A BIOLOGICAL FRAMEWORK FOR UNDERSTANDING FARMERS' PLANT ... · A BIOLOGICAL FRAMEWORK FOR...

A BIOLOGICAL FRAMEWORK FOR UNDERSTANDING FARMERS'PLANT BREEDING1

DAVID A. CLEVELAND, DANIELA SOLERI, AND STEVEN E. SMITH

Cleveland, David A. (Department of Anthropology and the Environmental Studies Program,University of California, Santa Barbara 93106-3210, USA, [email protected]), DanielaSoleri (Office of Arid Lands Studies, University of Arizona, 1955 E 6 th St, Tucson, AZ 85719,USA; current address, Center for People, Food and Environment, 340 Arboleda, Santa Bar-bara, CA 93110, USA, and Institute for Social Behavioral and Economic Research, Universityof California, Santa Barbara, CA, 93106, USA, [email protected]), and Steven E. Smith(School of Renewable Natural Resources, University of Arizona, 301 Biological Sciences East,Tucson, AZ 85721, USA, [email protected]). A BIOLOGICAL FRAMEWORK FOR UNDER-STANDING FARMERS' PLANT BREEDING. Economic Botany 54(3):377-394, 2000. We present aframework for understanding farmer plant breeding (including both choice of varieties andpopulations and plant selection) in terms of the basic biological model of scientific plant breed-ing, focusing on three key components of that model: 1) genetic variation, 2) environmentalvariation and variation of genotype-by-environment interaction, and 3) plant selection. Foreach of these concepts we suggest questions for research on farmers' plant breeding (farmers'knowledge, practice, and crop varieties and growing environments). A sample of recent re-search shows a range of explicit and implicit answers to these questions which are oftencontradictory, suggesting that generalizations based on experience with specific varieties, en-vironments or farmers may not be valid. They also suggest that farmers' practice reflects anunderstanding of their crop varieties and populations that is in many ways fimdamentallysimilar to that of plant breeders; yet, is also different, in part because the details of theirexperiences are different. Further research based on this framework should be valuable forparticipatory or collaborative plant breeding that is currently being proposed to reunite farmerand scientic plant breeding.

UN MARCO BIOL6GICO PARA ENTENDER EL FITOMEJORAMIENTO DE Los AGRICULTORES. Se pre-senta un marco teorico para un mas claro entendimiento delfitomejoramiento de los agricul-tores (se incluye tanto la selecci6n o identificacion de variedades, poblaciones, o plantas in-dividuales) desde la 6ptica de un modelo biol6gico bdsico. Dicho modelo trata 1) la variaci6ngenetica 2) la variaci6n ambiental, la variaci6n de la interaci6n genotipo-ambiente y 3) laselecci6n de plantas. Para cada uno de los conceptos anteriormente expresados se sugierenpreguntas para investigar el fitomejoramiento de los agricultores (conocimiento de los agri-cultores, prdctica, variedades de cultivo y sus ambientes). Una muestra de la reciente inves-tigaci6n demostr6 un rango de implicitas y explicitas respuestas para las preguntasformuladas,las cuales son en ocasiones contradictorias, lo que sugiere que la generalizaci6n de las ex-periencias basada con espec(ficas variedades, ambientes o agricultores pudieran no ser vdlida.Se plantea que las prdcticas de los campesinos reflejan un entendimiento de sus variedades ypoblaciones que tienen en parte cierta similitud con los fitomejoreadores convencionales,aunque en parte es tambien diferente ya que los detalles de las experiencias de agricultores yfitomejoradores convencionales son distintas. Otras investigaciones basados en este marcopudieran contribuir al fitomejoramiento colaborativo o participativo, lo cual actualmente hasido propuesto para reunificar a los agricultores y los cientificos delfitomejoramiento de plan-tas.

Key Words: collaborative plant breeding; farmer plant breeding; farmer crop varieties, ge-netic variation; genotype-by-environment interaction; landraces; plant selection.

Since the first domestications of wild plants have been responsible for the development ofabout 12 000 years ago, fanner plant breeders thousands of crop varieties in hundreds of spe-

cies (Harlan 1992). Plant breeding as a special-Received 7 July 1999; accepted 2 February 2000. ized activity began about 200 years ago in in-

Economic Botany 54(3) pp. 377-394. 2000I) 2000 by The New York Botanical Garden Press, Bronx, NY 10458-5126 U.S.A.

ECONOMIC BOTANY

dustrial countries (Simmonds 1979). Modernprofessional plant breeding developed in the ear-ly part of the 20th century, based on Darwin'stheory of evolution through selection and the ge-netic mechanisms of evolution developed byMendel, Johannsen, Nilsson-Ehle, East, and oth-ers (Allard 1999; Simmonds 1979). Plant breed-ing by modem, scientific, professional plantbreeders (hereafter simply plant breeders) hasbecome increasingly separated from plant breed-ing by farmers, especially small-scale farmers inhigh-stress growing environments with limitedaccess to extemal inputs (hereafter simply farm-ers) (Berg 1996). This separation is also true ofprofessional and farmer seed-supply systems(Cromwell, Wiggins, and Wentzel 1993). AsSimmonds succinctly stated, "'the current stageof crop evolution is rapidly passing into thehands of professional plant breeders, a trendwhich has been at least locally apparent for 200years" (Simmonds 1979:11). At the same time,many plant breeders consider farmers also to becapable of plant breeding. According to Allard,"The consensus is that even the earliest farmerswere competent biologists who carefully select-ed as parents those individuals ... with the abil-ity to live and reproduce in the local environ-ment, as well as with superior usefulness to localconsumers" (1999:29). Stoskopf et al. wrote thatin the period before scientific plant breedingthere was "considerable progress in plant im-provement" that "can by definition, be said tobe plant breeding" (1993:1).

The emphasis of professional plant breedingtypically has been on developing modem vari-eties (MVs) with geographically wide adaptationto optimal (relatively low stress and uniform)growing environments, and high yield in theseenvironments (Evans 1993; Fischer 1996). Al-though there has also been attention to breedingfor stress tolerance, this attention has focused onrelatively large-scale environments and com-mercial farmers who can afford to purchaseseed, not on the farmers who are the topic ofthis paper (Banziger, Edmeades, and Lafitte1999; Ceccarelli et al. 1994; Heisey and Edmea-des 1999) This contrasts with farmer breedingand farmers' local varieties (FVs, which includelandraces, locally adapted MVs, and progenyfrom crosses between landraces and MVs),which are usually assumed to have more narrowgeographical adaptation to marginal (relativelyhigh stress, and variable) growing environments,

and high yield stability (low variance across en-vironments including years and locations) andmoderate yield in those environments (Harlan1992; Zeven 1998). Landraces are often definedas "geographically or ecologically distinctivepopulations which are conspicuously diverse intheir genetic composition both between popula-tions and within them" (Brown 1978:145).

Collaborative or participatory plant breeding(CPB) is an attempt to bring farmers and plantbreeders together to develop new crop varietiesto meet farmers' needs (Hardon 1996; Smale etal. 1998; Witcombe et al. 1996). An importantimpetus for CPB comes from increasing aware-ness among plant breeders of the need to in-crease the sustainability of agriculture in the faceof environmental deterioration and growing de-mand for production, by placing greater empha-sis on

(I) increasing yields and yield stability inmarginal environments, both (a) thosethat have been high yielding, but whereinputs are being reduced to reduce pro-duction costs and negative environmentalimpacts, and (b) those of many of theworld's farmers who have not adoptedMVs, but whose FVs have inadequateyields. and

(2) conserving the base of genetic diversityon which all plant breeding depends, andwhich is threatened by the loss of FVs asthe area planted to FVs and the numberof farmers growing them declines (Cal-laway and Francis 1993; Ceccarelli 1996;Cooper and Byth 1996; Evans 1997; Fi-scher 1996; Heisey and Edmeades 1999;Sleper, Barker, and Bramel-Cox 1991).

CPB is based on the assumption that modem,scientific plant breeding can be adapted to localsociocultural and biophysical conditions and canbe integrated with farmer plant breeding. How-ever, few data are available for comparing thetwo systems, particularly in farmers' environ-ments and with farmers' practices. Most of theresearch with farmers has been done by socialscientists who have not used a plant-breedingframework for analysis, and most research onthe biological aspects of farmers' plant-breedingsystems has been done by plant breeders or bi-ologists who have not systematically investigat-ed farmers' knowledge or practice, if at all. Theresult is that plant breeders in CPB projects may

378 [VOL. 54

CLEVELAND ET AL.: UNDERSTANDING FARMERS' PLANT BREEDING

be forced to rely on assumptions about farmers'knowledge, practices, and crop varieties thathave not been tested and may not be valid forthe situation they are working in.

