Participatory Research Methods for Technology Evaluation Bellon Cymmit prm_al… · Participatory...

Participatory Research

Methods for Technology

Evaluation:

A Manual for Scientists Workingwith Farmers

Mauricio R. Bellon

CIMMYT® (www.cimmyt.cgiar.org) is aninternationally funded, non-profit scientific researchand training organization. Headquartered in Mexico,

the Center works with agricultural research institutions worldwide to improve theproductivity, profitability, and sustainability of maize and wheat systems for poor farmers in

developing countries. It is one of 16 food and environmental organizationknown at the Future Harvest Centers. The centers, located around theworld, conduct research in partnership with farmers, scientists, and

policymakers to help alleviate poverty and increase food security while protecting naturalresources. They are principally funded through the nearly 60 countries, private foundations,and regional and international organizations that make up the Consultative Group onInternational Agricultural Research (CGIAR) (www.cgiar.org). Financial support forCIMMYT’s research agenda also comes from many other sources, including foundations,development banks, and public and private agencies.

International Maize and Wheat Improvement Center (CIMMYT) 2001. All rights reserved.The opinions expressed in this publication are sole responsibility of the authors. Thedesignations employed in the presentation of material in this publication do not imply theexpressions of any opinion whatsoever on the part of CIMMYT or contributory organizationsconcerning the legal status of any country, territory, city, or area, or of its authorities, orconcerning the delimitation of its frontiers or boundaries. CIMMYT encourages fair use of thismaterial. Proper citation is requested.

Correct citation: Bellon, M.R. 2001. Participatory Research Methods for Technology Evaluation: AManual for Scientists Working with Farmers. Mexico, D.F.: CIMMYT.

Abstract: This manual presents methods that enable agricultural scientist and farmers toevaluate technologies/practices jointly. The methods are specifically designed forparticipatory research on germplasm and soil fertility technologies, and they are illustratedwith actual examples from three research projects. The manual begins by reviewingconceptual issues that are important in participatory research and presents information toassist researchers in selecting research sites and fieldwork participants. Next, the manualdescribes the rationale and associated methods for each major activity in farmer participatoryresearch: diagnosing farmers’ conditions, evaluating current and new technologies/practices,and assessing their impact. Goals, procedures, advantages, and limitations of each method areoutlined. The manual also presents detailed information on analyzing data gathered throughparticipatory methods, discusses differences between gathering data through participatorymethods and more traditional structured farm surveys, and offers examples, based on fieldexperience, of the choices and strategies involved in applying these methods.

ISBN: 970-648-066-8

AGRIS descriptors: Agricultural development; innovation adoption; technology transfer;evaluation; research methods; research projects; germplasm; soil fertility; crop management;Zimbabwe; Mexico.

AGRIS category codes: E14 Development Economics and PoliciesU30 Research Methods

Additional Keywords: CIMMYT; genetic resources; farmer participatory research

Dewey decimal classification: 338.16

Printed in Mexico

ContentsPage

iv Tablesvi Figures

vii Forewordviii Preface

1 An Introduction to Farmer Participatory Research2 Farmers’ Local Knowledge4 Farmers’ Experiments6 Farmers’ Exchange of Information and Technologies

8 An Overview of the Projects Used as Examples in This Manual8 The Oaxaca Project: Conserving Maize Diversity

11 The Chihota Project: Improving Soil Fertility14 The Chiapas Project: Linking Farmers’ Local Knowledge and Crop Management Decisions14 A Structure for a Participatory Research Project and Some Caveats

16 Participation: Identifying the Places, People, and Procedures for Research16 Where to Work: Site Selection18 Who to Work With: The Selection of Participants (Informants/Experimenters)21 How to Interact: Types of Interviews/Interactions22 Gender

24 Diagnosis of Farmers’ Conditions25 Local Classification of Farmers29 Wealth Ranking33 Minimum Set of Socioeconomic Indicators36 Calendar of Activities37 Local Taxonomies of Soils41 Local Classifications of Climate43 Local Crop Taxonomies46 Identifying Points of Intervention

49 Evaluation of Current and New Technological Options50 Eliciting Farmers’ Perceptions of Technological Options54 Comparing Different Technological Options64 Eliciting the Constraints on Using a Technology66 Demonstration Fields and Field Days69 Carrying Out Experiments with Farmers

73 Assessing the Impact of New Technologies73 The Complexity of Assessing Impacts73 The Impact Assessment Process

78 Conclusions

79 References

81 Appendix 1. Farmers’ Classification of Themselves, Chihota, Zimbabwe84 Appendix 2. Examples of the Cards Used to Depict Variety Characteristics, Oaxaca

Project (demand and supply of characteristics)85 Appendix 3. Examples of the Data Used for Analyzing the Supply and Demand of

Characteristics89 Appendix 4. Using an Attainment Index in Farmer Participatory Research93 Appendix 5. An Example of the Modified Stability Analysis

Page

5 Table 1. Levels of interaction between farmers and scientists and possibleoutputs

26 Table 2. Data collected in an exercise to elicit farmers’ classification ofthemselves, Chihota, Zimbabwe

27 Table 3. Farmers’ classification of themselves and their characteristics,Chihota, Zimbabwe

32 Table 4. Comparison of farmer characteristics by wealth rank, Chiapas,Mexico

34 Table 5. Field day participants in Oaxaca, Mexico, characterized byagricultural activity, gender, and other variables

35 Table 6. Selected personal and household characteristics of participants infield days and sample survey, Oaxaca, Mexico

40 Table 7. Farmers’ soil taxonomy, Chihota, Zimbabwe41 Table 8. Soil chemical properties by farmer soil class, Chiapas, Mexico42 Table 9. Underlying factors defining “good” and “bad” seasons according

to farmers, Chihota, Zimbabwe44 Table 10. Maize types and their characteristics in Santa Ana Zegache,

Oaxaca, Mexico51 Table 11. Perceived advantages and disadvantages of maize types,

Oaxaca, Mexico52 Table 12. Characteristics and criteria used to judge maize types,

Oaxaca, Mexico53 Table 13. Perceived advantages and disadvantages of soil fertility

improvement technologies, Chihota, Zimbabwe53 Table 14. Characteristics and criteria used to judge soil fertility

improvement technologies, Chihota, Zimbabwe58 Table 15. Average ratings of importance of maize characteristics by males

and females, Santa Ana Zegache, Oaxaca, Mexico60 Table 16. Average ratings of importance of maize characteristics by wealth

rank for males and females, Santa Ana Zegache, Oaxaca, Mexico62 Table 17. Average rating of the performance of different maize types, for

several characteristics of importance to male and female farmers,Santa Ana Zegache, Oaxaca, Mexico

65 Table 18. Technological options available to farmers in Chihota, Zimbabweto improve their soils, and the constraints they face, by localsoil type

Tables

Page

76 Table 19. Impact indicators identified by farmers and scientists in aparticipatory research project, Oaxaca, Mexico

86 Table A3.1. Ratings of importance for each characteristic for men (demand ofcharacteristics), Santa Ana Zegache, Oaxaca, Mexico

87 Table A3.2. Ratings of importance for each characteristic (demand ofcharacteristics) for women, Santa Ana Zegache, Oaxaca, Mexico

88 Table A3.3. Ratings of performance of each maize type for each farmer withrespect to each characteristic (supply of characteristics), Santa AnaZegache, Oaxaca, Mexico

91 Table A4.1. Demand and supply ratings for several characteristics and twomaize types grown by the man in household 4 used forcalculating an attainment index, Santa Ana Zegache, Oaxaca,Mexico

v

Figures

Page

17 Figure 1. Hypothetical matrix to classify villages18 Figure 2. Classification of survey sites by source of income, ethnicity, and

maize potential29 Figure 3. Causal diagram of the factors that affect yields based on those

identified by farmers’ classification of themselves in Chihota,Zimbabwe

38 Figure 4. An example of a calendar of activities, Santa Ana Zegache, Oaxaca,Mexico

45 Figure 5. Classification of maize types in Vicente Guerrero, Chiapas, Mexico56 Figure 6. Hypothetical example of cards rating the importance of maize

characteristics56 Figure 7. Example of a card layout to rate characteristics68 Figure 8. Layout of a demonstration field, Oaxaca Project69 Figure 9. Layout of a demonstration field with two factors, Chihota Project90 Figure A4.1. Matrix of scores for an attainment index94 Figure A5.1. Yield response to the environmental index in six communities of

the Central Valleys of Oaxaca, Mexico

Foreword

This manual on farmer participatory research continues a tradition in the CIMMYTEconomics Program of developing practical, instructive guides that are based ondirect experience in field research. The methods presented here have been tested andrevised in rural communities over the course of many years, and they lend themselvesto fieldwork in a wide range of settings.

I am pleased that the Economics Program’s evolving experience in working withfarmers can be made available to a wider audience through this manual. Althoughthese methods may not be suited to every situation that researchers are likely toencounter—after all, each rural community, household, farmer, and researcher isdifferent—I believe that readers will certainly benefit from the advice and experiencedistilled here, just as we shall benefit from their recommendations after they haveused this manual in their own fieldwork.

It is important for readers to understand that this publication does not pretend to offerthe final word in farmer participatory experimentation. Participatory researchmethods will continue to develop as researchers and farmers continue to learn fromeach other. For the present, however, Mauricio Bellon has given us a valuable guide tothe insights as well as the uncertainties that agricultural scientists often experience asthey seek to make the research process more inclusive—and ultimately morerewarding—for all who participate.

PRABHU L. PINGALI

Director, CIMMYT Economics Program

vii

This manual presents a set of methods for agricultural scientists and farmers toevaluate technologies/practices jointly. It is intended for agricultural scientists whowork on the development, adaptation, or diffusion of agricultural technology andwant to incorporate a participatory approach in their work. It focuses specifically onmethods that can be applied to germplasm and soil fertility technologies.

The manual describes how to collect, analyze, and use information for participatoryresearch. The user of this manual should pick and choose the relevant methodologiesfor his/her work rather than launching into some pre-determined scheme. Themethods are presented under three main activities in farmer participatory research:diagnosing farmers’ conditions, evaluating current and new technologies andpractices, and assessing the impact of new technologies and practices. Ideally theseactivities should fit into a coherent plan for developing technology, rather than beingone-off exercises.

The methods presented here are illustrated with examples from three researchprojects. The first project involves participatory conservation and improvement ofmaize landraces in the Central Valleys of Oaxaca, Mexico. The second concernsparticipatory evaluation of soil fertility improvement technologies in Chihota,Zimbabwe. The third project is a more conventional study in a community of centralChiapas, Mexico, where participatory methodologies were used to understand therelationship between farmers’ local knowledge of maize diversity and soils and theircrop management decisions.

The manual begins with an introduction to participatory research, especially to someof the conceptual issues that are important in this kind of research. This introductorysection is followed by an overview of the three projects used as examples throughoutthe manual so that the reader understands the context of the examples. Next, threecentral concerns of participatory research are explored: Where should this kind ofresearch be undertaken? Who should participate? How should the participants worktogether?

The next sections of the manual describe the participatory methods associated withthe three main activities mentioned above: diagnosing farmers’ conditions, evaluatingcurrent and new technologies/practices, and assessing their impact. First, therationale behind each participatory research activity is given. (For example, why doresearchers need to conduct a participatory diagnosis of farmers’ conditions?)Afterward, the methods corresponding to each activity are explained. The goal of eachmethod is outlined and the procedures are described. Methods are illustrated withexamples from the projects mentioned earlier. Occasionally the examples present the

Preface

work of other researchers and, very occasionally, consist of hypothetical situations.Comments on each method, such as a discussion of its limitations and advice for itsapplication, are presented as well.

This manual is not a comprehensive exposition of all methods available for farmerparticipatory research. It deals with the methods that I and my colleagues haveexperience in using. The strength of this approach is that the manual can providesound examples of how the methods were applied, including their advantages andlimitations in a variety of situations.

I wish to acknowledge the participants and funding agencies that supported the threeprojects used as examples in this manual. The project in Oaxaca, formally named “CGMaize Diversity Conservation: A Farmer-Scientist Collaborative Approach,” has beenimplemented jointly by the International Maize and Wheat Improvement Center(CIMMYT) and Mexico’s National Institute of Forestry, Agriculture, and LivestockResearch (INIFAP), under a grant from the International Development ResearchCentre (IDRC), Ottawa, Canada. The author gratefully acknowledges the work of theproject team (Melinda Smale, José Alfonso Aguirre Gómez, Julien Berthaud, SuketoshiTaba, Flavio Aragón, Irma Manuel Rosas, and Jorge Mendoza). The project in Chihota,formally named “Chihota Soil Fertility Project,” has been implemented jointly byCIMMYT and by Zimbabwe’s Department of Agricultural, Technical and ExtensionServices (Agritex) and Department of Research and Specialist Services (DR&SS). Theproject in Chihota is one activity of the Soil Fertility Management and Policy Networkfor Maize-Based Farming Systems (Soil Fert Net), funded by the RockefellerFoundation. The list of participants is too long for all to be acknowledged, but I wishto recognize the work of Stephen R. Waddington, Peter Gambara, Tendai Gatsi,Timothy E. Machemedze, Christine Kuwaza, Johannes Karigwindi, PhilipTawuyandago, and Obert Maminimini. Finally, the project in Chiapas was funded by agrant from CIMMYT and implemented jointly by the author and Jean Risopoulos.

I hope that researchers who are interested in using a participatory approach in theirwork find this manual useful, and I would appreciate any suggestions on how toimprove it. Finally, I would like to express my appreciation to the farmers andresearchers who, over the years, have contributed to the teaching and learningexperiences distilled in the pages that follow. I particularly wish to thank José AlfonsoAguirre Gómez for sharing his ideas on farmer experimentation, Angel Pita andXóchitl Juárez for providing one of the examples used, Stephen R. Waddington,Malcolm Blackie, Robert Tripp, Jeffrey B. Bentley, Michael Morris, and JanetLauderdale for comments on earlier drafts, Prabhu Pingali for his encouragement towrite this document, Kelly Cassaday for editorial assistance, Miguel Mellado fordesign, and Marcelo Ortiz for production.

MAURICIO R. BELLON

CIMMYT, December 2000

ix

1

Farmer participation in agriculturalresearch is more than talking to sixfarmers or putting ten experiments intheir fields. Above all, it is a systematicdialogue between farmers and scientiststo solve problems related to agricultureand ultimately increase the impact ofagricultural research. By respondingclosely to farmers’ concerns andconditions, researchers can developtechnologies that are adopted morewidely and that respond to importantsocial issues such as equity andsustainability.

Developing this dialogue betweenfarmers and scientists is not as simple asone might think, because farmers andscientists have different needs, worldviews, knowledge systems, methods, andtools. When it is successful, dialoguebetween farmers and scientists can leadto more productive, stable, equitable, andsustainable agricultural systems.Achieving this goal should be good forfarmers, because it enhances theirwelfare; for scientists, because it increasestheir job efficiency; and for society ingeneral, because it adds to the foodsupply and encourages the conservationof natural resources for futuregenerations.

Farmer participatory research has beendefined as “the collaboration of farmersand scientists in agricultural research anddevelopment” (Bentley 1994:140). Theneed to improve our understanding offarmers’ conditions and incorporate theirperspectives into the development andtesting of new agricultural technology isnot new. The current interest in farmerparticipation is related in large part tofarming systems research (FSR) (Tripp1989). The FSR perspective recognizesthat most small farms are an integrationof multiple enterprises that require themanagement of diverse householdresources to meet a range of subsistence,income, and community goals. A farmingsystems perspective also implies acommitment to include farmers’ criteriaand goals when setting research priorities(Tripp and Woolley 1990).

What, then, is new about farmerparticipatory research (FPR)methodologies? How do they differ fromthe FSR approach? It is useful at thispoint to consider the four approaches tofarmer participation described by Biggs(1989:3):

• Contractual: Scientists contract withfarmers to provide land or services.

• Consultative: Scientists consult farmersabout their problems and then developsolutions.

An Introduction toFarmer ParticipatoryResearch

2• Collaborative: Scientists and farmers

collaborate as partners in the researchprocess.

• Collegiate: Scientists work to strengthenfarmers’ informal research anddevelopment systems in rural areas.

Farming systems research has focusedmainly on the first two approaches,whereas FPR stresses the third and, to alesser extent, the fourth. Furthermore,FPR emphasizes three aspects of farmerparticipation, which are recognized byFSR but not given such importance:

1) Most farmers have extensive, well-developed knowledge of theirenvironment, crops, and croppingpractices.

2) Many farmers carry out experiments ontheir own and generate innovations.

3) Farmers actively exchange informationand technologies.

A short review of each of these aspects ofFPR follows.

Farmers’ LocalKnowledgeAs anyone who has worked closely withfarmers knows, they possess knowledgeabout their crops, their farmingenvironment, and their socioeconomicconditions. In many instances they canclearly articulate the rationale behindtheir management practices and theirdecisions. This knowledge, which hasbeen documented formally by numeroussocial and biological scientists, includestheir soils and productive environments(Bellon and Taylor 1993; Kamangira 1997;Edwards 1987), their crops and cropvarieties (Richards 1986; Sperling et al.1993), insects and pests (Bentley 1992;Bentley and Rodriguez 2001), and soiland water management practices (Wilken1987; Lamers and Feil 1995).

Understanding this knowledge is afundamental step towards generating adialogue between farmers and scientists.It is a key reference point that farmers useto make decisions and to communicateamong themselves. Scientists need tounderstand farmers’ knowledge if theywant to contribute to farmers’ welfare byproviding new information to them, bydeveloping appropriate technologies withthem, or communicating effectively withthem. To understand farmers’knowledge, scientists must first elicit andanalyze it.

Farmers’ knowledge can be classified intothree categories: perceptions, taxonomies,and rules of thumb. Distinct methods arerequired to elicit and analyze thesedifferent kinds of knowledge.

Perceptions are mental images obtainedthrough the senses. Perceptions may ormay not be shared widely among a groupof farmers. In some cases, they can beidiosyncratic, be particular to anindividual, and bear little or no relationto the perceptions of other members of agroup. In other cases, they may be widelyshared and agreed upon.

For our purposes, farmers’ perceptionsabout alternative technologies are veryimportant, particularly the characteristicsthey identify to assess whethertechnologies are appropriate for them.This assessment of whether a technologyis appropriate does not necessarilyconsist of an absolute “yes” or “no”answer. It usually consists of a ranking oftechnologies from more to lessappropriate. Knowing how to elicit theseperceptions, translate them into criteriafor evaluating a technology, and use themto rank alternative technologies isimportant for working with farmers todevelop and assess agriculturaltechnologies.

3

Taxonomies are the abstraction ofperceptions into categories with namesand defined properties. Taxonomies areorganized in a hierarchical fashion. Theyare usually widely shared, and a givenpopulation will show a high degree ofagreement about them. Among farmers,the most widely studied taxonomies areassociated with soils. For example,Kamangira (1997:43) reports that farmersin the Songani catchment area of Malawihave ten local soil classes, mainlyreferring to soil texture and color.Kamangira also demonstrates howfarmers’ soil knowledge can becombined with scientific views about soilclasses. Sandor and Furbee (1996) showhow soil classes are organized in ataxonomic tree, and they comparefarmers’ local knowledge with soilphysiochemical properties and theirscientific classification.

Rules of thumb are logical propositionsthat relate two events in a cause-and-effect relationship: “If this occurs (or if Ido this), then that happens.” These rulesmay or may not be widely shared oragreed upon within a group of farmers.

In many cases, rules of thumb relatetaxonomies to specific behaviors. Afarmer may think that if a modern maizevariety is not weeded early in the season,its yield will decrease significantly, buthe or she may not believe this to be thecase for a traditional variety (e.g., Bellon1991). The farmer may therefore have therule of thumb that if he/she can ensureaccess to sufficient labor early on, he/sheshould plant a modern variety;otherwise, a traditional one should beplanted.

The elicitation of rules of thumb andtheir organization into behavioraldecision models for the adoption of

specific technologies has been developedby social scientists; see, for example,Gladwin (1979) for timing of fertilizerapplication and Franzel (1984) for theadoption of an improved maize variety.These methods are particularly complex,however, and they can be timeconsuming.

Farmers’ knowledge should not bedismissed or, conversely, idealized. Asmentioned previously, farmers knowmany things about farming and theirconditions, but there are many otheraspects of farming that they do not knowor misunderstand. Farmers’ knowledgeis well developed for phenomena thatcan be observed readily and for relativelystraightforward cause-and-effectrelationships. Their knowledge of soilsand potential productivity is usually welldeveloped, as is their knowledge ofweeds and their impact on cropdevelopment. For phenomena that aredifficult to observe or that have multipleand sometimes interacting causal factors,farmers’ knowledge is often less precise,or incorrect, or non-existent. Forexample, farmers’ knowledge of pestsand diseases is usually deficient or non-existent. Smallholders lack themicroscopes or sophisticated equipmentthat would allow them to make finer ordeeper observations beyond the capacityof the naked eye, and they also lack thebasic scientific concepts, such asknowledge of microorganisms orgenetics, that would allow them tointerpret many of their observations(Bentley 1994). Furthermore, farmers’knowledge may be inadequate in thepresence of extremely rapid technicalchange, since farmers may not haveenough experience with a technology tohave understood all its dimensions.

4One should not assume that farmers’knowledge leads to specific behaviors orvice versa, just as knowing that smokingis harmful certainly does not deter manypeople from smoking. As anthropologistsfrequently find, actual behavior deviatesfrom and often contradicts the culturalrules of appropriate behavior (Johnson1974). People often make unarticulateddecisions that, though subconscious, havea definite impact on their behavior(Gladwin and Murtaugh 1980). Inparticipatory research, we are particularlyinterested in how knowledge affectsbehavior and how behavior affectsknowledge. For this reason, it is notenough simply to elicit and appreciatefarmers’ knowledge; we also need to linkthat knowledge to specific behaviors andvice versa. When interacting withfarmers, scientists should always askthemselves, “If what they are telling us istrue, what should we expect to see in theirbehavior?” and, if possible, probe for it.This attitude of scientists towards farmersshould not be interpreted as arrogant anddistrustful but rather should be seen as adesire to understand farmers better.Understanding evolves from testingperceptions and expectations againstreality. Scientists should also keep inmind that many farmers may have asimilar attitude towards scientists.

Finally, it should be pointed out thatfarmers’ knowledge is dynamic. Farmersincorporate new information andconcepts from extension, schools, inputsuppliers, the media, and others into theirknowledge base and abandon otherknowledge. They are particularly likely tocreate new categories or terms that reflectchanges springing from newly adoptedtechnologies. The response of varieties tonew inputs such as fertilizers orherbicides may generate local conceptssuch as “sturdy” and “delicate” varieties:sturdy varieties can withstand delays in

weeding and/or fertilizer applicationwithout a substantial decrease in yield,whereas delicate ones cannot (Bellon1991). In some cases, knowledge thatproved correct under earliercircumstances may now lead to poordecisions. For example, fire is a commonmeans of managing crop residues in thetropics. Fire is essential in swiddenagriculture, and it may not be harmfulprovided that fallow periods are longenough to allow natural vegetation toregenerate and restore lost nutrients.Nevertheless, in an intensive tropicalagricultural system its use is questionableat best and disastrous at worst. In thesesystems, using fire to recycle nutrientsoften results in their depletion; althoughnutrients become readily available, theefficiency of nutrient release is low (Lal1987). It is important to identify suchknowledge and try to modify it, althoughthis may be difficult.

Farmers’ ExperimentsThe fact that small-scale farmers in thedeveloping world conduct experimentson their own is well documented(Johnson 1972; Richards 1986) and hasbecome a pillar of farmer participatoryresearch (e.g., Ashby et al. 1995; Buckles1993). Farmers’ experiments areimportant because they promoteknowledge and evaluation of new andunproven technologies withoutjeopardizing farmers’ livelihoods orscarce resources. These experiments maybe farmers’ basis for generating andadapting new technological options thatfit their specific needs and conditions.

Farmers conduct different types ofexperiments (Rhodes and Bebbington1988; Scoones et al. 1996):

• curiosity experiments—just to see whathappens;

5

• problem-solving experiments—toaddress a specific problem they face;

• adaptation experiments—to adapt newtechnologies to known environments orknown technologies to newenvironments; and

• fortuitous experiments—chance eventsthat lead to changes in practices, which inturn lead to a new learning experience.

In the farmer-scientist interaction, theadaptation and problem-solvingexperiments are, respectively, the mostrelevant. The most common experimentsinvolve comparing a new variety with afamiliar one by planting a few rows ofthe former next to the latter (adaptationexperiment). Scoones et al. (1996) reportan example of a problem-solvingexperiment in which a farmer testedvarious planting strategies to improvesunflower germination.

Farmers’ and scientists’ experimentsoften differ (Bentley 1994; Perales 1993).Three key differences are:

1. Farmers’ experiments commonly lack acontrol treatment.1 Scoones et al.(1996:135) say that the farmer may carrythe control “in the head.”

2. In the fields where farmers’ experimentsare located, many factors may bemodified simultaneously, or extraneousfactors may not be controlled for.

3. Farmers usually do not replicate anexperiment, although it is often said thatthey do so over time. For example, theymay compare the current season’s resultswith those of previous seasons.

From a scientist’s point of view, thesecharacteristics make farmers’experiments hard to analyze andinterpret. As mentioned previously, themain source of data and evidence forfarmers is their own observations; theyusually lack instruments to observe suchphenomena as nematodes or lackconceptual tools such as statistics toisolate one event from another. Thisdifference serves to emphasize the pointthat many of scientists’ experiments maynot make sense to farmers, who probablylack the instruments and conceptualbackground to employ the scientificmethod.

Farmers and scientists can have differentdegrees of interaction or involvement inthe design, management, and analysis ofexperiments. Different degrees ofinvolvement are appropriate foraccomplishing different objectives(Table 1). On one end of the continuum,the experiment is located in the farmer’sfield, but the scientist designs, manages,and analyzes it. This strategy may beeffective for developing a basic

Table 1. Levels of interaction between farmers and scientists and possible outputs

Degree of interaction

Scientist Farmer Possible output for which interaction is appropriate

Designs, manages, analyzes Provides the field An understanding of processes, components of new technologyunder farmers’ biophysical conditions

Designs, analyzes Manages, provides input An understanding of processes, components of new technologyinto the analysis under farmers’ biophysical conditions and their management

Designs, manages, analyzes Designs, manages, analyzes Joint evaluation and modification of a new technology

Training, guidelines, technical Designs, manages, analyzes Capacity building, empowermentsupport

1 In many cases farmers conduct simple experiments, however, varying one factor at a time and comparing the results with their normalpractice. These experiments are easier to interpret and comparable to those made by scientists.

6understanding of processes or of thecomponents of a new technology in thebiophysical conditions where farmersoperate. Further along the continuum,the experiment is located in the farmer’sfield; the scientist designs it and analyzesthe results, but the farmer manages itand provides the scientist with someinput to interpret the results. This formof experimentation brings the farmer’sactual management into consideration.

Even further along the continuum, thefarmer and scientist jointly design,manage, and analyze the experiment.This approach may be particularly usefulfor jointly evaluating a new technologyand modifying it. Finally, at the otherend of the continuum, the farmerdesigns, manages, and analyzes theexperiment, which is located in one of hisor her own fields. The scientist helps toimprove the farmer’s experimentalmethodology through training, providessome basic guidelines, and, in the earlystages, offers technical support.Eventually, however, the farmerperforms these tasks completelyindependently. This approach isappropriate for building capacity,empowering farmers, and fostering aprocess that perhaps can continuewithout the scientist’s long-terminvolvement. We will return to thesethemes later in the discussion ofguidelines for farmer-scientistexperiments.

Farmers’ Exchange ofInformation andTechnologiesFarmers constantly share informationabout things that are important to them.These exchanges have been particularlywell documented for seeds of different

crops and varieties (Cromwell 1990;Sperling and Loevinsohn 1993). Manyinnovations have spread from farmer tofarmer without the intervention of anyformal agricultural extension services,such as the diffusion of the moldboardplow in many parts of Africa (M. Blackie,personal communication) or thediffusion of velvetbean (Mucuna spp.) inMesoamerica (Buckles 1995). Farmersexchange information as well. Forexample, farmers in Portugal exchangeda great deal of information through theirlaborers, who worked for differentfarmers and shared ideas such as usingsilage cutting machines (Bentley,personal communication).

