland rice in Indonesia is cultivated bypoor farmers on problem soils. The national averageyield of upland rice in 2000 was 1.9 t ha-1, comparedto that of lowland rice, 4.6 t ha-1.

In the islands of Sumatra, Kalimantan, and IranJaya, where most of the upland rice-growing area isfound, the major constraints are diseases includingblast, brown spot, and sheath blight, and soil problemsincluding acidity, aluminium toxicity, and phosphorusdeficiency. In the semi-arid zones of East Java, Westand East Nusa Tenggara, drought is the majorconstraint. In intercropping systems, which arewidespread in these areas, shade is also a constraint.

Some improved varieties of upland rice havebeen released; however, most farmers still cultivatetraditional varieties. The severe blast pressure in high-rainfall areas of Sumatra and Kalimantan is the singlemost severe constraint to the adoption of improvedvarieties. The blast resistance of improved varietiesoften breaks down after being cultivated for only twoor three planting seasons. Another reason whyimproved varieties are not adopted by farmers couldbe that their characteristics do not meet theirpreferences. Participatory selection, in which farmersare encouraged to select breeding lines based on theirown preferences in comparison with their varieties,was designed to overcome this second problem. An

experiment on selection of promising breeding linessuitable for an upland cultivated area in Sumatrahas been conducted using mother-baby trials.

Materials and methods

The experiment was conducted in Rumbia andSeputih Banyak villages of Central LampungProvince, Sumatra Island. The trial included eightpromising lines: B9071F-TB-7, B8503E-TB-19-13-3, TB177E-TB-28-B-3, TB47H-MR-10, IR30176B-1-B-2-2-MR-2, BIO528B-TB-12-1-1, BIO511B-TB-61-2-4, BIO530A-5-14-2-2-8, and threeimproved varieties of upland rice; Cirata, WayRarem, and Limboto (L).

The mother trial was laid out as a randomisedcomplete block design with four replications andan individual plot size of 5 x 4 m. Cultivationmethods followed those recommended by theextension services, including a 30 x 15 cm spacingand 150-125-100 kg ha-1 of Urea-SP36-KClfertilizer application. For the baby trial, 0.5 kg seedof each entry was given to the selected farmers. Toreduce the resource requirement and also reduce thecomplexity of the trials for farmers, six entries weredistributed to each farmer from the total of 18.Farmers were asked to cultivate the materials usingtheir own methods. The seeds were given along witha questionnaire with the characters to be scored incomparison to their variety. These charactersincluded vigour, competitiveness with weeds,

reaction to brown spot, reaction to blast, lodging,maturity, yield, grain shape, eating quality,threshability, overall farmer opinion and the reason.Scores given to the characters were:, 1 very good; 2good; 3 medium (equal to farmer variety); 4 bad; 5very bad.

Results and discussion

Limboto, a newly released variety, and B9071F-TB-7 appeared to have the best phenotypic acceptabilitywith an average score of 3.5 (Table 1). All thepromising breeding lines were selected for goodagronomic characteristics and resistance to blast,whereas Cirata and Way Rarem are high yieldingimproved varieties. In the experiment, however,these two improved varieties could not yield up totheir potential mainly due to a heavy infestation ofblast. Almost all plants of the line TB177E-TB-28-B-3 were killed by the disease and no agronomicdata could be recorded. This illustrates thebreakdown of the resistance and the variability ofthe pathogen. Heavy rainfall during the floweringstage caused high neck blast pressure.

Most of the lines were susceptible and had manyunfilled grains. Limboto had the fewest unfilledgrains and, along with TB47H-MR-10 andIR30176B-1-B-2-2-MR-2, produced the most filledgrains. Limboto and TB47H-MR-10 yielded themost (2.5 t ha-1) in the mother trial and B9071F-TB-7 was not significantly lower than this (Table2).

No variability was found among entries forcompetitiveness with weeds, lodging, andthreshability, and this was reflected in the farmers’scores, so none of these data are presented here.All of the farmers scored the vigour of the newreleased variety Limboto as good and most of themalso scored lines B9071F-TB-7 and TB47H-MR-10 as having good vigour in the baby trials. Thescores made by farmers for vigour and for theiroverall opinion on the entries were closelycorrelated (r = 0.91). However, the correlationcoefficients of the scores for vigour with yield (r =-0.77) and overall opinion with yield ( r= -0.66) werelower.

Yield is usually the most important characterdetermining farmers’ preference for a rice variety,

Line/variety Grains/panicle

Height (cm)

Tiller no.

Maturity (days)

Pheno. Accep.1 Filled Unfilled

B9071F-TB-7

85 14 112 3.5 82* 79*

B85036E-TB-19-13-3

114* 14 105 4.5 75* 66*

TB177E-TB-28-B-3

Killed by leaf blast

TB47H-MR-10

89 14 114 4.0 92 62*

IR30176B-1-B-2-2 MR-2

87 13 108 5.0 84 41

BIO528B-TB-12-1-1

89 13 112 4.5 58* 56*

BIO511B-TB-61-2-4

64* 14 115 6.0* 71* 46

BIO530A-5-14-2-2-8

76* 14 114 6.0* 65* 54*

Cirata

79* 14 109 5.5* 70* 37

Way Rarem

85 14 116 5.0 65* 50

Limboto

88 14 114 3.5 89 38

Table 1. Performance of some breeding lines and varieties of upland rice in the mother trialconducted in Lampung, West Sumatra, 2001-2002.

1Mean score for phenotypic acceptability 1-9 = very good - very bad; *significantly differs from the check varietyLimboto based on LSD 0.05.

Average score Average yield (t ha-1)

Line/variety Farmer no.

Vigour Opinion Baby Mother

B9071F-TB-7

9 2.4 3.0 2.80* 2.40

B85036E-TB-19-13-3

9 3.8 4.0 2.71* 2.23*

TB177E-TB-28-B-3

9 5.0 4.0 0.09* 0.00*

TB47H-MR-10

10 2.2 2.4 2.86* 2.50

IR30176B-1-B-2-2-MR-2

9 3.0 3.1 2.68* 2.26*

BIO528B-TB-12-1-1

9 3.3 3.1 2.97* 1.09*

BIO511B-TB-61-2-4

9 3.0 2.9 3.24 2.19*

BIO530A-5-14-2-2-8

9 3.6 3.3 3.10 1.29*

Cirata

10 3.6 3.3 2.63* 1.64*

Way Rarem

10 3.1 2.9 2.97* 1.79*

Limboto

15 2.0 2.3 3.49 2.50

Table 2. Mean vigour scores, overall scores of farmer opinion, and yield of breeding lines andvarieties of upland rice in the Mother-Baby trial conducted in Lampung, West Sumatra,2001/2002.

but in this experiment the farmers had been askedto score the entries before harvesting. The trial washeavily infected by neck blast, which made itdifficult for the farmers to predict the yield. Thismight account for a lower correlation coefficientbetween yield and overall opinion than betweenvigour and overall opinion.

Moreover, the notes made by farmers (Table 2)indicated that other characters influencedpreferences such as blast resistance, panicle size,grain shape, and grain size.

Most of the farmers scored the tested materialas bad in comparison with their variety because ofsusceptibility to blast. This indicated that blast isthe most important constraint on upland ricecultivation in the area. Some of the farmers' varietieswere also attacked by the disease, but most were

not. There were at least 15 different local varietiescultivated by farmers in the region where the babytrial was conducted. The high genetic variability ofthe local varieties is a good defence again thepathogen. Introducing many different breeding linesto substitute for the diverse array of local varietiescould maintain this high genetic variability, andmight be a good approach for increasing rice yieldwhile maintaining blast resistance. Since breedinglines are continuously developed in the breedingprogramme, similar trials could be conducted witha new set of breeding lines in the following growingseasons. The continuous deployment of breedinglines might develop genetic variability and blastresistance.

Table 3. Number of scores for farmer opinion given to each lines and variety and the correspondingnotes made by the farmers.

Number of score Mean Note Line/variety 1 2 3 4 5

B9071F-TB-7 - 4 1 4 - 3.0 Round grain; not uniform flowering

B85036E-TB-19-13-3

- - - 9 - 4.0 Less vigorous. Susceptible to blast

TB177E-TB-28-B-3

- - - 9 - 4.0 Susceptible to blast

TB47H-MR-10

- 7 - 2 - 2.4 -

IR30176B-1-B-2-2-MR-2

- 4 - 5 - 3.1 Small panicle;

BIO528B-TB-12-1-1

- 4 - 5 - 3.1 Susceptible to neck blast

BIO511B-TB-61-2-4

- 4 2 3 - 2.9 Susceptible to neck blast

BIO530A-5-14-2-2-8

- 2 2 5 - 3.3 Susceptible to neck blast

Cirata

- 2 3 5 - 3.3 Susceptible to neck blast

Way Rarem

- 4 3 3 - 2.9 Susceptible to neck blast

Limboto - 13 2 - 2.3 Round grain

Introduction

Breeders in Bangladesh are developing new ricevarieties through on-station trials and without theinvolvement of farmers. The bottlenecks of thisapproach could be realized by observing thepopularity of a variety produced by this system in oneregion but not in others. Moreover, the value of thesevarieties for marginal, risk-prone and stressedenvironments is limited. Hence, evaluation ofgenotypes in the real, target environment and theparticipation of the end users in the process of varietydevelopment have been emphasized. PVS wasinitiated to identify high-yielding rice varieties forthe coastal saline areas of Bangladesh.