We agree with Cooper and Byth's suggestionthat plant breeders' work on-farm "needs to befirmly incorporated into the overall paradigm ofcrop improvement, and not allowed to developas an independent thrust," yet adapting plantbreeding to farmers' conditions may require newdevelopments in theory and method, for exam-ple, in moving from wide to specific adaptationas a breeding goal (Cooper and Byth 1996:18,20). Thus, adapting professional plant breedingfor CPB may require plant breeders to better un-derstand farner plant breeding and crop man-agement. In addition, to the extent that CPB en-tails a change in the knowledge or practices offarmers, successful change may require thatfarmers understand the reasons for change interms of professional plant breeding. This papersuggests a framework for carrying out researchto enhance this mutual understanding.

METHODS

DEFINING PLANT BREEDING

CPB involves farmers and plant breedersworking together in some part of the plantbreeding system to develop crop varieties thatmeet farmers' needs. We define this system toinclude not only the crop varieties and growingenvironments, and the behavior or practice thatchanges them, but also individual and groupknowledge, which includes values, empiricaldata, and theories. We assume that the primarygoal of CPB is varieties which, in combinationwith specified growing environments, will opti-mize benefit to the farmer.

Plant breeding practice includes both:

(1) The development of new varietiesthrough artificial selection of plants byfarmers and breeders within segregatingplant populations, which changes the ge-netic make up of the population. Artificialselection is both indirect, a result of theenvironments created in farmers' fieldsand plant breeders' plots, and direct, a re-sult of human selection of planting ma-terial. Direct artificial selection can beboth conscious (based on explicit crite-ria), the result of decisions to select forcertain traits, or unconscious (based on

implicit criteria), when no conscious de-cision is made about the trait selected for,as when large seeds are automatically se-lected because they are easier to handle.(There is some confusion over terms inthe literature. Indirect artificial selectionis sometimes defined as (a) "'natural" se-lection (Sirnmonds 1979:14-15), (b) thesame as conscious selection (Allard 1999:19, 26), or (c) entirely "unconscious" se-lection (Poehlman and Sleper 1995:9).)

(2) The choice of germplasm that determinesthe genetic diversity available within acrop as a basis for selection. Farners andplant breeders make choices between va-rieties and populations, especially in theinitial stages of the selection processwhen choosing germplasm for makingcrosses, and in the final stages whenchoosing among populations/varieties(Hallauer and Miranda 1988) for furthertesting, or for planting (farmers) or re-lease (plant breeders). Farmers' choiceswhen saving seed for planting, in seedprocurement, and in allocating differentvarieties to different growing environ-ments affects the genetic diversity of thecrops they plant, and determines the di-versity on which future selection will bebased.

THE BIOLOGICAL MODEL

As a framework for evaluating farmer breed-ing we use the elementary biological model onwhich plant breeding is based, as it is presentedin standard texts (e.g., Falconer 1989; Sim-monds 1979). First, variation in population phe-notype (Vp) on which choice and selection arebased is determined by genetic variation (VG),environmental variation (VE), and variation ingenotype-by-environment interaction (VGXE) VP= VG + VE + VGXE. Broad sense heritability (H)is the proportion of V, due to genetic variance(VG/VP), whereas narrow sense heritability (h2)is the proportion of V, due to additive geneticvariance (VA), that is, the proportion of VG di-rectly transmnissible to offspring (VA/VP), andtherefore of primary interest to breeders.

Second, response to selection (R) is the dif-ference for the traits measured between themean of the whole population from which theparents were selected and the mean of the nextgeneration that is produced by planting those se-

2000] 379

ECONOMIC BOTANY

lected seeds under the same conditions. R is theproduct of two different factors (R = h2S).where S is the selection differential, the differ-ence between the mean of the selected group andthe mean of whole original population selectedfrom. Expression of S in standard deviation units(the standardized selection differential, Falconer,1989:192), permits comparison of selectionsamong populations with different amounts ortypes of variation. The results of selecting for agiven trait improve as the proportion of V, con-tributed by VG (especially VA) increases.

The biological relationships described in thesesimple equations underlie plant breeders' under-standing of even the most complex phenomenathey encounter (Cooper and Hammer 1996;DeLacy et al. 1996). For example, two highlyrespected plant breeding texts state that the re-lationship between genotype and phenotype is"perhaps the most basic concept of genetics andplant breeding" (Allard 1999:48), and of R =h2S, that "If there were such a thing as a fun-damental equation in plant breeding this wouldbe it .. ." (Simmonds 1979:100).

ASSUMFiONSAlthough we use a framework based on the

basic biological model of plant breeding, we donot assume that when there are differences be-tween farmers and breeders, that the farner iswrong, nor do we assume that outsiders have notbeen diligent enough in rationalizing farmerknowledge and practice in their own terms (seeScoones and Thompson 1993; Uphoff 1992). Weacknowledge that successful plant breeding byeither farmers or professional breeders does notdepend on a complete empirical or theoreticalunderstanding of the biological mechanisms in-volved (Duvick 1996; Simmonds 1979).

We assume that an essential element for suc-cessful CPB is increased understanding of pro-fessional plant breeding. We recognize that inthe elaboration and application of the basic mod-el of plant breeding there are many differencesamong plant breeders, and this may be espe-cially true when extending conventional breed-ing to farmers' environments, such as envi-sioned by CPB (see e.g., Ceccarelli, Grando, andImpiglia 1998). Another essential element forsuccessful CPB may be increased understandingof the basis for variation in plant breeders'knowledge and practice in terms of the geno-types and environments they work with, and of

their sociocultural environment, and how thesechange in the context of CPB (Cleveland n.d.).

Better understanding of farmers' plant breed-ing in terms of the biological principles onwhich scientific plant breeding is based shouldincrease the success of CPB by facilitating col-laboration in the use of scientific plant breed-ing's emphasis on data collection and dissemi-nation, theoretical and analytical tools, and glob-al access to genetic resources, and by facilitatingthe use of farmers' understanding of their com-plex farming systems within which they practiceplant breeding. This is especially likely whenCPB involves changes in farmers' knowledge orpractice, but not as likely when it doesn't, e.g.,when a new variety is introduced that is similarto existing farmers' varieties, and which theycan adopt without changes in selection, choice,or production practices.

The next three sections are on three centralaspects of farmer plant breeding from the view-point of the biological model: genetic variation.environmental and genotype-by-environmentvariation, and plant selection. For each of these,we propose key research questions about farm-ers' knowledge, practices, and crop varieties andgrowing environments. To explore answers tothese questions, we use examples from the re-search literature and, for those topics where littleother research exists, from our own work in Oa-xaca, Mexico.

GENETIC VARIATIONGenetic variability (often referred to generally

as genetic diversity) is the basis for genetic im-provement of farmers' crop varieties and croprepertoires through plant selection or varietalchoice.

FARMER KNOWLEDGE

How are Farmers' Criteria for CropClassification Related to Perceptions of

VG?

A major controversy within ethnobiology hasbeen whether classification is the result of theuniversal structure in nature that imposes itselfon the human mind, perhaps facilitated by uni-versals in human cognition (the intellectualistview), or whether it is the result of culture-de-pendent differences in goals, values, and theo-ries (the utilitarian view) (Medin and Atran1999). Boster's work with Aguaruna farmers inthe Amazon supports the first view: cassava

380 tVOL. 54


(Manihot esculenta Crantz) classification reflectsa tendency to create and classify the smallestdistinct taxonomic unit, in patterns similar tothose of scientists (Boster 1985, 1996). This the-ory is supported by data on Axi observation,classification, and selection of ensete (Enseteventricosum [Welw.] Cheeseman) in Ethiopia, inwhich primary importance is placed on morpho-logical traits not directly related to practical use(Shigeta 1996).

Support for the utilitarian view is a more com-mon research finding. For example, for the Men-de of Sierra Leone, growth duration is a majorcriterion for classifying rice (Oryza glaberrimaSteudel and 0. sativa L.) varieties, with a mix-ture of varieties of different durations managedand planted to avoid labor bottlenecks and in-terharvest food shortages (Richards 1996).Farmers also classify and choose varieties basedon ceremonial and religious values, as Hopi (So-leri and Cleveland 1993) and Quechua (Zim-merer 1996) farmers do with maize. Farmerclassification of utilitarian traits can be simnilarto scientific classification. Farners in Cuzalapa,Mexico classified their maize FVs using traitssuch as plant height and weight and diameter ofcob, in the same pattern as did statistical traitanalysis of ear characteristics (Louette, Charrier,and Berthaud 1997). There is also evidence sug-gesting that farmers may be aware of the valueof the genetic potential of hybrids, such as seedsharvested from the edges of plots, where hy-bridization is more likely to occur, both for self-pollinated (rice in Sierra Leone, Richards 1986),and cross-pollinated crops (maize in Mexico,Louette and Smale 1998).