Information and technology commonlyare diffused through a social network,which can be defined as a group ofpeople who share certain bonds, usuallyas a result of family or traditional socialobligations. Social networks may play afundamental role in the adoption of newtechnologies, particularly if they requirecollective action, such as constructingcontour dikes for soil erosion and watercontrol, which cannot be accomplishedby a single individual (Smale and Ruttan1997). Social networks also affect theflow of farmers’ own experimentalinformation. For example, the propensityof rice farmers in Sierra Leone to discussnew rice varieties (Richards 1986)contrasts with the concern of manyGhanaian farmers that excessive interestin a neighbor’s farming activities may belinked to witchcraft (Tripp 2000).

Exchanges of information andtechnology may extend beyond thesenetworks to include casual contactsmade through travel, migration, and off-farm labor. Social barriers to theseexchanges also exist, however; socialnetworks may include only members of

7

one village, ethnic group, or social class.Diffusion of information and innovationsoutside the network may be difficult, andthe network itself in some cases may actas a barrier rather than a conduit. Forexample, seed flows may take placemainly within a village, with few flowsoccurring between villages (Smale et al.1999). Another interesting case isreported by Scoones et al. (1996), whopoint out that fear of witchcraft or of

generating envy from others may promptfarmers to conceal their knowledge andinnovations.

It is important to keep in mind thatfarmers are not isolated individuals butmembers of social networks, and thatthese networks can play an importantrole in the diffusion, or lack thereof, ofinformation and technology.

8

Throughout this manual, experiencesdrawn from three participatory researchprojects provide examples of how themethods presented here have beenapplied in the field. The projects haveinvolved smallholder farmingcommunities in transition andconcentrated on issues related to maizegermplasm management and/or soilfertility. As of the writing of this manual,all of the projects are at different stagesof completion. An overview of eachproject is presented here to give thereader an understanding of the variouscontexts from which the examples aretaken. The specific methods used in eachproject and described in this manualare italicized.

The Oaxaca Project:Conserving MaizeDiversityIn the Central Valleys of Oaxaca, Mexico,a project has been undertaken todetermine whether it is possible toimprove maize productivity whilemaintaining or enhancing geneticdiversity. (“Maize productivity” isdefined broadly in terms of yield, yieldstability, and other characteristics thatinterest farmers.) The project developsand compares participatory interventions

An Overview of theProjects Used asExamples in This Manual

with small-scale farmers in sixcommunities in the Central Valleys.Through the project, farmers gain accessto the diversity of maize landraces in theregion, are trained in seed selection andmanagement techniques, and learnprinciples to assist them in maintainingthe characteristics of landraces they value.

Researchers selected the Central Valleysfor this project for a number of reasons.One of the most important reasons is thatfarmers in the region have a longtradition of cultivating maize and havemaintained the diversity of theirlandraces to the present. These landraceshave considerable value for agriculturebeyond the Central Valleys, because theyhave contributed to the development ofimproved, drought-tolerant maizecultivars that are popular elsewhere inMexico. Modern maize varieties bred byresearchers have had an almostnegligible impact in the Central Valleys,and while their virtual absence may ormay not have helped to conserve maizediversity in the region, it is a signal thatscientific research has not providedfarmers with new agricultural options.

The region is also ethnically diverse andagroecologically heterogeneous, anddespite economic changes in recentyears, Central Valley communitiescontinue to place a recognizable

9

emphasis on their indigenous culture,including culinary practices for maize.There is no guarantee that farmers willremain interested in maintaining thediversity of their maize cultivars,however, so it is important to startexploring options for supporting thisinterest, including scientific research thatresponds closely to farmers’ interests,needs, and constraints.

The Oaxaca Project is divided into threecomponents: 1) diagnosis, 2) thedevelopment and evaluation ofinterventions, and 3) impact assessment.

The diagnosis comprised severalactivities that made use of participatoryresearch methodologies. As a startingpoint, researchers collected samples ofmaize landraces that were thought torepresent the spectrum of maizediversity in the Central Valleys.Landraces were collected in 15communities that scientists chose fortheir range of agroecological andsocioeconomic conditions and ethnic andcultural multiplicity. The researcherswere also guided by some priorknowledge of the distribution of maizediversity. In each community, thescientists relied on eliciting the local croptaxonomy from a set of key informants toidentify the diversity of landracespresent and locate farmers who werewilling to donate samples. Although alack of funding prevented in-depthparticipatory research from beingconducted in these 15 communities, a siteselection exercise was done to choose asubset of six communities where most ofthe research would take place.

To assess the heterogeneity of farminghouseholds in the six communities andgain a clearer understanding of theirgoals, resources, and constraints, as wellas the spatial and temporal variability

that affected their agriculture, a set ofparticipatory methodologies wasemployed, mainly based on focus groupdiscussions and key informants. Thesemethodologies included the elicitation ofthe local soil taxonomy, local crop taxonomy,local classification of farmers, localclassification of climate, and wealth ranking.

Additionally, a baseline survey with arandom sample of 40 households percommunity was done to obtain arepresentative sample of households inthe communities. This sample wouldserve as a control group for checking orcomparing information obtained throughparticipatory methods; it would alsomake it possible to perform the impactassessment when the project ended. Thebaseline survey included a systematicevaluation of the characteristics farmersconsidered important (derived from thelocal crop taxonomy) in maize landracesand how those characteristics weredistributed among the landraces theygrew (demand and supply of characteristics).

To evaluate the agronomic performanceand morphological diversity of thecollected landraces (information that wasparticularly important to the scientists),trials including all of the landraces wereestablished in the 15 communities wherethey were collected. The trials wereplanted in farmers’ fields but managedby scientists (contractual approach). Sixfield days were organized so farmerscould view three of the trials: three fielddays were held when the maize plantsreached physiological maturity, andthree at harvest. At each field day,farmers were invited to see the landracesand to “vote” for the ones they liked. Thefarmers walked through the trial andrecorded the numbers of all of the plotscontaining the landraces they liked.Researchers viewed the participants’

10choices as “votes” and assumed that thehigher the percentage of farmers votingfor a maize landrace, the morepotentially valuable it was to them. Thepurpose of this exercise was to obtain a“quick and dirty” sorting of the maizesamples into a gradient of farmers’interest. A minimum set of socioeconomicindicators was collected from participantsso that researchers would have someidea of who participated in the fielddays. Based on the data from theagronomic evaluation and farmers’votes, 16 landraces and one improvedvariety were chosen for the secondcomponent of the project—the“interventions” component.

An important issue in this kind ofresearch is how to move from thediagnosis to the selection of specificinterventions. In the Central Valleys, thediagnosis showed that farmers valuedmany characteristics in their maizelandraces, especially traits related toconsumption. The field days, whichshowed the diversity of maize collectedin the region, drew much attention andparticipation from farmers, and thevoting exercise suggested that there wasno “best” or “ideal” variety. Instead,farmers appeared to want a range ofvarieties (i.e., a range of diversity).Although the collection of local landracesencompassed many different maizetypes, farmers actually planted only amean of 1.6 varieties per household, andresearchers concluded that farmerswanted access to diversity. They learnedwhich specific traits farmers valued mostin a maize variety: it tolerated drought,resisted insects in storage, and produced“something” even in bad years. Giventhe resources available to the project,none of these traits were easily amenablefor breeding interventions, but theycould be addressed through practices,

such as improved storage and seedselection practices. The diagnosisshowed that current storage and seedselection practices were not meetingfarmers’ needs and that training couldplay an important role in modifyingfarmers’ current practices. The trainingwas based on understanding farmers’knowledge about these issues and tryingto provide general principles thatfarmers themselves could use, followingBentley’s ideas about the interactionbetween local and scientific knowledge(Bentley 1994).

The interventions consisted of givingfarmers in the six communities access tothe diversity of maize landraces presentin the region (the 17 materials selected inthe field days), training them in seedselection and management techniques,and teaching principles to help themmaintain the characteristics of landracesthey valued. These interventions wereavailable to anybody who wanted toparticipate, and open invitations andpublicity encouraged farmers toparticipate. Researchers used thisapproach because they were interested inunderstanding who participates, theincentives for participation, who benefitsfrom participation, and how they benefit.

To give participants access to thediversity of maize landraces,demonstration plots were established inthe six communities and more field dayswere organized. During the field days,participants saw the plants and ears ofthe maize landraces being offered andreceived information on theirperformance in the field. After visitingthe demonstration plots, farmers couldpurchase seed of any material theywanted. The idea of giving access to thisdiversity was to facilitate farmerexperimentation with the landraces. With

11

a subset of farmers who were skepticalbut also highly motivated, researchersestablished a set of farmers’ experiments.

To train and teach farmers, five trainingsessions were offered in theircommunities, starting with a discussionof their knowledge about maizereproduction and perceptions of maizeimprovement. Additional sessions taughtbasic principles of maize reproduction,principles of seed selection in the fieldand in the household (including hands-on exercises in the field), and principlesand techniques for storing seed and grain.

The third component of the project,impact assessment, includes the baselinesurvey (described earlier) and themonitoring of a sample of farmers whoparticipated in each intervention.Monitoring consists of systematic, yearlyinterviews with this sample of farmers;the interviews cover their participationand their perceptions of the advantagesand disadvantages of their participation.A set of impact indicators was alsoestablished by scientists andparticipating farmers. To assess thedistribution of participants and impactsby socioeconomic status, a wealth rankingwas done for all participants.

To date, results of the project indicatethat participating farmers in the studyarea demand access to diversity,particularly to relatively scarce maizetypes. Farmers value many differentcharacteristics in their maize landraces,especially traits related to consumption.Among women, colored maize types andrarer types are in particular demand, anddiversity is enhanced when thesepreferences are taken into account. Thesubset of maize types jointly selected byfarmers and scientists for distributionwas a success. In the project’s first year

(1999), 804 kg of seed were sold in 197purchase events (a farmer purchasingseed of a landrace), with a total of 123farmers purchasing seed (the samefarmer could purchase seed of morethan one landrace). The trainingactivities showed that participatingfarmers often did not understand certainaspects of maize reproduction, but oncethis knowledge was provided, at leastsome of them were keen to try newmanagement techniques. Farmers whoparticipated in the joint experimentsverified that the “experimental” maizetypes worked well under theircircumstances, and some of the typeswere considered to be even better thantheir own landraces, used as controls inthe experiments.

The Chihota Project:Improving Soil FertilityIn Zimbabwe, the Chihota Soil FertilityProject seeks to expose farmers to a setof technologies for improving soilfertility and to gain farmers’ assessmentof those technologies in their own terms.Based on this assessment, projectparticipants are identifying the potentialfor farmers to adopt each technologyand the constraints that could impedeadoption. Participants are also identifyingany modifications required in thetechnologies or in institutionalconditions (i.e., market circumstances,policies) that could diminish oreliminate those constraints. The soilfertility technologies being assessed inChihota were developed by a networkof agricultural scientists in Zimbabweand Malawi (the Soil FertilityManagement and Policy Network forMaize-Based Farming Systems, alsoknown as Soil Fert Net).

12Infertile soils are a major constraint tofood production in Southern Africa,particularly in the communal areas ofZimbabwe, where smallholders with fewresources rely on agriculture to survive.For these farm households, thedevelopment and adoption of newtechnologies to enhance soil fertility arean important means of improvingfood security.

Chihota, a communal area in subhumidnortheastern Zimbabwe, was selected asthe site of this pilot project because it hasvery infertile soils, maize is the mostimportant crop, and the governmentagricultural extension service (theDepartment of Agricultural, Technical,and Extension Services, known asAgritex) has an active presence in thearea. Chihota is located in MaronderaDistrict, Mashonaland East Province, andhas nine wards, each with five or sixvillages. Contrasting conditions prevail inChihota with regard to farmers’experience with soil fertilitytechnologies: farmers in some of thewards have been exposed to soil fertilityresearch, but farmers in other wardshave not.

Like the Oaxaca Project, the ChihotaProject has three components:1) diagnosis, 2) interventions, and3) impact assessment.

The diagnosis component comprisedseveral activities in which participatoryresearch methodologies were used. Toassess the heterogeneity of farminghouseholds in Chihota and gain a clearerunderstanding of their goals, resources,and constraints, as well as the spatial andtemporal variability that affected theiragriculture, a set of participatorymethodologies was employed. Fourwards were selected for the diagnosis;soil fertility research had been conducted

only in two of them. In each ward, focusgroup discussions were organized withfarmers working closely with Agritex(altogether, ten focus groupsparticipated). The group discussionswere used to elicit the local soil taxonomy,local classification of farmers, and localclassification of climate. Theseclassifications were used as a frameworkfor discussing and identifying thetechnological options available toimprove soil conditions and theconstraints to their use (elicitingconstraints on using a technology).Research collected a minimum set ofsocioeconomic indicators from allparticipants to gain an idea of who theparticipants were.

Additionally, a baseline survey with arandom sample of 258 households wasdone to obtain a representative sample ofall nine wards in Chihota. The surveywas designed specifically to addressmany of the issues identified in theparticipatory diagnosis, particularly thetype and amount of knowledge thatfarmers have about soil improvementpractices. The survey helped researchersenhance their understanding of farmers’problems and perceptions. The sampleserves as a control group for checking orcomparing information obtained throughparticipatory methods; it also makes itpossible to perform the impact assessmentwhen the project ends. The baselinesurvey included a systematic evaluationof farmers’ knowledge of different soilimprovement technologies.

As noted, an important issue in this kindof research is how to move from thediagnosis to the selection of specificinterventions. The Chihota diagnosisrevealed that farmers were concernedabout many issues related to soilimprovement technologies and

13

suggested that knowledge of suchtechnologies was a particularlyimportant issue for farmers. Farmersneeded to be exposed to the technologiesand learn more about them, so theinterventions focused on enablingfarmers to try soil improvementtechnologies under their owncircumstances, using their own criteria.

The implementation component of theChihota Project consisted of organizingmany demonstration plots with farmers intheir fields and of organizing field days togenerate discussion and feedback amongfarmers and scientists.

The demonstration plots were set up andmanaged by groups of farmers in theircommunities in association with anAgritex extension worker. The plots werenot only a demonstration but played therole of farmer experiments so thatparticipating farmers could assess thetechnologies, which were:

• lime in combination with fertilizer;

• velvet bean (Mucuna pruriens) andsunnhemp (Crotalaria sp.), used as agreen manure sole crop or intercrop withmaize; and

• cereal legume rotations.

These technologies were chosen from alarger set of potential interventions bymatching probable solutions fromprevious on-farm soil fertility research tothe problems that Chihota farmersidentified.

During the middle and end of the maizegrowing season, field days wereorganized. At the field days, farmersfrom the communities where thedemonstrations were established visitedthem and discussed the pros and cons ofthe technologies with the farmers incharge of the demonstration plots.Agritex officers and scientists also

participated in the discussions. Thediscussions were documented to providefeedback to scientists. An importantfocus of the discussions was to identifythe criteria (in other words, thecharacteristics) that farmers used tojudge the technologies and tounderstand how farmers assessed thetechnologies (eliciting farmers’ perceptionsof technologies). A small, individualsurvey was done to quantify theperceptions of 85 farmers who belongedto the groups that helped conduct thedemonstrations.

The impact assessment component of theChihota Project remains to beimplemented, except for the baselinesurvey. The impact assessment will entailmonitoring a sample of farmers whoparticipated in demonstration plots, whoattended field days, and who did notparticipate at all. These farmers will besystematically interviewed about theirparticipation, their perceptions of theadvantages and disadvantages of theirparticipation, and their perceptions ofthe advantages and disadvantages of thetechnologies. (The feedback exercise heldduring the field days was also a form ofmonitoring.) A set of impact indicatorswill also be established by scientists andparticipating farmers.

To date, results of the Chihota Projectindicate that farmers who haveevaluated the soil fertility improvementtechnologies regard them very positively.However, farmers perceive that pooraccess to inputs and a lack of specializedknowledge are the most bindingconstraints to adopting the technologies.This finding suggests that a fundamentalstep toward promoting adoption of thetechnologies would be to developmechanisms for providing knowledgeand inputs. As knowledge and inputconstraints loosen, labor and land

14constraints may become more important.Given farmers’ limited ability to generatesurpluses (and cash) from farming, andgiven the alternative uses of thosesurpluses, there is a need to understandhow farmers can finance the adoption ofthe technologies. Poor availability andaccessibility of implements may also be aconstraining factor that establishes theupper ceiling to adoption.

The Chiapas Project:Linking Farmers’ LocalKnowledge and CropManagement DecisionsIn central Chiapas, Mexico, the ChiapasProject aimed to understand therelationship between farmers’ localknowledge of maize varieties and soiltypes and their crop managementdecisions, including decisions aboutwhich varieties to plant, where to plantthem, and how to manage them in termsof soil preparation, fertilizer use, andweeding. This project, unlike the twoprojects discussed previously, includedno intervention and therefore no impactassessment per se. Many participatorydiagnostic methodologies wereemployed, however, and the project hadan important emphasis on eliciting andunderstanding farmers’ local knowledge.

This fieldwork for the project wasconducted during two periods, 1988-89and 1998. Key informants wereinterviewed to elicit the local crop (maize)and soil taxonomies. Focus groups alsodiscussed the taxonomies and how theywere related (advantages and disadvantagesof different soil types and maize varieties,what variety to plant in which type ofsoil, and so on). A questionnaire wasapplied to a random sample of farmers inboth periods. In the second period, the

questionnaire included a systematicevaluation of the characteristics thatfarmers considered important (derivedfrom the local crop taxonomy) in theirmaize and sought information on howthose characteristics were distributedamong the maize varieties they planted(demand and supply of characteristics). Allfarmers in the sample were classifiedusing a wealth ranking methodology. Soilsamples were collected (based on thelocal soil taxonomy), and samples of earsfor each maize variety (based on the localmaize taxonomy) were collected as well.

The Chiapas Project had severalimportant results related to the use ofparticipatory methodologies:

• the local soil taxonomy reflected objectivesoil properties;

• the wealth ranking reflected statisticallysignificant differences in the possessionof assets and sources of income amongthe wealth classes;

• the local soil taxonomy and the wealthranking helped explain which specificmaize varieties were planted and wherethey were planted; and

• farmers modified improved maizevarieties to suit their needs better.

Many of these findings and methods willbe discussed in later sections of thismanual.

A Structure for aParticipatory ResearchProject and SomeCaveatsThe projects in Oaxaca and Chihota sharea similar structure based on threecomponents. First, in the diagnosticcomponent, scientists identify theconditions in which farmers operate,particularly from the farmers’ own point

15

of view and relative to their ownknowledge system. Second, based on thediagnosis, farmers and scientists identifya set of interventions and put them intopractice. Third, through the diagnosisand interventions, an impact assessmentcomponent is built into the project toassess any changes that farminghouseholds perceive to have resultedfrom the interventions. This descriptionof the three-component structure shouldnot lead readers to construe that theimplementation of a participatoryresearch project is a linear process,however. It is just described in a linearmanner here for clarity of exposition.During the intervention, or even duringthe impact assessment component of aproject, new understanding can begenerated and interventions can bemodified or changed. For example, in theOaxaca Project, joint experiments withfarmers were not originally planned, butthey were incorporated as researchersperceived farmers’ skepticism and triedto address it. In the Chihota Project, thelayout of demonstrations with farmerswas modified as researchers learned thatthe original layouts were too complexand lacked some controls for simpleinterpretation, a lesson that isincorporated in this manual.

It is important to point out that in thethree projects described earlier, theobjectives were established by scientistsbased on their own assessments of theneed to conduct research on particularissues, such as the improvement andconservation of maize genetic diversityor the development of new soil fertilityimprovement technologies. Thosespecific objectives were set because theywere important to strategic research and

not necessarily because they metimportant needs expressed byparticipating farmers. Through thechoice of location and dialogue withfarmers, however, it became clear thatthe objectives of the projects were also ofgreat interest to farmers. Aside from thespecific benefits they held forparticipating farmers, the projects had acommon interest in drawing lessons thatwould be widely applicable to otherplaces and other farmers. In someinstances, the issues addressed in theproject may not appear to be of directimportance to farmers (for example, theassessment of different strategies forconserving genetic resources in Oaxaca).These issues and their relatedinterventions did have to be explored ina real context, however, and the challengefor scientists is to find commonalitieswith farmers and make these issuesimportant and interesting to them as well.

There are other approaches and ways ofdoing participatory research. Theapproach presented here is not the onlyone and not necessarily the best one forall situations, which is why this manualexplicitly outlines the context(exemplified by the three projects) inwhich researchers and farmers have usedthe methods described in this manual.Some purists of participatory work mayconsider this approach too “top down”because it does not start from a specificassessment of the needs of specificfarmers or households. Although most ofthe methods described here can be usedin other contexts, many users of thismanual will be operating undercircumstances similar to those of theprojects in Oaxaca, Chihota, andChiapas.

16

Participation: Identifying thePlaces, People, andProcedures for Research

Three key decisions for a scientist usinga participatory approach are decidingwhere to work (in other words, selectinga site), who to work with (whoparticipates?), and how to work withthem. These decisions dependfundamentally on what researchers, inconjunction with farmers, are trying toachieve (i.e., the research objectives).These decisions are critical because thescientists will rely on the selectedpersons to provide information aboutproblems, resources, and constraints; toelicit local knowledge effectively; and tocollaborate in conducting experiments.The selection of the site for fieldworkwill to a great extent define thecomparisons and lessons that can bedrawn, and it will influence whetherthey are local comparisons and lessonsor may be generalized to other regionsor conditions. The method ofinteraction between scientists andfarmers will delineate the types ofanalysis that can be performed, becausethe interaction will define the degree ofaggregation of the data.

Farmers and their households, evenwhen they are part of the samecommunity, are not homogeneous. Byfailing to recognize the differencesamong farmers, scientists may end upworking with a small subset of farmers,

unaware of how they relate to the rest ofthe farming population in the study area.Working with a subset of farmers is notnecessarily incorrect, but ignoring theirrelationship to the rest of the communitycan lead to erroneous generalizationsand limit the scope of research and itsresults. For example, working only withfarmers who own cattle, who can applymanure to their fields, and who can useox-drawn implements may result in thedevelopment of technologies that areirrelevant for farmers who do notown cattle.

Where to Work: SiteSelectionThe first step in deciding which farmersto work with is deciding where to work.In many cases this decision is pre-ordained for administrative, political, orlogistical reasons. It may be possible,however, to choose villages orcommunities within a given region andthus select sites with particularcharacteristics that can enable theresearcher to make generalizations fromthe results. The key is to select sites tomaximize the possibility of meaningfulcomparisons based on a few keyexogenous factors that are hypothesizedto influence farmers’ conditions and/or

17

decisions. The choice of these factorsmay vary according to the specifics ofthe country, the region, the farmers, thetechnologies of interest, and othervariables, but the choice is usually basedon researchers’ prior understanding ofthe specific situation.

There are limits, however, to the numberof factors that can be consideredexplicitly (usually no more than three).Within a region, for example, villageswith contrasting infrastructures (accessto markets) and population sizes can beselected. These two variables areimportant because they influence accessto information, access to inputs, and theavailability of land, labor, and capital.For example, population size withrespect to available agricultural landplays a key role in the intensification ofagricultural production. Theagroecological environment, such asareas with contrasting soils or rainfallpatterns, is another major variable.

All of these important exogenousconditions influence farmers’ decisions,and scientists may want to know theirrelative importance while maintainingother factors constant. Scientists maythink, for example, that the adoption ofgreen manures is more attractive tofarmers located in isolated areas (withless access to purchased inputs andfewer opportunities for off-farm labor)where population density is increasing(in other words, fallows are becomingshorter and more labor is available). Bylocating research sites in areas with thesecharacteristics, researchers can test thesehypotheses. Furthermore, discussionswith farmers in such areas can confirmor dispute the hypotheses.

Villages in the region to be studied canbe classified into a matrix2 (Figure 1)through consultation with local experts(local officials, scientists, or extensionworkers). Another option for siteselection is to use secondary informationif it is available, including previousstudies, older diagnostic reports, or acensus. If the number of villages is nottoo large, yet another option is toconduct a short survey with the localauthorities, focusing on villagecharacteristics such as population,infrastructure (schools, electricity, roads,stores), sources of income, animals,and crops.

By locating research in villages withcontrasting conditions, it may be possibleto assess the impact of different factorswhile maintaining the others constant.For example, it may be possible to assessthe importance of the availability offamily labor and land versus theavailability of purchased inputs and paidlabor in the adoption of green manures.The village selection process can bethought of as a quasi-experimental

Figure 1. Hypothetical matrix to classify villages.

Population density of a region

Low High

Mar

ket i

nteg

ratio

n

High

Low

2 Obviously, the specific matrix may vary from one situation to another according to the specific exogenous factors selected. A matrixsuch as this was used by Pingali et al. (1987) to locate the sites for their study of mechanization in Africa.

18design that ensures enough variation inthe sample to make meaningfulcomparisons. Often cells in the matrixmay be void, which indicates thatexogenous factors are correlated (e.g.,villages with high population densitieshave good infrastructure and viceversa). In this case, although the effectof population density cannot bedisentangled from that ofinfrastructure, at least we know thatthis is the case.

Example: For the Oaxaca Project,researchers had to decide where towork in the Central Valleys. The regionencompassed many villages andthousands of people. Though maizelandraces were collected from 15communities, a smaller subset ofcommunities had to be selected becausethe project lacked resources to covereven this limited number. Researchersconsulted local authorities in eachcommunity to gain an idea of itsgeneral socioeconomic characteristics.These authorities estimated the numberof households in each community, themajor sources of income, supplies ofinfrastructure and transportation, andtypes of markets.

Little variation was apparent betweencommunities in distance to markets orbasic physical infrastructure. Localauthorities were then asked to classify aset of different sources of income (i.e.,crop production, animal husbandry, off-farm labor—agricultural and non-agricultural—and remittances fromwithin and outside Mexico) into threecategories according to theirimportance to the village economy (i.e.,very important, moderately important,and not important). An analysis of thisclassification showed pronounced

differences in the extent to which thevillages relied on non-farm income andremittance income from migrants. Thisinformation was combined with data onethnicity, derived from census data, andon maize yield potential, derived fromprevious work by the nationalagricultural research organization. The15 communities were located in a matrixof these variables, and six communitiesrepresenting contrasting circumstanceswere selected (Figure 2). The horizontalaxis in Figure 2 represents increasingdependence on local sources of income(local agricultural and off-farm labor)versus non-local sources of income(remittances from within and outsideMexico). The vertical axis representslocation in zones of increasing maizeyield potential, which also correspond toa gradient of rainfall (from low to high).

Who to Work With: TheSelection of Participants(Informants/Experimenters)In participatory research we alwayswork with informants and experimenters.The informants are farmers, understoodin the broadest sense as all members of a

Figure 2. Classification of survey sites by source ofincome, ethnicity, and maize potential.Source: Smale et al. (1999).

Note: * > 30% indigenous population.

Increasing dependence on local vs. non-localsources of income

High

Low

Valdeflores San Lorenzo HuitzoAlbarradas*

Amatengo Santa Ana Santo TomásZegache* MazaltepecM

aize

yiel

d po

tent

ial

19

farming household, whom scientistsquery about their knowledge, practices,needs, priorities, and resources. Theexperimenters are the farmers withwhom scientists perform experimentsand evaluations. The central question ishow to select these informants and/orexperimenters. (Note that usually anexperimenter is first an informant butthat not all informants becomeexperimenters.) The content and qualityof the information gathered, and theexperimental results obtained byscientists and farmers together, dependfundamentally on the people scientistswork with and therefore on how theyselect informants and experimenters.

Participants can be categorized into atleast four types:

1) Incidental: Persons that researchersencounter who are willing to talk tothem, without any a priori effort on theresearchers’ part to identify them.

2) Key: Persons researchers select based onwell-defined, pre-established criteria.Key participants are usually selected withthe help of local contacts who know thecommunities of interest well. Thesecontacts include local authorities,extension workers, health workers,teachers, and secular and/or religiousleaders.

3) Randomly selected: Persons who arechosen following statistical samplingprocedures.

4) Self-selected: Persons who volunteer toparticipate.

Incidental participants are usually easy tofind; an incidental participant can be afarmer who gets a lift from a scientist orthe owner of a store where researchersbuy supplies. The information collectedfrom such people should be treated withcaution, since researchers do not knowwho these people are in the context ofthe community (which socioeconomic,

political, or religious group they belongto), what interests they represent, orwhat biases they may have. Incidentalparticipants can provide a starting pointfor scientists’ interactions within acommunity, however, and they may givescientists an initial set of hypothesesabout the local farmers and community.

Key participants are selectedsystematically. They should have certainwell-defined characteristics that provideeither an idea of the variation within acommunity or information about aparticular group. Selection criteria couldinclude:

• farmers who plant many crop varieties,

• farmers who have a reputation for goodworkmanship or for having aninquisitive turn of mind,

• young or old farmers,

• male or female farmers, or

• farmers with large or small landholdings.