Materials and Methods

Fourteen advanced breeding lines of BRRI (BR) andIRRI (IR) origin, including two BR varieties knownto be salt tolerant at 6-8 dS/m, were selected for thePVS trial. The trial was conducted at six sites in thecoastal areas of Feni (Sonagazi), Noakhali, Laxmipur(Ramgati), Kaliganj, Shaymnagor and Satkhira(Figure 1). One popular farmers’ variety was used ineach site in this trial. There was a single-replicatetrial at each site with an individual plot size of 2 m x8 m. Fertilizers were applied (80:60:40 kg NPK ha-1)with a split application of N at 15, 40 and 60 daysafter transplanting. Insecticide was applied twice to

protect the crops from stem borer infestation. Zincsulphate was applied at 10 kg ha-1 at the tilleringstage of the crop. The farmer (owner of the trial plots)was involved for crop husbandry including weedingthe plots twice.

PVS was organized at four sites: Sonagazi,Ramgati, Noakhali and Satkhira. At crop maturity,farmers from three or four villages at each site wereinvited in cooperation of the Department ofAgricultural Extension (DAE). Ten grassroot-levelextension personnel (block supervisors) around eachsite, local NGOs, leaders, administrators and theDeputy Director of Agricultural Extension (DDAE)of the respective districts attended the function.

The 30-40 PVS farmers divided into threegroups of 10-15 (Figure 2). Scientific personnelguided each group. Each farmer was given a PVSsheet (Figure 3). The purpose of the trial was

Participatory variety selection (PVS): an initiation in Bangladesh

MA SALAM, MR ISLAM, MAR BHUIYAN, AA KHAN AND MAB FARUQUEI

Plant Breeding Division, Bangladesh Rice Research institute (BRRI), Gazipur-1701, Bangladesh.

AbstractBreeders are habituated to develop rice varieties using the centralized breeding approach in Bangladesh.Recommended varieties produced by this method are popular in favourable rice production systems butlocal (farmer-preferred) varieties still dominate in less-favourable environments. Efforts were initiatedto evaluate 14 salt-tolerant genotypes at four sites in the coastal saline area through a farmerparticipatory approach in the rainy season of 2001. Researcher-managed trials were conducted andconsultative farmer participation was employed at crop maturity. Thirty to 40 farmers took part in thePVS (participatory varietal selection) activity at each site and selected 10 genotypes that included twovarieties recommended for coastal wetlands. The farmers reported that, for the first time, they had theopportunity to select varieties according to their own judgement from a basket of new genotypes madeavailable to them. All the PVS farmers agreed upon to grow their selected varieties under their ownmanagement in the next season. Deputy directors from the agricultural extension service attended thePVS activities at the four sites and noted that it was a bottom-up approach. The direct involvement offarmers to choose variety(s) for their needs would be an excellent method for variety selection. Theyrecommended supporting this effort since it effectively brought farmers, NGOs and extension agentstogether.

N

Sonagazi Ramgati Satkhira

Kaliganj

Shayamnagar Noakhali

Figure 1. Coastal districts of Bangladesh where the PVS sites were situated.

explained to create enthusiasm among theparticipating farmers. The PVS sheets on which thevarietal preferences were recorded were collectedfrom the farmers and a discussion was held on theirreactions to the PVS exercise.

Results and Discussion

It was observed that none of the farmers at any sitewas ranked excellent, since all the participating

farmers selected one-four genotypes/varieties at eachsite (Table 1). It was agreed that 1 kg of seed of theselected PVS entries would be supplied to therespective farmers. All the farmers will grow theirselections under their own management in the nextrainy season.

The farmers appreciated having a basket ofnew varieties/ genotypes available and theopportunity to choose among them using their ownjudgement. They reported that the PVS genotypes:• Had lodging tolerance;

• Had acceptable grain shape;

• Were suitable for local conditions;

• Had a higher grain yield than local varieties.

The Deputy Director of DAE reported that:• DAE’s extension policy was top-down and

therefore adoption of new technologies waslow.

• Farmers’ need-based technology would be developed by PVS and thus rapiddissemination of the technology would beexpected.

• Farmers would have the ready-made opportunity of having seeds of improvedvarieties for increased production.

• Farmers, NGOs and extension agents should sincerely support this effort.

Figure 2. Farmer groups in action for PVS at Benarpota (Satkhira) on 6 November, 2001.

Figure 3. Farmer groups are in action for PVS along with PVS sheets at Char Jabbar (Noakhali) on 20 November, 2001

Number of farmers Variety Sonagazi Noakhali Laxmipur Satkhira Total BR5778-156-1-3-HR1 5 2 -- 16 23 BR5778-156-1-3-HR14 15 8 -- 28 51 BR5778-156-1-3-HR15 16 4 -- 16 36 IR66401-2B-14-1-1 3 -- 1 9 13 BR5999-82-3-2-HR1 18 3 14 3 38 BR5999-82-3-2-HR10 10 3 13 5 31 BR5999-82-3-2-HR16 6 13 28 8 55 BR5333-34-4-6 14 -- -- -- 14 BRRI Dhan40 16 19 19 9 63 BRRI Dhan41 28 18 28 12 86 Total 131 70 103 106 410

Table 1. Advanced lines and varieties selected by PVS farmers at four sites in the coastal districts of Bangladesh

Introduction

Cambodia has very diverse rice environments. About91% of its 2 million ha rice area are cultivated in thewet season and consists of 86% rainfed lowland, 4%deepwater, and 1% upland. Most of the wet-seasonrice areas are unfavorable because of several abioticand biotic stresses.

The Cambodia-IRRI-Australia Project (CIAP)started in the late 1980s with the goal of increasingCambodia's rice productivity. In August 1999, theCambodian Agricultural Research and DevelopmentInstitute (CARDI) was created. In 2000, it started tofunction and assume the leadership for rice researchin the country.

Methods

A comprehensive variety testing program was startedby CIAP in 1989. It involves the following sequentialstages of on-station variety evaluation - observationalyield trial, preliminary yield trial, and advanced yieldtrial (AYT). This is followed by participatory varietyselection (PVS), which is conducted infarmers’fields. Varieties are released based on theresults of AYT or the combined results of AYT andon-farm trials. PVS involves testing 1-3 varieties thatcan be either released or promising.

There are two components of PVS. The firstone is known as the on-farm variety adaptive trial(OFAT-V). Its objectives are:

1. To monitor the performance of released and promising varieties in farmers’ fields.

2. To allow farmers to select the best varieties under their own management practices.

3. To obtain feedback from farmers regarding the test entries.

4. To serve as a source of good seed for farmers.

5. To provide a demonstration field for released and promising varieties.

The trial has the farmer’s best varieties as localchecks. The amount of seed for each test entry is 1-1.5 kg. The suggested plot size is 100 m2 (20 m x 5m or 10 m x10 m). There is a common field layout forall sites. Sites should be accessible so that they can beseen by many farmers for monitoring purposes.Farmers use their own resources and managementpractices. The following data are collected: culturalpractices employed, yield, farmer’s preference andreason, and problems encountered.

The other type of on-farm trial is called thevariety-nutrient-pest monitoring trial (OFAT-VNP).Its objectives are as follows:

1. To monitor the performance of recommended varieties under different levels of fertilizer and soil type.

2. To determine the incidence of insect pests and diseases under different levels of fertilizer andsoil type.

3. To provide a demonstration field for CIAP’s current technology.

4. To serve as a basis for improving technology recommendation.

Two types of factorial treatment design areused. Design I involves a combination of twovarieties (recommended, farmer’s variety) and threefertilizer treatments (0, 1, and 1.5 times therecommended fertilizer rate). Design II has threevarieties (two recommended varieties and a farmer’svariety) and two fertilizer treatments (0 andrecommended fertilizer rate). For both designs, a trialis composed of six plots. Trials are conducted indifferent soil types. More data are collected in OFAT-VNP than in OFAT-V. Data collected were grain

Large-scale participatory varietal selection in Cambodia

EL JAVIER1 AND M SAROM2

1. INGER Coordinator, International Rice Research Institute, DAPO Box 7777, Metro Manila, Philippines.2. Director, Cambodian Agricultural Research and Development Institute (CARDI), PO Box 1, Phnom Pehn,

Cambodia. AbstractParticipatory varietal selection (PVS) has contributed substantially to the spread of new varieties of ricein Cambodia. In over 5,000 trials between 1990 and 2000, 91% were successful. Some trials failedbecause of drought, flood or other factors.

yield, straw yield, harvest index, DTF, plant height,tiller number panicle number, and cultural practicesemployed.

CIAP/CARDI has GO and NGO cooperatorsthat liaise with the participating farmers. Cooperatorsreceive protocol for trials; select farmer-participants;assist in site selection, field layout and datacollection; and monitor trials (seeding, transplanting,reproductive stage, harvesting). CIAP/CARDI staffalso monitor the trials. For OFAT-V, each cooperatorreceives $35-50 per trial. Cooperators for OFAT-VNP receive more than this per trial. NGOs fund theOFAT that they conduct.

Results

A total of 5,139 OFAT-V and OFAT-VNP wereconducted from 1990 to 2000. Most of them testedearly-duration varieties, and trials for upland ricevarieties were the least numerous (Table 1). Thepercentage success ranged from 18% (upland) to91% (aromatic group). From 1997-1999, 357 OFAT-VNP were conducted. The percentage success was77%. Some trials failed because of drought, flood,submergence, rats, birds, crabs, rabbits, securityproblems, poor layout and wrong choice of testentries for certain environments.

Three released early-duration varieties (IR66,IR72 and Kru) were tested in irrigated to partiallyirrigated fields of farmers from 1990 to 1993. The

average yields of released varieties were substantiallyhigher than those of the variable local checks (Table2). IR 66 had the highest yield and the was mostpreferred variety by farmers. The farmers’ ownvarieties were least preferred.

In the following years, IR66 was replaced by anewly released variety IR Kesar. The average yieldsof Kru and IR Kesar were higher than that of farmers’varieties from 1993 to 1995 (Table 3). Both

recommended varieties had a higher preference thanthe local check varieties. However, more farmersgave first preference to their own varieties comparedto the previous year, because they used IR66 as theirlocal check variety, reflecting the adoption ofreleased varieties.