Differences in conclusions about the basis forclassification systems may be due not only to thebias of the initial orientation of the researchers,but to differences in the nature of the crops andenvironments involved. For example, the patternof phenotypic expression of qualitative traits ina clonally propagated crop (cassava, ensete) ismuch different than for quantitative traits in sex-ually propagated crops (rice), especially cross-pollinated ones (maize). Fanners may simplyenjoy "playing" with diversity (Berg 1996), yettheir perceptions of genetic variation (to the ex-tent revealed in plant phenotypes and across en-vironments) depends on their physical ability toobserve it, determined in turn by the scale atwhich it occurs, the extent to which it is hiddenby VE, and also on how important it is to them.

Farmers' recognition of VG between maize (Zeamays L.) FVs and between and within maizepopulations in Oaxaca, Mexico, is influenced bythe relative contributions of VG and VB to VP,that is the heritability of the particular trait (So-leri, Smith, and Cleveland n.d.).

There is also significant variation in distribu-tion of farmer knowledge about VG as the resultof social factors including age, gender, socialstatus and affiliation, kinship, personal experi-ence, and intelligence (Berlin 1992), accompa-nied by variation in the way farmers conceiveof their intellectual ownership of these resources(Cleveland and Murray 1997). There may alsobe a range in the consistency of classificationschemes within communities; for example, po-tato variety (Solanum tuberosum L.) classifica-tion in the Andes appears to be consistent (Zim-merer 1996), whereas cassava variety namesamong the Amuesha of Peru have a high levelof inconsistency, with the same common nameapplied to different phenotypes (Salick, Cel-linese, and Knap 1997).

FARMER PRACTICE

Farmers can affect genetic variation in twomajor ways: at the intraspecific level by addingand deleting varieties, and at the intravarietallevel by consciously and unconsciously encour-aging genetic recombination through hybridiza-tion.

What Farmers' Practices can AffectIntraspecific VG?

Farmers' annual choices about what varietiesthey will plant in which locations and at whichtimes, including the abandonment of varietiesand the acquisition of new ones, affects VG atthe intraspecific level. In general, it appears thatfarmers add or delete a variety when changes inthe local biophysical or sociocultural environ-ment alter the importance of varietal traits foradaptation to those environments (Soleri andCleveland 1993; see also Bellon 1996; Louette,Charrier, and Berthaud 1997; Richards 1986).The interaction between these factors in deter-mining the fate of a particular FV may be com-plex, as in the case of changing Hopi blue maizevarieties where several varieties seem to be col-lapsing into one as a result of the decreasedamount of time available for maintaining varie-ties, increased availability of non-Hopi varietiesand food products that fill similar needs, and be-

2000] 381

ECONOMIC BOTANY

cause of changing social conditions that reducethe importance of unique characteristics (e.g.,the introduction of machine grinding reduced theimportance of the softer blue corn variety) (So-leri and Cleveland 1993).

Most research has focused on individual farm-ers making decisions about individual varieties.However, this may lead to assumptions that arenot justified. First, it is obvious that individualdecisions are made within a sociocultural con-text, and networks affecting farmers' access toseeds are not experienced equally by all. For ex-ample, a study in Rwanda found that seed net-works were socially limited, with poorer house-holds having the most limited access (Sperlingand Loevinsohn 1993). A common observationis that there are individuals in a communityknown for the number of varieties that theymaintain, e.g., male shamans and some womenfarmers in the Peruvian Amazon (Salick, Cel-linese, and Knap 1997). There has been almostno analysis until recently of the effect on a farm-er's varietal repertoire of his/her decisions basedon perceptions of other farmers' management oftheir varietal repertoires and his/her access tothem. Initial findings suggest that farmers' per-ceptions of changes in the maize varietal diver-sity in their communities can affect the numberof varieties they maintain (Smale, Bellon, andAguirre Gomez 1999).

Second, researchers may have focused unjus-tifiably on individual varieties. For example,contrary to the assumptions of other researchersin the area, Zimmerer found that some groupsof Andean farmers choose varietal mixtures ofpotato for planting en masse, only rarely or nev-er selecting individual FVs as components (Zim-merer 1996).

What Farmers' Practices can AffectIntravarietal VG

Hybridization, the crossing of two distinct ge-notypes (species, varieties, or populations), isprobably the main way in which farmers' choiceaffects the VG of FVs. Hybridization can resultfrom activities that unconsciously affect the lev-el of reproductive isolation, such as allocationof planting material (cropping patterns), or farm-ers may have hybridization as a conscious goal.

There are reports in the literature of encour-aged or tolerated hybridization between species,e.g.. wild squash (Cucurbita argyrosperma C.Huber subsp. sororia (L. H. Bailey) L. Merrick

& D. M. Bates) with cultivated species (C. ar-gyrospenna subsp. argyrosperma and C. mos-chata) (McKnight Foundation 1998). One of themost well-known examples is of teosinte (Zeamays L. subsp. mexicana) hybridizing withmaize (Zea mays subsp. mays) (Benz, Sanchez-Velasquez, and Santana Michel 1990; Wilkes1989), although the extent and significance ofthis has been challenged (e.g., Kato Y 1997).Hybridization between species is probably re-sponsible for a small, although potentially im-portant, portion of genetic variation in FVs, inpart because it is infrequent, and also because itis likely often managed as introgression.

Reports of hybridization between varieties orpopulations of a species are more common.Farmers may also manage this level of hybrid-ization primarily as introgression, although thisis probably less likely because of the smallergenetic distance involved. In vegetatively prop-agated crops, seed may be produced from oc-casional spontaneous hybridization between va-rieties and sought out by farmers, e.g., byAmuesha cassava farmers in Peru (Salick, Cel-linese, and Knap 1997), and Quechua potatofarmers in the Andes (Zimmerer 1996).

In self-pollinating species, e.g., rice, the pos-sibility for cross pollination between varietiesmay be increased when a fanner plants differentvarieties of the same duration class in adjacentplots, when adjacent plots contain differentfarmers' varieties of the same duration class, andwhen farmers use a large plot communally, aswith rice in Sierra Leone (Richards 1986, 1996).In cross-pollinating crops the levels of hybrid-ization potentially are much higher. In a CPBproject in Rajasthan. India with pearl millet(Pennisetum glaucumn (L.) R. Br.), farmers fre-quently planted seed of foreign varieties savedfrom variety trials with their own FVs, whichresulted in increased variability in the next gen-eration, and intense discussion by farmers aboutselection (Weltzien R. et al. 1998).

GENETIC VARIATION OF FARMERS'VARIETIES

What is the ILvel of VG in FVs?

Most of the evidence and discussion regardingVc; in FVs refers specifically to landrace popu-lations. Evidence suggests that since domesti-cation there has been increasing intraspecific di-versity in the form of landraces until modern

382 [VOL. 54


plant breeding "drastically restricted ... the in-traspecific diversity of crop species" (Frankeland Soule 1981:179). The genetic diversity oflandraces is probably one of the most researchedcomponents of farmers' plant breeding, althoughthis has rarely taken the broader populationstructure of what are now recognized as opengenetic systems into account (e.g., Louette et al.1997). A number of studies suggest that landra-ces have a large amount of allelic variation(Frankel, Brown, and Burdon 1995). Species'mating systems affect the structure of geneticvariation, with cross-pollinating species havingmore allelic diversity within as compared to be-tween populations, with the opposite being trueof self-pollinating species (Hamrick and Godt1997). Yet there is significant overlap, for ex-ample, about 19% of loci are polymorphic inlentil, which may be typical of self-pollinatingspecies (Erskine 1997). Significant intravarietalvariation in morphology and phenology has alsobeen documented for cross-pollinating species,for example, in two Hopi maize FVs (Soleri andSmith 1995).

How is VG Affected by Fanner Practice?

The net effect of the adoption and abandon-ment of varieties may either increase or decreaseVG, because there are "infinite combinations be-tween the variability of existing crops and thenew variability of cultivars that partially or com-pletely replace them" (Witcombe et al. 1996:456). Although adoption of MVs due to theirsuperior performance has been documented todecrease on-farm genetic diversity through theloss of FVs, there may be a limit to this loss inthe later stages of adoption (e.g., for potato inPeru, Brush, Taylor, and Bellon 1992). However,there are few genetic data on the effect of chang-ing varietal repertoires (including MV adoption)on allelic diversity (number and evenness) at thefarm, community, or regional levels.

Similarly, although it is known that farmerpractices can increase hybridization, there arefew data on its genetic effect. In Cuzalapa, Jal-isco, Mexico, farmers regularly mix maize pop-ulations together by classifying seed obtainedfrom widespread, diverse sources in the samevariety (Louette, Charrier, and Berthaud 1997).This practice, together with the planting pat-tems, leads to a 1-2% level of gene flow be-tween adjacent maize plots during one crop cy-cle, which probably has a significant effect on

genetic composition over several crop cycles(Louette 1997). Evidence of a morphologicaland genetic continuum across the four major lo-cal varieties suggests that traits from a varietyintroduced 40 years ago have introgressed intothe other varieties. Several studies have docu-mented introgression of maize MVs into FVs inMexico (Castillo-Gonzales and Goodman 1997).Although evidence from the Cuzalapa study sug-gests that gene flow "probably leads to a modestdegree of heterosis among all cultivar types,"this may not always be the case. In fact, whethergene flow increases or decreases VG, and whateffects this may have on adaptation is unknownfor FVs. In conservation biology gene flow isoften considered beneficial because it can pre-vent inbreeding depression and loss of VG insmall populations, but it can also reduce VG,leading to outbreeding depression and reducedadaptedness, and the effect may depend critical-ly on population size (Ellstrand and Elam 1993).