These criteria are defined by the type ofinformation scientists seek. Criteria maybe established to avoid or at leastdiminish biases (that is, to avoid focusingon one group and ignoring others), or,when different communities are beingcompared, to ensure that informants areas similar as possible and thereforecomparable. To focus on one group is notnecessarily wrong, but to generalize fromone group to others may be. Clearly, theprocess of selecting key informantsdepends on other informants(researchers’ contacts in the community),but by establishing criteria researchersminimize their contacts’ ability to choosewhoever they want, without researchers’knowledge.

20Each time scientists arrive in acommunity or contact a group offarmers, the scientists must notify andpossibly obtain authorization from localauthorities, such as the chief/headmanor leaders of farmer groups. Frequentlyresearchers already know useful peoplewho are familiar with the communityand its members, although theythemselves may be outsiders, includingextension workers, health workers, orteachers. These contacts are a primarysource of information for identifying keyinformants.

Randomly selected participants provide thebest perspective on a community offarmers in terms of theirrepresentativness. The probability ofincluding all the subgroups that mayexist depends on how common they are,not necessarily on the views of particularinformants, and random selection canhelp minimize biases. The informationcollected from these informants can beanalyzed statistically, allowing us tomake inferences with a defined level ofprobabilistic confidence about thefarmers with whom we work. However,when a research project is directed at aparticular group of people with specificcharacteristics, this selection method maynot be the best or most cost-effective,because many people of no interest to theresearch objectives may be included.

Statistical sampling procedures also haveproblems, however. Before the samplecan be drawn, ideally a census of thetarget community or communities mustbe conducted, but a census may notalways be feasible. The census can bedone using lists of farmers orhouseholds, compiled for otherpurposes, or by mapping all of acommunity’s dwellings. If lists of farmers

or households exist, it is important tonote that they may be biased. They mayfocus exclusively on a specific groupwithin the community, such as farmerswith irrigation, or farmers who growcash crops, or farmers who participate ingovernment programs. By combiningdifferent independently compiled lists,however, scientists can produce acomprehensive list. If the community ismapped, it is still possible to misspeople, particularly in sparselypopulated areas. Even thoughgenerating lists or maps may require alot of time and money, it can produceaccurate and comprehensiveinformation. It is also possible that therandomly selected case sometimes turnsinto the self-selected case (especially inmethods that require more than a briefmeeting or interview) because of drop-outs and refusals.

Self-selected participants are usuallyhighly motivated and may perceiveadvantages in participating, such aslearning new techniques and gettingaccess to new technologies. Theirmotivation may make them easier towork with, but researchers should becareful not to assume that they knowtheir motivations. These people maychoose to participate because they expecta political favor, whereas researchersthink they are interested in acquiringnew information. As usual, scientistsshould ensure that participants’expectations are explicit and that falseimpressions are not created. It isessential to know who these farmers arein the context of the community (i.e.,which socioeconomic, political, orreligious group they belong to andtherefore which interests they representor which biases they may have).

21

How to Interact:Types of Interviews/InteractionsOnce the informants/participants havebeen identified, two forms of interviewsor interactions can take place: individualor group exchanges. The individualinteraction consists of a one-to-oneinteraction between the interviewer andthe informant, while the groupinteraction brings together a group ofinformants, and the interviewerprovides a series of questions or topicsof discussion. In an individualinteraction, the person with whomscientists interact is well defined andhis/her characteristics (age, education,household resources, and so forth) caneasily be established. The outcomes ofthe interaction can be related to thesecharacteristics in a relativelystraightforward manner. If manyindividual interactions take place,researchers can relate the variability ofoutcomes more specifically to thediversity of individuals participatingand their conditions. In a group settingthis is much more difficult to do,because it is harder to disaggregate thespecific relationships between outputsand participants. A group settingprovides a broader and morecomprehensive perspective on theissues, however, and allows agreementsand disagreements among individualsto be identified relatively rapidly.Individual interactions are relativelymore suitable for generating ananalysis, whereas group interactions arerelatively more suitable for generating asynthesis, although results of each typeof interaction can be used for analysis aswell synthesis.

With respect to practical guidelines forthe individual interaction, researchersshould be sure that the informantunderstands the questions being asked.Researchers should be careful to usephrases, words, and examples that theinformant readily understands.(Providing examples also enhancesunderstanding.) The questions shouldbe pre-tested for vocabulary andcontent and modified accordingly.

Some common problems with theseinterviews should be avoided. Friendsand family members are frequentlypresent during an interview,volunteering information or answeringinstead of the informant. In thissituation, researchers have no controlover—or background information on—the people providing the information,which later complicates itsinterpretation. It is the informant’sinformation that researchers want. Inmany cultures, when a woman isinterviewed in the presence of herhusband, son, or father, she may beinhibited to answer questions freely, orthe men may answer for her. Again, thissituation should be avoided, becausethe information of interest is hers, and itshould be as truthful and open aspossible. It is particularly important toget women’s unhampered point ofview, since researchers will want toavoid gender biases in the informationthey collect.

In group interviews, it is important tolimit the number of questions. This typeof interview is excellent for generatinginventories of things or issues (e.g., soilor crop types, problems, activities, andtechnologies) or for generatingdiscussions among participants. In thelatter case, however, scientists should

22be careful not to impose a falseconsensus by forcing participants toagree on something when they find itdifficult to do so. It may be unrealistic toexpect consensus on many issues if thegroup is truly heterogeneous. Instead,the interviewer should aim to identifythe points of agreement anddisagreement among informants,especially the disagreements, which areof great value because they allow theinterviewer to probe into the informants’differences. It is very important to try toestablish the basis for the disagreementsand to relate them to specificcharacteristics of the informants (e.g.,poor versus wealthy, young versus old,men versus women). Backgroundinformation on the informants cantherefore be extremely useful; recall thatsuch information should have beencollected when the informants wereselected. Another point to bear in mindwith group interviews is that sometimesthey provide information on whatparticipants think “should be” ratherthan on what “actually is.” Researchersshould be careful in interpreting theresults and should probe to establishwhether the group is referring to an idealrather than an actual situation.

Like individual interviews, groupinterviews have some problems thatshould be avoided. Often a fewinformants tend to dominate thediscussions. They may be of highersocial status or belong to a certain ethnicor politically dominant group, and theycan give a biased view of the issues,while the perspectives of other groupmembers are completely ignored. Toavoid this situation, ask the quietermembers of the group for their opinions.In many cases, they will not respondfreely, since they may feel intimidated by

the dominant members. If necessary, theinterviewer should talk to themindividually or separately. Theinterviewer can also split the group intodominant and quieter members andrepeat the group interview separately.Distinguishing among informants in agroup is particularly important when oneis trying to rank problems or solutions.Different groups within a communitymay have different problems andsolutions or attach distinct levels ofimportance to them.

GenderAny participatory research methodologyshould consider the importance ofgender. From a practical point of view,this means that researchers should besure to include participants who playdifferent roles within households, such aswomen, children, spouses, parents, andfemale heads of households. This alsomeans paying special attention tointeractions among household members.Depending on where the research isbeing done, it may be necessary to formsame-sex groups (i.e., groups of only menor only women), since in mixed groupswomen may not participate at all. Inother contexts, however, mixed groupsmay provide an excellent opportunity toelicit gender differences and concerns.Even in individual interactions it may benecessary for men to interview or interactonly with men, and for women to interactonly with women.

In the past, agricultural research focusedmainly on male farmers and assumedthat all household members shared thesame goals, had the same access toresources and outputs, and faced similarconstraints. Now it is clear that in mostcases this view is incorrect. Just as

23

differences between farmers andhouseholds may be attributed todifferences in access to resources,knowledge, and information, differenceswithin households also exist and may beattributed to different factors. Householdmembers may have diverseresponsibilities, perform differentactivities, and have varying work loadsand access to resources. They may alsohave conflicting interests. Thesedifferences can be particularly striking inAfrica,3 where household organizationcan be extremely complex (for example,with polygamy or with members of the

3 Doss (1999) presents an excellent review and discussion of gender and agricultural technology issues for Africa.

same sex in a household there may behierarchies—the first wife, second wife,the mother in-law, and so on). Regardlessof where the research is beingundertaken, however, genderconsiderations are always important andrelevant. Researchers must also becareful to go beyond a simple concernwith females or female-headedhouseholds and to look carefully at theway household members are organizedand interact.

24

Diagnosis ofFarmers’ Conditions

Farmer participatory research involvesmore than identifying researchparticipants, of course. It also involvesidentifying “users” or “clients.” Theseare the farmers whom researchers wishto reach with technologies and practices,and they are not necessarily participantsin the research (either informants orexperimenters).

Scientists may think that “all farmers arethe same” or that they are working with“typical” or “representative” farmers,but unless researchers systematicallyaddress this issue from the start, theymay be making a critical mistake. Asdiscussed earlier, farmers and theirhouseholds often are not homogeneous,even within a community. Farminghouseholds in a community have accessto different resources. Some have moreland, labor, or capital than others.Knowledge and information are notshared equally, either. Therefore, goals,resources, and constraints differ betweenfarming households. Variability—spatialand temporal—is another fact of life forevery farmer and his/her household.Soils and topography vary and seasonschange. Because this variabilityinfluences what farmers can and wish todo, it is fundamental that researchersunderstand how resources and constraintsare distributed in time and space.

In failing to recognize differencesbetween farmers and households,researchers may overestimate the

potential impact of technologies orpractices, because researchers may endup working with a smaller and possiblyunrepresentative subset of the farmersthey hope to serve, or they may have avery static view of farmers’ resourcesand/or constraints. In other words,researchers run the risk of developingtechnologies that are adopted by a morerestricted number of farmers thanexpected or desired, resulting in a lowerimpact than anticipated. It is crucial toidentify and characterize groups offarmers who share similar goals,resources, and constraints in theirsocioeconomic and the biophysicalenvironment, because these farmers willshare similar problems and requirecomparable solutions (technologies/practices).

Many methods have been developed todescribe and analyze socioeconomic andbiophysical variability, but in aparticipatory approach our goal is todiscover how the farmer and his/herhousehold view this variability. Methodsfor achieving this goal for thesocioeconomic environment includefarmers’ own classification of farmers,wealth ranking, a minimum set ofsocioeconomic indicators, and a calendarof activities. Methods for understandingfarmers’ view of variability in thebiophysical environment include localclassifications of soils and climate. Eachof these methods is described in thesections that follow.

25

Local Classification ofFarmers

Goal: Identify the socioeconomiccategories and characteristics that farmersfind relevant.

Rationale: Farmers have their owncategories for classifying themselves. Byeliciting these categories and theirstrengths and weaknesses, researchersshould be able understand what isimportant about these farmers in an openbut systematic way, without imposingtheir own views on the farmers. Thisinformation can be used to generatehypotheses about how farmers’conditions and technologies interact,identify factors that affect technologyadoption, and define groups facingsimilar conditions regarding technologicalneeds or constraints (e.g.,recommendation domains).

Method: Researchers assemble a group ofinformants from a community, ideally amixture of people of different ages,resources, and genders. The interviewerexplains the objective of the exercise to theparticipants: researchers want to gain abetter understanding of which types offarmer exist in their community, includingthe strengths and weaknesses of eachtype. The interviewer should also explainthat this information will help researchersto understand farmers’ problems, developpossible solutions, and guide them, asscientists, to interact better with farmers.

The interviewer poses the question: Whattypes of farmers are present in yourcommunity?

The group makes a list of each of the types.

For each type, the interviewer asks thefollowing questions:

What are the characteristics of this type offarmer? (In some cases these are obviousfrom the name of the category, but inothers they may have to be described ingreater detail.)

What are the strengths of this type of farmer?(In many cases, strengths can beinterpreted as the resources available.)

What are the weaknesses of this type offarmer? (In many cases, weaknesses canbe interpreted as the constraints faced.)

Table 2 shows the types of data that canbe gathered using this method. It isimportant to identify responses that referto the same concept, since people mayexpress their ideas in different forms.This requires some judgement on thepart of the scientist, but usually it is notdifficult. The farmer types usually referto the presence, absence, or extent of anattribute, such as the ownership or lackof an asset (e.g., owns cattle, does notown cattle, owns a few cattle but not alot). The number of types can be verylarge, and some of them are likely to becorrelated. For example, one type may be“farmers with cattle” and another“farmers with manure.” Clearly, farmerswho have cattle are also likely to havemanure.

Implicit in the farmer types are “themes”or wider categories, which make itpossible to group different types within alarger theme or category. These themesor categories become the basis foranalyzing the classification. They tellresearchers which factors farmersconsider important and in many caseshow the factors are related. These factorsand their relationships can be used togroup farmers in homogeneous groupsand/or to generate hypotheses abouthow these factors influence farmers’decision-making (see example).

26

Example: This method was used in thediagnosis component of the ChihotaProject to assess and understand theheterogeneity of farming households andto identify some of the socioeconomicvariables underlying this heterogeneity.Agritex extension officers organizedfocus group discussions with farmersworking closely with them. There werethree types of groups: male, female,and mixed, for a total of ten groups.

The groups identified 29 farmer types.This number may seem excessive and thetypes ad hoc, but analysis of the typesrevealed that they could be grouped intoeight themes or categories. (The data inTable 2 came out of that exercise.) Table 3shows the types grouped by category.Some of the types refer to personalcharacteristics such as age and sex. Mostinvolve the ownership or lack of an asset

such as cattle, or access to income orknowledge. The themes or categoriesrefer to common socioeconomic variablessuch as age, gender, wealth, and access toinputs and knowledge. Although manyof the results presented below may seemobvious, the reader should bear in mindthat there was no a priori reason why thisshould have been so, and that thisinformation was collected in only fourdays of fieldwork. For somebody notfamiliar with the system, this informationmay be very valuable to provide a firstset of hypotheses about thesocioeconomic factors that are important.At least it can serve as a check thatfarmers also attach importance to factorsthat scientists believe are important.

Based on the strengths and weaknesses4

associated with each farmer type, thefollowing picture emerged from Chihota

Table 2. Data collected in an exercise to elicit farmers’ classification of themselves, Chihota, Zimbabwe

Farmer type Strengths Weaknesses

Farmers who plan Farm operations done on time Crops can be eaten by livestockGood crop stands Crops wilt if rains come late

Farmers who do not plan – Extensive farmersNo rotationsLack resources

Farmers with cattle Have manure Do not have grazing areasHave resources

Farmers without cattle Borrow in time Delayed farming operationsProvide labor for others Lack resources, lazy at times

Farmers with manure Crop stands are good and yields high –

Farmers without manure – Crops are of poor quality and thereforeyields are low

Field farmers Plan well ahead of time Seasonal farmerHigh volume of output for storage Take risks because production is seasonal

Garden farmers Stable income because production is perennial Do not help the needy

Resource-rich farmers Sell produce to others Do not give implements for freeStable incomeFarm operations done on time

Source: Gambara et al. (1998).

4 Presented at length in Appendix 1. This appendix provides a good sense of the “raw” data collected in this type of exercise.

27

farmers: Age is a category associatedwith the ownership of assets, access tofamily labor, and knowledge. In general,younger farmers are considered worseoff than older farmers. Gender isassociated with control over labor, assets,and income. Male farmers are in control.Not surprisingly, there seems to betension between male and femalefarmers. For example, females considerthat they are not rewarded for their laborand that their fields are prepared last.

The ownership of assets in general islinked with the timing of farmingoperations, the ease of performing them,and the crop yield achieved. Owners areconsidered to perform operations ontime and easily, and therefore to gethigher yields than non-owners. Aparticularly important asset is theownership of gardens. Gardens werementioned in very positive terms. Theyprovide a stable income and are lesssubject to drought compared to drylands, where income is more seasonal,less stable, and production is moreexposed to drought. The size oflandholding is another interesting case.Farmers consider that farmers owninglarger fields tend to spread inputs thinly,while those with smaller fieldsconcentrate inputs. Cultivating as largean area as possible is a practice that hasbeen observed in marginal environmentsin Africa, and it may be a riskmanagement strategy or a means toestablish or maintain property rightsover the land.

Labor allocation refers to a process bywhich farmers with skills to workelsewhere substitute hired local labor fortheir own labor, which highlights theincreased integration of these farmersinto the market economy. Another aspectof labor is organized labor; farmersworking in a group cooperate by sharinglabor as well as knowledge, and they canbuy inputs together. Working in a groupmay be more common among farmerswho work closely with extension, sinceextension staff often favor grouparrangements.

A puzzling classification is the one thatidentifies farmers as “lazy” or“industrious.” It is not clear whether“lazy” farmers are truly lazy or if theyare classified this way because they are

Table 3. Farmers’ classification of themselves andtheir characteristics, Chihota, Zimbabwe

Numberof groups

Socioeconomic mentioningcategory Farmer type type

Age Young 3Old 3

Gender Male 3Female 3

Ownership of, Draft animals 3access to inputs Cattle 3

Manure 1Implements 4Garden 6Dry lands 6Large fields 1Small fields 1Own fields 1Fenced fields 1

Labor allocation Works outside the area 1Works in groups 2Works individually 2Industrious 4Lazy 4

Access to cash, wealth Adequate cash for farming 3Rich 2Poor 2

Knowledge Has knowledge 5Has Master Farmer Certificate 1

Linkage to market Sells produce 1Farms for subsistence 1

Synthetic (combines Performs operations on time 2different categories) Attains high yields 1

Plans operations 1

Source: Bellon et al. (1999).

28poor or sick. For example, participantsrecognized that “lazy” farmers were agood source of labor for others, whichraises the question of why these farmersare working for others, if they are solazy. Puzzling results such as this may bethe product of rapid research, and alonger stay and interaction with farmersmay reveal the factors that explain thepuzzle. At least two hypotheses aboutthese types of farmers can be considered.One is that there are lazy people in anysociety, and these farmers are indeedlazy; the second is that the farmersparticipating in the group classificationexercise are of a higher social status andconsider people of lower status to belazy, even if clearly they are not, sincethey work for them.

Access to cash is linked with the timingof farm operations and with the ability topurchase inputs and hire labor. Thosewith access to cash were considered to bein a better position that those without it.

Farmers who possess knowledge areviewed very positively. The groupsprovided a long list of strengths for thosewho have knowledge and a long list ofweaknesses for those who do not.Knowledge is associated with timelyoperations, high yields, and croprotations. The emphasis on knowledgemay also be related to the fact thatalmost all participants work with theextension service. Therefore they valueaccess to knowledge and have beenexposed to the message that knowledgeis important.

Linkage to the market captures thedifferences between those who sell theirproduce and those who are subsistencefarmers. This distinction may not beabsolute, since it is most likely that manyfarmers produce crops for sale as well assubsistence.

Finally, three types appear again andagain, frequently together, as attributesthroughout the farmer classification:timely performance of farmingoperations, high yield of crops, andplanning of operations. These attributesare highly correlated. As farmers see it,the ownership of assets, access to cash,and possession of knowledge lead togood planning and timely operations,which in turn lead to high yields.

The classification provides researcherswith a set of variables that can be used togroup farmers in homogeneous groups:by age, gender, ownership of assets,labor allocation strategy, and access toknowledge. For example, the mostcontrasting groups can be seen as1) young females with few assets, whodo not work off of the farm and havepoor access to knowledge, and 2) oldermales with many assets, who work off ofthe farm and have good access toknowledge. Obviously these groups mayhave different goals and resources, facedistinct constraints, and require differenttechnologies.

The classification also providesresearchers with a set of hypothesesabout the problems that these farmersface and their possible causes. It shouldbe pointed out, however, that in manycases these classifications provideresearchers with associations betweenfactors and not necessarily with relationsof causality, which researchers have todeduce. For example, the followinghypotheses derived from the examplecan be postulated:

• Female farmers get low yields becausetheir fields are plowed late by malefarmers who control the oxen and theimplements.

• Male farmers who own oxen get higheryields because they perform operationson time.

29

• Farmers who own cattle get higher yieldsbecause they have access to manure toapply to their crops.

• Farmers working in groups get higheryields because they gain better access toinputs by pooling their resources.

These hypotheses can be expressed in acausal diagram that provides a model ofhow different factors interact (Figure 3).This figure illustrates the relationshipbetween factors identified in theclassification, particularly in relationshipwith the timing of the performance ofagricultural operations and yields.

Comments: The types elicited fromfarmers may, in some cases, be self-serving and value-laden. For example, itis not clear whether the qualities oflaziness and industriousness refer totruly personal characteristics, describe aposition within a social hierarchy, orrepresent a value judgment by one groupof people regarding others. Ininterpreting the data, researchers should

always be careful to recognize the implicitvalue judgments and the social relationspresent in these types.

Another example of how informants’judgments can be value-laden or self-serving comes from the list of strengthsassociated with farmers who have no cattle(Appendix 1). According to informants,those farmers have the following strengths:they borrow money, provide labor forothers, are cattle herders, and buy cattlefrom others. Clearly, these qualities areviewed as strengths by those who benefitfrom farmers without cattle: people wholend them money, hire their labor, or sellanimals to them (e.g., cattle owners).Furthermore, an examination of theweaknesses listed for farmers who have nocattle (such as cruelty to cattle or gainingwhen crops are unintentionally destroyedby livestock) confirms that these viewscome from people who have cattle.

Wealth RankingGoal: Classify farmers in a communityinto wealth categories.

Rationale: Wealth is an important socialcategory in most societies, although itsspecific definition will vary not only fromone culture to another but sometimes fromone village to the next. Wealth is a relativecategory that depends on the veryparticular circumstances of farmers. Unlikethe local classification of farmers discussedpreviously, wealth ranking establishescertain predetermined concepts andcategories (e.g., ”wealth,” “rich,” “poor”).The specific definitions of “wealth” and ofwhat constitutes a “rich” or a “poor”farmer depend on the local conceptions ofthese terms. Members of a communityusually are keenly aware of their positionsand those of others in the community. Thismethod is based on that knowledge. The

Figure 3. Causal diagram of the factors that affectyields based on those identified by farmers’classification of themselves in Chihota, Zimbabwe.a Indicates a factor associated with a farmer type; the rest

indicate factors identified as strengths or weaknessesassociated with a farmer type.

gender wealth

cash

family implementsa draft access to seed,labor animals fertilizer

agea hiredlabor

timeliness ofoperationsa yields

cattlea

planning

workinggroupsa knowledgea

30wealth ranking provides a way to groupfarmers and analyze their preferences.Clearly what is appropriate or desirablefor one wealth group may not be so foranother. Furthermore, the constraints toadopting a new technology or practicemay be completely different amongwealth categories, since each may controldifferent sets and amounts of resources.A wealth ranking is also a tool to analyzethe potential and actual distribution ofbenefits and costs of a technology (see“Comparing Different TechnologicalOptions,” p. 54, for an example of howwealth ranking can be used).

Method:5 This method assumes thatresearchers have compiled a list ofhouseholds in a well-definedcommunity/village, which they want torank. The researchers will need to definewhat constitutes a household in theparticular place where they are working.Although a household is often defined asa group of people who live together andeat from the same pot, this definitionmay not be useful in certain societieswith extended families, and researchersshould establish what constitutes ahousehold through discussions withlocal informants. A few (one to four)reliable informants with goodknowledge of the people in theircommunity should be identified. It maybe a good idea to include male andfemale informants. Informants can beinterviewed either together or separately.The former strategy provides aconsensus ranking, while the lattermakes it possible to test the consistencyof the rankings. In the case of multipleindividual rankings, the rankings of

several informants should be highlycorrelated. If they are not, the lack ofcorrelation indicates that there is aproblem. Perhaps the informants do notknow the households well, have accessto different information, have useddifferent criteria to do the ranking, orhave simply not provided accurateinformation.

First, the interviewer asks theinformant(s) to define what “wealth” isin the community. After identifying thelocal word(s) for wealth, theinterviewer and informant(s) candiscuss what a “wealthy” or richhousehold/farmer is, focusingparticularly on its characteristics. Thenthe characteristics of poor farmers canbe discussed; following that, thecharacteristics of the group that falls inbetween (not rich or poor) can beidentified. After the distinctivecharacteristics of each group aredefined, the interviewer writes thecharacteristics corresponding to eachgroup on a large, easily visible piece ofpaper. The interviewer checks with theinformants to see that everyone agreeswith the characteristics. Next, theinterviewer reads the names of thefarmers from the list of households andasks the informants to indicate thegroup to which they belong.Alternatively, researchers can preparecards, each one with the name of ahousehold, and ask the informants toput them into one of three piles, eachrepresenting a wealth rank.

When this exercise is done with a groupof informants rather than an individual,

5 Some of the ideas described here are based on Grandin’s (1988) work on wealth ranking, although the method presented here differssomewhat from her approach. In Grandin’s method, informants make as many groups as they wish by creating piles of cardscontaining the names of households that in their view belong in the same wealth rank. Then each household is given a score based onthe pile where it was classified, and the scores of several informants are averaged out. This average is used to do the final ranking.For the specifics of the method consult Grandin (1988).

31

the group may discuss the classification.If there is disagreement about theclassification, the interviewer should askabout the reasons for the disagreementand note the contrasting rankings thatinformants provide for farmers whoclassification was disputed. When theexercise is done with several individualinformants, the researcher shouldcompare the characteristics associatedwith the rich, intermediate, and poor, aswell as the rankings, once the exercisehas finished. If discrepancies areidentified, ideally the researcher shouldgo back to the informants forclarification, and this information shouldbe noted as well.

If in addition to (and independently of)the wealth ranking the researcher hascollected other qualitative and/orquantitative socioeconomic data on thehouseholds that have been ranked, theresearcher can test for an associationbetween those variables and the wealthranking. Ideally this association shouldbe significant, as indicated by statisticalanalyses of the quantitative andqualitative data, using the wealth classesas a grouping factor by village. Asignificant association providesindependent evidence of the validity6 ofthe wealth ranking. In most cases,however, such data are not available,which is one reason why a wealthranking may be done. It should bestressed that a wealth ranking is faster,cheaper, and easier to implement than afull-scale survey.

Even if socioeconomic data are available,it may still be desirable to perform awealth ranking. The wealth ranking is

based on the knowledge of local peoplewho may be aware of assets andrelationships that may not even havebeen captured by survey data. Theseinclude initiative, entrepreneurial ability,experience, and social or politicalrelationships.

Example: This methodology was used inthe Chiapas Project to rank all theparticipating households and test theextent to which different types of farmersadopted various maize varieties. Theinformants defined wealth based oncertain characteristics such as ownershipof a pair of bullocks, cattle, a motorvehicle, or appliances; the type of house;and total landholdings. Poor householdshad houses made of wattle and daub oradobe, without cement floors andplastered walls, and possessed almost noappliances. Fewer households owned apair of bullocks and some cattle. Nonehad privately held land or a motorvehicle. Households that wereintermediate between rich and poor hadadobe houses with cement floors andplastered, painted walls. Ownership of atelevision set, gas stove, and even arefrigerator was common. Manyhouseholds owned a pair of bullocks andcattle, and a few had privately held land.Rich households owned brick or adobehouses with cement floors and plastered,painted walls. These households hadtelevision sets, refrigerators, gas stoves,and even videocassette recorders. Someowned a pair of bullocks but others didnot, since they had a tractor or could payto rent one. Many had privately heldland and some had motor vehicles.7

6 Validity denotes the extent to which a measurement tool is measuring what it was designed to measure (Adams et al. 1997).7 Contrast this ranking in Chiapas with one in Malawi (Smale and Phiri 1998), where well-to-do households produced enough maize to

last from harvest to harvest; owned some livestock, an oxcart or other farm machinery, and several changes of clothing; and had ahouse with an iron roof and brick walls.

32Independently of the wealth ranking, ahousehold survey was done for allhouseholds. The survey includedquestions on socioeconomic variablessuch as landholding, ownership oflivestock and other assets, performanceof off-farm labor, and reception ofremittances. Therefore the validity of thewealth ranking could be tested. To do so,an analysis of variance (ANOVA) wasdone to compare the means of each of thewealth classes (rich, intermediate, andpoor) for key quantitative socioeconomicvariables, and a Chi-square test ofassociation was done to assess therelationship between wealth rank andeach qualitative variable. Table 4presents the results.

The wealth ranking was consistent withobjective, independently measuredcharacteristics of the households. Ingeneral the trends in ownership among

wealth ranks were what one wouldexpect: the rich had more assets than theintermediate class, and the intermediateclass had more assets than the poor.These results corroborate the validity ofthe wealth ranking. Furthermore, thewealth ranking was included in aregression analysis that showed thatthose classified as poor planted onaverage a smaller area to improvedvarieties and a larger one to landracesthan the rest (Bellon and Risopoulos2001). These results show howcombining participatory methodologieswith more conventional analytical toolscan enhance the analysis.