In 1996, a very high yielding line (IR84725) instation trials was tested against IR Kesar and Kru infarmers’ fields (Table 4). IR84725 showed yield

superiority over released and local check varieties. Ithad the highest preference rating. However, it wasdropped for further testing due to poor grain qualities.There were no progenies from this line with good

Table 1. Number of PVS trials conducted and their success, 1990-2000

Trial Number of trials

% success

OFAT-V Early duration 1952 78 Medium duration 824 61 Medium duration 938 87 Late duration 990 83 Deepwater 326 57 Upland 76 18 Aromatic 33 91 OFAT-VNP 357 77

Table 2. Yield of recommended early duration varieties in 232 wet season and 200 dry season on-farm trials and farmers’ preferences, 1990-1993

Entry Yield (t ha-1)

Farmers’ preference (no.)

1st 2nd

IR66 3.6 220 87 IR72 3.4 72 157 Kru 3.4 99 144 Local checks

3.6 19 16

Table 3. Yield of recommended early duration varieties in 235 wet season and 230 dry season on-farm trials and farmers’ preferences, 1993-1995

Entry Yield (t ha-1)

Farmers’ preference (no.)

1st 2nd

Kru 3.2 136 148 IR Kesar 3.3 204 108 Local check

3.0 67 135

Table 4. Yield of recommended early duration test entries in 71 wet season and 381 dry season on-farm trials and farmers’ preferences, 1996-1997

Entry Yield (t ha-1)

Farmers’ preference (no.)

1st 2nd

IR Kesar 3.5 105 167 Kru 3.4 44 112 IR84725 3.7 187 96 Local check

3.4 69 68

grain characteristics. In other on-farm trials, somepromising lines performing well were discarded laterbecause of grain quality problems. From then on,promising lines should have passed a grain qualitytest before they are tested in farmers’ fields .

In 1997, on-farm VNP trials were conducted invarious soil types using released pureline selectionsfrom Cambodian traditional varieties. In Bakan soiltype, released varieties CAR 1, CAR 2 and CAR 3and the local checks had comparable mean yields(Table 5) without fertilizer. However, releasedvarieties gave higher yield than the local varietiesunder recommend nutrient management. In a verypoor soil (Table 5), local check varieties were betterthan released varieties regardless of nutrientmanagement employed. These information help plantbreeders in defining more clearly their targetenvironments.

Conclusions

In general, on-farm trials have met their objectives.They have contributed substantially in the spread ofnew varieties in Cambodia. NGOs have playedsignificant role in the success of PVS.

Table 5. Yield of entries (t ha-1) in on-farm VNP trial conducted in two soil types, 1997

Bakhan soil1 Koktrap soil2

Zero fertilizer

Recommended fertilizer

Zero fertilizer

Recommended fertilizer

CAR 1 1.4 2.0 0.4 0.9 Local check 1.2 1.8 0.9 1.3 CAR 2 2.0 2.7 0.0 0.9 Local check 2.1 2.5 0.7 1.1 CAR 3 1.8 2.5 0.0 0.6 Local check 1.8 2.3 0.8 1.2

1CAR 1, CAR 2 and CAR 3 were evaluated in 2, 6, and 4 trials, respectively2CAR 1, CAR 2 and CAR 3 were evaluated in 2, 4, and 4 trials, respectively

Introduction

We explored means of analysing and interpreting datafrom simple on-farm experiments, and have foundmany existing statistical methods that are directlyrelevant to participatory varietal trials. This paperaims at demonstrating the application of variousstatistical methods for analysing quantitative andqualitative data from PVS trials.

Creating spreadsheets

PVS data are often poorly documented. A researchershould be able to prepare abstracts of information inmany ways and subject data to statistical analyseswithout much extra effort. To do so, data from

Mother and Baby trials should be entered inspreadsheets so that they can be analysed without re-entry of data. This requires data to be organised incolumns where each column is a single descriptor, ortrait, so that a two-way ANOVA is possible instatistical programmes such as Minitab (Table 1).

Design of PVS trials and type of data

A simplified Mother-Baby trial design (Snapp, 1999;J de Meyer and M Bänziger [pers. comm.]) that doesnot require complex layout plans is used for PVStrials (Annexe 1).

1. Mother trials (single-replicate all varieties)

An introduction to data management and analysis for participatory varietalselection trials

DS VIRK AND JR WITCOMBE

Centre for Arid Zone Studies, University of Wales, Bangor, Gwynedd, LL57 2UW, UK.

AbstractData from on-farm varietal trials are usually not considered as amenable to statistical analysis astreatment of the data from research station experiments. This is because variability is believed to behigher in farmers’ fields; and the data from simple experiments are considered to be inappropriate forstatistical analyses. However, in this paper, we show, with examples, how data from simple participatoryvarietal selection (PVS) experiments using Mother-Baby designs can be usefully analysed using existingstatistical methods. The data collected on quantitative traits and farmers’ perceptions demand theapplication of parametric and non-parametric methods.

Farmer

Village

Cluster

Var.

Area (m2)

Plant ht (cm)

Grain yd (kg/plot)

Straw yd (kg/plot)

Grain yd (kg/ha)

Straw yd (kg/ha)

1 1 1 1 50 95 8.7 8 1740 1600 1 1 1 2 50 82 7.3 7 1460 1400 1 1 1 3 50 87 3.7 4 740 800 1 1 1 4 50 82 3 3.1 600 620 1 1 1 5 50 87 7.4 7 1480 1400 1 1 1 6 50 70 3.4 3.5 680 700 2 1 1 1 50 79 11.3 11 2260 2200 . . . . . . . . . . 2 1 1 6 50 76 5.2 5 1040 1000 3 2 2 1 54 100 7.3 8.1 1352 1500 . . . . . . . . . 3 2 2 6 54 97 6 6.7 1111 1241 4 2 2 1 48 94 7.2 7.5 1500 1563 . . . . . . . . . . 4 2 2 6 48 94 6.3 6.5 1313 1354

Table 1. Extract from a spreadsheet for a rice mother trial with 3 farmers in 2 villages of 2 clusters in GVT east in 2001. Data for varieties 2 to 5 not shown for farmers 2 to 4). The data require no reformatting for Minitab.

allow direct comparison among all varieties. Farmers(sites) serve as replications. Mother trials are kept toa minimum, as they are research intensive. Thefollowing types of data are usually collected in thesetrials:• Quantitative data on traits such as grain yield,

straw yield, days to flowering.

• Ranking of all the varieties for traits such as cooking quality, and market price.

2. For the Baby trials (single-replicate, single-variety) each farmer provides a replication and thereare more baby trials than Mother trials. Householdlevel questionnaires (HLQs) are collected from eachfarmer to record their perceptions about a new varietyas ‘more’, ‘same’ or ‘less’ in comparison to the localcheck for many traits including yield and quality(Appendix 1). Sometimes quantitative data for yieldare obtained, although considerable practicaldifficulties have to be overcome.

1. Analysis of data from Mother trials

Analysis of quantitative data.The analysis of a two-way data table, with varietieson one side and farmers (replications) on the other, isstraightforward and can be computed as a randomisedcomplete block design using standard statisticalpackages.

We analysed the data of single-replicate all-entry Mother trials of six varieties of upland riceconducted by 12 farmers in Jharkhand, in the rainyseason of 2001. There was significant variationamong varieties and farmers (Table 2). Thesesignificances are conservative, as the error term totest the main effects is the varieties x farmers'interaction that includes a large field-to-fieldvariation.

Analysis of matrix ranksTwo-way ANOVA. The six varieties in the nine-replicate mother trial were ranked for grain yield by

consensus in focus group discussions. The data wereanalysed by a two-way ANOVA having ninereplications and six varieties. The ANOVA revealedsignificant differences among varieties and farmersfor ranks at P<0.001.

The mean ranks for varieties of interest were:Ashoka 200F = 4.3; Ashoka 228 = 4.7; Kalinga III =3.0; and BG 102 = 2.4. The lsd (5%) = 0.71 wascomputed. The mean scores of Ashoka 200F andAshoka 228 do not differ but both have significantlyhigher score than the check varieties Kalinga III andBG 102.

Friedman’s method for randomised blocks.Alternatively, Friedman’s non-parametric method canbe used. The matrix scores of the six varieties withineach block (farmer) are assigned ranks (1=lowest,6=highest) for making the paired comparisons.Friedman's method yields a c2 value for eachcomparison. The results are similar to that of theANOVA, since c2 for 1 d.f. for Ashoka 200F vs.Ashoka 228 = 1.78 (non-significant); and for Ashoka200F vs. Kalinga III = 9.00 (P<0.01); Ashoka 200Fvs. BG 102 = 9.00 (P<0.01); Ashoka 228 vs. KalingaIII = 5.44 (P<0.05); and Ashoka 228 vs. BG102 =9.00 (P<0.01)

2. Analysis of data from Baby trials

Analysis of quantitative dataGrain yield data when collected can be analysed asfollows:

Analysis of paired-plots by randomised block designor paired t-test: In Baby trials, a pair is composed ofnew and local cultivars sharing the common field.These data can by analysed as a two-way ANOVA.This design is a special case of the randomised blockdesign where there are only two treatments and the nfarmers represent the blocks.

An alternative analysis of paired plots isprovided by the paired t-test (Snedecor and Cochran,1973). Which of these two analyses should be

preferred varies with the interest ofthe researcher. For example, whenit is desired to assess if the varietieswere tested over variable fields,then the ANOVA technique willprovide a separate sum of squaresfor testing variation amongfarmers’ fields that is not possiblewith a paired t-test.