Is VG "Optimal" for Farmers'Environments?

Much of the research on genetic diversity inFVs has been done with the aim of assessing itspotential for MV breeding programs (e.g., Ouen-deba et al. 1995), not in terms of adaptation tofarmers' growing environments. Available datasuggest that the genetic variation of FVs appearsto buffer the effects of variable, high-stress en-vironments, and provide genetic potential for se-lection of superior material (Frankel, Brown,and Burdon 1995). However, the variation with-in FVs is not always optimal for the farmers whoare using them in a given environment, for ex-ample when resistance to an important stressfactor is absent (Trutmann and Pyndji 1994).One reason for low or less than optimal VG with-in a FV is genetic drift due to a founder effect(Barrett and Husband 1989), and this may bemost likely when crops are introduced outsideof their areas of origin and/or diversity.

VG among varieties is also important. A num-ber of studies illustrate the greater stability andproductivity of mixtures of populations (lines)for self-pollinated crops. For example in theGreat Lakes region of East Africa farmers plantmixtures of many varieties of common bean(Phaseolus vulgaris L.) that are site specific,with varying resistances to multiple diseases, astrategy that may be the most effective for op-timizing yield stability. A similar situation can

2000] 383

ECONOMIC BOTANY

occur for cross-pollinated crops, and an Ethio-pian study found that a mixture of maize culti-vars with different durations increased yield andstability in experiments under variable rainfalland drought (Tilahun 1995).

ENVIRONMENTAL DIVERSITY ANDGENOTYPE-BY-ENVIRONNIENT

INTERACTION

Environmental diversity or variation can bepartitioned into several components: VE = VL +VT + VM (VL = variance due to location, e.g.,soil and climatic variables; VT = variance dueto time, e.g., season or year; and VM = variancedue to breeder or farmer management). VGXE rep-resents the degree to which genotypes behaveconsistently across a number of environments.Low quantitative GxE means relatively littlechange in performance over environments. Highquantitative GxE is characterized by markedchanges in performance with changes in envi-ronmental factors and is associated with reducedstability of performance (defined as varianceacross environments) of an individual genotype.Qualitative GxE between two or more varietiesmeans that they change rank across environ-ments, and this is often referred to as a crossoverbecause the regression lines for yield (or othertraits) cross over at some point.

FARMER KNOWLEDGEHow do Farners Perceive GxE in

Relation to Their Choice of a Variety fora Given Location?

The choice of a variety (A) with higher meanyield but larger regression slope (lower yieldstability) over all locations compared with an-other variety (B) can be unproblematic in somecontexts because this describes a situation inwhich there is only quantitative GxE, i.e., nocrossovers (Tripp 1996). However, in terms offarmers' understanding, this may depend on theway in which environment is defined. When afarmer is choosing between varieties to grow ina given location, his/her choice may depend onhow variation in yield (and income) is perceivedover time for that location (VG,X), as well asmean yield. If variety A has larger mean yieldand lower yield variance than variety B throughtime in a given location, then the choice wouldbe A. However, if variety A has larger meanyield but also a larger variance, then the choicebetween the two varieties will depend on his/her

attitudes toward risk and ability to manage it,and he/she may be willing to sacrifice meanyield in order have a more stable yield, or a'smoother income stream" through time.

(Walker 1989).Few data exist on farmer risk perception in

terms of VGXT. One study from Malawi on thedynamics of MV hybrid maize adoption, FVmaize retention, and commercial fertilizer appli-cation, suggests that the ratio of coefficients ofvariation of MVs to those of FVs, based onfarmers' subjective yield estimates, is negativelyrelated to area planted to MVs and to fertilizerapplication rate on MVs (Smale, Heisey, andLeathers 1995). That is, the lower the yield sta-bility through time of MVs compared to the FV,the less land and fertilizer are allocated to theMVs.

How do Farmers Perceive GxE inDeciding Whether to Have One or MoreThan One Variety for a Set of Locations?

Another situation that appears to be commonfor farmers is the choice of varieties for a rangeof locations for a given planting season (i.e.,time is held constant). If farmers do not perceivequalitative GxL (crossovers) differences in per-forrmance for varieties between locations, thenthere may be no reason for them to grow differ-ent varieties in these locations. When farners doperceive crossovers between varieties for twoenvironments, then they may have to decidewhether to grow one variety in both environ-ments, or if the extra yield obtained by growingtwo different varieties in the two environmentscompared with the extra effort required, willproduce a net benefit.

One sample of Rwandan farmers chose fromanmong plant breeders' varieties of common beanfor home testing. They based their choices onperformance under bananas, on poorer soils, andin heavy rain, in addition to high yield. Re-searchers judged those farmers to be well awareof the responses of different genotypes in dif-ferent environments, and thus of GxE, thoughno details of farmer knowledge were reported(Sperling, Loevinsohn, and Ntabomvura 1993).Rajasthani pearl millet farmers, realizing thatthere is a trade-off between panicle size and ti]-leling ability, prefer larger panicles in the leaststressful environment, and high tillering abilityin the most stressful (Weltzien R. et al. 1998),suggesting that in the most variable environ-

384 [VOL, 54


ments, traits contributing to yield stability aremore important than those contributing.to moreyield. There are similar results from research inCambodia, where rice farmers' criteria for va-rietal choice differs according to level of envi-ronmental stress, so that yield stability (droughtand flood tolerance) was more important inchoosing early- and late-maturing varietiesgrown in higher stress environments, and yieldand eating quality were more important for me-dium-maturing varieties grown in lower stressenvironments (Lando and Mak 1994).

Farmers' perceptions of qualitative GxE dif-ferences appear to depend on the range of en-vironmental and/or genotypic diversity theymust consider in making a decision about allo-cation of land to varieties. In our study in Oa-xaca, farmers in one community (A) do notmaintain distinct varieties of white maize iden-tified for allocation to specific environments(e.g., soil type, elevation, water availability,planting season), and instead identify two vari-eties based on ear and kernel phenotypes withno consensus regarding a relationship to in-fieldperformance (Soleri and Cleveland n.d.). In con-trast, farmers in a nearby community (B) withless favorable growing conditions classify andmaintain violento (short growth duration) andtard6n (long growth duration) varieties of whitemaize. Whether or not distinct duration varietiesof white maize are maintained appears to dependon the importance to farmers of duration formanaging drought. In community B, farmers seeVE between location-year combinations as beinggreater than do farners in community A, andthey believe that violento performs better inyears with late season drought, and tard6n inyears with early season drought. These findingssuggest that farmers in community B perceiveVE between environments (characterized accord-ing to within and between year variation in pre-cipitation) as greater than farmers in communityA, and also perceive that the net benefits justifiesmaintaining separate varieties for those environ-ments.

FARMER PRACTICE

What Growing Environment Variables areCorrelated with Farmers' Choice of

Varieties?

Some FVs appear to be managed for narrowlydefined environments, for example, in East Af-

rica varietal mixtures of common bean selectedfor different on-farm environments are oftenmaintained separately (Trutmann 1996), andMende farmers in Sierra Leone maintain a largeand constantly changing collection of rice FVs,with selection primarily for traits, including du-ration, and adaptation to a variety of specificmoisture regimes (Richards 1986). In contrast,the common assumption that the large numberof potato varieties maintained by Andean farm-ers is managed by allocating each variety to spe-cific fields or within field environments is con-tradicted by the finding that these varieties areharvested and planted by some groups of thesefarners as bulk mixtures across all environments(Zimrnmerer 1996).

What Environments do Farmers Choosefor Testing New Material, and How are

They Related to Their TargetEnvironments?

Research findings on farmer practice regard-ing new accessions are also contradictory. Farm-ers may test new varieties in optimal environ-ments, for example, in home gardens, wherethey evaluate them for later planting in specific,more marginal environments, and researchersassume that they do this to reduce risk of loosingseedstock (Ashby et al. 1995; Soleri and Cleve-land 1993). However, rice farners in Nepal of-ten plant new varieties on their worst land,which has been interpreted as a risk aversionstrategy (Sthapit, Joshi, and Witcombe 1996),with the implication that farners are looking forvarieties adapted across a wide range of loca-tions. It has been suggested that this is a com-mon practice (Witcombe 1998).

FARMERS' GROWING ENVIRONMENTs ANDVARIETIES

Are FVs Adapted to a Narrow or WideRange of Environments?