Comments: This method is mostappropriate for a single village orcommunity since it relies on theknowledge that the informants have offellow community members, and thedefinition of wealth classes is relative to

Table 4. Comparison of farmer characteristics by wealth rank, Chiapas, Mexico

Wealth rank

Variable Poor Medium Rich Overall P-valuea

Number of farmers 50 32 16 98 –Ownership of assets

Total land holdings (ha/farmer) 6.2 10.6 14.5 9.0 .000Cattle (% farmers own) 18.0 37.5 68.8 32.7 .001Cattle (head/farmer) 1 3 11.2 2.9 .000Pair of oxen (% farmers own) 44.0 50.0 56.3 48.0 .668Horses (% farmers own) 58.0 75.0 87.5 68.4 .054Pigs (% farmers own) 64.0 84.4 75.00 72.5 .127Pick-up truck (% farmers own) 0.0 3.1 68.8 12.2 naTractor (% farmers own) 0.0 0.0 6.3 1.0 na

Sources of incomeOff-farm labor by farmer (% performing) 68.0 43.8 37.5 55.1 .030Type of labor (% performing) .009

Agriculture 50.0 42.9 16.7 44.4 –Construction 29.4 57.1 16.7 35.2 –Commerce 0.0 0.0 16.7 1.9 –Other 20.6 0.0 50.0 18.5 –

Off-farm labor by other family members (% performing) 40.0 75.0 37.5 51.0 .004Remittances (% receiving) 10.0 28.1 18.8 17.4 .106

Use of hired labor (% hiring) 58.0 68.8 93.8 67.4 .029

a P-value associated with a Chi-square test of association for qualitative variables and one-way ANOVA for quantitative variables; na = not applicable (too many blank cells).

33

the members of the community.Comparisons across communities may bemore difficult, because the definition of“rich” or “poor” in one place may bedifferent in another. There is goodevidence, however, that at least in certaincircumstances wealth rankings are validacross regions (Adams et al. 1997). Evenif this is not the case, the characteristicsthat informants use for the classificationmay provide a rough basis forcomparisons across villages. If additionalquantitative and/or qualitativesocioeconomic information is available,then researchers can compare the wealthclasses among villages.

Minimum Set ofSocioeconomicIndicatorsGoal: Identify key characteristics ofparticipants (informants/experimenters).If possible, compare them to thepopulation of users/clients, and thusestablish whether they are representative(or at least make any bias explicit).

Rationale: One problem withparticipatory work is that usually itinvolves a self-selected group of people(i.e., the people who choose toparticipate). This group does notnecessarily reflect the conditions andinterests of all farmers in a region, so it isimportant to know the participants. Thecontent and quality of the informationelicited and the joint outputs obtaineddepend on the people with whomresearchers work. To assess the degree towhich participants are representative ofall farmers in the region of interest, theresearcher should compare theparticipants’ characteristics withcharacteristics of the population ofhouseholds in the region.

Method: Develop a short questionnairethat includes a few, mostly qualitativequestions. The questionnaire should befilled in 5 to 10 minutes with allparticipants in an activity or (if thenumber is too large) with a sample ofthem (e.g., one out of four). The questionsshould be simple and easy to answer.Usually they will deal with characteristicsthat reflect the participants’ resources,constraints, and goals. Ideally, theinformation gathered should becomparable to other information that isrepresentative of the households in theregion of interest, such as a census or arepresentative survey. The questionnairemay request information on:

• gender;

• age;

• ability to read and/or write;

• number of years of formal educationcompleted;

• number of years of independent farming(farming experience);

• size of land holdings by tenurialarrangement (this requires previousknowledge of the land tenure regime);

• crops grown;

• types and number of animals owned;

• agricultural off-farm labor;

• non-agricultural off-farm labor; and

• remittances from family members workingelsewhere.

The researcher may decide to includeother key characteristics identified fromfarmers’ own classification. It is importantto clarify whether the questions refer tothe respondent as an individual or to thehousehold in which he or she lives. Forexample, the enumerator should carefullyspecify whether the question refers to landowned or controlled by the respondent asan individual or to land that is owned orcontrolled by the respondent’s household.

34The same specificity is needed whenasking about animals and sources ofincome.

Example: In the Oaxaca Project, field dayswere organized for farmers to vote on thelandraces collected so that they could besorted into a gradient of interest. Duringthese field days a questionnaire was usedto obtain a minimum set of socioeconomicindicators. The purpose of thequestionnaire was to get an idea ofparticipants’ characteristics and toseparate farmers’ votes from the votes ofthe other field day participants. Thepurpose of the voting exercise was togauge farmers’ interest in the landraces,so researchers were not concerned withthe votes of other participants. Thequestionnaire showed that of 306 personswho attended the field days, only 213individuals were involved in maize

farming, and 54% of these were women.Only the votes of those 213 individualswere taken into account.

The questionnaire also revealed importantdifferences between male and femalefarmers who participated in the field days(Table 5). Compared to the men, the womenwere younger, had less farming experienceand more formal education, and planted amuch smaller area to maize on average.More women received remittances, andfewer worked off of the farm. Notsurprisingly—given that women planted asmaller area to maize than men—a higherproportion of women purchased maize anda lower proportion sold it. Almost all of thewomen said that they grew maize to beself-sufficient in that commodity, comparedto a still important, but smaller, percentageof men. The percentages of male and femaleparticipants who said that they grew maize

Table 5. Field day participants in Oaxaca, Mexico, characterized by agricultural activity, gender, and othervariables

Characteristic All Male Female

Number of participants 213 97 116Age (yr) 43.6 49.7 38.4Mother tongue (% speaking)

Spanish 88.0 87.9 88.0Zapotec 11.6 12.1 11.1Other 0.4 0.0 0.9

Education (mode) Elementary, No education Elementary,not completed not completed

Experience in farming (yr) 19.7 24.1 15.9Area planted to maize (ha) 2.7 4.3 1.3Remittances (% receiving) 44.0 40.4 47.0Off-farm labor (% performing) 47.2 57.6 38.5Purchase maize (%) 55.1 39.4 68.4Sell maize (%) 28.7 38.4 20.5Goals of maize production (%)

Home consumption 94.0 88.9 98.3Sale 24.1 33.3 16.2

Livestock ownership (%)Bullocks 31.5 49.5 16.2Cattle 31.0 39.4 23.9Pigs 59.3 48.5 68.4Poultry 71.8 70.7 72.7Goats, sheep 38.6 36.7 40.2


35

for sale differed dramatically. Twice asmany men as women engaged incommercial maize production. Anotherdifference between female and maleparticipants was that a higher percentageof men tended to own bullocks andcattle, whereas women tended to ownpigs. These data suggest that while self-sufficiency was a fundamental goal of allfarmers, men tended to be morecommercially oriented, to produce moremaize because they planted a larger area,to depend more on off-farm labor andless on remittances, and to raise differenttypes of livestock. These findings, inturn, suggest that men and women mayvalue the characteristics of maizevarieties differently.

Aside from the questionnaire for fieldday participants, researchers surveyed asample of farmers in the Oaxaca studysites (baseline survey). The random,representative sample of the farmingpopulation in the region enabledresearchers to determine the extent towhich field day participants wererepresentative of the farming populationin the area. A few questions asked offield day participants were not includedin the sample survey, although thesample survey retained the questionsrelated to personal characteristics,sources of income, and agriculturalassets. Table 6 compares some of thepersonal and household characteristics ofparticipants in field days with

Table 6. Selected personal and household characteristics of participants in field days and sample survey,Oaxaca, Mexico

Females Males Households

Field Sample Field Sample Field SampleCharacteristic days survey days survey days survey

Participants (no.) 116 240 97 240 213 240Age (yr) 38.3 48.1+++ 50.1 54.2++

Education (% reporting)No formal education 8.6 31.3*** 5.2 16.7*** – –Elementary, not completed 36.2 40.0 38.1 53.8 – –Elementary, completed 38.8 22.5 33.0 22.9 – –Junior high school 9.5 3.8 10.3 3.8 – –High school or technical school 5.2 1.7 3.1 2.1 – –College 1.7 0.8 10.3 0.8 – –

Literacy (%) 92.2 67.9*** 94.8 82.1*** –Mother tongue Spanish (%) 87.9 74.6*** 87.6 68.3*** – –Non-farm sources of income (%)

No off-farm labor or remittances – – – – 25.4 26.3 ns

Off-farm labor only – – – – 30.5 37.5Remittances only – – – – 28.2 24.2Off-farm labor and remittances – – – – 16.0 12.1

Maize area (ha) – – – – 1.8 3.0+++

Ownership (%)Pair of bullocks – – – – 31.5 59.6***Cattle – – – – 30.5 37.9*Pigs – – – – 59.2 50.0*Horses, mules – – – – 45.1 76.7***Goats, sheep – – – – 38.0 40.4

Source: Bellon et al. (2000).Note: ++ (+++) indicate t-test, significant at the .05 (.01) level; * (**) *** indicate chi-square test of homogeneity, significant at

the 0.1 (.05) .01 level; ns = not significant. In the case of education and sources of income, the statistical test applies to allcategories.

36information from the random sample offarmers in the study sites. These datamake it possible to test whether therewas a bias between field day participantsand a representative sample of thepopulation of farmers in the area.

The results show that men and womenwho participated in the field days wereyounger and better educated than theaverage for the region. A higherpercentage of field day participants hadSpanish as their mother tonguecompared to the respondents in thesample survey. In terms of non-farmsources of income, there was nodifference between field day participantsand respondents in the sample survey,although the survey respondents farmeda larger maize area and a higherpercentage of them owned bullocks,cattle, horses, and mules. These data donot necessarily mean that field dayparticipants are poorer than the surveyrespondents. Since field day participantsgenerally have more years of formaleducation, farming may havecontributed less to their livelihoods thanit did for farmers in the region as awhole. Field day participants seem to bea biased sample of the overall farmingpopulation of the region, but regardlessof the reason for the bias, maize farmingis clearly still important for field dayparticipants, as demonstrated by theirattendance at the field days.

Ideally the researcher would like arepresentative sample of farmers toparticipate in the research activity, butparticipation is a voluntary endeavor,and farmers cannot be forced toparticipate purely for “representation.”

Comments: One problem with theminimum set of indicators is that if theychange from one group to another orfrom one situation to another, it may be

difficult to compare results. As moreinformation becomes available, forexample, a researcher may wish tochange the indicators to fit the newknowledge, but this should be avoidedto the extent possible. If changes areunavoidable, the researcher should atleast retain as many common indicatorsas possible. Ideally, researchers shouldinclude questions that elicit informationthat is comparable to information fromother sources, such as census or othersurvey data, so that results can becompared and if possible extrapolatedacross different groups or settings.

Calendar of ActivitiesGoal: Identify how productive andleisure activities are organized andinteract during the year in a community.

Rationale: Households in a community,and individuals within them, carry outdifferent activities during the year.These activities may be complementary,may compete with one another, or maynot interact at all. Competition in theallocation of time among activities is animportant consideration for anyhousehold because it has implicationsfor the household economy. It isparticularly important to identify anylabor bottlenecks and when they occur.Researchers should be especially carefulto develop separate calendars for malesand females within the same household,because their activities may differsubstantially.

Method: The method presented here isto develop a generic calendar ofactivities for a community. The methodfocuses on all of the activities carried outby all households within a community,rather than on the activities of onespecific household, because an

37

individual household may pursue only asubset of the activities carried out acrossa community. Identifying specificcombinations of activities for specifichouseholds will provide an idea of thedifferent livelihood strategies present inthe community.

A group of key informants is assembledand asked to list all activities in whichmales and females engage. First thegroup is asked about productiveactivities, which in the case of agriculturewould include crops grown and types oflivestock kept. Informants are also askedabout the kinds of off-farm labor inwhich they engage, such as day labor inagriculture, construction work, andworking as a mechanic or carpenter.Second, the group is asked to listactivities necessary for the household tofunction, including food preparation,going to the market to purchase food,repairing the house, cleaning the house,and studying with children. Third, thegroup is asked to list activities conductedfor the community, such as repairing theroads or irrigation system or organizingand participating in religiouscelebrations. Finally, informants arequestioned about their leisure activities,including time spent resting.

Once the list has been compiled for eachtype of activity, informants are asked topoint out the months during the yearwhen they take place and to specifywhich household members participate.

Activities of particular interest may bedisaggregated by subactivities. Forexample, maize production can bedisaggregated by land preparation,number of weedings, number of fertilizerapplications, harvest, storage, and sale,and informants can identify the month ofthe year in which each subactivitytakes place.

Example: Figure 4 is a one-year calendarof activities for Santa Ana Zegache, acommunity in the Oaxaca Project.Activities related to crops and animalsare listed first, followed by off-farmlabor, community work, and religiouscelebrations. This calendar shows theconflicts between tending one’s owncrops of maize and beans andperforming agricultural labor off of thehousehold farm. Taking care of sheepand goats is a year-long activity, whilecaring for cattle has a better definedperiod. To analyze the potential impact(timing conflicts and opportunity cost) ofa new activity, such as growing a newcrop or building contours for erosioncontrol, the labor demand for the newactivity should be overlaid on thiscalendar.

Comments: One common mistake withthis method is that researchers develop acalendar only for agricultural activities,ignoring off-farm labor, communitywork, and religious celebrations. Such acalendar may omit activities that are asimportant—or even more important—than the agricultural ones.

One limitation of this method is that itprovides information only on the timingof activities and not on the intensity oflabor use. The researcher knows when anactivity takes place but not how muchtime and labor it requires (informationthat can be difficult and time consumingto obtain).

Local Taxonomies of SoilsGoal: Identify the soil types farmersrecognize and the characteristics theyfind relevant for each type.

Rationale: Farmers have their owncategories for classifying soils. They mayrecognize different problems in each

38

DJ F M A M J J A S O N

Beans Males/Females

Maize / squash Males/Females

Alfalfa (irrigated) Males

Castor beans Females

Vegetable production Males

Peanuts Males/Females

Chickpeas Males

Backyard garden Females

Oxen Males

Cattle Males

Pigs Females

Sheep / goats Children/Females

Poultry Females

Fieldhand Males

Constrution work Males

Other off-farm work Males/Females

Community work Males/Females

Temporary migration Males

Religious festivities Males/Females

Seeding period Growing period

Growing period Growing period

Growing periodGrowing period

Growing periodSeeding period

Grown and harvested throughout the year

Seeding period

Harvest period

Harvest period

Harvest period

Post-harvest

Sell / buy

Sell / buy

Sell

Sell

Buy

Buy

Fatten

Fatten

Sell / buy BuyFatten

Grown and harvested throughout the year

An ongoing process of buying, selling, and growing

Primarily police and army (males), house help, hand crafts, sell tortillas in the market (females)

Patron saint festivity Day of the DeadHoly Week

Figure 4. An example of a calendar of activities, Santa Ana Zegache, Oaxaca, Mexico.

39

type, such as waterlogging. They maytailor their crops, varieties, ormanagement practices to the specific soiltypes, perhaps by applying differentamounts or types of inputs or plantingparticular varieties. Therefore, explicitlytaking into account the variability farmersrecognize in their soils may be animportant factor for the developmentand/or adoption of agriculturaltechnologies. These taxonomies may beuseful in defining where certaintechnologies may or may not beappropriate (i.e., recommendationdomains). Furthermore, for scientists itmay be also important to know thistaxonomy to communicate moreeffectively with farmers.

Method: A group of informants from acommunity is assembled, ideally a mixtureof people of different ages, resources, andgenders. Researchers explain that theywant to learn about the types of soils thatexist in the community, including theirpositive and negative characteristics.Researchers explain that this knowledge isvital for understanding and developingsolutions for soil problems faced byfarmers.

The interviewer poses the question: Whattypes of soils are present in your community?

The group lists each soil type. For eachtype, the interviewer should checkwhether there are subtypes by askingwhether all soils of that type are the same,or whether there are different classes forthat type. Once subtypes are identified,the interviewer asks the followingquestions for each type:

How do you identify this soil type?

What are its positive characteristics(advantages)?

What are its negative characteristics(disadvantages)?

It is important to identify responses thatrefer to the same concept, since peoplemay express their ideas in differentforms. This requires some judgment onthe part of the scientist, but usually it isnot difficult. As with the farmerclassification, the responses may refer tosome underlying property of the soil,which should be identified. Researchersthen use this information to generate atable that synthesizes all of the data.

Example: This method was used in theChihota Project to identify the soil typesfarmers recognized. Afterwards, their soiltaxonomy was the basis for identifyingand discussing the technological optionsthey used to cope with soil infertility andwhether these options were targeted tospecific soil types or not. Farmers listedten types of soil for agricultural use.Table 7 describes the four most importanttypes. The descriptions are based ontexture (i.e., particle size), fertility status,and color (the latter is used to distinguishsubclasses). The advantages anddisadvantages listed for each soil typerefer particularly to its water-holdingcapacity, ease of work, inherent fertility,response to fertilizers and manure,tendency to become waterlogged; to itsparticular uses, such as use in gardens;and its appropriateness as a buildingmaterial.

The two most common soil classes formaize production were the lightertextured soils, Jecha and Shapa. Jecha is asandy soil of low fertility and poor water-holding capacity, which can easilybecome waterlogged, is easy to work, andis good for building. Shapa is a sandyloam soil of low to average fertility. Yieldsof crops grown on this type of soil may below unless additional inputs are applied,but Shapa soils have better water-holdingcapacity than Jecha soils. Although theycan also become waterlogged, Shapa soils

40

are easy to work but are not good forgrowing groundnut. The subclasses ofShapa soils depend on the position of thesoil in the toposequence. The darkersubclass, which is considered morefertile, is located lower in thetoposequence; the whitish subclass is inthe intermediate parts of thetopsequence; and the grayish and leastfertile subclass is found at the top.Although agronomists and soil scientistsworking in the area knew many of thesecharacteristics, they did not know howfarmers referred specifically to these

soils. Therefore at a minimum thisexercise enhanced the communicationbetween scientists and farmers at a lowcost to both.

The underlying soil properties of thetaxonomy are texture, color, water-holding capacity, ease of work, inherentfertility, response to fertilizers andmanure, and proneness to waterlogging.Aside from actual soil properties,particular uses (e.g., in gardens and asbuilding material) were important in thetaxonomy.

Table 7. Farmers’ soil taxonomy, Chihota, Zimbabwe

Soil class Subclasses Description Advantages Disadvantages

Jecha White Sandy soil, Responds to manure application Low fertilityBlackish coarse-grained, Can get good yields, even with Low water-holding capacityGrayish low fertility, inadequate rains Erodes easily

used for building Easy to work Becomes waterlogged easilyGood for building Can get very hot

Difficult to farm, because ofneed to apply more inputs

Shapa Black (dema) Sandy-loam soil, Produces good yield, even with Low to average fertilityWhite (nhuke) easy to cultivate, inadequate rains No yield unless inputs added

low fertility Average water-holding capacity Gets waterlogged under heavyCan hold water for long periods rainOne can grow any crop Crops fail if little rainResponds well to manure and Maize wilts easily when hotfertilizer Not good for growing groundnutsEasy to workCan be worked by hand

Rukangarahwe Reddish Gravel, mixture Resists erosion InfertileWhitish of fine and coarse- Good yields if rains are good Blunts farming implements

grained sands Does not get waterlogged Difficult to work (to plow, weed)Good for road construction Poor water-holding capacityGood for fruit tree production Crops wilt with reduced moisture

Difficult to plow deeplyNeeds too much waterMany plants are cut duringcultivationHarbors termites

Churu/Rechuru Makura (upland Termite mound Can be used to improved soil Hard to digsoil, type of soil, heavy texture, High fertility Crops wilt with slighttermite mound) sticks when wet Good yields if rains are good moisture stressBani (fley soil, type and cracks when dry Used for molding and plastering Requires a lot of water toof termite mound) Used as graveyards support plant growth

Difficult to plow


41

It is possible to study the relationshipbetween the local taxonomy andobjective soil properties. Researchers cansample each of the soil types that farmersidentify and conduct physical andchemical laboratory analyses. Forexample, in the Chiapas Project, farmersidentified five soil types: Tierra Negra,Tierra Baya, Tierra Colorada, TierraColorada Arenosa, and Tierra Cascajosa.Researchers sampled 104 fields thatincluded the five soil types and analyzedthe samples’ chemical and physicalproperties. An analysis of variance usingthe soil classes as the grouping factor(Table 8) indicated that farmers’ soiltaxonomy discriminated among objectiveproperties in their soils and that objectiveproperties were consistent with farmers’perceptions.

Comments: In working with farmers’soil taxonomies, as with any other typeof local knowledge, researchers must becautious about making generalizations toother people or areas. Specific soil classesmay change from one community to thenext. Even within a community,researchers should check with farmerswho did not participate in the taxonomyexercise to see whether they hold thesame ideas about the soil classes andproperties and to probe for additionalclasses. When researchers work in morethan one community and similar soilnames recur, they should always check tosee whether the names refer to the same

underlying soil or soil property or tosomething different.

Local Classifications ofClimateGoal: Identify factors relevant to farmersthat define the climate during thegrowing season.

Rationale: Farmers recognize favorableand unfavorable climatic conditions forcrop production. These conditions areassociated with particular climatic eventsand conditions. Many of farmers’ riskmanagement strategies are ways ofcoping with these events and conditions,so identifying farmers’ views of theseevents and conditions and theirinteraction is fundamental tounderstanding those strategies anddesigning technologies that arecompatible with farmers’ currentpractices. To a great extent, these factorsreflect a value judgment, not a value-freedescription of a phenomenon. Farmersoften refer to a “good” or a “bad” seasonfor the crop of interest, and there aremany different ways in which a badseason occurs.

Method: A group of informants from acommunity is assembled, ideally amixture of people of different ages,resources, and genders. Researchersexplain that they want a better

Table 8. Soil chemical properties by farmer soil class, Chiapas, Mexico

Tierra Tierra Baya, Tierra Colorada- TierraProperty Mean Negra Tierra Colorada Arenosa Cascajosa F-statistic P-value

Organic matter (%) 6.1 8.7 5.9 3.3 1.7 9.7 .0000pH 6.6 6.7 6.4 6.1 7.3 8.1 .0001Sand (%) 49.0 38.4 48.9 65.0 68.1 9.7 .0000Clay (%) 28.0 36.2 26.2 22.0 14.0 6.7 .0004Observations (no.) 97 33 44 10 10 – –

Source: Bellon and Taylor (1993).

42understanding of which climaticcharacteristics constitute a “good” and a“bad” cropping season.

The interviewer poses the questions:

What are the characteristics of a “good”season?

What are the characteristics of a “bad”season?

Usually these characteristics refer tounderlying climatic factors or events.These factors can be combined to createdifferent “types” of seasons, some“good” and some “bad.” Not alltheoretical combinations are real orappear frequently. Researchers may needto relate the factors identified by farmersto actual rainfall data to identify relevant“types” of seasons in terms that aremeaningful to farmers.

Example: This method was used in theChihota Project as a framework for alater discussion of risk managementstrategies. Farmers were asked about thecharacteristics of “good” seasons and of“bad” seasons. Their answers reflectedfive underlying factors (Table 9): theonset of the rains, the end of the rains,drought in the middle of the croppingseason, distribution of rainfall, andquantity of rainfall. By combining thesefactors, types of seasons can beidentified. For example, one seasonbegins in November, finishes in March,

and has a mid-season drought. Anotherstarts in mid-October, finishes in April,and has no break in rainfall in themiddle of the season. These types ofseasons can be used to discuss differentmanagement options to cope withclimate-related cropping problems or toexplore how climatic factors mightaffect a new technology (for example,how the late onset of the rains mightaffect the application of lime or thechoice of a new variety).

Comments: Local classifications aremore complex for climate than for soils,because climate is much more dynamic,changing from one year to the next,whereas soils change very slowly.Developing a classification of climatealso requires a higher level ofabstraction, because participants aretrying to identify common aspects inclimate patterns that occur throughoutrelatively long periods. People arenotoriously bad at judging long-termtrends. A classification of climate clearlyentails more limitations than otherclassifications, but it can still be usefulto systematize and discuss key aspectsof climate and their impact onagriculture and other elements offarmers’ livelihoods. It should be notedthat the method presented here is notconcerned with eliciting farmers’perception of climate data (see, forexample, Gill 1991) but with identifying

Table 9. Underlying factors defining “good” and “bad” seasons according to farmers, Chihota, Zimbabwe

Underlying factor Good season Bad season

Onset of rains Mid-October After OctoberEnd of rains April December, MarchMid-season drought – Rains break for three weeks in mid-seasonDistribution of rains Even throughout season; High rainfall in April, low rainfall during

allows periods of sunlight grain filling stageQuantity of rain Rains give time to work in field Excessive rains cause waterlogging,

very long rainy season

43

conditions and events that farmers use toclassify a season with regard to its impacton crop production.

Local Crop TaxonomiesGoal: Identify the different types (orfarmer varieties8 ) that farmers recognizein one crop species, and identify the traitsfarmers find relevant for each type. (Thismethod may also be used for differentcrop species rather than for varieties of asingle crop species.)

Rationale: Small-scale farmers usuallyplant more than one variety of a crop,particularly if it is one of their mostimportant crops, and they have their owncategories for the different varieties ortypes. Each of these varieties has specificcharacteristics, some positive and somenegative. By identifying the differentvarieties and their advantages anddisadvantages, it is possible to recognizethe crop characteristics that farmers valueand how these are distributed across thevarieties they plant. This information isvaluable for improving breedingstrategies (for example, by pinpointingwhich traits to improve) or for identifyingnew varieties that may interest farmers.Additionally, this information may bevaluable for understanding farmers’incentives to maintain crop diversity onthe farm, an approach to conservinggenetic resources and biodiversity that isbecoming more important.

Method: A group of informants from acommunity is assembled, ideally maleand female farmers with a reputation forplanting many different varieties.Researchers explain that they want a

better understanding of the various typesof a particular crop that exist in thecommunity, including the positive andnegative characteristics of each type.Researchers explain that this informationis important for understanding theproblems that farmers have with thisparticular crop and their possiblesolutions.

The interviewer poses the question: Whattypes or varieties of crop X (e.g., maize) doesyour community plant?

Each of the types is listed. Theinterviewer checks whether each type issubdivided into finer categories and askswhether these categories are subdividedas well. The interviewer continues untilthere are no finer categories. Once allcategories have been elicited, theinterviewer asks the following questionsfor each one:

How do you tell this variety apart fromother ones?

What are its positive characteristics(advantages)?

What are its negative characteristics(disadvantages)?

It is important to identify responses thatrefer to the same concept, since peoplemay express their ideas in differentforms. This requires some judgment onthe part of the scientist, but usually it isnot difficult. As with the other farmerclassifications, the responses may refer tosome underlying characteristic orproperty, and therefore it is important toidentify them. Researchers can use thisinformation to generate a table thatsynthesizes all the data.

8 Farmer varieties (referred to as “varieties” in this manuscript) are the crop populations that a group of farmers recognize as distinctunits. Each of these varieties combines a particular set of characteristics that farmers recognize, such as a certain yield potential,growing cycle, particular performance under biotic and abiotic stresses, response to management, or culinary and storage properties.

44Example: This method was used in theOaxaca Project to identify the diversity ofmaize types grown by farmers, and theresults formed the basis for an analysis ofthe supply and demand of characteristics(a method presented in the next sectionof this manual). For simplicity, thisexample will focus only on the exercisecarried out in one of the communities inthe project, Santa Ana Zegache. Thisexercise was conducted with a group ofeight farmers (two women, six men).They identified four types of maize,based on grain color: Blanco (white),Amarillo (yellow), Negro (black), andBelatove (red). They did not recognizedivisions within these classes. Theadvantages and disadvantages of eachtype are presented in Table 10. Theunderlying characteristics of the varietytaxonomy are yield, duration, ease ofsale, consumption quality, and suitabilityas animal feed.

During the discussion, it emerged thatplanting date—and therefore theuncertainty of the duration of thegrowing season—was very important.Although in the earlier part of thisexercise farmers did not identify anydisadvantage associated with whitemaize, a key disadvantage became clear:white maize had a high yield, multiple

uses, and was easy to sell, but it had thelongest growing cycle. Its longerduration was a negative characteristic ifthe rains were delayed and it had to beplanted late, because then the croprisked being exposed to drought and tofrost. The other maize types had shortergrowing cycles (white > yellow > black> red) and provided farmers with theflexibility to respond to the uncertainonset of the rains. If the rains arrivedlate, farmers could plant a shorterduration maize type. Farmersrecognized the trade-off betweenduration and yield, and grain color wasan indicator of this relationship.Although women particularlyappreciated the colored maize types,they were difficult if not impossible tosell, which was not a great problem intheir subsistence-oriented farmingsystem. These insights emphasize thatthere is no “best” or “ideal” variety;farmers need and want diversity. Eventhe very desirable white type hadproblems. The results from Santa AnaZegache confirm the idea that plantingdifferent maize types is, at least in part,a risk management strategy. They alsoshow that grain color is an important“marker” that farmers use to makeplanting decisions.