We analysed grain yield t ha-1 of the upland rice varietyAshoka 200F tested alongside

Table 2. Analysis of variance for grain yield (t ha-1) of six varieties of upland rice tested in mother trials in Jharkhand (India) in the rainy season of 2001.

Standard error of difference of variety mean yield = 0.04 t ha-1; lsd = 0.07 t ha-1.

Source df SS MS F P

Sites (s, Farmers) (s-1) =11 2973 270.27 35.95 <0.001 Varieties (a) (a-1) =5 1046 209.20 27.83 <0.001 Sites x Varieties (error)

(s-1) (a-1) =55 413 7.51

Total (as-1)

the local check variety, BG102, in Jharkhand, India,in the rainy season of 2001 in 22 Baby trials.

Both analyses similarly concluded that therewere significant differences between varieties forgrain yield (Table 3). Since there is 1 degree offreedom for varieties the computed F=t2. Theadvantage of the ANOVA technique is clear as itdiscerns significant differences among the farmers’fields over which the varieties were tested. Note thatthe F-test assumes the farmers x varieties interactionto be non-significant and it cannot be tested in thepresent case. The paired t-test assumes that the pair ofobservations is not independent but related; if onemember of the pair is high, so is the other - i.e. theyare positively correlated. This can be verified byperforming a correlation or regression analysis.

Regression analysis: Regression of grain yield (t ha-1) of Ashoka 200F on to the grain yield of thelocal variety was significant (a = 0.123 ± 0.106; b =1.44 ± 0.08) with a very high R2 = 0.94. Theregression coefficient of the new variety is more thanone indicating its increased response to improvingenvironmental conditions. However, the yield of thenew variety in the poor environment is still higherthan the local check. The new variety is thusadaptable to the sample of target environments andcan be a potential replacement for the local variety.

Non-parametric methods: We also investigated theuse of some alternative non-parametric methods thatmake no assumptions for specific distribution for thepopulation.

Friedman’s method for randomised blocks: This is analternative to the two-way ANOVA. In this methodthe varieties (a = 2) are ranked within each block orfarmer; b = 22 in our case. If there are ties, they aretreated in the conventional manner (by averageranks). The ranks are summed for each of the varietyand the statistic S or c2 is computed (Sokal andRohlf, 1995). c2 is distributed as c2(a-1), where a isthe number of varieties. Our computed value of c2 =21.98 is significant at P<0.001. We conclude, as for

the ANOVA and the paired t-test, that the grain yield ofAshoka 200F is significantlyhigher than the local variety.

Wilcoxon's signed-ranks test:This method is more widelyused than the Friedman’s testfor paired-plots. Thedifferences between the grainyields of the varieties across

farmers are computed and ranked without regard tothe signs so that the smallest absolute difference isranked 1 and the largest is ranked 22. Tied ranks arecomputed as averages as usual. The original signs ofdifferences are then assigned to the correspondingranks. The sum of the positive and the negative ranksare computed, and whichever is smaller in absolutevalue is labelled as Ts. This is compared with thecritical value of T in statistical tables for thecorresponding sample size.

In our case the value of Ts = 0. Since this isless than the table value at the two-tailed 1% level(see Table A9 in Snedecor and Cochran, 1991) thenull hypothesis is rejected. Thus Ashoka 200Fsignificantly yields more than the local variety.

Analysis of perception (qualitative) dataZ-test for comparing per cent perceptions: Standarderrors can be computed for per cent perceptions andcompared for testing the significance of differences.In HLQs returned by 40 farmers comparing theAshoka 200F variety of rice to the local check in2000 in Jharkhand, 78% perceived Ashoka 200F tohave ‘more’, 23% to have ‘same’, and none ‘less’yield than the local variety. In this case N = 40;variance of proportion (p) is = pq/n where q = (1-p);SE for ‘more’ = v(pq/n) = v{(0.78 x 0.23)/40} =±0.07. We may compare ‘more’ with ‘same’ bycomputing, Z = (p1-p2)/ v(p1q1/n1 + p2q2/n2). Thecomputed Z = 0.78-0.23/ v2{( 0.78 x 0.23)/40} =5.81**. The null hypothesis is rejected and we canconclude that a significantly larger number of farmersperceived Ashoka 200F to be higher yielding than thecheck variety.

Chi-squared test: Alternatively, a chi-squared testmay be applied for the perception of 40 farmers ofwhich 31 recorded ‘more’, nine the ‘same’, and 0‘less’ for the grain yield of Ashoka 200F over thelocal check. A c2(2 d.f.) = 38.15 (P<0.01) wascomputed on a 1:1:1 hypothesis. It shows thatsignificantly more farmers perceive Ashoka 200F tobe higher yielding than the local variety.

ANOVA Paired t-test Source df MS F P Mean

A200F Mean Local

t P

Farmers (F) 21 2728 18.4 <0.001 1.43 0.91 4.52 <0.001 Varieties (V)

1 3028 20.4 <0.001 -- - - -

F x V 21 148 - - - - - -

Table 3. ANOVA and paired t-test for grain yield (t ha-1) for 22 baby trials of A200F vs. local in Jharkhand, India, rainy

Discussion

We have shown how simple experiments (Mother-Baby design) by farmers provide data that can bestatistically analysed by existing, robust parametricand non-parametric statistical techniques.

We recommend that researchers:• Conduct some Mother trials and a large

number of Baby trials.

• Analyse the data from Mother trials tocompare all the new varieties among themselves.

• Analyse the matrix rank data from Mother trials using parametric or non-parametric methods.

• Analyse Baby trials as paired-plots, i.e. use a paired t-test or, preferably, a randomised blockdesign analysis.

• Do a regression analysis of paired-variety datato test the response of the new variety in relation to the check variety.

• Alternatively, non-parametric methods such as Friedman’s test, Wilcoxon’s signed-rank testcan be used to compare paired varieties.

• Analyse farmers’ perceptions from HLQs by non-parametric methods.

Implications for drought-prone station trials

While performing a combined analysis ofmultilocational research station trials, only balancedsets are usually considered. The trials from drought-prone locations where only a few entries may survivein the trial are rejected as ‘failed trials’. However, thefew entries that survive, in fact, provide the mostvaluable information for marginal areas. Thisdemands modification of multilocational analysis. Ifdata from partly successful trials are analysed by thepaired-plot techniques, described in this paper, thereseems little need to reject a drought prone-site fromthe overall analysis.

Acknowledgement

This publication is an output from projects R7281and R7542 funded by the UK Department forInternational Development (DFID) Plant SciencesResearch Programme for the benefit of developingcountries. The views expressed are not necessarily

those of DFID.

References

Snapp, S., 1999. Mother and Baby trials: a novel trialdesign being tried out in Malawi. In: TARGET. TheNewsletter of the Soil Fertility Research Networkfor Maize-Based Cropping Systems in Malawi andZimbabwe. January 1999 issue. CIMMYT,Zimbabwe.

Snedecor, G.W. and Cochran, W.G. 1973. Statisticalmethods. 6th edition. Iowa State University Press,Ames, USA.

Sokal, R.R. and Rohlf, F.J. 1995. Biometry- The principalsand practice of statistics in biological research. 3rdedition. W.H. Freeman and company, New York.

Chair: Bhuwon SthapitDiscussants: A Subedi and T ParisRapporteur: GN Atlin

Participants in this session were primarily socialscientists, and represented both national programsand NGOs in the Philippines, India, and Nepal.

The group identified lack of social sciencecapacity as the primary constraint to the integrationof social science expertise in PPB/PVS programs.The shortage of skilled social scientists is acute inmost national programs. This situation is unlikely tochange in the near future. There was a consensus thattraining plant breeders in robust methods forparticipatory rural appraisal (PRA), the study of seednetworks, and preference analysis is likely to be themost feasible approach to integrating social sciencein participatory crop improvement.

Best practices

It was generally agreed that the involvement of socialscience professionals in PPB/PVS is most importantat the goal-setting stage. However, the group felt thatsimple diagnostic methods and tools can givebreeders and agronomists the information needed toset breeding priorities and identify target groups offarmers.

It was also agreed that the use of ‘keyinformants’, ‘nodal farmers’, and ‘leadingfarmers’can increase impact. It was noted that thesedifferent community roles are not always filled by thesame person. Some concerns were expressed aboutrelying too heavily on ‘leading farmers’.

Support was expressed by some members ofthe group for the use of positive approaches toinformation collection and PRA, notably theAppreciative Inquiry technique, where the focus is onidentifying the favourable features of farmerpractices and varieties.

Some participants cautioned that long-termrelationships with farmers can negatively affect dataquality by narrowing the sample of farmerscontacted, or by the initiation of a feedback loopbetween farmers and breeders, in which farmers maybe influenced by the priorities and methods ofresearcher. This was not a consensus point, however.Some participants noted that long-term relationshipswith farmers can increase their capacity andeffectiveness as research partners, and can allowfarmers to participate in planning and evaluationresearch.

Training needs

Because of the lack of professional social scientistsworking in agricultural research in nationalprogrammes, the group agreed that there is an urgentneed for breeders and agronomists to be trained inPRA. Most agronomists and breeders also needtraining at a very basic level in communicating withfarmers and in cultural sensitivity. Training inparticipatory methods must be institutionalized in thetraining of plant breeders

Research needs

There was a consensus on the need for research onhow to quickly and efficiently identify nodal andsocial networks for seed dissemination, and how tolink with these networks to increase the adoption ofsuperior varieties. Information is also needed on thestability of farmer networks over time and crops.

There is also a critical need for research onmethods for assessing rice quality in the context of aPVS or PPB program. Quality is recognized as thekey element influencing farmer preference in manysituations, yet effective, large-scale methods forparticipatory quality evaluation have not beendeveloped.