Commonly used definitions of FVs includethe statement that they are adapted to a narrowrange of environments (Frankel, Brown, andBurdon 1995; Zeven 1998). However, discus-sions of whether a particular variety is narrowlyor widely adapted are often confused by failureto distinguish temporal, locational, and manage-ment aspects of the environment, and sometimesconflate these with geographical extent of ad-aptation (Souza, Myers, and Scully 1993), es-

2000] 385

ECONOMIC BOTANY

pecially when discussing farmers' systems(Bjornstad 1997).

Some FVs are adapted to specific environ-ments for which they show high GxE, a frequentexample being phenological adaptation to cli-mate patterns, e.g., with drought patterns andpearl millet in Rajasthan (van Oosterom. Whi-taker, and Weltzien R. 1996). One study in Ethi-opia found that 13 wheat (Triticum turgidum L.)FVs showed qualitative GxE for four locationswhere they are grown, but low correlations be-tween yield and stability measures (Tesemma etal. 1998). At the regional level, varieties of lentil(Lens culinaris Medick) appear to have specificadaptation to locations, temporal patterns, andmanagement levels (Erskine 1997).

On the other hand, some FVs may be geo-graphically widely adapted, and be plantedacross a wide range (Witcombe 1999). Thewidespread exchange of crop varieties by farm-ers suggests such wide adaptation (Wood andLenne 1997), although the range of environmen-tal variation and its components and the degreeof genetic variation between FV populationsgrown by different groups are largely unknown.In southwestern North America maize varietieshave been shared frequently between tribes,even though each tribe usually considers a givenvariety to be its own unique FV, and has its ownname for it (Soleri and Cleveland 1993). Thehigh level of variation within farmers' fieldsplanted to a single FV also suggests that suchFVs can be widely adapted to a large range oflocations within a small geographical area (So-leri, Smith, and Cleveland n.d.).

How do FVs Compare with MVs in Rangeof Adaptation?

MVs that are widely adapted geographicallymay be narrowly adapted to the high-yielding,low-stress conditions of locations and manage-ment environments used by many plant breedersin selection (Banziger, Edmeades, and Lafitte1999; Ceccarelli 1989; Witcombe 1999). Quali-tative GxE interaction of varieties across envi-ronments when the range is wide enough arewell known (Evans 1993 and Ceccarelli 1996summarize the data) and may explain the lackof adoption of some MVs for farmers-theymay be out-yielded by FVs in farmers' high-stress environments (Bdnziger, Edmeades, andLafitte 1999; Ceccarelli, Grando, and Impiglia1998; Weltzien and Fischbeck 1990).

On the other hand, MVs may be widely adapt-ed to a range of locations, including not onlylow-stress locations, but also high-stress oneswhere they out yield FVs. For example, the ex-perience of the International Maize and WheatImprovement Center (CIMMYT) in wheatbreeding using large numbers of crosses, inter-national testing of advanced lines, and continu-ous alternating selection cycles in environmentsthat differ but allow expression of high yield(shuttle breeding) have led to wheat MVs thatappear widely adapted and are higher yieldingthan local varieties in high-stress environments,such as Western Australia (Rajaram, Braun, andvan Ginkel 1997; Romagosa and Fox 1993). InZimbabwe, maize hybrid MVs have had highadoption rates among limited-resource farmersin more marginal environments (Heisey et al.1998).

PLANT SELECTIONSelection of plants from a heterogeneous pop-

ulation to obtain planting material for the nextgeneration can affect allelic frequencies throughtime, and thus genetic gain, or R. Mass selectionappears to be the most common form of selec-tion used by farmers. It involves the identifica-tion of superior individuals in the form of plantsand/or propagules from a population and thebulking of seed or other planting material toform the planting stock for the next generation.This approach requires only a single season andrelatively little effort compared with other selec-tion methods. If practiced season after seasonwith the same seed stock, mass selection has thepotential to maintain or even improve a croppopulation, depending upon the extent to whichthe selected trait is heritable, GxE for the trait,the proportion of the population selected (selec-tion intensity), and gene flow in the form of pol-len or seeds into the population.

FARMER KNOWLEDGE

What are Farmers' Explicit SelectionCriteria?

Farmers stated selection criteria are oftencomplex. In our Oaxacan case study the mostimportant criteria appear to be those related toseed viability-all maize ears with evidence ofpest or disease damage to the seed or cob areusually discarded (Soleri, Smith, and Clevelandn.d.). The next category includes traits that con-tribute to large ears and large kernels, especially

386 [VOL. 54


ear length and weight. The final category en-compasses a number of traits that define a va-rietal type. Our sample included traits like graintype (e.g., flinty vs. starchy), grain form (roundvs. flat), and cob and husk color. Although cri-teria in the third category varied between house-holds and communities, the first two categorieswere universally the most important. Explicitcriteria in the Cuzalapa case study were similar(Louette and Smale 1998).

Selection criteria can vary within a commu-nity. In a panicle selection exercise in Rajasthan,India, farmers with good land or more land se-lected a wide range of pearl millet panicle typesfor seed because they said they wanted seedstocks useful for a broad range of planting con-ditions, including variations in soil fertility aswell as in rainfall, and they frequently purchasedseed. In contrast, farmers with poor land choseonly one panicle type, one rejected by the for-mer group, and were proud of saving their ownseed for 100 years (Weltzien R. et al. -1998).

What are Fartners' Explicit SelectionGoals?

Weltzien et al. noted that surprisingly little re-search has been done on selection goals consid-ering their importance for the selection process,especially for marginal environments and forfarners (1998). The implicit assumption oftenhas been that farmers must be attempting direc-tional selection for quantitative, relatively lowheritability traits like yield, the main goal ofplant breeders. However, there appear to be rel-atively few data demonstrating that farmers havedirectional selection for quantitative traits as aconscious goal, in contrast with data on farmers'conscious choice of new varieties.

As researchers' understanding of the com-plexity of farners' knowledge and traditionalfarming systems increases, it seems possible thatdisruptive, stabilizing, and random selectionmay either be an explicit (conscious) goal offarmers, or occur as an unintended result of se-lection practices. In informal interviews in Cuz-alapa, Mexico, farmers indicated that they donot see seed selection as a way of changing orimproving their maize varieties, but of protect-ing the "legitimacy" of a variety, i.e., of main-taining varietal ideotypes (Louette and Smale1998).

Results of our study in Oaxaca were similarto those in Cuzalapa. They suggest that for traits

with low heritability, farmers generally did nothope to change a variety (Soleri, Smith, andCleveland n.d.). Both the lack of expectationsfor change and the concem with maintenance ofcurrent traits appear to be a pragmatic recogni-tion of the substantial VE and large amounts ofgene flow via cross-pollination that must occurunder local conditions: areas of vast-in somecases year-round-maize cultivation, often infields as narrow as 11 m. Nevertheless, their an-swers indicated an awareness of selection andthe ability to use it when they felt it desirableand possible. These farmers typically have lowexpectations for change regarding traits thatcomprise their seed selection criteria. They at-tributed their low expectations to cross-pollina-tion and their understanding of the influence ofV. on plant phenotypes in their fields (h2 ofthose traits). Interpreted as such, their expecta-tions appear to reflect two observations made byresearchers: 1) lack of control over pollen sourc-es (extensive cross-pollination) effectively re-duces h2 of phenotypes by as much as one halfin comparison to its level under biparental con-trol, and 2) in traits with medium to low h2

(<0.5), the progeny of selected individuals willtend to reflect more the mean of the entire pop-ulation from which the parents were selectedthan the mean of the parents (Simmonds 1979).In contrast, the Oaxacan farmers we workedwith clearly understood qualitative, relativelyhighly heritable traits like tassel color different-ly. They perceived of the possibility of direc-tional selection for this trait, and showed us ex-amples of the successful results of such selectionin their fields.

Farmers' perceptions of the potential to im-prove their populations via selection-and thustheir selection goals-will be influenced notonly by their understanding of genetic variationin the population and h2 for traits of interest, butalso of altemative uses of their time and labor.If they do not believe population improvementto be possible or cost-effective, one altemativemay be to choose different varieties or popula-tions or infuse their own varieties or populationswith new genetic variation as discussed above.

FARMER PRACTICEAt what Stages in the Plant Life Cycle,

and by Which Fanners, is SelectionCarried Out?

The stage of the plant life cycle at which se-lection occurs can have a strong effect on its

2000] 387

ECONOMIC BOTANY

efficiency in terms of genetic gain, and differentpersons may have different selection goals, con-scious and unconscious. In some crops, for ex-ample, small-seeded crops like rice, farmers aremore likely to carry out in-field selection onplants using a range of criteria (Richards 1986;Sthapit, Joshi. and Witcombe 1996), althoughmajor selection criteria may also include post-harvest traits (Sthapit, Joshi, and Witcombe1996). In other crops, for example, large-seededcrops like maize, selection may be almost en-tirely post-harvest, as in Mexico (Smale et al.1998). In a study in Columbia, 31% of farmersbegan selection of bean seed in-field by select-ing areas where plants had abundant foliage andlow disease incidence, whereas the remainingfarmers selected entirely postharvest (Janssen,Adolfo Luna, and Duque 1992).