Table 10. Maize types and their characteristics in Santa Ana Zegache, Oaxaca, Mexico

Maize type Characteristic Advantages Disadvantages

Belatove Red grain Grows very fast Low yieldNot a lot of animal feed

Amarillo Yellow grain Good yield Not widely consumedFaster growing Difficult to sell

Negro Black grain Fast growing Very difficult to sellLower yield

Blanco White grain Good for consumption No disadvantage(tortilla, atole)Used for everythingEasy to sell

45

In the case of Santa Ana Zegache, theclassification and number of maizetypes was simple, but this is not alwaysso. Figure 5 shows the complexity of themaize taxonomy produced by farmersin the Chiapas Project, which stands insharp contrast to the simplicity of theOaxacan taxonomy. Farmers in Chiapasgrouped their maize varieties into threemajor classes: landrace (criolla),improved, and “creolized”9 (acriollada)varieties. Each class comprised severalmaize types. Some landraces werefurther divided by grain color. Thedifferences between the taxonomiesfrom Oaxaca and Chiapas are partlyexplained by the fact that farmers inChiapas are much more commerciallyoriented, even though subsistenceproduction is also important. Althoughthey had landraces with desirablecharacteristics, the farmers in theChiapas Project had also been exposedto improved varieties that were welladapted to their conditions, and in factthey had modified some of theimproved varieties to suit their needs(the creolized varieties).

Comments: Even within onecommunity, the information elicited just

from one group may be incomplete. It isnecessary to probe further with otherfarmers or groups. Ideally, researchersshould ask farmers to bring samples ofthe different crop varieties theyrecognize to the group discussion andask them to classify the varietiestogether.

Farmers’ classification of varieties maynot necessarily coincide withresearchers’ classification. In Santa AnaZegache in Oaxaca, genetic resourcespecialists collected samples of tentypes of maize, including all four graincolors. Based on agromorphologicalcharacteristics, these types wereclassified into three classes (one classcould include more than one grain color).

As with other types of localtaxonomies, a local crop taxonomy maybe valid just for the community whereit was elicited. The same name mayrefer to different biological entities fromone community to the next. It may bemisleading to compare varieties fromdifferent communities using localtaxonomies. “Maíz Blanco” fromcommunity A may not be the same as“Maíz Blanco” from community B.

9 Creolized maize varieties are scientifically improved varieties that have been in the hands of farmers for several growing seasons andhave been modified by them. These varieties usually are appreciated because they combine desirable traits of improved varieties withthose of landraces.

Maize

Improved Landraces Creolized

Tuxpeño Olotillo Jilguero Higuera Napalu Crema Tuxpeño Rocamex HíbridoCriollo Amarillo

524 526 534 Pioneer Amarillo Blanco Crema

Figure 5. Classification of maize types in Vicente Guerrero, Chiapas, Mexico.

46

Identifying Points ofInterventionGoal: Identify the technologies/practicesto be developed and/or tested withfarmers.

Rationale: The diagnosis of farmers’conditions usually reveals a large set ofproblems or constraints that farmersconfront. The classification of farmersmay show socioeconomic constraints,soil taxonomies may indicate problemswith soils, and so on. Many of theseproblems cannot be resolved by research.If patterns of land inheritancediscriminate against women, forexample, there is little that an agronomistor a soil scientist can do, aside fromnoting the problem and considering howit may affect the technical solutions thatcan be offered to farmers to improve soilfertility.

Among the spectrum of problemsuncovered in the diagnosis, it isfundamental to identify the areas ofintervention where interaction betweenscientists and farmers can provideappropriate solutions through newtechnologies or practices. Obviously theparticular expertise of the scientistsworking with the farmers will influencewhich problems can be addressed. Evenso, the specific problems that should beaddressed (and therefore the specificareas of intervention) are not necessarilyeasy to identify.

Method: A group of informants from acommunity is assembled, ideally amixture of people of different ages,resources, and genders. Researchersexplain that they want a betterunderstanding of the informants’problems.

The interviewer poses the question: Whatare your problems?

The interviewer lists the responses. Sincethe informants’ answers may refer to thesame problem in different ways, once allproblems have been identified, theyshould be grouped by similarity. Forexample, someone may say, “The cropdoes not produce,” and someone elsemay say, “We get bad production.” Bothresponses refer to low yields. Responsesshould be grouped in consultation withthe informants, by saying, for example,“Do you agree that the statements ‘Thecrop does not produce’ and ‘We get badproduction’ refer to the same problem? Ifso, let’s agree on a common way ofexpressing it.”

Once the problems have beenconsolidated, the interviewer asks theinformants to rank them by askinginformants which problem they considerto be the most important, which issecond most important, and so on. Theremay not be consensus; differentinformants may rank problemsdifferently. The interviewer notes thedifferent rankings for each problem.Alternatively, the interviewer can askeach informant to rank the problems, andthen use the average ranking or the mostfrequent ranking to order the problemsby importance. Another strategy is to askinformants to vote on the importance ofeach problem.

This exercise helps researchers identifythe general areas of intervention wherethey can make a contribution. It alsohelps researchers to gauge the potentialimportance of each intervention, becausethey can see the entire range of problemsthat farmers face and the importance ofeach. The farmers’ answers may range over avery broad range of topics, including all sortsof things that agricultural research can donothing about, and they may raise people’sexpectations. Therefore, researchers should be

47

extremely careful to be clear to farmers aboutwhat the researchers can and cannot do. Theunderstanding that researchers may havegained with the use of farmers’classification of themselves may behelpful in guiding and focusing thediscussion.10

Once the general areas of interventionhave been identified, researchers shouldrepeat the exercise to identify and rankthe specific problems that are suitable forresearch. At this stage, it is fundamentalthat researchers keep the discussion focusedon areas where they can make a contribution.Keep the discussion as specific as possible.For example, the general concern of “lowyields” may consist of more specificproblems, including late planting, insectattack, lack of irrigation, and difficulty inpurchasing fertilizers.

After problems have been identified andranked, the group of informants andresearchers should discuss possibleoptions for addressing them.

The interviewers asks the informants:What do you think can be done to improve/solve this problem?

The pros and cons of the differentoptions identified can be discussed andthe group can agree how to proceed. It isimportant that the responsibilities offarmers and scientists regarding futureaction are defined very clearly in termsof what each will and will not do.

Example: A maize agronomist and arural sociologist used the methoddescribed above to query a group of verypoor, subsistence-oriented, indigenousfarmers in a small community in thestate of Puebla, Mexico, about their

problems. 11 The group comprised 100farmers, 40 of them female, ranging inage from 20 to 60. This number ofparticipants is unusually high andreflects a high degree of socialorganization within the community.After a long discussion in whichresearchers used their knowledge of thearea and the communities to encouragefarmers to focus on specific issues, thegroup mentioned the followingproblems:

• low prices for coffee and pepper;

• lack of labor for harvesting coffee;

• lack of infrastructure to dry and processcoffee, which led to marketing problemsbecause farmers could sell coffee only asberries, not beans;

• poor transportation infrastructure;

• insufficient maize production to covertheir needs;

• difficulty selling other agriculturalproducts, such as tropical fruit (priceswere so low that was not worthharvesting the fruit);

• lack of sufficient drinking water duringthe dry season; and

• lack of doctors and medicines, althoughthe community had a health center.

The group was asked to rank theproblems in order of importance.Problems associated with coffee andmaize were equally important, followedby the lack of services (water and health),the lack of transportation infrastructure,and the difficulty of marketing tropicalfruit. The scientists participating in theexercise explained to the group that theirexpertise was in maize, andunfortunately they could give littleassistance with problems related to

10 Other methodologies can be used to address these issues in a very focused manner, such as causal analysis with farmers (Tripp andWoolley 1989).

11 This example was kindly supplied by Angel Pita and Xóchitl Juárez from the Universidad Autónoma de Chapingo, Mexico.

48coffee, pepper trees, services, orinfrastructure. The remainder of theexercise focused on insufficient maizeproduction.

Participants were asked about theirspecific problems in maize production.They mentioned that while their localmaize varieties were good, the varietiesnevertheless had some problems. Themain problem was that the varieties weretall and vulnerable to lodging, and theparticipants wished to test new maizevarieties. The group mentioned highstorage losses as another problem, aswell as losses to pests in the field (whitegrubs and fall armyworm). They alsowanted to know about other types offertilizer. The fertilizer formulations theyused for maize were originally providedfor coffee production and had lownutrient concentrations (e.g., 18-12-6 N-P-K). The ranking of these problems inorder of importance was: 1) maizevarieties, 2) fertilizers, 3) storage losses,and 4) field pests.

Based on this exercise, several specificareas of intervention were defined:1) evaluating new maize varieties, bothlocal and external, with farmers;2) conducting simple experiments withdifferent fertilizer types and rates; and3) evaluating the use of metal silos forstoring maize. Although farmers wantedto evaluate the use of pesticides, theydecided against it when they learned ofthe expense and of the need for specialhandling to avoid health risks.

Comments: An important role forscientists participating in this type ofexercise is to provide their analyticalskills to identify the causes behind theproblems and propose solutions thatmay not be apparent to farmers. Farmersknow their environment andcircumstances better than anyone, but inmany instances the causes of many oftheir problems may not be evident tothem, and scientists can explain thosecauses. For example, farmers may notunderstand the workings of supply anddemand. When they see that the price ofa crop increases, they may all plant it thenext season, perhaps increasing supplyso greatly relative to demand that theprice falls substantially. Nor may farmersunderstand decreasing marginal returnsto an input. They may believe thatapplying double the amount of fertilizerwill increase production twofold, whichmay lead them to waste the inputwithout obtaining the expected results.

In summary, researchers can proposenew options that may be unknown tofarmers, such as conservation tillage forareas with erosion or where soilpreparation is a constraint. Researcherscan also provide new knowledge to helpfarmers understand problems better;with pest control, for instance,researchers can provide knowledgeabout pests’ reproductive cycles or therole of beneficial insects.

49

Any technology or practice used byfarmers represents a particular way tosolve one or several problems. Eachtechnology or practice responds tofarmers’ concerns in specific ways, whichmay be regarded as the traits orcharacteristics that define the technologyor practice. Farmers can view somecharacteristics as positive oradvantageous (i.e., as benefits) and othersas negative or disadvantageous (i.e., ascosts).

Any practice or technology entails trade-offs between its positive and negativetraits. As a farmer from Chiapas once saidwhen discussing maize varieties, “Witheach variety you gain in certain things butlose in others.” He explained that with amodern variety, farmers gained higheryields, shorter duration, and less lodging,but they also lost something, because thevariety required more inputs and morecareful management. The choice of onetechnology/practice over others is greatlyinfluenced by the balance between itspositive and negative characteristics.Depending on the preferences, resources,and constraints that individual farmersface, a beneficial characteristic for onefarmer may be a negative one for another,or the balance between positive andnegative traits may be acceptable for onefarmer but not for another.

Evaluation of Currentand NewTechnological Options

Any new technology presented tofarmers will either improve orsubstitute for the technological optionsthey currently have. It is fundamentalto identify these options andunderstand perceptions about theadvantages and disadvantages of eachone. Only then will researchers be ableto assess the appropriateness ofpotential new technologies orpractices, evaluate the likelihood thatthey will be adopted, and if necessarymodify them to suit farmers’ needsbetter. To identify gaps in knowledgeand perceptions among those involvedin the process of technological change,it is vital to understand not onlyfarmers’ perceptions but also those ofother stakeholders in the researchprocess, mainly the scientists andtechnicians proposing these newtechnologies.

This section presents several methodsfor identifying technologies thatfarmers presently use, eliciting andanalyzing farmers’ perceptions of theircosts and benefits, and enablingfarmers and researchers to evaluatenew technologies together.

50

Eliciting Farmers’Perceptions ofTechnological OptionsGoal: Identify the criteria used byfarmers to assess available technologicaloptions.

Rationale: Farmers have severaltechnological options at their disposal.They have perceptions of theiradvantages and disadvantages andtherefore their trade-offs. Inherent inthese perceptions are the criteria thatfarmers use to judge these technologiesand most likely any new ones. It isimportant to know and understand thesecriteria if researchers are to identify newtechnological options of interest tofarmers, including improvements tocurrent ones.

Method: Define the problem to beaddressed, such as inappropriategermplasm, infertile soil, or problemswith pest management or crop storage. Agroup of informants from a communityis assembled, ideally a mixture of peopleof different ages, resources, and genders.The first step is to identify thetechnological options that farmersrecognize to deal with the problem ofinterest. For germplasm, this is relativelyeasy because the local crop taxonomyprovides this information. For otherproblems—soil infertility, a pest, storage,and so on—it may be necessary to ask,for example:

What can you do to deal with this problem?

Or ask specifically: What can you do toimprove your soils? What do you do tocontrol a particular pest? What do you do toprotect your stored maize?

The answers to this question are theoptions recognized by farmers.Researchers should try to be as inclusive

as possible, and to elicit as many answers(options) as possible. At this stage it is notimportant to establish how importantthese are, but to have the mostcomprehensive list.

Then for each option identified, theinterviewer asks:

What are its advantages?

What are its disadvantages?

The interviewer records all answers. It isimportant to identify responses that referto the same concept, since people mayexpress their ideas in different forms. Thisrequires some judgement on the part ofthe scientist, but usually it is not difficult.(This is similar to what has been done forthe local soil and crop taxonomies). Oncethis has been done, researchers shouldidentify the underlying properties,characteristics, or concerns implied infarmers’ responses. This last activity is afundamental part of this method, sincethese characteristics, properties, orconcerns are the basis for the criteria. It isimportant to express the criteria in termsthat make sense to farmers. The followingexamples show how this method isapplied for germplasm and soil fertilitymanagement.

Example for germplasm: The OaxacaProject included a collection of landracesrepresenting the maize diversity present inthe region. As indicated earlier, thecollection was based on the local croptaxonomy elicited from key informants inall communities sampled. Farmersdonating the maize samples were askedabout the advantages and disadvantagesof each landrace they donated. Table 11presents the local maize taxonomy and theadvantages and disadvantages for the sixcommunities that are the focus of theproject. Note for example how farmersrefer to an advantage with different terms:

51

Table 11. Perceived advantages and disadvantages of maize types, Oaxaca, Mexico

Type Blanco Amarillo Negro Belatove(white kernel) (yellow kernel) (black kernel) (red kernel) Pinto

Subtypes Tempranero Delatoba Olote delgado Blanco Amarillo(early) (thin cob) (generic) (generic) Tepecente None None None

Advantages Early Heavy Yields by volume Good tortillas Weight Withstands pests Good tortillas Good tostadas Good adaptationGood tortillas Yields by volume Easy to shell Good production Good tortilla A lot of grain Early Very earlyThin cob Good storage Not too delicate Tasty tortilla Tasty tortillas Grows very fastYields by volume Soft pasture Heavy Withstands drought ColorWhite tortilla Yields by volume Yields by volume Tasty tlayuda

Low ear rot Withstands weeds Sweet atoleGood pasture Early Withstands coldGood for sale Good yieldWithstand drought Good storageEasy to shell Tasty atoleGood for consumption Good yield

Grows fast

Disadvantages Low yield Low yield Poor storage Attacked by pests Low yieldSmall ear High ear rot Not widely consumed Low yield Little pasture

Poor storage Difficult to sell Very difficult to sell

52“early” and “grows fast.” The taxonomyhas five maize types, based on graincolor: Blanco (white), Amarillo (yellow),Negro (black), Belatove (red), and Pinto(multicolored). The white and yellowtypes were subdivided further into fourand two subtypes, respectively. Allanswers can be grouped as characteristicsrelated to a set of concerns: consumption,yield, sale, duration, adaptation, andresponse to biotic and abiotic stresses.These advantages and disadvantageswere used to identify the criteria thatfarmers use to judge their maize. Table 12presents these characteristics grouped byconcern and then expressed as thecriteria. The data show how importantconsumption characteristics are for thesefarmers. These criteria will be used laterfor comparing different varieties/technological options.

Example for soil fertility management:The Chihota Project included feedbackfrom farmers who had been evaluatingthree soil fertility improvementtechnologies: lime in combination withfertilizer; green manuring (velvetbean andsunnhemp), sole or intercropped withmaize; and cereal legume rotations.

During these feedback sessions, farmerswere asked about the advantages anddisadvantages they perceived in thesetechnologies (Table 13). All answers can begrouped as characteristics related to a setof concerns: impacts on soil fertility,fertilizer use efficiency, productivity, costs,labor and inputs, alternative uses for thecrops, rainfall, and biotic stresses. Theseadvantages and disadvantages were usedto identify the criteria that farmers use tojudge the technologies. Table 14 presentsthese characteristics grouped by concernand then expressed as the criteria.

Table 12. Characteristics and criteria used to judge maize types, Oaxaca, Mexico

Concern Advantages Disadvantages Criteria

Consumption Tasty tortillas Taste of tortillasTasty/sweet atole Taste of atoleTasty tlayuda Taste of tlayudasTasty tostada Taste of tostadasEasy to shell Ease of shellingGood storage Poor storage Storage propertiesGood pasture Little pasture Production of pastureSoft husk (totomoxtle) Husk quality

Yield Good yield–weight Yield by weightGood yield–volume Small ear Yield by volumeGood yield (generic) Low yield

Duration Early/ fast growing Duration

Sale Easy to sale Difficult to sell Ease of sale

Adaptation Good adaptation Adaptation

Abiotic stress Withstands drought Withstands droughtWithstands cold Withstands cold

Biotic stress Withstands weeds Withstands weedsWithstands pests Attacked by pests Withstands pestsLow ear rot High ear rot Susceptibility to ear rot

53

Table 13. Perceived advantages and disadvantages of soil fertility improvement technologies, Chihota, Zimbabwe

Lime with fertilizers Cereal/legume rotations Green manures

Advantages Improves yields Residual fertility Improves soil fertilityCrops grows well Reduced fertilizer use Cheaper than fertilizersImproves soil structure High yields Increase yieldsImproves soil fertility Increased crop diversity Can be used to feed cattleCorrects pH Multipurpose use of legumes NoneIncreases fertilizer efficiency Disease controlNot expensive Early to assessCut costs of fertilizersSupresses weedsEarly to assess

Disadvantages Needs adequate rains Legumes affected by disease Not for human consumptionCrops suffers if rains are late Poor germination Seed unavailableDamage soil if over-used Still assessing Labor intensiveCan be washed away by wind None NoneStill assessingNone

Table 14. Characteristics and criteria used to judge soil fertility improvement technologies, Chihota, Zimbabwe

Concern Advantages Disadvantages Criteria

Soil fertility Improves soil fertility Impact on soil fertilityCorrects pH Impact on pHResidual fertility Impact on residual fertilityImproves soil structure Impact on soil structure

Damage soil if over-used Impact on soil if over-used

Fertilizer efficiency Increases fertilizer efficiency Impact on fertilizer efficiencyReduced fertilizer use

Costs Not expensive Cost vis-à-vis inorganic fertilizersCut costs of fertilizersCheaper than fertilizers

Inputs Seed unavailable Ease of accessing inputsCan be washed away by wind Chances of input loss

Labor demands Labor intensive Impact on available labor

Productivity Improves yields Impact on yieldCrops grows wellHigh yieldsIncreases yields

Alternative uses for crops Increased crop diversity Not for human consumption Alternative usesMultipurpose use of legumesCan be used to feed cattle

Rainfall Needs adequate rains Interaction with rainfallCrops suffer if rains are late

Weeds Suppresses weeds Impact on weeds

Germination Poor germination Impact on germination

Diseases Disease control Legumes affected by disease Impact on/from disease

54Comments: This method—whether it isapplied to germplasm or another kind oftechnology—only provides an inventoryof characteristics that farmers use toassess the technological options theyknow, although it is likely that theywould use the same criteria to judge newoptions. This method is only descriptive.Researchers cannot assess whichcharacteristics are more important for thefarmer. Nor can they assess theimportance of the characteristics inrelation to the technological optionsavailable, particularly for thosecharacteristics that can be delivered byseveral technologies (in other words,researchers cannot tell how much of acharacteristic of interest, such as yield, issupplied by a particular variety orinput). This information is important fordetermining which characteristics shouldbe improved or to evaluate newtechnological options compared tocurrent ones.

Comparing DifferentTechnological OptionsGoal: Systematically compare andanalyze farmers’ perceptions oftechnological options.12

Rationale: The previous method helps toelicit information on the advantages anddisadvantages of technologies, on theimplicit characteristics that farmers valuein those technologies, and therefore onfarmers’ criteria for judgingtechnological options. To compare andevaluate technological options in asystematic manner, however, it isnecessary to assess the importance ofeach of these characteristics relative to

each other (i.e., farmers’ demand forthese characteristics) and the extent towhich each technology provides thesecharacteristics (i.e., the technologysupply of characteristics). A techno-logical option that is better at supplyingcharacteristics that farmers considermore important is more valuable thanone that is inferior. Furthermore, evenwhen a technological option is good atsupplying certain characteristics, if theseare not very important, its value isdiminished.

Method: The previous method provideda list of characteristics that farmersvalued, but it could not clarify howmuch the individual characteristics werevalued relative to each other or howspecific technological options providedeach characteristic. In many cases,scientists add characteristics to the list ofcharacteristics identified previously,based on their experience, even thoughfarmers may not have identified them. Insome instances, certain issues are notmentioned because they are obvious toinformants, or informants simply fail toarticulate or mention them. Clearly, thescientists’ experience and common senseshould complement the farmers’.

The exercise described here can be donewith a group of farmers or individualfarmers, a choice that has implicationsfor the analysis (see comments below). Itassumes that the relevant technologicaloptions have already been identified(e.g., maize varieties, soil improvementtechnologies, and so on).

First, researchers explain the objective ofthe exercise to the participants. Theresearchers make it clear that, in

12 The emphasis here is on the evaluation of technologies based on their characteristics. This method is particularly suited for cropvarieties, which is the focus of the work described here. The reader is referred to CIMMYT (1988) for methods using other factors intechnology evaluation.

55

discussions with them or with otherfarmers, they have identified a set ofcharacteristics or issues that farmers findimportant in their technological options.Now they wish to know how importantthose characteristics are to farmers, sincesome are likely to be more important thanothers. Researchers can provide anexample to make this point. Then theyshould note that not all of farmers’technological options perform equallywell with respect to each of thosecharacteristics (here another example maybe useful). Therefore, researchers alsowant to know how good or bad farmersconsider each of these options to be withrespect to each characteristic.

Second, the interviewer asks farmers torate the importance of each of theidentified characteristics (in other words,to assess the demand of characteristics)by asking:

How do you consider this characteristic (e.g.,yield, drought resistance) to be: veryimportant, somewhat important, or notimportant?

This question is repeated for allcharacteristics identified as important. It ishighly desirable to make cards, eachillustrating one of the characteristics, andask farmers to place each card in a pilecorresponding to the rating they considerappropriate (very important, somewhatimportant, not important). Figure 6presents a hypothetical example withcards. (Appendix 2 shows examples ofcards used in the Oaxaca Project).

Third, farmers are asked to rate theperformance of each technological optionwith respect to each characteristic as: very

good, intermediate/acceptable, or poor13

(assess the supply of characteristics). Todo this, the interviewer asks:

How do you consider this option (e.g.,variety A, velvet bean, lime) to be in termsof its performance with respect to thischaracteristic (e.g., drought resistance,increasing soil fertility): very good,intermediate, or poor?

This question is repeated for all optionsand a given characteristic. Then theprocess is repeated for the nextcharacteristic, and so on until no morecharacteristics remain to be discussed. Itis desirable, since it simplifies theprocess, to use the cards from the secondstep in this method. Put them in a row(Figure 7). Above them, place three cardsdepicting the rating of performance(poor, intermediate, very good), perhapsshown as a frowning, straight, or smilingface, or thumbs up/thumbs down, orsome other image adapted to the placewhere the research is being done. Asshown in Figure 7, this card placementcreates a matrix in which the first rowdisplays the characteristics and thecolumns display the possibleperformance ratings.

Using cards that name or depict theoptions (with varieties you can use actualears or panicles of specific varieties),researchers ask the farmers to place thecard with the option (or the ear) in therow with the card with the characteristicand under the column for theappropriate performance rating. Theresults are noted.

The results from these ratings can becompared across different types or

13 The rating has to be adapted to the characteristic. In some cases, “very good,” “intermediate,” or “poor” may not be the mostappropriate way to rate a characteristic. If the characteristic of interest is the labor required for the technology to work properly, amore suitable rating may be “high,” “intermediate,” or “low.”

56

Figure 6. Hypothetical example of cardsrating the importance of maizecharacteristics.Note: No order of importance is implied within acolum. Each column represents a pile of cardsassociated with the importance rating.

Figure 7. Example of a card layoutto rate characteristics.

Not important Somewhat important Very important

Withstands cold Withstands wind Withstands drought

Easy to shell Invest labor Cash investment

Good for tejate Taste of tortilla Good for nixtamal

Poor Intermediate Very good

Good for nixtamal

Invest labor

Withstands drought

57

groups of farmers and/or varieties/technological options using the averageratings. These average ratings can be usedto compare and rank the importance ofdifferent characteristics to farmers(demand of characteristics) or to compareand rank the performance of differentoptions with respect to each characteristic(supply of characteristics).

As mentioned previously, this ratingexercise can be done with a group offarmers or individual farmers. The groupstrategy may produce a consensus on theratings. There is no guarantee, however,that a consensus may be reached. If thegroup is heterogeneous, it is very likelythat farmers may not agree on theimportance of many characteristics,because each farmers faces differentproblems, may have different priorities,and therefore may value characteristicsdifferently. In fact, identifying thedisagreements and discussing them mayprovide important information onfarmers’ diverse priorities. Furthermore,in a group setting it may be difficult toanalyze variation among individuals withdifferent goals, resources, and constraints,and researchers will be more limited intheir ability to generalize the results toother farmers. One strategy for gainingthis information is to ask for a show ofhands (voting) and record how eachmember of the group rates the importanceof a characteristic and the performance ofa technological option with respect to acharacteristic. It may be useful to recordthe votes disaggregated by gender. Thisprocedure provides a better idea of thevariability in ratings across group members.

A second strategy allows statistical testsand inferences to be made if researchershave a random, representative samplefrom a population of farmers. The ratingscan be combined with a typology offarmers, such as the wealth ranking, to

analyze how different types or groups offarmers rate the characteristics (forexample, which characteristics areimportant for poor or rich farmers, maleor female farmers, farmers with andwithout machinery). The performance ofdifferent technologies with respect to eachcharacteristic can be assessed statistically,which offers a better idea of the trade-offsinvolved (see example below).

A third strategy can be used if manygroups are interviewed. Each group canbe treated as an “individual” and theaverage ratings can be calculated acrossgroups. Alternatively, if a show of handsis asked for each group and the results arerecorded, individual votes within a groupcould be used for the analysis. Since thereare many groups, this would lead to alarge number of ratings. Researchersshould be careful in applying statisticalinferences to these techniques, however. Ifthe sample of informants is not randomlychosen, researchers may violate theassumptions of the tests they want toapply, invalidating their results. However,these approaches provide a better idea ofthe variability present and still permitsome basic parameters to be calculated,such as the average rating or percentagefor each rating, at least for the participantsand without claiming widerrepresentation.

Example: This method was used in theOaxaca Project to compare different maizelandraces, based on the categoriesidentified by eliciting the local croptaxonomy (presented earlier). The resultsfor only one community, Santa AnaZegache, are presented here for simplicityand because the results differed acrosscommunities.

The rating exercise was done as part ofthe baseline survey with a random sampleof 40 farming households in the

58community. Male and female membersof each household were interviewedseparately. The list of characteristicsincluded all the ones identified across theregion. The reader should note that thislist of 25 characteristics includedcharacteristics that were not identifiedexplicitly by farmers using the method toelicit their criteria. The additionalcharacteristics were included becauseresearchers thought that they would beimportant (in fact they were). Theadditional characteristics included yieldstability (“produces something even in a

bad season”), yield of tortillas by kilogramof dough, and suitability for all usesidentified in the region (special dishes andpreparations).