Integrating social science into PPB/PVS

Chair: P CraufurdDiscussants: KD Joshi and PK Sinha

Eleven researchers, representing India, Ghana,Nepal, and the Philippines participated in thediscussion. It was generally agreed at the outset thatthe availability of seed of farmer-preferred varietiesis one of the major constraints to their spread. Thediscussion focused on PVS and seed systems forreleased and unreleased varieties in the context ofrainfed upland systems.

Best practices

Participants identified several practices that had metwith some success in promoting dissemination offarmer-preferred varieties. These included:

Farmer groups or community-based seedmultiplication (CBSM) programmesFarmer or CBSM groups have been successful insome contexts. Farmer training and outsidefacilitation is needed to make these groupssustainable and effective.

Seed-producing villagesThe Central Rainfed Upland Rice Research Station(CRURRS) in Jharkand, India, is experimenting witha system in which whole villages specialize in seedproduction of a single variety, with assistance fromthe research station, which provides seed andtechnical backstopping. The producing village thengenerates income by selling seed to farmers in thesurrounding areas. This approach, which is similar tocontract commercial seed production, has not beenwidely tested, and is likely to be best suited to theinitial introduction of new varieties, with substantialresearcher involvement. Centralization of seedproduction in this manner would facilitate seedcertification systems based on the ‘TruthfullyLabelled’ or ‘Quality Declared’ designations.

Use of nodal farmers to encourage farmer-to-farmerspread of preferred varietiesNodal farmers are central links in local seed-exchange networks. If farmers who are known andrespected sources of seed for the community areidentified and integrated in PVS and seed distributionprograms, farmer-to-farmer spread can be enhanced.Seed/diversity fairs

Farmer-to-farmer seed exchange can be promoted

through seed or diversity displays and contests atagricultural fairs.

Large-scale PVS programmesLarge scale PVS programmes using Mother-Baby orInformal Research and Development (IRD) designshave demonstrated effectiveness in facilitatingextensive variety testing, with substantial adoptionand dissemination of some varieties in the trials. Butimportant questions remain as to how often thesetrials should be conducted in a given regions, andhow the programmes should be designed tomaximize their effectiveness as dissemination tools.

Practices that do not work

Private sector seed distributionIn general, the private sector will not find it profitableto distribute seed in marginal areas. Particularly inthe uplands, farmers use little purchased seed andminimize investment in all inputs.

Seed systems and PVS - scaling up for impact

Introduction

IRRI’s participatory crop improvement research isfocused on increasing the impact of rainfed ricebreeding programmes. Particularly in upland rice,these breeding programmes have had only limitedsuccess in:• Identifying the traits that are important to

farmers;

• Identifying lines that outperform traditionalvarieties in the most stressful environments;

• Giving farmers access to new varieties.

IRRI is collaborating with nationalprogrammes in India, Laos, and Indonesia in PVSresearch that addresses the problem of the limitedpower of on-station experimentation to predictfarmer adoption of new lines, and on reducing thehigh costs of introducing effective participatoryresearch into formal breeding programmes. Thisresearch recognizes that random genotype xenvironment interaction (GxEI), or variation inranking among experimental lines across trials, even

within narrowly-defined target environments, is largein rainfed rice (Cooper et al. 1999). The only reliableway to minimize the effects of random GxEI is toevaluate new cultivars over many farms and severalseasons (Atlin et al. 2001).

The experience of IRRI and its nationalpartners indicates that linking small rainfed ricebreeding programmes into collaborative breeding andvariety testing networks is an effective way toachieve the environmental replication needed tomake genetic gains in stressful environments. Thispaper summarizes the research results of IRRI and itscollaborators on approaches to integrating andscaling up PVS within national rainfed riceprogrammes.

Key points for introducing participation

Experience has identified three key points whereparticipation is needed in breeding programmes:

Identifying traits valued by farmers throughparticipatory assessment

It is important to identify at the outset quality and

Integrating conventional and participatory crop improvement in rainfed rice

GN ATLIN1, T PARIS1, B LINQUIST1, S PHENGCHANG2, K CHONGYIKANGUTOR2, A SINGH3, VN SINGH3, JL DWIVEDI3, S PANDEY1, P CENAS1, M LAZA1, PK SINHA4, NP MANDAL4, AND SUWARNO5

1. International Rice Research Institute (IRRI), DAPO Box 7777, Makati, Philippines.2. National Agricultural and Forestry Research Institute, Luang Prabang, Lao PDR.3. Narendra Devra University of Agriculture and Technology, Faizabad, U.P., India.4. Central Rainfed Upland Rice Research Station, Hazaribag, Jharkand, India.5. Central Research Institute for Food Crops, Bogor, Indonesia.

AbstractTo increase the impact of rainfed rice breeding programmes, participatory varietal selection (PVS) mustbe institutionalised in the varietal screening process. PVS can be conveniently integrated in two steps:(1) preliminary PVS, in which a relatively large number of lines is screened for adaptation andacceptability on a few farms, with a high level of researcher involvement per farm; (2) advanced PVS,in which a small set of varieties is screened by many farmers, with minimal researcher involvement perfarm.

Use of farmer ratings for yield and other agronomic traits, collected via surveys during and afterthe growing season instead of through crop cuts and researcher measurements, permits even smallbreeding programmes to scale up advanced PVS, sampling the large numbers of farmers andenvironments needed to reliably detect differences among varieties. Research in rainfed rice systems inIndia, Laos, Indonesia, and the Philippines has shown that farmer pre-harvest yield ratings are highlycorrelated with researcher-measured yields and with farmer preference ratings taken at the same stage.Little cultivar x farm interaction is observed in rainfed rice PVS trials, indicating that broadly adaptedcultivars can be identified.

production traits preferred by farmers, in order toallow the selection of genotypes with preferred traitsearly in the on-station evaluation process. Failure todetermine and screen for traits critical to farmerpreference means that resources are expended ondeveloping and increasing lines that have no hope ofbeing adopted. This is particularly important inupland systems where farmer quality and plant-typepreferences are often not well understood bybreeders, but it is also a necessary prerequisite to thescaling up of PVS. This is because large-scale PVS,in which dozens or even hundreds of farmers areasked to assess the performance of new lines, mustrely on quantitative survey instruments involvingfixed questions and numerical rating scales thatpermit statistical summary and analysis, rather thanon open-ended questionnaires. The questionsincluded in quantitative surveys must therefore bedesigned to capture information on the traits mostvalued by farmers.

We have found that several participatorytechniques can be used to elicit farmer priorities andpreferences. Among the most effective of these ispreference analysis undertaken with individuals orsmall farmer groups in the field. In these exercises,farmers view a variety trial (usually just beforematurity), and discuss the strengths and weaknessesof particular varieties by explaining why they like, ordislike, a particular variety.

Preliminary farmer-managed screening of fixedlines: PVS with many genotypes

A second point at which farmer participation isneeded is the preliminary evaluation of fixed linesthat have been developed and evaluatedagronomically on-station. Typically, rainfed ricebreeding programmes in Asia promote about 20 suchlines annually from a preliminary on-station yieldtrial to advanced yield testing in multi-environmenttrials (MET) or in replicated on-station trials.Information on farmer acceptability and performanceunder farmer management is needed on this relativelylarge set of lines. Without it, there is a highprobability that the lines will fail in wide-scale PVS.

PVS strategies permitting the evaluation ofmany genotypes on few farms are useful for thispurpose. We have found that the Mother-Baby PVStrial design can permit a set of up to about 20varieties to be evaluated under farmer management.An incomplete-block design is used, with three toeight varieties tested per farm, and each variety testedon three to five farms. This design requiresconsiderable researcher involvement, but is useful ingetting a preliminary farmer assessment of a large

number of varieties, allowing varieties that performwell on-station but fail under farmer management tobe quickly eliminated.

The importance of this step was confirmed in aPVS programme conducted by the Indonesian uplandrice programme in the wet season of 2002, in which11 varieties were evaluated on six farms (Suwarno etal., 2002, this volume). Although all the linesperformed well on-station, several failed on-farm dueto blast or brown-spot disease, which had notappeared with severity in screening trials on-station,probably because of differences in fertilitymanagement. Differences in plant density and weedpressure between the station and farmers’ fields mayalso have contributed to the poor performance ofsome lines on-farm.

It is important for the credibility and success ofadvanced PVS trials that lines with a high probabilityof non-acceptance by farmers be excluded. This PVSstep is at least as important as replicated on-station orMET testing. Most formal breeding programmes candirectly manage such trials on a small number offarms, but may need to shift resources into them andout of other on-station activities.

Advanced PVS and the use of farmer ratings

The products of the preliminary PVS trials describedabove must be evaluated in advanced PVS trials bylarge numbers of farmers in all the conditions thatoccur within the target environment. Such trials,which may require the participation of hundreds offarmers, must be designed to require limitedresearcher involvement per trial. A modification ofthe informal research and development (IRD) designdescribed by Witcombe et al. (2001), in which eachparticipating farmer evaluates only one test variety, issuitable for this purpose.

Farmer assessments are collected throughsurveys in which farmers use a simple numericalscale to rate test cultivars as better than, the same as,or worse than their own cultivar for agronomic andquality traits. Survey questions must be carefullydesigned to target the characteristics that areimportant to farmers, and administered in a way thatfacilitates quantitative analysis. Questions can bedeveloped using information obtained in the stepsabove.

Some breeding programmes will be unable tomanage direct farmer contacts on the scale needed foradvanced PVS, but will need to link with extension ordevelopment organizations to recruit farmers,distribute seed, and administer surveys. However,advanced PVS can be efficiently undertaken directlyby researchers linked in multi-station breeding

networks, if each station manages the programme inseveral nearby villages.

The use of farmer ratings to replace agronomicmeasurements in advanced PVS

The relationship between farmer yield ratings andresearcher-measured yield

Successful implementation of large-scale PVSprogrammes in rainfed rice will require the use ofmethods that minimize researcher involvement pertrial. In general, this will involve the use of farmerratings rather than quantitative measurements ofyield and other traits.