In the Sierra Santa Marta of Veracruz, Mexi-co, detailed and repeated interviews with bothmen and women in the same farm householdshowed that selection occurs in four or five stag-es, most of which women participate in (Rice,Smale. and Blanco 1998). In Mexico, it has beena common finding of researchers, and an as-sumption of development workers, that men areresponsible for seed selection, but new findingssuggest that this conclusion may be the result ofthe methods employed and that women play animportant role in seed selection (Smale et al.1998).

What are Farmers' Implicit SelectionCriteria?

Farmers implicit (unconscious) selection cri-teria can be ascertained by comparing values ofphenotypic traits in selected material with thosein nonselected material, but this has rarely beendone. In Cuzalapa, Mexico, farmers implicit cri-teria reflected their explicit criteria (Louette andSmale 1998). Although all ear descriptors mea-sured showed a significantly higher level in theselected set compared with the population se-lected from, the greatest differences were for thecriteria farmers said were most important: earweight, ear length, length of ear presenting ker-nels, total number of kernels, and kernel filling.

In Oaxaca, Mexico we carried out selectionexercises with farmers on maize ears, in whichthey selected the best 10 ears for planting seedfrom a random sample of 100 ears taken fromplots in their own or neighboring fields (Soleri,Smith, and Cleveland n.d.). Values for standard-

ized S were not significant for ear traits such askernel row number and shelling ratio, and therewere only occasional but no consistent signifi-cant differences for some plant morphologicaland phenological traits. However, for the corre-lated characteristics of ear length and weight,selections were significantly different than the100 ear sample, with standardized S values (alsoreferred to as intensities, Falconer 1989:192) of0.48-1.33 and 0.73-1.81 respectively (comparedwith an intensity of 1.8 typically sought bybreeders in directional selection of a 10% sam-ple, Hallauer and Miranda 1988). Thus their ex-plicit selection criteria accurately reflect thetraits that farmers actually seek when selectingseed for planting.

FARMERS' VARIETIES AND THE RESULTS OFFARMER SELECTION

What is the Heritability of FVs inFarmers' Selection Environments?

The success of farmer selection will dependto a great extent on the heritability of the traitsincluded in their implicit selection criteria in thecrop populations and environments they aremanaging. A few studies have been done on her-itability of FVs in experimental plots, for ex-ample, in Ethiopia, research with wheat FVsfound intermediate to high heritabilities formany traits, including grain yield (Belay et al.1993). Assessment of heritabilities in farmers'selection environments are even more rare. Oneexample is our study in the Central Valleys ofOaxaca, Mexico (Soleri and Smith n.d.), usinga new method for estimating H in farmers' fields(Smith et al. 1998). Although less precise thanconventional methods, this approach appears toprovide a useful initial orientation to H and thusselection potential in areas and with populationsfor which estimates are rare. Overall, H esti-niates calculated in this study indicate that re-sponse to mass selection as practiced by farmersand as advocated by some CPB projects (Rice,Smale, and Blanco 1998 describe one example),will be negligible or low. It also suggests a num-ber of traits of interest to farmers with H valuesshowing potential for significant response (av-erage H = 0.65 for days to anthesis, 0.74 for earheight, and 0.63 for ear length) if mass selectionwas improved, for example, through in-field se-lection with stratification. However, this is trueonly if whole plant traits are relevant to fanners'

388 [VOL. 54


selection goals, and, when only in-field selectionis used, if the overall gain will be greater thanfor postharvest selection.

What is R for Farmers' SelectionPractices?

Although mass selection has clearly been ef-fective over time for some traits, genetic gain(R) from selection year to year may be smallbecause heritabilities for many selection criteriasuch as yield are generally low (< 0.30). Theresults of farmer selection (R) may be differentthan their implicit goals (S), however, R has sel-dom been measured. For example, farmers mayintend to perform directional selection for a par-ticular trait based on the phenotypes in a popu-lation, but if the intrapopulation V, on whichthey are selecting is predominantly the productof VE, then the result of their efforts may berandom selection, i.e., R = 0. This may be thecase for farners in our Oaxaca case study be-cause there were no significant differences be-tween the means of random samples from thewhole population and samples of farmer-select-ed seed derived from these as represented bytheir progeny generations, and significant differ-ences were not observed among random samplesof the same population over generations for thetraits evaluated (Soleri, Smith, and Clevelandn.d.). However, a more accurate interpretationmay be that these fanners perceive some traitsas part of a group of nonheritable selection cri-teria related to seed quality and seedling perfor-mance. As such, their interest in and expecta-tions for these traits may not be related to theirinheritance, and S values would reflect theweighing of seed stock quality along with anyother selection criteria in farmers' selection, andnot the sole goal of directional selection as sofrequently assumed.

It seems clear that farmers, like plant breed-ers, observe that the effectiveness of selectiondepends to a great extent on the heritability (h2)of the traits constituting the selection criteria.Thus, and in contrast to the findings in Oaxaca,the Cuzalapa study shows that farmers canachieve discemable R in the face of high levelsof gene flow from morphologically contrastingvarieties into their black maize variety (Louetteand Smale 1998). Specific morphological and tosome extent isozyme data show fanners' selec-tion maintains the phenotype characteristic oftheir varieties (ear traits and linked phenological

traits that define the ideotype) in the face of geneflow, even though other characteristics not visi-ble to farmers (isozymes) continue to evolve ge-netically. These farmers are apparently seekingto maintain varietal integrity, and base their se-lection on traits such as kemel row number, ker-nel width, and color, all with medium to highaverage h2. In contrast, the selection documentedin Oaxaca showed, for example, that kemel rownumber is not a criterion of interest, with se-lected populations not significantly differentfrom nonselected ones for both standardized Sand R. The selection in the Cuzalapa examplemay be most aptly described as stabilizing,though this would require further investigation.

There is overwhelming indirect evidence fordirectional selection by fanners achieving sig-nificant genetic gain over long periods of timefor the quantitative agronomic traits that are thefocus of most professional plant breeding, suchas yield. However, there appears to be little ev-idence from research with contemporary farmersfor positive R as result of conscious or uncon-scious directional selection for such traits. Thepossibility of directional selection for such traitsby farmers, and the resulting R, is an area inneed of more research.

CONCLUSION

In this paper we have presented a biologicalframework, based on scientific plant breedingtheory, for evaluating farmers' plant breeding(their knowledge, practice, crop varieties, andgrowing environments). Building on this frame-work, we have posed key questions that mayneed to be answered if collaboration betweenfarmers and breeders is to be successful. Ourexamination of the research literature (not an ex-haustive review) lead to the following conclu-sions.

1. The data needed to answer these key ques-tions is often scant or nonexistent, and fur-ther empirical testing of specific hypothe-ses based on the questions presented hereis needed, especially in the context of CPBprojects.

2. The explicit and implicit answers to thesequestions that are available in the literatureare often different and even contradictory,and may be based on unexamined andeven unrecognized assumptions.

3. Farmers' knowledge of their genotypes

2000] 389

ECONOMIC BOTANY

and environments is in many ways funda-mentally similar to that of plant breeders'understanding of their genotypes and en-vironments, in terms of the biologicalmodel of plant breeding.

4. Farmers' knowledge may also be differentthan that of plant breeders, in part becausethe genotypes and environments they dealwith are also different in important waysthan those commonly dealt with by plantbreeders.

5. Farmers' practice show a wide range of ef-fectiveness in meeting their explicit selec-tion goals.

6. The search for generalizations about farm-er plant breeding is valid, but we need tobe careful about making them at too su-perficial a level-generalizations based onconventional plant breeding, or on experi-ence with specific farmers' and their vari-eties and growing environments, may notbe a valid foundation on which to base thedevelopment of CPB.

7. Adapting professional plant breeding toCPB will require further research on farm-er plant breeding, and further examnination,adaptation and development of plantbreeding theories and methodologies.

Ultimately, it will be necessary to greatly ex-pand the framework we present to include moresociocultural and economic variables at both lo-cal and regional levels, and to include analysisof professional plant breeding along with farmerplant breeding. It seems that with more criticalqualitative and quantitative research on farmerand scientist plant breeding, it is likely that whatare frequently assumed to be either different orsimilar systems of plant breeding, will tum outto be two complex systems with many similar-ities as well as differences.

It will also be important for CPB projects toinclude consideration of whether altemativeways of investing scarce resources, such as man-aging soil organic matter, or helping farmers togain more control over political processes, mayimprove farmers' well-being more effectivelythan CPB.