Analyzing the demand ofcharacteristics

Table 15 compares the ratings for theimportance of maize characteristics bymen and women in farming households.The table reports the average rating, basedon the following scale: 1 = very important,2 = somewhat important, and 3 = notimportant.14 A Wilcoxon matched-pairs

Table 15. Average ratings of importance of maize characteristics by males and females, Santa Ana Zegache,Oaxaca, Mexico

Average rating Top 5 characteristics

Concern Characteristic Males Females P-valuea Males Females

Consumption Taste of tortillas 1.78 1.38 0.01 – –Good for atole 1.80 1.55 ns – –Good for tlayudas 2.23 1.63 0.00 – –Ease of shelling 2.08 2.68 0.00 – –Good for storage 1.08 1.50 0.00 2 –Good pasture 1.90 1.70 ns – –Good feed 1.20 1.53 0.02 5 –Nixtamal quality 2.05 1.33 0.00 – 5Good for tamales 2.25 2.23 ns – –Good for tejate 2.73 2.38 0.01 – –Good for pozole 2.95 2.80 0.03 – –Good for nicoatole 2.90 2.70 0.02 – –

Yield Yield by weight 1.25 1.05 0.03 – 2Yield by volume 1.28 2.03 0.00 – –Yield of tortillas 1.98 1.45 0.00 – –Yield stability 1.13 1.03 0.10 4 1

Duration Duration 1.40 1.55 ns – –

Sale Ease of sale 1.85 1.53 0.03 – –

Abiotic stress Withstands drought 1.03 1.08 ns 1 3Withstands wind 2.55 1.88 0.00 – –Withstands cold 2.75 2.30 0.00 – –

Biotic stress Withstands weeds 2.45 2.35 ns – –Withstands pests 2.40 1.60 0.00 – –

Management Produced with little labor 1.40 1.85 0.01 – –Produced with little money 1.10 1.18 ns 3 4

Note: ns = not significant.a P-value associated with a Wilcoxon signed ranks test for two related samples.

14 Appendix 3 shows what data for the demand and supply of characteristics look like.

59

signed ranks test (a non-parametricstatistical procedure) was used to test forstatistically significant differencesbetween males’ and females’ ratings for acharacteristic.15

A comparison of men’s and women’sratings shows highly significantdifferences for most characteristics. Ofthe 25 characteristics, only seven had nostatistically different ratings. Of the fivetop-rated characteristics, however, menand women coincided in three: toleranceto drought, yield stability, and low cashinvestment. Men also included storageproperties and suitability as feed in thetop five characteristics, and womenincluded yield by weight and nixtamal16

quality. These results also show that menand women value many characteristics:the average ratings for 14 and 17characteristics for men and women,respectively, were between “very” and“somewhat important.”

These results show important genderdifferences in the demand for maizecharacteristics. Failure to recognize thesedifferences would lead to biasedinterventions. In the Oaxaca Project, ifmales alone had participated in thevoting exercise that identified landracesto be distributed, it is very likely that thechoices would have been of interest tothem but less so for women. Theseresults also have implications forbreeding. Improvements in yield stabilityor tolerance to drought would bebeneficial for both men and women, but

any improvements that come at the costof decreasing nixtamal quality couldnegatively affect women more than men,since women value nixtamal qualitymuch more than men do.

The large number of characteristics ratedas “very” or “somewhat important” alsosuggests that both men and womendemand diversity, since it is unlikely thatone maize type will be good at supplyingall of the characteristics they value.Therefore there may not be a “best” or“ideal” maize type. These farmersrequire a range of maize types, and thisfact motivates the intervention ofproviding farmers with access todiversity in the Oaxaca Project.

Similar analyses can be done using anygrouping or classification of farmers,such as a wealth ranking. Table 16groups men and women separately bywealth rank and reports the averagerating for each wealth rank (i.e., rich,medium, poor), based on the followingscale: 1 = very important, 2 = somewhatimportant, and 3 = not important. AKruskal Wallis one-way analysis ofvariance by ranks (a non-parametricstatistical procedure) was used to testwhether there were differences in theratings—in other words, whether eachrating for a characteristic was statisticallyequal or not among the three wealthgroups.17

The ratings of characteristics among thewealth groups were not statistically

15 The table reports the mean or average rating, from which it is easier to identify differences and trends, but the test is based on thenull hypothesis that the median (not the mean) of the population of differences is zero (Daniel 1978:135-9). A non-parametric test,such as the one used here, is more appropriate because the ratings are ordinal and their underlying distribution is unknown and is notlikely to be normal. In this case, this test is used because males and females were not selected independently of each other but weremembers of the same household (they were related).

16 Nixtamal is the dough used to make tortillas, which requires that the milled maize be soaked in water with lime.17 The table reports the mean or average rating, from which it is easier to identify differences and trends, but the test is based on the

null hypothesis that the three population distribution functions are identical against the alternative hypothesis that they do not allhave the same median (Daniel 1978:200-5).

60Table 16. Average ratings of importance of maize characteristics by wealth rank for males and females, Santa Ana Zegache, Oaxaca, Mexico

Males by wealth rank Females by wealth rank

Concern Characteristic Rich Medium Poor Total P-valuea Rich Medium Poor Total P-valuea

Consumption Taste of tortillas 1.79 1.83 1.83 1.81 ns 1.38 1.54 1.00 1.38 nsGood for atole 1.64 1.92 1.67 1.75 ns 1.38 1.69 1.33 1.50 nsGood for tlayudas 2.21 2.42 2.17 2.28 ns 1.62 1.54 1.67 1.59 nsEase of shelling 2.21 2.00 2.00 2.09 ns 2.54 2.77 2.67 2.66 nsStorage properties 1.14 1.08 1.00 1.09 ns 1.31 1.62 1.50 1.47 nsGood pasture 1.93 2.00 1.50 1.88 ns 1.46 1.92 2.00 1.75 nsGood feed 1.29 1.17 1.00 1.19 ns 1.46 1.54 1.67 1.53 nsNixtamal quality 2.07 2.08 2.17 2.09 ns 1.46 1.31 1.00 1.31 nsGood for tamales 2.50 2.25 1.83 2.28 0.06 2.46 2.08 2.17 2.25 nsGood for tejate 2.86 2.75 2.67 2.78 ns 2.54 2.23 2.33 2.38 nsGood for pozole 3.00 2.92 2.83 2.94 ns 2.85 2.85 2.67 2.81 nsGood for nicoatole 2.86 3.00 2.83 2.91 ns 2.69 2.69 2.50 2.66 ns

Yield Yield by weight 1.36 1.08 1.33 1.25 ns 1.15 1.00 1.00 1.06 nsYield by volume 1.29 1.50 1.17 1.34 ns 2.15 1.85 2.00 2.00 nsYield of tortillas 1.93 2.00 2.00 1.97 ns 1.62 1.54 1.17 1.50 nsYield stability 1.14 1.00 1.00 1.06 ns 1.08 1.00 1.00 1.03 ns

Duration Duration 1.29 1.58 1.50 1.44 ns 1.46 1.54 1.50 1.50 ns

Sale Ease of sale 1.71 2.00 1.83 1.84 ns 1.31 1.85 1.83 1.63 ns

Abiotic stress Withstands drought 1.00 1.00 1.00 1.00 ns 1.00 1.15 1.17 1.09 nsWithstands wind 2.43 2.58 3.00 2.59 ns 2.08 1.69 2.00 1.91 nsWithstands cold 2.71 2.50 3.00 2.69 ns 2.31 2.38 2.17 2.31 ns

Biotic stress Withstands weeds 2.14 2.67 2.50 2.41 ns 2.15 2.31 2.67 2.31 nsWithstands pests 2.36 2.33 2.67 2.41 ns 1.31 1.85 1.50 1.56 ns

Management Produce with little labor 1.36 1.42 1.50 1.41 ns 1.92 1.77 1.67 1.81 nsProduce with little money 1.07 1.08 1.00 1.06 ns 1.15 1.23 1.17 1.19 ns

Note: ns = not significant.a P-valule associated with a Kruskal Wallis one-way analysis of variance by ranks for males and females separately.

61

different.18 Not surprisingly, for the topfive characteristics all wealth ranks amongmen and among women coincided on thefollowing: yield stability, tolerance todrought, and low cash investment. Menacross all wealth categories coincided onstorage properties. Women across wealthcategories agreed on yield by weight; forpoor women, taste of tortillas andnixtamal quality were also particularlyimportant.

These results suggest that improvementsin any of the traits may benefit all farmersequally. If differences between wealthgroups had emerged for certaincharacteristics, however, the improvementof those characteristics would havebenefited some groups more than others.It is also important to note that losses insome characteristics may be more negativefor some groups than for others. Forexample, if resistance to lodging is ratedsignificantly higher by the “rich” group,the introduction of a new variety moreresistant to lodging may benefit themmore than the other groups. On the otherhand, if the “poor” group rates resistanceto storage pests significantly higher, and anew variety has substantially lowerresistance to these pests, the cost ofadopting the new variety will be higherfor the poor group than for the othergroups.

By analyzing the ratings of thesecharacteristics as shown here, researchersgain a method to predict how the costsand benefits of introducing a newtechnology are likely to be distributedamong different groups of farmers and/ormembers of farming households.

Analyzing the supply ofcharacteristics

Table 17 compares farmers’ ratings of theperformance of Blanco (white), Amarillo(yellow), Negro (black), and Belatove(red) maize types by gender group. Foreach characteristic identified earlier, eachmaize type was rated based on thefollowing scale: 1 = very good,2 = intermediate, or 3 = poor. For thecharacteristics related to labor and cashinvestments, the rating scale was:1 = little, 2= intermediate, 3 = a lot. Thetable reports the average rating permaize type,19 except for yield by weight,yield by volume, yield of tortillas,anthesis (days to male flowering), anddays to be ready for harvest (an indicatorof duration), for which the means ofestimates provided by farmers in theappropriate units are used. A non-parametric Kruskal Wallis one-wayanalysis of variance by ranks for theratings and a parametric one-wayanalysis of variance for the continuousvariables were used to test for statisticaldifferences across the different maizetypes for each characteristic.

Men’s assessments of the four typesshowed statistically significantdifferences for most characteristics. TheBlanco type is superior to the other typesfor all characteristics, except for havingthe longest duration. On the other end ofthe spectrum, the Belatove type isinferior to all other types, except forhaving the shortest duration. Amarilloand Negro are intermediate. Theassessment shows a gradient ofperformance from Blanco to Amarillo,Negro, and Belatove. These results

18 Except for the case of “good for tamales” among men, where the poor rated it higher than the rest.19 As with the demand of characteristics (Table 16), Table 17 for supply of characteristics reports the mean or average rating, which makes

it easier to identify differences and trends, but the test used in each table is based on the null hypothesis that the three populationdistribution functions are identical against the alternative hypothesis that they do not all have the same median (Daniel 1978:200-5).

62Table 17. Average rating of the performance of different maize types for several characteristics of importance to male and female farmers, Santa Ana Zegache, Oaxaca, Mexico

Males Females

Concern Characteristic Blanco Amarillo Negro Belatove Total Signif.a Blanco Amarillo Negro Belatove Total P-valuea

Consumption Taste of tortillas 1.00 1.11 1.00 1.33 1.04 0.01 1.03 1.07 1.00 1.00 1.03 nsGood for atole 1.00 1.47 2.46 2.33 1.42 0.00 1.00 1.33 2.40 3.00 1.32 0.00Food for tlayudas 1.00 1.17 1.00 1.00 1.04 0.09 1.00 1.00 1.00 1.00 1.00 nsNixtamal quality 1.00 1.22 1.29 1.67 1.13 0.00 1.00 1.07 1.00 1.00 1.02 nsGood for tamales 1.00 1.06 1.93 2.33 1.24 0.00 1.00 1.07 1.10 1.00 1.03 nsGood for tejate 1.00 2.00 2.36 2.33 1.55 0.00 1.03 1.80 2.20 2.00 1.39 0.00Good for pozole 1.00 1.83 2.43 2.33 1.52 0.00 1.03 1.20 1.80 1.00 1.18 0.00Good for nicoatole 1.00 2.11 1.50 3.00 1.44 0.00 1.00 1.87 2.50 3.00 1.46 0.00Ease of shelling 1.05 1.11 1.36 1.00 1.12 ns 1.45 1.07 1.00 1.00 1.29 0.01Storage properties 1.75 2.06 2.71 3.00 2.05 0.00 1.85 2.20 2.90 3.00 2.11 0.00Good pasture 1.00 1.00 1.93 2.33 1.23 0.00 1.08 1.07 1.90 3.00 1.23 0.00Good feed 1.00 1.00 1.07 1.00 1.01 ns 1.00 1.00 1.00 1.00 1.00 ns

Yield Yield by weight b 653.8 544.9 520.4 461.3 595.1 0.01 395.8 296.0 230.0 156.7 346.9 0.01Yield by volume c 4.00 3.99 3.99 4.00 3.99 ns 3.97 3.97 3.98 4.00 3.97 nsYield of tortillas d 38.37 38.78 39.14 39.00 38.64 ns 36.05 36.80 38.00 40.00 36.58 nsYield stability 1.08 1.56 1.86 2.00 1.37 0.00 1.63 1.33 1.20 1.00 1.48 0.04

Duration Anthesis e 79.9 74.6 62.9 60.0 74.6 0.00 74.0 65.9 53.5 45.0 68.9 0.00Harvest f 121.9 116.2 97.4 95.0 114.9 0.00 127.5 118.3 97.1 96.0 120.5 0.00

Sale Ease of sale 1.00 1.28 2.00 2.00 1.29 0.00 1.00 1.20 1.80 2.00 1.18 0.00

Abiotic stress Withstands drought 1.35 1.89 2.64 2.33 1.76 0.00 1.54 1.47 1.60 2.00 1.54 nsWithstands wind 1.25 1.33 1.21 1.33 1.27 ns 1.48 1.60 1.20 2.00 1.47 nsWithstands cold 1.13 1.11 1.14 1.00 1.12 ns 1.25 1.47 1.40 1.00 1.32 ns

Biotic stress Withstands weeds 1.63 2.06 2.00 1.67 1.80 0.01 1.80 1.93 1.60 1.00 1.79 nsWithstands pests 1.45 1.56 1.71 1.33 1.52 ns 1.58 2.07 2.11 3.00 1.78 0.00

Management Produced with little labor 2.50 2.33 2.50 2.00 2.44 ns 2.30 2.33 2.40 2.00 2.32 nsProduced with few purchased inputs 2.58 2.56 2.57 2.00 2.55 ns 2.33 2.40 2.40 2.00 2.35 ns

Note: ns = not significant.a P-value associated with a Kruskal-Wallis ANOVA test for the ratings, except for yield by weight, yield by volume, yield of tortillas, anthesis, and harvest, which are associated with a parametric ANOVA.b Expected yield (kg/ha) calculated from the best, worst, and more frequent yield declared by farmers for each maize type, following the method of the triangular distribution (Hardaker et al. 1997).c In kg/local unit of volume (almud).d Number of tortillas/almude Number of days to anthesis (male flowering)f Number of days for the crop to be ready for harvest

63

suggest a trade-off between duration andgood performance for other traits. Alltypes, however, are consideredparticularly inferior for storageproperties. These results are consistentwith those obtained from the folk maizetaxonomy exercise, in which farmersexpressed that planting date—andtherefore the uncertainty of the durationof the growing season—was veryimportant. While Blanco maize had ahigh yield, multiple uses, and was easyto sell, it also had the longest growingcycle. Its longer duration was a negativecharacteristic if the rains were delayedand it had to be planted late, becausethen the crop risked being exposed todrought and to frost. As noted, the othermaize types had shorter growing cycles(white > yellow > black > red) andprovided farmers with the flexibility torespond to the uncertain onset of therains, even though they were inferior forother characteristics.

Women’s assessments of the four maizetypes showed statistically significantdifferences for a lower number ofcharacteristics than men’s assessments.For example, unlike men, women did notconsider differences for consumptionqualities such as taste of tortillas,nixtamal quality, tlayudas, and tamales,but they did for ease of shelling. All ofthese characteristics have to do withaspects of maize preparations they areresponsible for making. Womenprovided much lower estimates for yieldby weight and duration, but theirordering of these characteristics wassimilar to men’s. An important differenceis that they considered that Amarillo,Negro, and Belatove had higher stabilitythan Blanco. In general they rated

colored maize types much better than mendid. In particular, women perceivedcolored maize types to perform bettercompared to Blanco than men did, so thetrade-off between good performance andduration was not as strong among womenas among men. Colored maize types maybe more important for females than formales, and women may be playing animportant role in their conservation.

The performance of any new varietyintroduced into this area of Oaxaca couldbe rated with respect to thesecharacteristics by a panel of farmers topredict how the variety might fit into theproduction system, which varieties itmight displace, and how it wouldcomplement other varieties. For example,a shorter duration white maize type equalin other respects to the white typecurrently in use could displace the coloredmaize types since it would decrease thetrade-off between desirability andduration. On the other hand, improvingthe storage quality of colored maize typesmay encourage their conservation.

Attainment index

Ideally these two types of ratings(demand and supply of characteristics)could be combined into a single measureto indicate how well a particular varietyor technological option meets all of theinterests and needs of a farmer or group offarmers. This attainment index20 wouldaggregate the performance of a variety ortechnological option over allcharacteristics that are important to afarmer, while taking into considerationthat the importance of the characteristicsis not equal. Having very goodperformance for a characteristic that isvery important for a farmer—in other

20 This concept and term have been used in the economics literature to describe the extent to which a service-provider meets customerexpectations (Reed et al. 1991). The concept has also been used to explain the adoption of rice varieties (Sall et al. 1997).

64words, it meets his/her interests orneeds—is not the same as having verygood performance for a characteristicthat is only somewhat or not at allimportant. Generating an attainmentindex is a complex procedure that isrooted in economic theory and requiresresearchers to make assumptions aboutpreferences. Although methods forproducing an attainment index arebeyond the scope of this manual,interested readers are referred to Reed etal. (1991), and Appendix 4 provides someof the author’s personal reflections onthis very important subject.

Comments: This method is particularlywell suited for assessing crop varieties(as shown in Tables 16 and 17). In theoryit should also be useful for other types oftechnologies, although experiences of itsapplication to other technologies such assoil fertility improvement or pestmanagement options are scant. Thereforethe application of this method to thoseareas is still an open area of research.

The method described here used a scalewith three levels. A scale with morelevels (five, for example) could be usedfor the supply of characteristics. Such ascale could range from “very good” to“good,” “intermediate,” “poor,” and“very poor.” Going beyond five levelsmay be impractical, however. The morelevels used, the more precise the results,but the exercise may become moredifficult for farmers. Using a scale withmore levels becomes particularlyimportant when the technologicaloptions are very similar; it helps todistinguish among them.

This method is analogous to the matrixranking method commonly used inparticipatory research. Ranking is more

intuitive and easier to do with farmersthan rating (ordering items from more toless important or from better to worse).However, if the number of options to beranked is only one or is not the same forall informants, problems may arise. Ifthere is only one option (for example, afarmer plants or knows only onevariety), how can it be ranked? How canresearchers compare the rankings of twofarmers, one who grows two varietiesand another who grows five?21

Obviously this is not a problem ifinformants are presented with a similarnumber of options. Another potentialdifficulty is that several options can beranked, but the best may still beconsidered inferior or vice versa (i.e., alloptions are inferior, but this is the leastbad, or all options are very good, but thisis the best). These issues cannot beaddressed by ranking alternatives, so itmay be preferable to rate them. Themethod presented here also rankstechnological options, but it does soindirectly, based on the ratings.

Eliciting the Constraintson Using a TechnologyGoal: Identify the factors that farmersperceive as constraining the use of atechnology or practice.

Rationale: Even a well-known andappreciated technology may not be usedby all of the farmers who want to use it.Factors beyond the specificcharacteristics of the technology mayconstrain its use. Although comparisonsof different technologies provide someimportant information about thesefactors, it is useful to have a specificmethod to identify them.

21 There are methods to standardize the rankings from different numbers of options; see, for example, Smith et al. (2000).

65

Method: Researchers identify whichtechnologies or practices will be evaluated(see “Eliciting Farmers’ Perceptions ofTechnological options,” p. 50, for how todo this). An interviewer asks a set of keyinformants or focus groups:

What do you do, or what could you do, ifanything, to solve a particular problem (forexample, to improve a soil, cope withdrought, and store the harvest in a waythat protects it better from insects)?

The answers to this question provide a setof available technological options. Foreach of these options, the interviewer asks:

Has anyone among the group used the option?

Table 18. Technological options available to farmers in Chihota, Zimbabwe to improve their soils, and theconstraints they face, by local soil type

Local soil type

Rebani/ Mhukutu/ Churu/ Rondo/Technological option and constraint Jecha Shapa Rukangarahwe Doro Bukutu Rechuru Chinamwe Chidaka

Apply termite mound soilShortage of termite mounds x x x xShortage of labor to dig and move mound x x x x xLabor intensive x x xNo cart to move termite mound x xLow priority for the soil class xDigging mound causes erosion x

Apply manureNo cattle x x x xShortage of draft power x xGarden has priority for manure applications x x x x

Apply fertilizerNo cash to purchase x x x x x xLack of knowledge x x x xNo cash to hire labor x x x x

Apply limeNo cash to purchase x x x x x xLack of knowledge x x x x

Early plantingNo cash to hire labor x x x x

Deep plowingShortage of draft power x x

Early plowingShortage of draft power x x

Fallow the landShortage of arable land x

Raised bedsLabor intensive to raise beds x

Source: Adapted from Bellon et al. (1999).

What factors have limited your ability toapply the option?

If you did not apply that option, what werethe reasons?

The answers should be compiled andtabulated for analysis.

Example: This method was used in theChihota Project to understand theconstraints to technologies that farmersrecognize and could use to improve soilfertility. The technologies and theirconstraints were identified in the contextof farmers’ own soil taxonomy(presented earlier). Table 18 shows theresults of this exercise.

66The constraints reflect a number ofunderlying themes. The two mostcommon themes were 1) scarcity ofinputs and lack of access to them(including local inputs, such as manureand termite mound soil, and purchasedinputs, such as fertilizers and lime) and2) scarcity of labor to apply inputs,caused by the labor-intensive nature ofthe operations, by the lack of labor, or bythe lack of cash to hire labor. Otherthemes that emerged were the highpriority given to alternative uses forinputs (farmers preferred to applymanure to gardens rather than fieldplots) and the low priority given toimproving some soil classes (e.g.,Rukangarahwe). The lack of implementsand power were also cited as limitations,although these constraints relatedspecifically to the practices of deepplowing and application of soil fromtermite mounds. Farmers also noted thatthe lack of land limited the frequencyand duration of fallows. Several farmergroups mentioned that the lack ofknowledge about application rates forfertilizer and the use of lime was aconstraint.

Demonstration Fields andField DaysGoal: Expose farmers to newtechnologies, such as varieties, practices,and inputs, and get farmers’ feedback onthe new technologies.

Rationale: If scientists, extension agents,or some other external agent would likefarmers to evaluate or adopt newtechnologies, farmers need to getacquainted with these technologies in away that costs them little money, time,and risk. Even before farmers can decidewhether they want to experiment with a

new technology or practice, they need tosee it. Demonstration fields and field daysare organized to accomplish this goal. Thefield days can also give scientists andextension workers information in asystematic way about farmers’perceptions of new technologies.

Method: Researchers, extension agents, orother interested groups (e.g., staff of non-governmental organizations) establish oneor several demonstration fields, whichmay be located on farmers’ fields or onexperiment stations. The demonstrationsmay be established and managedexclusively by the researcher/extensionworker or together with farmers.

The demonstration field is divided intoplots containing the set of technologies tobe shown to farmers. The technologiesshould be presented in a way thatdistinguishes them from one another asclearly as possible (for example, bypartitioning the plots so that eachtechnology is obvious to observers). Thetechnologies should be laid out as simplyas possible. Avoid complex designs thatobscure the characteristics of eachtechnology. A demonstration field is not acomplex multifactorial experiment.

Good and sufficient information abouteach technology should be presented nextto the plot it occupies.

Demonstration fields showing cropvarieties are straightforward. Each varietyis planted in plots of a few short rows(e.g., four rows, each 6 m in length). Theplot is labeled with a sign giving the namethe of the variety, its duration, yield,performance under drought, and anyother information that may interestfarmers. It is advisable to includecommonly planted local varieties tofacilitate farmers’ comparisons betweennew and current varieties.

67

Showing the effects of inputs, rotations,and other agronomic practices indemonstration fields can be moredifficult than displaying new varieties.For example, different levels of pestattack with different cultural controlpractices must be shown with numbers,even if there is a demonstration plot. Thesimplest and most recommended way isto plant adjacent plots with and withoutthe input, rotation, or other practice,making it easier for farmers to judge theimpact of each practice.

Often a goal of demonstrations is toshow the impact of different rates of aninput or of different inputs together. Inthis case, demonstration plots should beorganized in an incremental way. Toshow the impact of different input rates,the first plot has the lowest input rate,the adjacent one the next rate, and so on.To help farmers compare the effects ofseveral inputs together, the first plotincludes just one input, the next includestwo, and so on, until the last plot has thefull package of inputs. The inputs shouldbe ordered from the one with the highestreturn to investment to the one with thelowest return. It is important toremember that in some cases the inputwith the highest return may be the mostexpensive or difficult for farmers toobtain. In that case, the order of inputsshould be adjusted from the one with thehighest return to investment to the onewith the lowest return, subject to theconstraints faced by farmers.22

Demonstrations with technologies thatinvolve impacts over more than oneseason (e.g., rotations, applications oflime) are even more complex to present,because the benefits do not accrue duringthe growing season when the

demonstration is established. This meansthat the demonstration will need to berepeated the next season, which must beplanned from the start.

Once the demonstration fields areestablished, field days can be organizedfor farmers to come and look at them. Thenumber of participants is an importantvariable for the way these days areorganized. If few farmers participate, amore in-depth discussion about each ofthe technologies can take place. With alarge number of participants this usuallyis not possible.

Example for germplasm: An interventionof the Oaxaca Project was to providefarmers with access to the diversity ofmaize landraces present in the region. Thisdiversity was represented by a set of 16landraces and one improved varietychosen by farmers and scientists.Demonstration plots with these 17materials were established in theparticipating communities. The aim of thedemonstration was to enable farmers fromeach community to see the 17 materials,especially their plant and earcharacteristics, and to purchase the onesthey wished to experiment with. Thevarieties included ten with white grain,three with yellow grain, three with blackgrain, and one with red grain. They wereplanted in small plots, each with fourrows, and grouped by color so farmerscould compare them. Each plot had a signgiving the identification number for thevariety and information on yield, plantheight, and drought resistance. Figure 8shows the layout of a demonstration fieldin the Oaxaca Project.

The demonstration plots were establishedunder irrigation during the dry season.This schedule meant that the field day

22 This presupposes an economic analysis of the inputs under farmers’ conditions.

68

could be held just before planting in therainy season, so farmers who purchased amaize variety for experimentation couldplant it soon after the field day.

For the field day, the two inner rows ofeach plot were harvested and theharvested ears were put next to the plot.Farmers were organized into smallgroups of five together with one guide(usually a student from the localagricultural school) to tour thedemonstration. They were toldbeforehand that the purpose of the fieldday was to show them an array of maizevarieties from the region and sell themany variety they found interesting.During the tour, the guide recordedfarmers’ opinions, positive and negative,regarding the varieties. Farmers wereencouraged to note the identificationnumbers of the varieties they wanted topurchase. At the end of the tour theyproceeded to a stand where the seed wasoffered for sale. The seed was sold at theprice of local maize seed in the region.Sales were recorded along with

information on the purchaser (name andaddress) so that researchers could followup on the impact of this process.

Example for soil fertility management:An intervention of the Chihota Project wasfor farmers and researchers to establish anumber of trials with new soil fertilityimprovement technologies developed bySoil Fert Net. Farmers managed the trials,and farmers and scientists designed themtogether. These trials were not typicalscientists’ trials but were simplified to fitfarmers’ interests and management. Theyhad a dual role. On the one hand, theywere a joint experiment between scientistsand farmers to assess the technologies; onthe other hand, they served asdemonstration plots to expose otherfarmers to the technologies.

One of the technologies assessed was theuse of lime together with nitrogenous (N)fertilizers, because low pH is an importantproblem in these soils. The trial/demonstration plot had a simple design inwhich a maize crop was planted in a fieldof 0.1 ha. Half of the plot was treated withlime and the other half was not. Themanagement was exactly the same for bothhalves of the plot in all other respects—variety, number, and timing of weedings,and fertilizer application.