It is important, therefore, to estimate thecorrelation between farmer ratings and quantitativemeasurements. This has been done for grain yield inPVS trials in several countries and ecosystems (Table1). In both upland PVS trials in the Philippines and arainfed lowland trial in eastern India, farmer ratingsof cultivar yield were highly correlated withresearcher-measured yields. Farmer perception ofyield captured with a simple rating system thereforeappears to be strongly associated with grain yieldmeasurements taken by researchers. It is likely thatyield ratings collected in surveys may substitute forcrop-cut measurements in large-scale PVS trials,greatly reducing their cost.

The relationship between grain yield and farmerpreference

Questions have been raised about the extent towhich rainfed rice producers, who grow ricemainly for food security, are concerned withgrain yield in comparison with other productionparameters. This question was addressed byexamining the correlation between farmerpreference ratings and grain yieldmeasurements or ratings in PVS trialsconducted in the Philippines, India, Indonesia,and Laos.

In all cases, although there were high-yielding cultivars that were not preferred, therewas a strong correlation between measuredgrain yield and preference ratings (Table 1).

The relationship between farmer yieldratings and farmer preference at harvest was atleast as strong as the relationship betweenresearcher-measured yield and preference.Grain yield, either as measured by researchersor as visually perceived by farmers, appears tobe an important criterion determining farmerpreference.

Genotype x environment interaction in PVS trials

One of the proposed effects of increased farmerparticipation in plant breeding is increased local orspecific adaptation of cultivars. This would implythat cultivars emerging from PPB and PVSprogrammes exhibit increased genotype xenvironment interaction. This hypothesis was testedin farmer-managed, multi-location PVS trials inJharkhand, India and in northern Laos. Little cultivarx village interaction was observed for grain yield inthe Indian trial (Table 2). No significant cultivar xprovince interaction was observed for grain yield inLaos, but a strong interaction was observed forfarmer preference among early cultivars (Table 3).The reasons for this province-specific preference areunknown.

Overall, these results indicate that superiorcultivars identified in PVS trials may be broadlyadapted and perform well at a regional level, afinding that has also been reported by Joshi et al.(2001), and is consistent with the broad farmer-to-farmer diffusion of preferred rainfed rice cultivarssuch as Mahsuri (Mackill et al., 1996).

However, local preferences for cultivars withparticular agronomic traits may also be detected inPVS trials with sufficient replication withinenvironmental subclasses such as regions ormanagement regimes to permit cultivar x classinteractions to be tested against the pooled cultivar x

Table 1. Correlations among researcher-measuredyields, farmer yield ratings, and farmerpreference ratings in on-farm PVS trials in India, Laos, and the Philippines.

Trial location Correlation Farmer yield ratings versus researcher-measured yields Batangas, Luzon, Philippines 0.72 Faisabad, Uttar Pradesh, India 0.90 Researcher-measured yields versus farmer preference ratings Tamanbogo, Sumatra, Indonesia 0.66 Faisabad, Uttar Pradesh, India 0.67 Luang Prabang, Laos (medium maturity)

0.82

Luang Prabang, Laos (early maturity) 0.54

Farmer yield ratings versus farmer preference ratings Faisabad, Uttar Pradesh, India 0.81 Batangas, Luzon, Philippines 0.80

farm variance within classes (Atlin et al., 2000).

In general, our results show that participatoryapproaches can be usefully integrated at severalpoints in rainfed rice breeding programmes,including identification of farmer priorities, initialelimination of unsuitable genotypes, and the finalstage of cultivar evaluation. Farmer ratings for grainyield are closely associated with both preference andmeasured yield, and can serve as the basis forselection decisions in large-scale PVS trials,permitting low-cost scale-up of participatoryselection. PVS products appear to be broadly adaptedacross farms within the target environments sampledin these studies.

Integrating PVS in rainfed rice breeding networks

Because of the sporadic nature of drought and

flooding stresses in rainfed rice productionenvironments, resulting in a large random GxEIvariance, large-scale, multi-location testing systemsthat link small breeding programmes together haveproven to be an effective approach for increasingbreeding progress. In the Eastern Indian RainfedLowland Shuttle Breeding Network, IRRIcollaborates with nine Indian breeding programmesat ICAR research stations and state agriculturaluniversities to jointly develop and evaluate linesacross the spectrum of rainfed rice productionenvironments.

A similar interstation network, involving eightsites, has been operating for two years in northeastThailand. These breeding networks annually evaluateover 130 lines in multi-location at high levels ofprecision, and are ideally structured to implementlarge-scale PVS as the final step in the varietalevaluation process. Networks in which eachindividual centre manages a preliminary Mother-Baby PVS trial requiring researcher layout in three orfour communities, with the participation of three tofive farmers per community, could, in collaboration,obtain information on varietal performance andfarmer preference from 70-150 farmers in a singleseason at a manageable cost in time and resources.

Large-scale advanced PVS networks relyingon the use of farmer ratings could sample manyhundreds of farmers in the target region, permittinglocal adaptation to be reliably detected if it exists.Lines selected from such networked PVS trials arelikely to be broadly adapted, stress tolerant, andacceptable to farmers.

References

Atlin GN, Cooper M, Bjørnstad Å, 2001. A comparison of formal and participatory breeding approaches usingselection theory. Euphytica 122: 463-475.

Atlin GN, Baker RJ, McRae KB, and Lu X, 2000. Theeffect of subdividing a target region on selectionresponse. Crop Science. 40: 1-6

Cooper M, Rajatasereekul S, Immark S, Fukai S, andBasnayake J, 1999a. Rainfed lowland rice breedingstrategies for northeast Thailand. I. Genotypicvariation and genotype x environment interactionsfor grain yield. Field Crops Research 64: 131-151.

Joshi KD, Sthapit BR, and Witcombe JR, 2001. Hownarrowly adapted are the products of decentralizedbreeding? The spread of rice varieties from aparticipatory plant breeding programme in Nepal.Euphytica 122:589-597.

Mackill DJ, Coffman WR, and Garrity DT, 1996. Rainfed

Source of variation Variance component

Cultivar 0.077 Cultivar x location 0 Cultivar x year 0 Cultivar x location x year 0.246

Table 2. Variance components for upland rice cultivars evaluated under farmer management in four villages over three years near Hazaribag, Jharkhand, India.

Trait Source of variation Cultivar Cultivar x

province

Yield of early cultivars

* ns

Yield of late cultivars

ns ns

Preference rating for early cultivars

ns *

Preference rating for late cultivars

* ns

Table 3. Significance level of F-tests of cultivarand cultivar x province interaction for eight early-maturing and eight late-maturing cultivars evaluated in Luang Prabang and Sayabouly provinces, Laos, 2001.

lowland rice improvement. International RiceResearch Institute, Los Baños.

Witcombe JR, Joshi KD, Rana RB, and Virk DS, (2001).Increasing genetic diversity by participatoryvarietal selection in high potential productionsystems in Nepal and India. Euphytica 122: 575-588.

Chair: Gary AtlinDiscussant: Dr Krishna JoshiRapporteur: Dr BN Singh

The discussion focused on describing ‘bestpractices’for rainfed rice breeding programmes.Steps in the process where integration is beneficialinclude the following:

Selection of the target environment and farmerpopulation, and setting breeding goals

PRA, focus group discussions, and otherparticipatory research tools need to be made moreaccessible to formal breeding programmes. Farmers,millers and consumers must be consulted in thesetting of breeding goals.

Selection of parents and donors for hybridization

Participatory crop improvement experience stressesthe need to use adapted local varieties as parents. Atthe same time, one of the strengths of conventionalplant breeding programmes is their access to a broadrange of germplasm from other regions. Exoticgermplasm has been used very successfully in theDFID upland rice PPB programme in India, whereIR64 has been used as a parent in upland targetedcrosses for stressful environments.

Conventional breeding programmes canefficiently produce heterogeneous populations forfarmer selection or lines for PVS. The importance ofcareful parent choice for small breeding programmeswas stressed, and the value of a strategy based onscreening large populations of a few well-chosencrosses was reiterated by J. Witcombe.

Screening for diseases, problem soils, drought,and submergence tolerance

There was a consensus that conventional breedingprogrammes can have an advantage in screeninglarge, heterogeneous or segregation populations fortolerance to well-characterized biotic and abioticstresses, notably disease and insect resistance,submergence, salinity, other problem soils, andpossibly drought.

Programmes relying solely on on-farmselection may not be able to apply consistentselection pressure for such resistances when stressesare sporadic in occurrence. Managed-stress nurserieson research stations are very useful in pre-screeningmaterial before it is exposed to farmer selection.

Post-harvest evaluation

Early evaluation of cooking and eating quality isessential to the success of breeding programmes.PVS programmes are very effective in collectinginformation from farmers on harvestability,storability, and quality of advanced breeding lines,but there was strong consensus that qualityevaluation must be conducted as early as possible inthe cultivar development process, and certainly nolater than the first year of yield evaluation.

J. Witcombe and collaborators described theiruse of micromilling and participatory approaches toorganoleptic testing to ensure that genotypes ofunacceptable quality are discarded before they aresubjected to expensive agronomic evaluation onstation or by farmers.

Integration of PVS in conventional breedingprogrammes and cultivar testing systems

Several strategies for integrating PVS into varietaltesting were discussed. It was suggested by severalparticipants that on-farm PVS of the Mother-Babymodel is best initiated during or following multi-environment yield testing, or during testing fornational release. G. Atlin noted that in many rainfedrice programmes it would be beneficial to introduce apreliminary PVS step to complement or replaceinitial on-station replicated yield testing, which isoften poorly predictive of farmer preference.