Our aim in this article is to encourage re-search useful for CPB work whose goal is tohelp fanners and plant breeders communicatemore effectively with each other, so that breed-ers' theory, knowledge, statistical design and

analysis, and access to a wide range of geneticdiversity, can be used collaboratively with farm-ers' knowledge of their crops and envirom-nents,and techniques for management. Such researchwill permit testing of the idea that reunitingfarmer and professional plant breeding, after 200years of increasing separation, can increase theeffectiveness of developing crop varieties thatbetter meet farmers' needs, conserve crop ge-netic diversity in situ, and thus contribute to sus-tainable agriculture.

ACKNOWLEDGMENTSWVe thank the small-scale Kusasi, Hopi, and Zuni farmers, and espe-

cially the maize farmers of Oaxaca, Mexico, who have helped us beginto lnderstand farmer plant breeding: Melinda Smale (with Michael Mor-ris and Mitch Renkow), David Beck, Trygve Berg, Salvatore Ceccarelli,and Eva Weltzien for comments on drafts of the CIMMYT EconomicsWorking Paper on which this paper is based (Cleveland, Soleri, and Smith1999); two anonymous reviewers for Econoink Botany for their helpfulcomments; Humberto Rios Labrada for translation of the Spanish ab-stract; Melinda Smale and CIMMYT for support of our work in Mexico;Association of American University Women, Association of Women inScience, US Agency for International Development-CGIAR LinkageGrunts, and US-Mexico Fulbright Commission for support to DS; Uni-versity of California, Santa Barbara, especially the Committee on Re-search, Faculty Senate for support to DAC; and the Arizona AgriculturalExperiment Station for support to SES.

LITERATURE CITED

AUlard, R. W. 1999. Principles of plant breeding, Sec-ond Edition. John Wiley & Sons. New York.

Ashby, J. A., T. Gracia, M. del P. Guerrero, C. A.Quir6s, J. I. Roa, and J. A. Beltran. 1995. Net-work Paper, 57: Institutionalising Farmer Partici-pation in Adaptive Technology Testing with the"CIAL." Agricultural Administration (Researchand Extension) Network, Overseas DevelopmentInstitute, London.

B.:inziger, M., G. 0. Edmeades, and H. R. Lafitte.1999. Selection for drought tolerance increasesmaize yields across a range of nitrogen levels. CropScience 39:1035-1040.

Barrett, S. C. H., and B. C. Husband. 1989. Thegenetics of plant migration and colonization. Pages254-277 in A. D. H. Brown, M. T Clegg, A. L.Kahler and B. S. Weir, eds., Plant population ge-netics, breeding, and genetic resources. Sinauer As-sociates Inc, Sunderland, MA.

Belay, G., T. Tesemma, H. C. Becker. and A. Mer-ker. 1993. Variation and interrelationships of ag-ronomic traits in Ethiopian tetraploid wheat landra-ces. Euphytica 71:181-188.

Bellon, M. R. 1996. The dynamics of crop infraspe-cific diversity: a conceptual framework at the farm-er level. Economic Botany 50:26-39.

Benz, B. F., L. R. Sanchez-Velasquez, and F. J. San-tana Michel. 1990. Ecology and ethnobotany ofZea diploperennis: Preliminary investigations.Maydica 35:85-98.

390 [VOL. 54


Berg, T. 1996. The compatibility of grassroots breed-ing and modem farming. Pages 31-36 in P. Eyza-guirre and M. Iwanaga, eds., Participatory plantbreeding. Proceedings of a workshop on participa-tory plant breeding, 26-29 July 1995, Wageningen,The Netherlands. International Plant Genetic Re-sources Institute, Rome.

Berlin, B. 1992. Ethnobiological classification: Prin-ciples of categorization of plants and animals intraditional societies. Princeton University Press,Princeton, NJ.

Bj0rnstad, A. 1997. Participatory and formal plantbreeding: the need for interaction. Lecture at theSecond CBDC-ITP Workshop, Uniao da Vitoria,Brazil, November 13, 1997.

Boster, J. S. 1985. Selection for perceptual distinc-tiveness: evidence from Aguaruna cultivars of Ma-nihot esculenta. Economic Botany 39:310-325.

Boster, J. 1996. Human cognition as a product andagent of evolution. Pages 269-289 in R. Ellen andK. Fukui, editors. Redefining nature: ecology, cul-ture and domestication. Berg, Oxford, UK.

Brown, A. H. D. 1978. Isozymes, plant populationgenetic structure and genetic conservation. Theo-retical and Applied Genetics 52:145-157.

Brush, S. B., J. E. Taylor, and M. R. Bellon. 1992.Technology adoption and biological diversity inAndean potato agriculture. Journal of DevelopmentEconomics 39:365-387.

Callaway, M. B., and C. A. Francis, eds. 1993. Cropimprovement for sustainable agriculture. Universityof Nebraska Press, Lincoln, NE.

Castillo-Gonzales, F., and M. M. Goodman. 1997.Research on gene flow between improved maizeand landraces. Pages 67-72 in J. A. Serratos, M.C. Willcox and E Castillo-Gonzales, eds., Geneflow among maize landraces, improved maize va-rieties, and teosinte: implications for transgenicmaize. C1MMYT, Mexico, D.E

Ceccarelli, S. 1989. Wide adaptation: How wide? Eu-phytica 40:197-205.

. 1996. Positive interpretation of genotype byenvironment interactions in relation to sustainabil-ity and biodiversity. Pages 467-486 in M. Cooperand G. L. Hammer, eds., Plant adaptation and cropimprovement. CAB International in associationwith IRRI and ICRISAT, Wallingford, Oxford, UK.

Ceccarelli, S., S. Grando, and A. Impiglia. 1998.Choice of selection strategy in breeding barley forstress environments. Euphytica 103:307-318.

Ceccarelli, S., W. Erskine, J. Hamblin, and S.Grando. 1994. Genotype by environment interac-tion and international breeding programmes. Ex-perimental Agriculture 30:177-187.

Cleveland, D. A. n.d. (Accepted) Plant breeding sci-ence as objective truth and social construction: Thecase of yield stability. Agriculture and Human Val-ues.

Cleveland, D. A., and S.C. Murray. 1997. Theworld's crop genetic resources and the rights of in-digenous farmers. Current Anthropology 38:477-515.

Cleveland, D. A., D. Soleri, and S. E. Smith. 1999.Economics Working Paper, 99-10: Farmer Plantbreeding from a biological perspective: implica-tions for collaborative plant breeding. CIMMYT,Mexico, D.E, Mexico.

Cooper, M., and D. E. Byth. 1996. Understandingplant adaptation to achieve systematic applied cropimprovement-a fundamental challenge. Pages 5-23 in M. Cooper and G. L. Hammer, eds., Plantadaptation and crop improvement. CAB Interna-tional in association with IRRI and ICRISAT, Wal-lingford, Oxford, UK.

Cooper, M., and G. L. Hammer. 1996. Synthesis ofstrategies for crop improvement. Pages 591-623 inM. Cooper and G. L. Hammer, eds., Plant adapta-tion and crop improvement. CAB International inassociation with IRRI and ICRISAT, Wallingford,Oxford, UK.

Cromwell, E., S. Wiggins, and S. Wentzel. 1993.Sowing Beyond the State. Overseas DevelopmentInstitute, London, U.K.

DeLacy, I. H., K. E. Basford, M. Cooper, J. K. Bull,and C. G. McLaren. 1996. Analysis of multi-en-vironment trials-an historical perspective. Pages39-124 in M. Cooper and G. L. Hanimer, eds.,Plant adaptation and crop improvement. CAB In-ternational in association with IRRI and ICRISAT,Wallingford, Oxford, UK.

Duvick, D. N. 1996. Plant breeding, an evolutionaryconcept. Crop Science 36:539-548.

Ellstrand, N. C., and D. R. Elam. 1993. Populationgenetic consequences of small population size: im-plications for plant conservation. Annual Reviewof Ecology and Systematics 24:217-242.

Erskine, W. 1997. Lessons for breeders from land rac-es of lentil. Euphytica 93:107-112.

Evans, L. T. 1993. Crop evolution, adaptation andyield. Cambridge University Press., Cambridge.

- . 1997. Adapting and improving crops: the end-less task. Philosophical Transactions of the RoyalSociety of London: Biological Sciences 352:901-906.

Falconer, D. S. 1989. Introduction to quantitative ge-netics, Third Edition. Longman Scientific & Tech-nical, Essex, UK.

Fischer, K. S. 1996. Research approaches for variablerainfed systems-thinking globally, acting locally.Pages 25-35 in M. Cooper and G. L. Hammer, eds.,Plant adaptation and crop improvement. CAB In-ternational in association with IRRI and ICRISAT,Wallingford, Oxford, UK.

Frankel, 0. H., A. H. D. Brown, and J. J. Burdon.1995. The conservation of plant biodiversity. Cam-bridge University Press, Cambridge, UK.

2000] 391

ECONOMIC BOTANY

Frankel, 0. H., and M. E. Soule. 1981. Conservationand evolution. Cambridge University Press, Cam-bridge.