Just before harvest, farmers from thevillage where this trial/demonstration plotwas established were invited to visit it. Thecriteria that farmers used to judge thedemonstration were the growth of theplant stand and how green the maizeplants looked. Farmers could readily seethe difference between applying and notapplying lime. During the field day aninteresting discussion took place abouthow to finance the purchase of lime.Farmers in the village were applying8 bags of N fertilizer per hectare(ammonium nitrate and Compound D), for

Figure 8. Layout of a demonstration field, Oaxaca Project.

Note: The color refers to the grain color of the maize planted in the plot.

White White White White White

White Yellow Yellow Yellow

Black Black Black Red

Starting point

End

White White White White

69

which they paid approximately Z$ 450/bag. One bag of lime cost Z$ 60 and 8bags were recommended. By sacrificingone bag of N fertilizer, farmers could payfor almost all the lime required. If there isa synergistic interaction between Nfertilizer and lime in these soils, it may beworthwhile for farmers to buy the lime. Itwas decided that the next demonstrationexperiment should test the substitution ofsome N fertilizer for lime.

Other demonstrations of the limingpractice were not so straightforward.They compared the farmers’ rate of Nfertilizer and lime with the managementpractices recommended by the extensionservice, which included a higherapplication of N fertilizers, potash,phosphate, and lime. Although thedifferences between plant stands in thetwo treatments were striking, it wasimpossible to identify how each inputcontributed to the overall result.Furthermore, farmers thought it would bedifficult to purchase all of the inputs. Forfarmers who had strong financialconstraints on purchasing inputs, thesecond type of demonstration ultimatelyproved less useful than the first one. Analternative for this type of demonstrationis a layout in which farmers’ practice andthe extension recommendation areseparated from the addition or lack oflime. This design should allow farmers toidentify the impact of lime independentlyof the impact of other fertilizers anddifferent fertilizer rates. Figure 9 shows alayout for this type of demonstration.Note that there are paths to access thedifferent treatments and a central place tosee the four treatments at the same time(the circle in the figure).

Comments: In many cases, experimentalplots designed to fulfill scientists’experimental needs are used as

demonstrations because research andextension systems lack the resources tomount special demonstrations. Theproblem with using experimental plots isthat their randomized layout and testingof several factors often make it verydifficult for farmers to draw lessons orconclusions about the technologies(treatments) displayed. To the extentpossible this should be avoided.

It is important to keep a list of whoattends a field day and to obtain at leastsome basic socioeconomic informationon the participants (see the earliersection, “Minimum Set of SocioeconomicIndicators,” p. 33). This information isuseful for following up on the impact offield days on participants and forunderstanding the distribution ofbenefits among them.

Carrying Out Experimentswith FarmersGoal: Help farmers improve their ownexperiments by providing some basictraining and guidelines (theexperimental agenda and the process arecompletely in farmers’ hands); or helpfarmers evaluate new technologies andpractices selected jointly by farmers andresearchers.

Lime (L) No lime (NL)

Farmers’ L + F NL + Fpratice (F)

Recommended L + R NL + Rpractice (R)

Figure 9. Layout of a demonstration field with twofactors, Chihota project.

70Rationale: Many farmers conductexperiments on their own, although theirexperiments usually differ from those ofscientists in three ways: farmers usuallydo not include a control treatment,farmers may not keep all factors constantaside from the experimental factor, andfarmers usually do not replicateexperiments. An additionalconsideration is that farmers rely only onsimple observation to judge the results ofan experiment. These characteristics canmake it difficult to interpret experimentalresults and derive clear conclusions.Scientists can help farmers improve theirexperiments.

Method: Rather than relying on aspecific methodology, this processinvolves training farmers to conduct andinterpret experiments, using theguidelines given in the followingparagraphs.23

Test only one factor at a time. Researchersshould not use a multifactorialexperiment if they want farmers to learnfrom it. Multifactorial experiments arefor scientists, not farmers. If there areseveral factors, each one should be testedindependently in a different field orsection of the field. Although somefarmers can work with multifactorialexperiments, they may need to be trainedto understand what these are and how tointerpret them.

Emphasize to the farmer the need for acontrol treatment. The scientist shouldexplain to the farmer that a controltreatment is important for interpretingthe results of an experiment. If there areseveral independent experiments indifferent fields, researchers should usethe same control treatment to facilitate

comparisons and interpretation of theresults. For example, if cattle manureapplied at a specific rate is the commoninput used to maintain soil fertility, andresearchers want to compare it to otherinputs such as inorganic fertilizer, leaflitter, and termitaria, they should ensurethat experiments with these alternativeinputs include cattle manure as a controltreatment.

Emphasize to the farmer that all conditions,except the experimental one, need to be keptequal in his/her field. The farmer candecide what those conditions should be,and they should be very clear in his/hermind. (Making a list of these factorstogether with the farmer is helpful.) Anagronomist may object to having anexperiment in which weeds are left togrow (if that is not its objective), but afarmer may consider that having acertain number of weeds reflects his/hernormal conditions. What the farmer andthe agronomist should understand is thathaving weeds is all right, as long as theexperimental and control treatmentshave a similar number of weeds andweedy plots are a condition relevant tothe farmer.

Establish which indicators and criteria willbe used to judge the outcome of theexperiment. Do not assume that farmersand scientists focus on the sameindicators or have the same criteria tojudge and interpret the outcome of anexperiment. A farmer may focus only onhow good or green a plant stand looks,while a researcher may want to look atthe yield at harvest. One approach is touse the characteristics identified byeliciting farmers’ perceptions of costsand benefits of a technology, described

23 The author thanks José Alfonso Aguirre Gómez for sharing his ideas on farmer experimentation, which are the basis for theseguidelines.

71

earlier. Another approach is for thescientist to discuss several questions withthe farmer:

What do you expect from the experiment?

What elements would you focus on to judge it?

Under what circumstances would you judgeone treatment to be better than the control?

The scientist may want to ask him/herself the same questions and comparethe answers with those of the farmer.Ideally, farmer and scientist should reachconsensus on these three issues, althoughdiffering perspectives are acceptable aslong as both parties are aware of them.

Replicate an experiment among farms, notnecessarily in the same field. As pointed outearlier, farmers usually do not replicateexperiments across space. It is oftenassumed that farmers replicateexperiments over time by comparing thepresent season’s results with those of theprevious season. From a researcher’sperspective this may be a poor practice,because conditions change from oneseason to the next and the comparisonmay not be valid. From a farmer’sperspective, it may not make sense toreplicate an experiment across fields.Replications may appear to be a waste ofresources, and in any event, farmers lackthe statistical tools to identifyrelevant factors.

If it is considered important to replicatean experiment, however, it may befeasible to do so across farms. Thisstrategy also entails problems, becauseexperimental conditions vary acrossfarms. One solution is for researchers toask farmers with replicates of the sameexperiment to agree on the conditionsthat will be maintained constant. Forexample, with all farmers carrying out aparticular experiment, reach an

agreement on the following factors: thesoil type (according to their localtaxonomy); the placement ofexperimental plots; and the number ofweedings, the method used, and theirtiming. The use of farmers’ local soiltaxonomy may help ensure that farmerswith replicates put them in similar soils.

An interesting way of combiningfarmers’ and scientists’ experiments isthe approach of the “mother-baby” trial(Snapp 1999; CIMMYT 2000). A research-managed trial is established in a centrallocation, usually a village, with all of thetechnological options to be tested andappropriate controls and replicationaccording to scientific standards (the“mother” part of the trial). Nearby,within easy access to farmers in thevillage, a set of farmers’ experiments isestablished (the “babies”). Theseexperiments include a subset of thetechnological options of the mother trial,they follow a simple design based on theguidelines presented earlier, and they areestablished and managed by farmers.The conditions and management of thebaby trials should reflect farmers’circumstances and interests. Thisexperimental layout yields results thatare of interest and have credibility forboth scientists and farmers.

Example: During field days in theOaxaca Project, researchers learned thatmany farmers were skeptical about howthe varieties would perform under theirown management conditions, soresearchers proposed a set of jointexperiments with the varieties. Theyfurnished the seed and a simpleexperimental design, and farmersprovided the fields and the management.Initially the idea was that four farmersfrom each community would participate,but certain communities had more

72

24 All 29 farmers received seed, but 3 did not plant the experiments and another 6 harvested nothing because of pests, early frost,excess water, and lodging.

25 Researchers should be careful not to read too much into these extrapolated yields, because there is great variation within fields. Theyields should be interpreted as indicative and compared only to others obtained through similar means.

volunteers; eventually 29 farmersparticipated.24 Each farmer agreed toplant three of the varieties included inthe field day, plus one of his/her ownvarieties, and to manage them in exactlythe same way. Each variety was plantedin four rows of approximately 10 m. Oneof the varieties was a common check.

Researchers imposed no managementconditions but systematically monitoredand recorded what farmers did. Duringthe growing season, researchersorganized visits by one group of farmersto other groups in different communitiesso that they could assess the performanceof the varieties under differentenvironmental and managementconditions and discuss the experimentswith the other participating farmers.

At the end of the season, researchers andfarmers harvested the plots together andmeasured the yield. The maize washarvested from two different 5 msections taken at random from each ofthe two inner rows of each plot.Researchers measured the distancebetween rows, the distance betweenplants, and the number of plants plantedper hole to determine planting densityand extrapolate yield per hectare.25

Farmers kept the harvested maize andrated the agronomic performance of thematerials with respect to a set of traits(see “Analyzing the Demand ofCharacteristics” and “Analyzing theSupply of Characteristics,” pp. 58-63).Farmers verified that the varietiesworked well under their managementand environmental conditions.

Comments: The main goal ofexperimenting with farmers is to addresstheir information needs about newtechnologies and solutions to problemsin a way that is relevant, cheap,systematic, and has low risk for them.Ideally, farmers’ and scientists’interactions regarding experimentationwill produce information that is crediblebut not too costly for all parties involvedin the experimental process. This meanssimplifying the experimental design sothat it does not take too much of farmers’time and labor, yet it produces resultsthat are relevant to farmers and ofinterest to scientists. Although there is acompelling need for simplicity and easeof interpretation, these experiments canbe useful for scientists and evensubjected to certain statistical analyses.For example, the data collected in farmerexperiments can be used to carry out amodified stability analysis (Hildebrand1984; Kamara et al. 1996) (see Appendix 2for an example).

Note, however, that some technologiesdo not lend themselves to this type oftrial because they are very complex andinteract with many different factorssimultaneously, so by focusing onindividual factors, one may not really getthe “big picture.”

Another approach for the interactionbetween farmers and scientists regardingfarmers’ experimentation is for scientiststo provide new techniques and scientificconcepts to fill key gaps in farmers’knowledge and not necessarily to changethe farmers’ own styles ofexperimentation.

73

• Impact ideally should involve measuring“objective” impacts (e.g., changes innutritional status, labor allocation,productivity, and income) as well asdetermining “subjective” impacts (e.g.,changes in perceptions of well-being withthe adoption of the new technology).

• The same technology may have acompletely different impact on thevarious members of a household.

• A new technology may have unintendedimpacts, both positive and negative.

• A new technology may affect people whowere never considered in itsdevelopment and implementation.

This manual presents methods forassessing the perceived impacts of a newtechnology on its intended beneficiaries,including different members of a farminghousehold. While the manual alsoattempts to deal with unintendedimpacts, it does not consider impacts onpeople who are not members of thetarget group, such as urban foodconsumers or farmers located outside thestudy area.

The Impact AssessmentProcessGoal: Assess the changes that thefarming household perceives to haveoccurred as a result of adopting a newtechnology or practice. These changes

Assessing theImpact of NewTechnologies

The adoption of a new technology or apractice changes the way that farminghouseholds operate, the costs they incur,and the benefits they generate and/orreceive. As pointed out earlier(“Evaluation of Current and NewTechnological Options,” p. 49), anytechnology represents a particular way ofsolving one or several problems, andideally it translates into an increase in thefarming household’s well-being. Inassessing the impact of a technology,researchers want to determine whetherthe new technology has really addressedthe needs and/or desires of the intendedbeneficiaries and whether it in fact hascontributed to increasing their well-being. A new technology may also havemany unintended consequences,including positive and/or negativeeffects on people that were not targetedoriginally, and it is important to learnabout these other impacts.

The Complexity ofAssessing ImpactsThe complex nature of impactassessment has several sources:

• It is often difficult to separate the changesbrought about by the adoption of a newtechnology from the effects of otherfactors that are unrelated to the newtechnology.

74may be positive or negative and may benot be the same for different householdmembers.

Rationale: Changes brought about by theadoption of a new technology or practiceideally should translate into increasedwell-being for all members of thefarming household, but unfortunatelythis is not always the case. For thisreason, it is important to establish whichchanges have been brought about by thenew technology/practice and the extentto which these changes have increased ordecreased the well-being of the membersof the household. Obviously, such anassessment depends on householdmembers’ perceptions of well-being.

Method: Given the complexity of impactassessment, this manual presents a set ofguidelines rather than a fixedmethodology for assessing impact.Although the focus is on the perceptionsof impacts rather than on the actualimpacts, the guidelines presented heremay be appropriate for both.

Establish a set of impact indicators. Impactindicators are a set of variables,conditions, and/or perceptions that bothfarmers and scientists expect to changewith the adoption of a certain newtechnology or practice. These indicatorsmay be different for farmers and scientists.

The first step is to identify indicators ofwell-being that are relevant for differentmembers of the farming household.Many of these indicators should havebeen identified in the diagnostic phasedescribed earlier, for example during theclassification of farmers or the wealthranking (see “Diagnosis of Farmers’Conditions,” p. 24). It is also possible tohave discussions with key informants orgroups of different types of householdsand household members to identify

which conditions signal that they aredoing well (e.g., they have no need tobuy food during the year, or they haveadditional time for new activitiesor leisure).

The second step is to identify indicatorsof the changes that may result fromusing the new technology. To do this,scientists and key informants or groupsof different types of households orhousehold members discuss thefollowing question:

If you adopt this technology, what do youexpect to be different?

This question may seem vague, but thepoint is to be as open and broad aspossible—in other words, to “cast the netwidely.” Besides identifying theindicators, the answers to this questionallow farmers and scientists to discusswhich indicators are reasonable andwhich are not, or, put another way, whatis reasonable to expect from atechnology. Far-fetched expectations maydisappoint farmers and create aperception of failure, even when atechnology may have had verypositive impacts.

Once the indicators have been identified,the next step is to relate the two sets ofindicators, since not all indicators ofwell-being may be relevant to the specifictechnology being adopted. Researchersshould also ask themselves the samequestions so that they develop one list ofindicators for farmers and another forthemselves, which may or may notcoincide but will be explicit.

Establish a baseline. Since impactassessment is based on an analysis ofchanges, it is fundamental to generate abaseline to which changes can becompared. The baseline describes and, ifpossible, measures the impact indicators

75

that have been identified, and anyassociated relevant conditions, before anew technology/practice is adopted. Therelevant conditions depend on thetechnology/practice to be adopted,particularly with respect to the currenttechnologies or practices that may bemodified or displaced by the new ones.Ideally, the baseline should be doneamong a random, representative set offarming households so thatgeneralizations can be made.Alternatively, it can be done among keyinformants or focus groups thatencompass the range of potentialbeneficiaries of a new technologyor practice.

Establish a monitoring system. Once theindicators have been established and thebaseline done, researchers should followup systematically among a sample orsubgroup of people who participated inthe baseline. A follow-up consists ofvisiting the sample or subgroup andcollecting information on the impactindicators from them. To identifyunintended impacts, the follow-up visitshould also feature an open-endeddiscussion of people’s views, positiveand negative, of the adoptedtechnology/practice. Obviously, thefollow up cannot be done immediately.Time (at least a year) has to pass betweenthe introduction of a new technology/practice and the first follow-up, andseveral follow-up visits may be made atsubsequent intervals. Unfortunately, lackof funding may constrain the ability tocarry out these visits, but a systemshould be in place so that if they do takeplace, the information collected is valuable.

Carry out a final assessment. At some pointafter a new technology or practice hasbeen introduced and (one hopes)adopted, a “final” assessment should be

done. The idea of a “final” assessment isslightly artificial, because the impacts of atechnology will probably continue tounfold after the impacts study has beencompleted, but funding considerations orthe closure of the research project make itimportant to choose a specific time to carryout this assessment. The final assessmentconsists of a dialogue that includesscientists, farmers who adopted the newtechnology/practice, and farmers who didnot. Ideally the discussion will includefarmers who participated in the baselineanalysis, but it need not be restrictedto them.

The dialogue is based on a discussion offarmers’ and scientists’ perceptions of thechanges that occurred in the impactindicators as the result of adopting thetechnology. The discussion should includean open-ended consideration of positive aswell as negative changes and shouldparticularly try to identify the unintendedimpacts of the technology. For example, thediscussion could be guided by thefollowing questions.

Earlier you said that you expected changes inthese things (refer to the indicatorspreviously identified).

Do you think that those changes have occurred?

Have they been positive or negative for you,and why?

Have changes occurred with the adoption of thistechnology that you did not expect or foresee?

Or, in more general terms:

What do you do you consider has changed inyour livelihood with the adoption of (name thetechnology/practice)?

Which of those changes do you consider to bepositive, and why?

Which of those changes do you consider to benegative, and why?

76This dialogue can be organized as aseries of group discussions betweenscientists and farmers, includingdifferent members of the household. Itmay also include a more formal survey,particularly with those in the baselinestudy. A more formal survey mayinclude techniques for evaluation of thenew technological options, such as therating methodologies explained above. Itmay also include specific questionsbased on the indicators identified byboth farmers and scientists, such as thenumbers and types of varieties nowplanted, adoption of the new soil fertilityimprovement technologies, knowledgeof new concepts, application of newtechniques, and so on. The results of thisdialogue should be documented andused to reassess the new technology/practice.

Example: The Oaxaca Project includes animpact assessment component. To assessthe impact of this project, farmers andscientists established a set of indicators(Table 19). Farmers’ indicators referredmainly to enhanced food security andaccess to landraces with valuable traits,either by recuperating old materials oracquiring new ones. Scientists’ indicatorsreferred to an increase in the diversity oflandraces grown at the household andcommunity levels, as well as the geneticdiversity present in those landraces.

The baseline study of a representativerandom sample of farming households inthe six communities provides a controlagainst which researchers and farmerscan eventually compare the changes

resulting from the Oaxaca Project (which,as of this writing, has not yet concluded).The baseline study included questions onmaize requirements, distribution ofyields, storage practices, numbers andtypes of maize currently and no longergrown, and a rating of traits for eachmaize type grown. All information wascollected for male and female membersin the household involved in maizeproduction, preparation, andconsumption. The baseline also includeda collection of the maize types grown bya subsample of these farmers.

During the different interventions(demonstrations and field days, trainingsessions, joint experiments), researchersrecorded the names and addresses ofparticipants and selected a randomsample of participants for monitoring. Atthe end of the growing season after thedemonstrations took place, these farmerswere interviewed about their ownsocioeconomic characteristics, the maizetypes they grew, and their perceptions ofthe landraces they bought, including asystematic rating of their characteristics.Additionally researchers and farmersparticipated in open-ended discussionsabout gaining access to these “new”landraces. The discussions yieldedinformation on unforeseen impacts. Forexample, the availability of a shortduration, red-grained maize (Belatove)had two advantages. First, it offeredsome farmers the possibility of growingtwo crops a year. Second, it gave othersthe opportunity to plant and harvest anearlier maturing crop that provided

Table 19. Impact indicators identified by farmers and scientists in a participatory research project, Oaxaca, Mexico

Farmers’ impact indicators Scientists’ impact indicators

Maize harvest does not get spoiled while stored Farmers grow new maize types with desirable traitsLess need to purchase maize Farmers grow more maize typesRecover desirable maize types that were lost Genetic diversity increases at the household levelIdentify new maize types with good consumption and/or sale characteristics Genetic diversity increases at the community level

77

maize for home consumption whenstores from the previous year werefinished, thus decreasing the need topurchase maize. Another of the landracesoffered was in great demand by women,who liked its purple husk for makingtamales (a special maize preparationwrapped in the husk). Before the project,this maize type was very rare, but now itseems to be diffusing rapidly throughoutthe region.

It is too early to provide an accuratepicture of the impacts of the OaxacaProject, but to date, the monitoringshows that the project is having animpact on farmers’ indicators and tosome extent on scientists’ indicators(although the impact on genetic diversityhas not yet been established). Farmershave shown great enthusiasm forpurchasing diverse sets of landraces.During the 1999 demonstrations, 804 kgof seed were sold in 197 purchase events(a farmer purchasing seed of a landrace),with a total of 123 farmers (27% female)purchasing seed (the same farmer couldpurchase seed of more than onelandrace). As part of the follow-up,researchers also interviewed farmerswho did not participate indemonstrations and experiments. Thesefarmers explained that they had chosennot to participate because they thoughtthat the landrace varieties offered wouldnot work under their conditions, theylacked time to participate in projectactivities, and they did not want to takethe risk of planting landraces that theydid not know.

The example provided here is not typicalof most new technologies offered tofarmers, because the technologies in thisinstance are sets of landraces. Morecommonly new technologies consist ofimproved varieties, inputs, andimproved crop management practices.However, the basic procedure illustratedabove is applicable to other kinds oftechnology.

Comments: As pointed out, impactassessment is complex and ideallyincludes subjective as well as objectiveindicators. Because subjectiveperceptions may not coincide withobjective conditions and vice versa, ifresearchers focus exclusively on one orthe other kind of impact, they willdevelop an incomplete picture of the trueimpacts of a new technology and/orpractice. It is also important to rememberthat externalities—unintended impactsthat go beyond the target group—shouldalso be taken into account in the impactassessment. There is no scope to discussthis subject in this manual, but it iscovered extensively in the literature onimpact assessment.

78

Conclusions

new technologies has dealt with the thirdone, and the section on impactassessment has dealt with the finalactivity. Although this manual hasexplained participatory researchactivities in a fixed order, by now readerswill understand that these activities arenot steps in a strictly linear process.During research on the interventions, oreven the impact assessment, newunderstanding can be generated andinterventions can be modified orchanged. Finally, readers should bereminded that they should pick andchoose the methods that are relevant fortheir work rather than launching intosome predetermined scheme. Thespecific methods selected should fit intoa coherent plan for technologygeneration, rather than being one-offexercises.

Farmer participatory research inagriculture is, above all, a systematicdialogue between farmers and scientiststo solve agricultural problems. Themethods presented in this book are toolsfor guiding and organizing thatdialogue. As the reader has seen, thismanual and the agricultural researchprojects it describes are structured toreflect a sequence of participatoryresearch activities that can besummarized as follows:

• learn from farmers;

• identify technological options to test;

• design a method to test them; and

• evaluate their impact.

Participating farmers and scientistsimplement all these activities jointly. Thesection on diagnosis in this manual hasdealt with the first two activities, thesection on the evaluation of current and

79

References

Adams, A.M., T.G. Evans, R. Mohammed, and J. Farnsworth. 1997.Socioeconomic stratification by wealth ranking: Is it valid? WorldDevelopment 25: 1165-1172.

Ashby, J.A., T. Garcia, M. Guerrero, C.A. Quiros, J.I. Roa, and J.A.Beltran. 1995. Institutionalising farmer participation inadaptive technology testing with the CIAL. ODI AgriculturalResearch and Extension Network Paper No. 57. London, UK:Overeseas Development Institute (ODI).

Bellon, M. R. 1991. The ethnoecology of maize variety management:A case study from Mexico. Human Ecology 19: 389-418.

Bellon, M.R., and J.E. Taylor. 1993. “Folk” soil taxonomy and thepartial adoption of new seed varieties. Economic Developmentand Cultural Change 41: 763-786.

Bellon, M.R., M. Smale, and F. Aragon. 1998. Identifying maizelandraces for participatory breeding: Does gender make adifference? Paper presented at the workshop on StrategicResearch on Gender Issues in Rice-Based Household Economy,August 1998, International Rice Research Institute, Los Baños,Philippines.

Bellon, M.R., M. Smale, A. Aguirre, S. Taba, F. Aragón, J. Díaz, andH. Castro. 2000. Identifying Appropriate Germplasm forParticipatory Breeding: An Example from the Central Valleysof Oaxaca, Mexico. CIMMYT Economics Working Paper 00-03.Mexico, D.F.: CIMMYT.

Bellon, M.R., P. Gambara, T. Gatsi, T.E. Machemedze, O. Maminiminiand S.R. Waddington. 1999. Farmers’ Taxonomies as aParticipatory Diagnostic Tool: Soil Fertility Management inChihota, Zimbabwe. CIMMYT Economics Working Paper 99-13.Mexico, D.F.: CIMMYT.

Bellon, M.R., and J. Risopoulos. 2001. Small-scale farmers expandthe benefits of improved maize germplasm: A case study fromChiapas, Mexico. World Development 29(5) (forthcoming).

Bentley, J.W. 1992. The epistemology of plant protection: Hondurancampesino knowledge of pests and natural enemies. In R.W.Gibson and A. Sweetmore (eds.), Proceedings of a Seminar onCrop Protection for Resource-poor Farmers. Chatham, UK: TCAand NRI. Pp. 107-118.

Bentley, J.W. 1994. Facts, fantasies, and failures of farmerparticipatory research. Agriculture and Human Values 11: 140-150.

Bentley, J.W., and G. Rodriguez. 2001. Honduran folk entomology.Current Anthropology (forthcoming).

Biggs, S.D. 1989. Resource-poor farmer participation in research:A synthesis of experiences from nine national agriculturalresearch systems. OFCOR Comparative Study Paper No. 3. TheHague, Netherlands: International Service for NationalAgricultural Research (ISNAR).

Buckles, D. (ed) 1993. Gorras y sombreros: Caminos hacia lacolaboración entre técnicos y campesinos. Mexico, D.F.: CIMMYT.

Buckles, D. 1995. Velvetbean: A “new” plant with a history. EconomicBotany 49: 13-25.

CIMMYT. 1988. From Agronomic Data to Farmer Recommendations:An Economics Training Manual. Completely revised edition.Mexico, D.F.: CIMMYT.

CIMMYT. 2000. CIMMYT-Zimbabwe: 2000 Research Highlights.Harare, Zimbabwe: CIMMYT.

Cromwell, E. 1990. Seed diffusion mechanisms in small farmercommunities: Lessons from Asia, Africa, and Latin America.Network Paper 21, Agricultural Administration (Research andExtension) Network. London, UK: Overseas DevelopmentInstitute (ODI).

Daniel, W.W. 1978. Applied Nonparametric Statistics. Boston,Massachusetts: Houghton Mifflin.

Doss, C. R. 1999. Twenty-Five Years of Research on Women Farmersin Africa: Lessons and Implications for Agricultural ResearchInstitutions; with an Annotated Bibliography. CIMMYT EconomicsProgram Paper 99-02. Mexico, D.F.: CIMMYT.

Edwards, R.J.A. 1987. Farmers’ knowledge: Utilization of farmers’soil and land classification in choice and evaluation of trials.Paper presented at the workshop, Farmers and AgriculturalResearch: Complementary Methods, 26-31 July, Institute ofDevelopment Studies, University of Sussex, UK.

Franzel, S.C. 1984. Modeling farmers’ decisions in a farming systemresearch exercise: The adoption of an improved maize varietyin Kirinyaga District, Kenya. Human Organization 43: 199-207.

Gambara, P., T. Gatsi, and O. Maminimini. 1998. Farmerparticipation in a soil analysis survey in Chihota CommunalArea, Marondera District. Agritex and Department of Researchand Specialist Services, Marondera, Zimbabwe. Unpublishedreport.

Gill, G.J. 1991. But How Does It Compare with REAL Data? RRANotes No. 14. London, UK: International Institute forEnvironment and Development (IIED).

Gladwin, C.H. 1979. Cognitive strategies and adoption decisions:A case study of nonadoption of an agronomicrecommendation. Economic Development and CulturalChange 28: 155-173.

Gladwin, H., and M. Murtaugh. 1980. The attentive-preattentivedistinction in agricultural decision-making. In P. F. Barlett(ed.), Agricultural Decision Making: AnthropologicalContributions to Rural Development. New York, New York:Academic Press. Pp. 115-136.

80Grandin, B.E. 1988. Wealth Ranking in Smallholder

Communities: A Field Manual. Rugby, UK: IntermediateTechnology Publications.

Hardaker, J.B., R.B.M. Huirne, and J.R. Anderson. 1997. Copingwith Risk in Agriculture. Wallingford, UK: CAB International.

Hildebrand, P.E. 1984. Modified stability analysis of farmermanaged, on-farm trials. Agronomy Journal 76: 271-274.