Seed production

Rate of adoption is strongly associated with seedavailability. Conventional breeding programs rarelyhave the capacity to produce and distribute largequantities of seed of new varieties. Participatory,community-based seed production approaches needto be pursued by formal-sector programs.

Varietal release

Greater use of PVS data by national releasecommittees to complement data from on-station trialsin making varietal release decisions is likely to resultin a higher proportion of released varieties beingacceptable to farmers.

Integrating conventional and participatory breeding

Chair: Thelma ParisDiscussant: VS TomarRapporteur: DN Singh

The group outlined the respective roles and potentialsynergies of the formal research institutions andagricultural development organizations (ADO,) invarietal development. Specific areas in which ADOcontributions are important include:

• Determining farmers’ needs and preferences, and helping breeders understand and prioritize these;

• Representing farmers’ interests in seed production organizations and varietal release committees;

• Help in scaling-up on-farm evaluation, especially for quality and sensory parameters;

• Help in involving farmers in selection in early generations;

• Influencing policy makers and other NGOs.

The formal breeding sector has critical contributionsto make in the following areas:

• Collection, characterization, evaluation, and dissemination of germplasm;

• Crossing and development of populations;

• Screening for disease, insect, and abiotic stress tolerances in managed-stress nurseries;

• Facilitating the varietal release process;

• Producing breeder seed;

• Packaging and transferring technology for faster delivery to farmers

Overall, the strengths of ADO center on their abilityto contact farmers, integrate them in the planning anddecision-making process, and facilitate large-scaletesting and seed distribution. Formal-sector breedingprograms have great capacity to access and screengermplasm under managed-stress conditions, and todevelop novel populations, but are constrained intheir ability to work directly with large numbers offarmers. Integrating the strengths of ADOs and

formal breeding programs can greatly speed thedevelopment and dissemination of improvedvarieties.

Several important factors in the success ofpartnerships between ADO and formal breedingprograms were identified:

• The ADO partner should have the resources, capacity, and interest needed to participate in research and development.

• A formal framework for collaboration, usually in the form of a Memorandum of Understanding, is needed between the ADO and research organization.

• Simple methods for experiment management and data collection are needed.

• The ADO and farmers need to be involve in dissemination of information and seed through informal channels.

• Credit for successes must be shared between the formal program and ADO.

Partnerships between breeding programmes and agricultural development

Introduction

This is an example of a mother-baby trial system(Snapp et al. 1999). This system is flexible, and someresearchers may choose to have more than onereplicate in the mother trial. Likewise, in somesystems, individual farmers may grow more than onevariety in the baby trial.

Mother trials

Single-replicate designs (mother trials) are used toassess the relative performance of varieties. Manyentries are grown together in the same field. Thetrials are researcher-designed but farmer-managed,and they are replicated across villages. They not onlyserve as demonstration plots or focal points fordiscussions, but are also specifically designed toprovide quantitative, analysable data on yield.

Baby trials

The mother and baby trial system recognises thedifficulty of obtaining reliable yield data from many,widely dispersed participatory trials, so in baby trialsonly farmers' perceptions on yield are collected. Afarmer grows one new variety alongside the localvariety under the customary management practices.Replication is across farmers, either in the samevillage or across villages. Even if there are manynew varieties, only one is tested by any one farmer.The trials are evaluated by participatory methods -farm walks, Focus Group Discussions and

Annexe 1A mother and baby trial system

JR WITCOMBECentre for Arid Zone Studies, University of Wales, Bangor, Gwynedd, LL57 2UW, UK

One or two new cultivars per farmer Compared to local cultivar or second new variety Farmer managed, farmer inputs Evaluation of farmer's perceptions

NEW LOCAL

NEW or

Figure 3. The baby trial

Table 1. The stages of popularisation.

Stage Process Initial Mother and baby trials

Adaptive* IRD

After recommendation by project or SAU

Seed sales and dissemination

Many cultivars, several locations, one replicate per location Researcher designed, farmer managed, farmer level of inputs Yield and maturity measured by researchers Consultative evaluation of other traits

Figure 2. The mother trial

IRD is a cheaper form of PVS because it uses lessintensive evaluation. In IRD programmes, the smallquantities of seed given to farmers are of the samerange of varieties used in the FAMPAR baby trials.

No monitoring or participatory evaluation is doneduring the growing season, but farmers' perceptionsare evaluated after harvest, using informal interviews.In IRD the subsequent adoption and farmer-to-farmer

household-level questionnaires. Qualitative data, (forexample, on yield) are obtained by project staff frominterviews with farmers.

Informal research and development trials

The mother and baby trials are followed by methodsof popularisation (Table 1). Large numbers of smallpackets of seed are distributed free of cost to farmers.Once a variety is popular or released then seed salesanddissemination follow. The differences between mothertrials, baby trials and informal research anddevelopment (IRD) trials are summarised in Table 2.

Research station trials and mother trialsmust include the GVT 'check' varieties. Thesechecks are essential entries in the trials andinclude farmer-preferred varieties andpromising entries. Unless these are included itis not possible to tell if a test variety is better orworse than the ones GVT are currentlypromoting in the formal and informal seedsupply system. Detailed protocols for the designand conduct of the mother trials are describedbelow.

Table 2. The differences between mother trials, baby trials and IRD

5 kg 1 kg 0.5 kg Maize Rice Chickpea† Wheat Chickpea Sunn hemp Wheat† Horsegram Niger Black gram

†When seed availability is limiting

Table 3. Seed quantities per trial

The mother trial

PurposeThe mother trial gives statistically analysable dataon yield per hectare.

DesignEach trial consists of a single replicate of arandomised complete block design. Hence everymother trial has a separate randomisation.The plotsshould have a border of the local variety whereverpossible. Plot size will vary with the land availablebut are larger than is normally used in researchstation trials.

Number and locationThere are only a few mother trials compared to thenumber of baby trials. As a minimum, three mothertrials are grown in each of two villages. The threemother trials in a village are grown by differentfarmers in different fields.

Colour coding of entriesNot required.

Selection of farmersThe mother trials will be conducted in villages inwhich GVT is well established with farmers thathave co-operated with the project in the past. Thereis no need to change the farmers and locations overyears so the trials will become easier to run.

Mother Baby

IRD

Obtain yield data

Obtain perception data Popularise

Few trials

Many trials More trials than baby trials

Researcher designed and supervised

Simple design - farmer supervised

No design

All entries, single-replicate design, small plots

One or two entries, simple design, large plots

One entry – the identified variety

Yield is measured

Yield is not measured Yield is not measured

Farmers’ perceptions usually measured by matrix ranking

Farmers’ perceptions measured by HLQ

Farmers’ perceptions measured informally (by anecdote)

Farmers’ management BUT seed priming and, if needed, more weeding

Farmers’ management Farmers’ management

Farmer can be compensated for growing the trial

Farmer bears the cost and risk (but has free seed)

Farmer has free seed and benefit

seed dissemination provides the evidence forthe degree of acceptance of each variety.Mother and baby trials satisfy the needs ofresearch, extension, and the farmer, while IRDis predominantly targeted at extension. Thequantities of seed to be given in baby trials andIRD can vary but it is recommended that, ingeneral, the quantities shown in Table 3 aregiven per trial.

ManagementAs farmer's practice. Two changes are allowed (seedpriming to obtain more uniform stand establishment,and additional weeding if required).

CostsOnly for mother trials can the farmer be compensatedfor the use of his or her land. This is done when (a)farmer is reluctant to devote such a large area ofland to the trial or (b) when the project wishes toretain the seed from the trial. GVT or the SAU canpay for the estimated amount of produce that wouldhave been obtained from the area of the trial. Theproject can also pay for purchased inputs (fertilizer)but the fertilizer dose must still be as farmer's usualpractice. Labour is supplied free of cost by thefarmer. The seed from the mother trial belongs tothe project when compensation has been given, butthe fodder from the trial belongs to the farmer.

Responsibilities• Layout of the trial: Project staff (SAU or GVT

as applicable).• Sowing of the trial: Supervised by project staff.• Growing and trial management: Farmer (but see

management above).• Trial monitoring and data collection: Project

staff.• Harvesting and threshing: Supervised by project

staff.• Plot measuring: Project staff.• Plot weighing: Harvest weighed by project staff

(whole plot harvest) to determine yield perhectare.

Pre-harvest participatory evaluationA matrix ranking of the performance of the varietiescan be done by a group of interested farmers whenthe varieties are near to maturity.

Post-harvest participatory evaluationOptional (because yield data is the primaryobjective). Can be done if the participating farmeris allowed to retain some of the harvest. A post-harvest interview to rank the varieties is useful.

The mother trial is repeated on the researchstation as a conventional three- or four-replicate,randomised, complete-block design (RCBD).Additional check and test varieties can be added tothe trial if desired.

Baby trials

PurposeThe baby trials give statistically analysable data on

farmers' perceptions and acceptance of varieties.

DesignDesign 1 - single test entry. Single variety per farmercompared to local control. (This is the method thathas been followed so far in GVT).Design 2 - two test entries.Two varieties per farmer†.This method is more powerful, and colour codinghelps farmers lay out the trials correctly without staffsupervision at sowing time.When there are two testentries there may, or may not be, a formal localcontrol. There will always be an informal localcontrol (the farmer's own crop, or a neighbour's crop,grown on an adjacent or nearby field, or the farmer'srecollection).

Number and locationThere are many more baby trials than mother trials.To avoid confusion it is simpler not to have anyfarmer growing both a mother and a baby trial. As aminimum, trials should be in four villages. Eachvariety should be tested a minimum of a total of sixtimes (and to get six successful trials more areinitially needed). Baby trials are not repeated on theresearch station.