Johnson, A. 1972. Individuality and experimentation intraditional agriculture. Human Ecology 1: 149-59.

Johnson, A. 1974. Ethnoecology and planting practices in aswidden agricultural system. American Ethnologist 1: 87-101.

Kamangira, J.B. 1997. Assessment of soil fertility status foragroforestry interventions using conventional andparticipatory methods. M.S. thesis, Bunda College ofAgriculture, University of Malawi.

Kamara, A., T. Defoer, and H. De Groote. 1996. Selection of newvarieties through participatory research, the case of corn inSouthern Mali. Tropicultura 14: 100-105.

Lamers, J.P.A., and P.R. Feil. 1995. Farmers’ knowledge andmanagement of spatial soil and crop growth variability inNiger, West Africa. Netherlands Journal of AgriculturalScience 43: 375-389.

Lal, R. 1987. Tropical Ecology and Physical Edaphology.Chichester, UK: John Wiley.

Pingali, P., Y. Bigot, and H. Binswanger 1987. AgriculturalMechanization and the Evolution of Farming Systems in Sub-Saharan Africa. Baltimore, Maryland: Johns Hopkins.

Perales, H. 1993. Experimentación campesina. In D. Buckles(ed.), Gorras y sombreros: Caminos hacia la colaboraciónentre técnicos y campesinos. Mexico, D.F.: CIMMYT. Pp. 9-16.

Reid, G.V., M.R. Binks, and C.T. Ennew. 1991. Matchingcharacteristics of a service to the preferences of customers.Managerial and Decision Economics 12: 231-240.

Rhodes, R., and A. Bebbington. 1988. Farmers Who Experiment:An Untapped Resource for Agricultural Development. Lima,Peru: International Potato Center (CIP).

Richards, P. 1986. Coping with Hunger: Hazard and Experimentin an African Rice-Farming System. London, UK: Allen andUnwin.

Sandor, J.A., and L. Furbee. 1996. Indigenous knowledge andclassification of soils in the Andes of Southern Peru. SoilScience Society of America Journal 60: 1502-1512.

Sall, S., D. Norman, and A.M. Featherstone. 1997. Adoption ofimproved rice varieties in the Casamance, Senegal: Farmers’preference. Poster presented at the 23rd InternationalAssociation of Agricultural Economists Conference, 10-16August, Sacramento, California.

Scoones, I., with C. Chibudu, S. Chikura, P. Jeranyama, D.Machaka, W. Machanja, B. Mavedzenge, B. Mombeshora, M.Mudhara, C. Mudwizo, F. Murimbarimba, and B. Zirereza.1996. Hazards and Opportunities. Farming Livelihoods inDryland Africa: Lessons from Zimbabwe. London, UK: ZedBooks and International Institute for Environment andDevelopment (IIED).

Smale, M., and V. Ruttan. 1997. Social capital and technicalchange: The groupements Naam of Burkina Faso. In C.Clague (ed.), Institutions and Economic Development.Baltimore, Maryland: Johns Hopkins. Pp. 183-200.

Smale, M., and A. Phiri, with contributions from G.A. Chikafa,P.W. Heisey, F. Mahatta, M.N.S. Mwanyongo, H.G. Sagawa,and H.A.C. Selemani. 1998. Institutional Change andDiscontinuities in Farmers’ Use of Hybrid Maize Seed andFertilizer in Malawi: Findings from the 1996-97 CIMMYT/MoALD Survey. Economics Working Paper 98-01. Mexico,D.F.: CIMMYT.

Smale, M., A. Aguirre, M. Bellon, J. Mendoza, and I. ManuelRosas. 1999. Farmer Management of Maize Diversity in theCentral Valleys of Oaxaca, Mexico: CIMMYT/INIFAP 1998Baseline Socioeconomic Survey. CIMMYT Economics WorkingPaper 99-09. Mexico, D.F.: CIMMYT.

Smith, K., C.B. Barrett, and P.W. Box. 2000. Participatory riskmapping for targeting research and assistance: With anexample from East African pastoralists. World Development28(11): 1945-1959.

Snapp, S. 1999. Mother and baby trials: A novel trial designbeing tried out in Malawi. Target, The Newsletter of the SoilFertility Research Network for Maize-Based CroppingSystems in Malawi and Zimbabwe 17:8.

Sperling, L., and M.E. Loevinsohn. 1993. The dynamics ofadoption: Distribution and mortality of bean varieties amongsmall farmers in Rwanda. Agricultural Systems 41: 441-453.

Sperling, L., M.E. Loevinsohn, and B. Ntabomvura. 1993.Rethinking the farmer’s role in plant breeding: Local beanexperts and on-station selection in Rwanda. ExperimentalAgriculture 29: 509-519.

Tripp, R. 1989. Farmer participation in agricultural research: Newdirections or old problems? IDS Discussion Paper 226.

Tripp, R., and J. Woolley. 1989. The Planning Stage of On-FarmResearch: Identifying Factors for Experimentation. Mexico,D.F., and Cali, Colombia: CIMMYT and Centro Internacionalde Agricultura Tropical (CIAT).

Tripp, R. 2000. The organization of farmer seed systems:Relevance for participatory plant breeding. Paper presentedfor the symposium on the Scientific Basis for ParticipatoryImprovement and Conservation of Crop Genetic Resources,8-14 October 2000, Oaxtepec, Mexico.

Wilken, G.C. 1987. Good Farmers: Traditional AgriculturalResource Management in Mexico and Central America.Berkeley, California: University of California Press.

81Appendix 1Farmers’ Classification of Themselves, Chihota, Zimbabwe


Young • Energy to conduct operations • Have no farming implements• Grow modern types and varieties of crops • Have no arable land• Opportunities to get new knowledge • Have no children to help with farming

Old • Have farming implements • Have less energy to conduct operations• Have arable land• Have children to help with farming• Grow traditional types and varieties of crops• Have old and new knowledge to impart to new generations

Male • Have better opportunities• Own farming implements; their lands are worked earlier• Are in charge, get the bulk of proceeds from crop sales• Decision-makers• Good planners• Own fields• Prefer garden crops

Female • Grow all types of crops • Do not own farming implements• Better farmers than males • Do not own fields• Do not go beer drinking • Their fields are prepared last• Do a lot of work in the fields • Fall under husbands, not rewarded for their labor• Plan operations in a timely fashion • Not leaders in farming issues• Devoted to work

Have draft animals • Prepare lands on time • Cattle destroy other people’s crops• Have manure • Some prepare fields of others before• Can use livestock for other purposes preparing their own• Operations are easy and timely• Get good yields

Have no draft animals • Late land preparation• Carry out operations late• Field operations are difficult• Have no access to manure• Get low yields

Have cattle • Have manure • Cause soil erosion• Have resources • Can destroy crops of other farmers• Timeliness of operations • Do not have grazing areas• Have milk, meat, and lobola (dowry)• Have access to cash

Have no cattle • Borrow • Delay operations• Provide labor for others • Little manure, low yields• Cattle herders • Lack of resources• Buy cattle from others • Lazy at times

• Cruel to livestock• Gain when their crops are unintentionally

destroyed by livestock

Have manure • Have good plant stands• Get high yields

Have no manure • Have crops of poor quality, low yields

Have implements • Able to perform timely operations• Get good yields• Have necessary implements• Operations are made easy

Have no implements • Delayed operations• Get low yields• Do not have cattle and other implements

Have garden • Stable income from year-round production • Cause soil erosion• Sell produce • Some are encroaching into grazing areas• Provide enough to family • Use large quantities of water• Well-paying enterprise • Do not help the needy• Help others • Small arable lands• Produce several crops a year • Workaholics

82

Have garden • Always busy (throughout year)(cont’d.) • More income than dryland farmers

• Not as affected by adverse weather (droughts)as dryland farmers

• Diversity of crops• Mainly grow horticultural crops

Dryland farmers • Plan ahead of time • Get only one crop per year• High volume of output for storage • Receive income once a year• Market their crops • No stable income because of seasonality• Maintain contours of production• Large landholding to be inherited • No adequate financial resources to• Big arable lands prepare/plan next season• Practice crop rotations • Expose to negative effects of adverse• Have time to relax weather (droughts)

Have large fields • Attain high yields • Spread inputs thinly over large area• Wide variety of crops

Have small fields • Low total harvest• Narrow variety of crops• Apply inputs to smaller piece of land, therefore

sometimes attain high yields

Own fields • Grow crops in both dryland and gardens

Do not own fields • Borrow fields temporarily from other farmers

Have fenced fields • Field protected from livestock

Have no fenced fields

Work outside the area • Can afford to hire labor

Do not work outside • Conduct operations on their own, do not use hired labor

Work in groups • Quick operations • Operations can be delayed by some group members• Team spirit• Share knowledge and experiences• Team up to buy inputs• Share labor• Encourage each other

Work individually • No-one will delay operations • No-one to encourage/correct another• Little knowledge • If sick, no-one will work in field• Slow with operations

Industrious • Work hard in their fields • Workaholics• Attain high yields • Always thinking• Feed visitors• Send children to school• Seek advice from other farmers• Healthy

Lazy • Source of labor for others • Do not perform timely operations• Good messengers • Want to rest more than work

• Beg for food• Feign sickness when rainy season starts• Always away from their farms• Do not feed their families• Do not send children to school• Do not care• Do not follow what others are doing

Adequate cash for • Can hire laborfarming • Can buy inputs early

• Plan well• Timely operations• Attain high yields• Can afford to sell farm produce

Inadequate cash • Buy from others • Operations always conducted late• Attain low yields• Cannot attain high yields• Cannot grow crops well• Not enough production for home consumption

Rich • Stable income • Do not give implements for free• Timely operations• Adequate farm implements


83

• Care about farm activities• Adequate food supply• Adequate inputs (freezer)

Poor • Cannot afford seed, fertilizer• Do not produce adequate amount of food• Delay operations• Lazy and sometimes cruel

Farmers with knowledge • Have resources• Have enough food• Sell to others• Attend Agritex (extension) lessons• Know how to plan for next season• Have good homestead• Know how to look after livestock• Timely operations• Good crops• Attain high yields• Practice crop rotations• Know when to establish crops• Use manure• Not jealous

Farmers without • Operations not timely even if usingknowledge own implements

• Delay operations• Work hard but get poor yields• Do not practice crop rotations• Jealous• Do not use manure• Cause erosion• Do not plan activities• Do not attend Agritex (extension) lessons• Do not know how to look after livestock• Use outdated farming methods

Have Master Farmer • Knowledgeable about farming operationscertificate

Do not have Master • Sometimes not sure of operationsFarmer Certificate

Sell their produce • Feed society • Little food for home consumption• Have cash• Good role model

Subsistence farmers • Produce enough for family • Do not help others• Honest, do not steal from others • Hate those who sell

• Do not send children to school

Perform operations • Have knowledgeon time • Attain high yields

• Have enough implements• Produce good crops

Do not perform operations • Give names to rains to explain their latenesson time • Do not have implements

• Do not attain high yields

Attain high yields • Have high yields every year • Arrogant• Sell to others• Plant early

Do not attain high yields • Lack resources• Attain low yields• Run out of food stocks before next harvest

Plan operations • Operation done on time • Crops can be eaten by livestock• Get good crop stands• Crops wilt if rains arrive late

Do not plan operations • Extensive farmers• No rotations• Lack resources


Rich (cont’d.)

84

Appendix 2

Examples of the Cards Used to Depict Variety Characteristics, OaxacaProject (demand and supply of characteristics)

Taste of tortilla Good for nixtamal Good for tejate Good for storage

Yield

Yield by weight Yield by volume Yield of tortillas Yield stability

Abiotic and biotic stress

Withstands drought Withstands wind Withstands pests Withstands weeds

Management, sales, and duration

Invest labor Cash investment Good for sale Duration

22 F63167473A

216 7 8 9 10 1112 13 14 15 16 1718 19 20 21 22 2324 25 26 27 28 29

3 54

Consumption

85

Appendix 3Examples of the Data Used for Analyzing the Supply and Demand ofCharacteristics

Tables A3.1 and A3.2 (for men andwomen in the same household,respectively) show the data that can beobtained by using the method to elicitthe importance of characteristics of avariety or other technology. Each row is ahousehold and each column is acharacteristic. This table came from aspreadsheet. To perform a statisticalanalysis like the one presented in theexample, the data can be imported into astatistical package such as SPSS (release7.5.3), which was used for this example.With SPSS, researchers used twononparametric tests: the Kruskal-Wallistest for “K–independent samples” andthe Wilcoxon matched-pairs signed rankstest for the “two related samples.” Thissecond test was used for comparing theratings of the importance ofcharacteristics between men and womenfrom the same household. Note that todo this test, the two tables should be putside by side with slightly different namesfor the characteristic, i.e. “withstandsdrought-men,” “withstands drought-women.”

Table A3.3 presents an example of thedata that would be obtained by using themethod to elicit the performance for eachcharacteristic by each variety ortechnological option from men (a similartable should be generated for women,but unlike the previous case the analysisis independent). Each row is acombination of a household and avariety grown by a male farmer, whileeach column is a characteristic. This tablecame from a spreadsheet. Note that eachfarmer may have more then one row,since he may plant more then onevariety. The names of the maize types arepresented, but they have also been codedinto numbers (1 to 4) in an adjacentcolumn. For the statistical analysis, thedata were imported into SPSS (release7.5.3). Researchers used the routinestatistics, nonparametric tests, and (forthe Kruskal Wallis analysis of varianceby ranks) the “K Independent Samples”option. This latter test is used forcomparing the ratings of theperformance for each characteristic acrossthe four maize types.

86Table A3.1. Ratings of importance for each characteristic for men (demand of characteristics), Santa Ana Zegache, Oaxaca, Mexico

Household Yield- Yield- Nixtamal Taste of Yield Ease of Withstands Withstands Withstands Cash LaborID, weight, volume, quality, tortilla, stability, shelling, drought, wind, weeds, investment, investment,men men men men men men men men men men men a men a

1 1 1 2 1 1 2 1 3 2 1 32 2 1 3 3 1 2 1 3 2 1 13 1 2 2 1 1 2 1 3 2 1 24 2 1 2 1 1 2 1 3 1 1 15 2 1 2 2 1 2 1 3 3 1 16 1 2 2 2 1 2 1 3 2 1 17 1 1 1 1 1 2 1 3 2 1 28 1 2 3 3 1 2 1 3 2 1 19 2 1 1 3 1 2 1 3 2 1 1

10 2 1 2 1 1 2 1 3 3 1 211 1 1 3 2 1 2 1 3 3 1 212 1 1 2 3 1 3 1 2 2 1 213 1 1 1 2 1 2 1 2 2 1 214 1 1 2 2 1 2 1 3 3 1 215 1 1 1 2 1 2 2 3 3 3 2

Note: 1 = very important, 2 =somewhat important, 3 = not important.a For cash and labor investment, 1= little, 2= regular, 3= a lot.

87

Table A3.2. Ratings of importance for each characteristic (demand of characteristics) for women, Santa Ana Zegache, Oaxaca, Mexico

Household Yield- Yield- Nixtamal Taste of Yeld Ease of Withstands Withstands Withstands Cash LaborID, weight, volume, quality, tortilla, stability, shelling, drought, wind, weeds, investment, investment,

women women women women women women women women women women women a women a

1 1 2 1 1 1 3 1 1 3 2 12 1 3 1 2 1 1 1 2 2 1 13 1 2 1 1 1 3 1 2 2 2 14 1 2 1 1 2 3 1 3 2 1 35 1 1 1 1 1 3 1 2 3 1 16 1 2 3 1 1 3 1 2 1 1 17 1 3 1 1 1 2 1 2 3 1 18 1 1 1 2 1 3 1 3 1 1 29 1 3 1 1 1 3 1 1 3 1 210 1 1 2 1 1 3 1 3 2 1 311 1 1 1 1 1 3 1 1 2 1 212 1 2 2 1 1 2 1 2 2 1 213 1 3 1 1 1 3 1 1 2 1 214 1 1 1 1 1 2 2 2 2 1 115 1 1 1 1 1 3 1 2 2 1 1

Note: 1 = very important, 2 =somewhat important, 3 = not important.a For cash and labor investment, 1= little, 2= regular, 3= a lot.

88Table A3.3. Ratings of performance of each maize type for each farmer with respect to each characteristic (supply of characteristics), Santa Ana Zegache, Oaxaca, Mexico

Household Number Maize Type Nixtamal Taste Yield Ease of Withstands Withstands Withstands Cash LaborID of maize types type code quality of tortilla stability shelling drought wind weeds investment investment

1 1 Blanco 1 1 1 1 2 1 1 2 2 22 1 Blanco 1 1 1 2 2 2 2 1 2 23 1 Blanco 1 1 1 1 1 2 1 1 2 23 2 Amarillo 2 1 1 1 1 1 1 1 2 24 1 Blanco 1 1 1 2 2 2 2 2 2 24 2 Amarillo 2 1 1 2 1 2 2 2 2 24 3 Negro 3 1 1 2 1 1 2 3 2 24 4 Belatove 4 1 1 2 1 1 2 3 2 25 1 Blanco 1 1 1 1 2 1 2 2 25 2 Amarillo 2 1 1 1 1 1 1 2 2 26 1 Blanco 1 1 1 2 1 1 2 2 2 27 1 Blanco 1 1 1 2 1 1 1 2 2 27 2 Amarillo 2 1 1 2 1 1 1 2 2 28 1 Blanco 1 1 1 2 2 2 2 2 2 28 2 Negro 3 1 1 1 1 1 1 1 2 29 1 Blanco 1 1 1 2 1 1 1 1 2 2

10 1 Blanco 1 1 1 1 1 1 1 2 2 210 2 Amarillo 2 1 1 1 1 1 2 2 2 210 3 Negro 3 1 1 1 1 2 2 1 2 210 4 Belatove 4 1 1 1 1 2 2 1 2 211 1 Blanco 1 1 1 2 2 2 2 2 3 312 1 Blanco 1 1 1 1 2 1 1 2 2 212 2 Amarillo 2 1 1 1 1 1 1 2 2 213 1 Blanco 1 1 1 2 2 1 2 1 2 213 2 Negro 3 1 1 2 1 2 2 2 2 214 1 Blanco 1 1 1 2 2 1 1 1 3 314 2 Negro 3 1 1 1 1 1 1 1 3 315 1 Blanco 1 1 1 2 2 2 1 2 3 3

Note: 1= very good, 2 = intermediate, 3= poor. Each farmer has a different number of maize types, e.g., Farmer 1 only has one type, while Farmer 4 has 4 types.

89

Appendix 4Using an Attainment Index in Farmer Participatory Research

The following discussion is based on theauthor’s intuition, and for that reason itis not included in the main text of thismanual. Although the approachdescribed here differs from the approachof Reed et al. (1991), which has beenused in the published literature, it maystimulate further thinking on thisimportant subject.

The attainment index is a measure of theextent to which the overall performanceof a particular variety or technologicaloption meets all of the interests andneeds of a farmer or group of farmers.Therefore an attainment index combinesthe two types of ratings—the demandand supply of characteristics—discussedpreviously.

It would seem intuitively obvious that avariety or other technology thatperforms very well for many importantcharacteristics should be more desirableoverall than one that performs very wellfor characteristics that are onlysomewhat important. Conversely, avariety or technology that performspoorly for many importantcharacteristics should be less desirablethan one that performs poorly for lessimportant characteristics. The question,however, is how to combine both typesof rating to generate an ordinal measure

that makes it possible to rank thedifferent varieties or technologies frommore to less desirable.

The first possibility that comes to mind issimply to multiply the supply anddemand ratings. The numbers associatedwith these ratings are in any casearbitrary, and what is important is theirorder, not their magnitude. Researcherscould code the ratings by 1 = veryimportant, 2 = somewhat important, 3 =not important, and 1 = very good, 2 =intermediate, and 3 = poor. Multiplyingthe ratings would give a scale between 1and 9 (best to worst) for each trait, and itwould be possible to sum across thecharacteristics. A drawback of this scaleis that it would have many ambiguities.For example, it would imply that thecombination “very important, poor”would be equal to “not important, verygood.” Obviously a variety that performspoorly for a very important characteristicis the worse case, and if a characteristic isnot important, it is irrelevant whether avariety performs very well or poorly, butwith this method the two cases would beequivalent. Furthermore, if a farmerconsiders many traits to be unimportant,the attainment index would be verylarge, indicating that he/she isdissatisfied with the variety, when in factthe opposite may be true.

90A second possibility is to assign arbitraryscores but with certain properties to bothtypes of rating. For ratings of theimportance of characteristics, the scores,could be between 1 and zero (1 for “veryimportant” and zero for “notimportant”). “Somewhat important” canbe assigned an intermediate score suchas 0.4. These scores maintain the order ofimportance, and the zero takes intoaccount that it does not matter how avariety performs for a characteristic thatis irrelevant. (The reason for choosing 0.4for the intermediate rating will beexplained later.)

For ratings of the performance of avariety for a characteristic, the scorescould be between 1 and –1 (1 for “verygood” and –1 for “poor”). The“intermediate/acceptable” rating can beassigned an intermediate score, such as0.5. These numbers maintain the order ofperformance, and the –1 takes intoaccount that a poor performance has anegative impact on the well-being ofa farmer.

Both ratings can be combined in a matrixthat produces an ordinal scale that runsfrom more to less desirable (Figure A4.1).For each cell in the matrix, the scores inthe column and row are multiplied,

generating an index that varies between1 and –1. The ordinal scale is as follows:

Very important–very good (1) > veryimportant–regular performance (0.5) >somewhat important–very goodperformance (0.4) > somewhatimportant-regular performance (0.20) >not important–any performance (0) >somewhat important–poor (–0.5) > veryimportant–poor (–1).

The score 0.4 was assigned to“intermediate importance” to producethe ordering shown above, following theassumption that it is more important ordesirable to have an intermediateperformance for a very importantcharacteristic than to have a very goodperformance for a characteristic that is“somewhat important.” Clearly it ismore desirable to have (1) a variety thathas an intermediate rather than a poorperformance for a very importantcharacteristic, rather than (2) a varietythat has a very good rather than anintermediate performance for asomewhat important characteristic, or avariety that has an intermediate ratherthan a poor performance for a somewhatimportant characteristic.

Alternatively, one could assign an equalscore to both intermediate ratings and

Figure A4.1. Matrix of scores for an attainment index.

Very Somewhat Notimportant important important

1 .4 0

Very good 1 1 .4 0

Intermediate 0.5 .5 .2 0

Poor -1 -1 -.4 0

Supply weights

Demand weights

91

Table A4.1 Demand and supply ratings for several characteristics and two maize types grown by the man inhousehold 4 used for calculating an attainment index, Santa Ana Zegache, Oaxaca, Mexico

Importance Performance

Characteristic Demand score Blanco Supply score Negro Supply score

nixtamal quality, 2 0.4 1 1.0 1 1.0taste of tortilla, 1 1.0 1 1.0 1 1.0yield stability, 1 1.0 2 0.5 2 0.5ease of shelling 2 0.4 2 0.5 1 1.0drought 1 1.0 2 0.5 1 1.0wind 3 0.0 2 0.5 2 0.5weeds 1 1.0 2 0.5 3 –1.0cash 1 1.0 2 0.5 2 0.5labor 1 1.0 2 0.5 2 0.5

Note: Demand and supply scores from Figure A4.1

assume that a farmer is indifferentbetween the two cases presented above(i.e., it is equally desirable to have anintermediate performance for a veryimportant characteristic, or a very goodperformance for a characteristic that is“somewhat important.”)

Then, for each particular variety, thescores for each characteristic can beadded to generate an overall weightedscore of performance—the attainmentindex. The index reflects the overalldesirability of a variety to the farmerwho rated it.

Some farmers may consider somecharacteristics to be unimportant(therefore they will have a zero score),whereas other farmers may not. To takethese differences into account, it isnecessary to normalize the index.Otherwise, when two scores arecompared, one may be very large—notbecause one of the varieties was moresatisfactory, but simply because thefarmer who rated it considered manytraits to be very or somewhat important,whereas another farmer rating the same

variety might consider fewer traits to beimportant (and may even have found thevariety to be more satisfactory). It isimportant to divide the score by a“perfect” score (i.e., the score that wouldhave been obtained if the variety hadscored very well for all relevantcharacteristics, weighted by theimportance of the characteristic). Thismeans that the perfect score is simply thesummation of all demand scores and thatunimportant characteristics are not takeninto account.

To get a measure of the desirability of acertain variety for a community as awhole, the attainment indices for thefarmers in the community can beaveraged. Researchers should be carefulnot to read too much into the actualscores, which are based on arbitrarynumbers. As noted, the important pointis the ordering of the varieties in terms oftheir desirability (ability to supply whatfarmers want).

An example of how to calculate thisindex using these scores follows. Thedata are taken from the man in

92household 4 of Tables A3.2 and A4.1 forthe maize types Blanco and Negro. TableA4.1 presents the data.

For the variety Blanco:

(.4x1)+(1x1)+(1x.5)+(.4x.5)+(1x.5)+(0x.5)+ (1x.5)+(1x.5)+(1x.5)= 4.1

The perfect score to be used fornormalization would be:

(.4x1)+(1x1)+(1x1)+(.4x1)+(1x1)+(0x1)+(1x1)+(1x1)+(1x1)= 6.8

Normalized score:

4.1/6.8= 0.603

For the variety Negro:

(.4x1)+(1x1)+(1x.5)+(.4x1)+(1x1)+(0x.5)+(1x-1)+(1x.5)+(1x.5)= 3.3

The perfect score to be used fornormalization would be:

(.4x1)+(1x1)+(1x1)+(.4x1)+(1x1)+(0x1)+(1x1)+(1x1)+(1x1)= 6.8

Normalized score:

3.3/6.8= 0.485

Hence, Blanco is superior to Negrooverall. However, it should also bepointed out that for ease of shelling andparticularly for drought tolerance Negrois better (although it is much worse attolerating weeds).

The normalized attainment index is moreimportant for comparing differentfarmers, who naturally will differ in theirdemand for certain characteristics.

93

Appendix 5An Example of the Modified Stability Analysis

The kind of data generated from farmers’experiments in the Oaxaca project—yielddata from several varieties grown ondifferent farms in the region—can beanalyzed with a modified stabilityanalysis.

Each experiment located on a farmwould be considered a trial. The averageyield of all varieties included in a giventrial, which is representative of theconditions of crop production at thatlocation (i.e., the environmental index), isplotted against the yield of each varietyin that trial. The relative height of theplotted line represents the general yieldof the variety; the slope represents itsadaptability to different environmentalconditions. A flat slope represents astable response, whereas a steep sloperepresents the opposite. Hildebrand(1984) recommends using a minimum of14 farms (trials) to gain an accurateestimate of treatment differences overenvironments, when there is need for awide range of environments. Clearly it isnot appropriate for farmers to participatein this kind of analysis, although it isbased on data generated by participatoryexperiments. Its results may be useful toscientists, however, and can be useful tofarmers when presented in a simplifiedmanner to discuss the appropriateness ofplanting the varieties tested in differentenvironments.

The dataset from the farmer experimentsin Oaxaca is small (3 to 6 farms, with tworeplicates per farm), but bearing thislimitation in mind, they can still be used toprovide an example of the possibleinterpretation of such an analysis. Yieldsof maize landraces were plotted againstthe environmental index for each farmwhere they were grown during the wetseason of 1999 (Figure A5.1). Asmentioned previously, the yield was theweight of the ears harvested in a 5 m stripchosen randomly in the inner two rows ofthe experimental plot. The six landracesincluded three with white grain, one withyellow grain, one with black grain, andone with red grain. Figure A5.1 shows thatthe red and yellow landraces (varieties 34and 40, respectively) are the most stable(i.e., they have the flattest slope), whereasthe white materials (118, 134, and 152)have a steeper slope. There is a crossoverpoint where the white maize types start toperform better than the other maize types.This crossover indicates that in “poor”environments, the other maize types maybe superior, whereas white maize mayperform better in “good” environments.(Remember that for farmers in Oaxaca,grain color is an indicator of other traits,particularly duration.) Kamara et al. (1996)provide another example of thismethodology for four maize varieties(three improved and one local) evaluatedin three locations of Mali.

2.0

1.5

1.0

.5

0.0.4 .6 .8 1.0 1.2 1.4 1.6

Environmental index

Figure A5.1 Yield response to the environmental index in six communities of the Central Valleys ofOaxaca, Mexico.

Yield

(kg) Variety

152

134

118

42

40

34

Date post:	04-Jul-2018
Category:	Documents
Upload:	buikhanh
View:	219 times
Download:	0 times

Participatory Research Methods for Technology Evaluation Bellon Cymmit prm_al… · Participatory...

Documents