Colour coding of entriesRequired. Varieties are allocated colours (e.g. red,blue, green, yellow, white) and supplied in cottonbags of that colour. The name of the variety, and theyear of supply, is also written on the bag. Each bagis supplied with four short bamboo pegs of the samecolour for the farmer to mark out the plot.

Selection of farmersBy group meeting of villagers and random allocationof varieties to farmers. All three wealth ranks shouldbe included.

ManagementStrictly as farmer's practice.

CostsFarmers pay all costs. All seed and fodder belongsto the farmer. Only subsidy is that the seed is providedfree of charge.†The number of varieties is low so all possible pairedcomparisons (n(n-1)/2 can be made. For example, withfour varieties there are six possible comparisons, with fivevarieties 10, and with six there are 16 comparisons.However, there is no need to compare every variety withevery variety, so when the number of varieties is larger asample of comparisons (e.g., comparisons where eachentry appears twice and comparisons are allocatedrandomly = n comparisons) can be made and validlyanalysed.

Responsibilities• Layout of the trial: Farmer (layout is discussed

in detail in the group meeting).• Sowing of the trial: Farmer.• Growing and trial management: Farmer.• Trial monitoring: Project staff complete page 1

of the HLQ (see below) that describes the trial.This needs to be completed whilst the trial is

inthe field - preferably after the last fertiliserapplication.

• Harvesting and threshing: Farmer.• Plot measuring: Not required.• Plot weighing: Not required.• Post-harvest evaluation: Essential. Project staffcomplete page 2 of HLQ and pages 3 and 4 as wellin the case of design 2 (see below). Data in theHLQ are collected on a 'more, same or less'qualitative evaluation. There are now powerfulstatistical methods to analyse such data. By notcollecting yield data, many more trials can beconducted to give a more reliable estimate of farmers'perceptions.

Non-participatory data - Mother trials

Yield data are collected by carefully measuring thearea of each plot and the total plot yield. Other traitssuch as plant height and days to flower or tassel arealso recorded. A complete data set (optional traitsin italics) will comprise the following:

Date of sowing Days to flowering Days to maturity Plant stand Yield per plot (kg)

Plot area (length and breadth) Straw yield (kg per plot) Incidence of diseases Incidence of insect-pests Comments on the trial (e.g. occurrence of drought)

Matrix ranking - Mother trials

Two examples are given of matrix ranking for mothertrials below - one for rice and one for maize. Not allof the characters need to be ranked, only thoseconsidered important by farmers. There is no pointin trying to rank a trait if farmers say there is littledifference between the varieties.

It will not be possible to matrix rank the varietiesfor post-harvest traits unless• The participating farmer has been allowed toretain the seed - i.e. it has not been purchased by theproject.• Sufficient time has elapsed after harvest for grainprocessing, consumption and sales.

The Household Level Questionnaire - baby trials

The baby trials are monitored by household levelquestionnaires. This consists of either two or fourpages:

For Design 1 For Design 2

Page 1 Trial details Page 1 Trial details

Page 2 Test entry versus local

Page 2 Test entry 1 versus Test entry 2

Page 3 Test entry 1 versus Local variety

Page 4 Test entry 2 versus Local variety

How to complete the question concerning the

cross section of the trial on page 1 of the HLQ maynot be obvious, so examples are given in Fig. 3.

Variety 1

Variety 2

Variety 1 Variety 2

Examples of 'cross section' of baby trials

ACCEPTABLE REJECT

Figure 3. Examples of cross section of the trialfrom page 1 of the HLQ of the baby trial.

Acknowledgements

JR Witcombe would like to thank DS Virk and allthe staff in the Gramin Vikas Trust West who havebeen working on PVS, and who helped to developthis system.

Parameter V1 V2 V3 V4 V5 V6 V7 V8 Germination Days to flowering Days to maturity Plant height Lodging resistance Disease resistance Insect resistance Ease of dehusking Grain yield Straw yield Grain colour Grain type Cooking quality Taste Fodder quality Market price Additional traits of farmers’ liking (1)

(2) (3)

Rank varieties on scale where Best = Total no. of varieties being evaluatede.g. 1-5 when there are 5 varieties with 5 = best and worst = 1.

Ranking to be done by group consensus.

Matrix ranking of varieties in FGDs for rice

Variety Name V1 V2 V3 V4 V5 V6 V7 V8

Matrix ranking of varieties in FGDs for maize

Rank varieties on scale where Best = Total no. of varieties being evaluatede.g. 1-5 when there are 5 varieties with 5 = best and worst = 1.

Ranking to be done by group consensus.

Variety Name V1 V2 V3 V4 V5 V6 V7 V8

Number of participants in FGD = Parameter V1 V2 V3 V4 V5 V6 V7 V8 Germination Days to silking Days to tasselling Tasselling silking interval Plant height No. of cobs Cob placement (height) Cob size Cob filling and husk cover Days to maturity Lodging resistance Disease resistance Insect resistance Grain yield Straw yield Grain colour Grain type Cooking quality Fodder quality Market price Additional traits (1)

(2) (3)

Page 1. Household Level Questionnaire for FAMPAR Baby Trial

The answers in the boxes on this page describe the trial

KEY INFORMATION

Farmer name: Village: Hamlet: State: Year: Season: kharif/rabi/summer Crop:

Interviewer: Date of interview:

Names of V1 and V2 Colour codes of V1 and V2 V2 is: Variety 1 (V1): V1 is a test entry Variety 2 (V2): local / test entry

TRIAL DETAILS

Variety 1 Variety 2 Soil local name: Date of sowing: Slope (score): level / gentle / steep Date of harvest: Slope (%): <10% / 10-20% / >20%

Map of the trial: Draw the trial layout with the positions of variety 1 and variety 2. Include field name:

Cross section of the trial: Draw the topography of the trial:

Any other comments on conduct of trial – e.g. is weeding, plant population, intercropping the same for Variety 1 and Variety 2?

FYM Urea DAP Variety 1: Variety 2:

Did variety 1 get: more/same/less more/same/less more/same/less FYM than Variety 2? Urea than Variety 2? DAP than Variety 2?

Was Variety 1 grown under the same management as Variety 2? Yes / No

Page 2. Compare Variety 1 with Variety 2

The answers in the boxes on this page determine if Variety 1 is better, about the same, or worse thanVariety 2

1. Establishment:

Variety 1 is better/same/worse for establishment than variety 2? 2. Flowering time (tasselling):

Variety 1 is earlier/same/later to flower than variety 2?

If flowering time (tasselling) is different then by how many days ? Variety 1 is days earlier than local (Either ‘days earlier’ or Variety 1 is days later than local ‘days later’ as applicable)

3. Yield: Variety 1 yields more/same/less than variety 2?

If the yield is different provide more information in the box below. Give the farmer’s description - e.g. in local units, as a proportion (e.g. half as much again), or a percentage.

4. Grain quality: V1 is better/same/worse in quality than variety 2? 5. Market price: V1 is higher/same/lower in price than variety 2?

6. Overall preference: V1 is better/same/worse overall than variety 2?

7. Intention variety 1: Will the farmer grow Variety 1 next year ? yes / no 8. Intention variety 2: Will the farmer grow Variety 2 next year ? yes / no

Question 8 applies only if V2 is a test entry. (Otherwise it is assumed the local variety will be grown next year)

Add any other information that the farmer thinks is important. This could include reaction to pests and diseases; particular quality traits such as taste, cooking quality, milling quality; and whether the farmer mentions that he or she has given seed to others.

Variety 1

Variety 2

Page 3. Only for Design 2 (two test entires). Compare Variety 1 with thelocal variety

The answers in the boxes on this page determine if Variety 1 is better, about the same, or worse than thelocal variety

1. Establishment: Variety 1 is better/same/worse for establishment than the local variety? 2. Flowering time (tasselling):

Variety 1 is earlier/same/later to flower than the local variety?

If flowering time (tasselling) is different then by how many days ? Variety 1 is days earlier than local Either ‘days earlier’ or Variety 1 is days later than local days later’ as applicable.)

3. Yield: Variety 1 yields more/same/less than the local variety?

If the yield is different provide more information in the box below. Give the farmer’s description e.g. in local measures, as a proportion (e.g. half as much again), or a percentage.

4. Grain quality: V1 is better/same/worse in quality than the local variety? 5. Market price: V1 is higher/same/lower in price than the local variety?

6. Overall preference: V1 is better/same/worse overall than the local variety?

Add any other information that the farmer thinks is important. This could include reaction to pests and diseases; particular quality traits such as taste, cooking quality, milling quality; and whether the farmer mentions that he or she has given seed to others. Comments on the local variety versus Variety 1

Page 4. Only for Design 2 (two test entries). Compare the variety withthe local variety.

The answers in the boxes on this page determine if Variety 2 is better, about the same, or worse than thelocal variety and ranks all 3 entries

1. Establishment: Variety 2 is better/same/worse for establishment than the local variety? 2. Flowering time (tasselling):

Variety 2 is earlier/same/later to flower than the local variety?

If flowering time (tasselling) is different then by how many days ? Variety 2 is days earlier than local (Either ‘days earlier’ or Variety 2 is days later than local days later’ as applicable.)

3. Yield: Variety 2 yields more/same/less than the local variety?

If the yield is different provide more information in the box below. Give the farmer’s description e.g. in local units, as a proportion (e.g. half as much again), or a percentage.

4. Grain quality: V2 is better/same/worse in quality than the local variety? 5. Market price: V2 is higher/same/lower in price than the local variety?

6. Overall preference: V2 is better/same/worse overall than the local variety?

Add any other information that the farmer thinks is important. This could include reaction to pests and diseases; particular quality traits such as taste, cooking quality, milling quality; and whether the farmer mentions that he or she has given seed to others.

Comments on the local variety versus Variety 2

7. Rank of three varieties: Rank V1, V2 and local variety in order of preference:

Rank Variety name 1 (Best) 2 3 (Worst)