Hallauer, A. R., and J. B. Miranda. 1988. Quanti-tative genetics in maize breeding, second Edition.Iowa State University, Ames, IA.

Hamrick, J. L., and M. J. W. Godt. 1997. Allozymediversity in cultivated crops. Crop Science 37:26-30.

Hardon, J. 1996. Introduction. Pages 1-2 in P. Eyza-guirre and M. lwanaga, eds., Participatory plantbreeding. Proceedings of a workshop on participa-tory plant breeding, 26-29 July 1995, Wageningen,The Netherlands. International Plant Genetic Re-sources Institute, Rome.

Harlan, J. R. 1992. Crops and man, Second Edition.American Society of Agronomy, Inc. and Crop Sci-ence Society of America, Inc., Madison, WI.

Heisey, P. W., and G. 0. Edmeades. 1999. Part l.Maize production in drought-stressed environ-ments: technical options and research resource al-location. Pages 1-36 in CIMMYT, ed., Worldmaize facts and trends 1997/98. CIMMYT, Mexico,D.E

Heisey, P. W.. M. L. Morris, D. Byerlee, and M. A.L6pez-Pereira. 1998. Economics of hybrid maizeadoption. Pages 143-158 in M. L. Morris, ed.,Maize seed industries in developing countries. Lyn-ne Rienner; CIMMYT. Boulder, CO; Mexico, D.E

Janssen, W., C. Adolfo Luna, and M. C. Duque.1992. Small-farmer behavior towards bean seed:evidence from Colombia. Journal of Applied SeedProduction 10:43-57.

Kato Y. T. A. 1997. Review of introgression betweenmaize and teosinte. Pages 44-53 in J. A. Serratos,M. C. Willcox, and E Castillo-Gonzales, eds., Geneflow among maize landraces, improved maize va-rieties. and teosinte: implications for transgenicmaize. CIMMYT. Mexico, D.E

Lando, R. P., and S. Mak. 1994. IRRI Research Pa-per Series, Number 154: Cambodian farmers' de-cision making in the choice of traditional rainfedlowland rice varieties. International Rice ResearchInstitute, Manila, Philippines.

Louette, D. 1997. Seed exchange among farmers andgene flow among maize varieties in traditional ag-ricultural systems. Pages 56-66 in J. A. Serratos,M. C. Willcox, and E Castillo-Gonzdles, eds., Geneflow among maize landraces, improved maize va-rieties, and teosinte: implications for transgenicmaize. CIMMYT, Mexico, D.E

Louette, D., A. Charrier, and J. Berthaud. 1997. Insitu conservation of maize in Mexico: genetic di-versity and maize seed management in a traditionalcommunity. Economic Botany 51:20-38.

Louette, D., and M. Smale. 1998. Economics Work-ing Paper 98-04: Farmers' seed selection practices

and maize variety characteristics in a traditionalMexican community. CIMMYT, Mexico, D.E

McKnight Foundation. 1998. McKnight FoundationCollaborative Crop Research Program 1998. Uni-versity of California. Davis, CA.

Medin, D. L., and S. Atran. 1999. Introduction. Pages1-15 in D. L. Medin and S. Atran, eds., Folkbiol-ogy. MIT Press, Cambridge, MA.

Ouendeba, B., G. Ejeta, W. W. Hanna, and A. K.Kumar. 1995. Diversity among African pearl mil-let landrace populations. Crop Science 35:919-924.

Poehlman, J. M., and D. A. Sleper. 1995. Breedingfield crops, Fourth Edition. Iowa State UniversityPress, Ames, IA.

Ra,jaram, S., H-J Braun, and M. van Ginkel. 1997.CIMMYT's approach to breed for drought toler-ance. Pages 161-67 in P.M.A. Tigerstedt, ed., Ad-aptation and plant breeding. Kluwer AcademicPublishers, Dordrecht, The Netherlands.

Rice, E., M. Smale, and J.-L. Blanco. 1998. Farmers'use of improved seed selection practices in Mexi-can maize: evidence and issues from the Sierra deSanta Marta. World Development 26:1625-1640.

Richards, P. 1986. Coping with Hunger. Allen andUnwin, London, U.K.

- . 1996. Culture and community values in theselection and maintenance of African rice. Pages209-229 in S. B. Brush and D. Stabinsky, eds.,Valuing local knowledge. Island Press. Washington,DC.

Romagosa, I., and P. N. Fox. 1993. Genotype x en-vironment interaction and adaptation. Pages 373-390 in M. D. Hayward. N. 0. Bosemark, and I.Romagosa. eds., Plant breeding series: plant breed-ing: principles and prospects, 1st Edition. Chapmanand Hall, London.

Salick, J., N. Cellinese, and S. Knap. 1997. Indige-nous diversity of cassava: generation, maintenance,use and loss among the Amuesha, Peruvian upperAmazon. Economic Botany 51:6-19.

Scoones, I., and J. Thompson. 1993. Discussion Pa-per, 332: Challenging the populist perspective: ruralpeople's knowledge, agricultural research and ex-tension practice. Institute of Development Studies,University of Sussex, Brighton, UK.

Shigeta, M. 1996. Creating landrace diversity: the caseof the Ari people and ensete (Ensete ventricosum)in Ethiopia. Pages 233-268 in R. Ellen and K. Fu-kui, eds., Redefining nature: ecology, culture anddomestication. Berg, Oxford, UK.

Simnmonds, N. W. 1979. Principles of crop improve-ment. Longman Group Ltd, London. UK.

Sleper, D. A., T. C. Barker, and P. J. Bramel-Cox,eds. 1991. Plant breeding and sustainable agricul-ture: considerations for objectives and methods.Crop Science Society of America, Inc.. AmericanSociety of Agronomy, Inc., Madison, WI.

Smale, M., M. R. Bellon, and J. A. Aguirre Gomez.

392 [VOL. 54


1999. Economics Working Paper, 99-08: The pri-vate and public characteristics of maize land racesand the area allocation decisions of farmers in acenter of crop diversity. CIMMYT. Mexico, D.E

Smale, M., P. Heisey, and H. D. Leathers. 1995.Maize of the ancestors and modern varieties: Themicroeconomics of high-yielding variety adoptionin Malawi. Economic Development and CulturalChange 43:351-368.

Smale, M., D. Soleri, D. A. Cleveland, D. Louette,E. Rice, J. L. Blanco, and A. Aguirre. 1998. Col-laborative plant breeding as an incentive for on-farm conservation of genetic resources: economicissues from studies in Mexico. Pages 239-257 inM. Smale, ed., Farmers, gene banks, and cropbreeding: economic analyses of diversity in wheat.maize, and rice. Kluwer Academic, Norwell, MA.

Smith, S. E., R. 0. Kuehl, I. M. Ray, R. Hui, andD. Soleri. 1998. Evaluation of simple methods forestimating broad-sense heritability in stands of ran-domly planted genotypes. Crop Science 38:1125-1129.

Soleri, D., and D. A. Cleveland. 1993. Hopi crop di-versity and change. Journal of Ethnobiology 13:203-231.

- n.d. (Accepted) Farmers' genetic perceptionsregarding their crop populations: an example withmaize in the Central Valleys of Oaxaca, Mexico.Economic Botany.

Soleri, D., and S. E. Smith. 1995. Morphological andPhenological comparisons of two Hopi maize va-rieties conserved in situ and ex situ. Economic Bot-any 49:56-77.

. n.d. (Submitted) Broad-sense heritability offarmer-managed maize populations in the CentralValleys of Oaxaca, Mexico and implications forimprovement.

Soleri, D., S. E. Smith, and D. A. Cleveland. n.d.(Accepted) An evaluation of the potential for farm-er and plant breeder collaboration: a case study offarner maize selection in Oaxaca, Mexico. Euphy-tica.

Souza, E., J. R. Myers, and B. T. Scully. 1993. Ge-notype by environment interaction in crop im-provement. Pages 192-233 in M. B. Callaway andC. A. Francis, eds., Crop improvement for sustain-able agriculture. University of Nebraska Press, Lin-coln, NE.

Sperling, L., and M. E. Loevinsohn. 1993. The Dy-namics of adoption: distribution and mortality ofbean varieties among small farmers in Rwanda.Agricultural Systems 41:441-453.

Sperling, L., M. E. Loevinsohn, and B. Ntabomvu-ra. 1993. Rethinking the farmer's role in plantbreeding: Local bean experts and on-station selec-tion in Rwanda. Experimental Agriculture 29:509-519.

Sthapit, B. R., K. D. Joshi, and J. R. Witcombe.

1996. Farmer participatory crop improvement. Im.Participatory plant breeding, a case for rice in Ne-pal. Experimental Agriculture 32:479-496.

Tesemma, T., S. Tsegaye, G. Belay, E. Bechere, andD. Mitiku. 1998. Stability of performance of tet-raploid wheat landraces in the Ethiopian highland.Euphytica 102:301-308.

Tilahun, A. 1995. Yield gain and risk minimization inmaize (Zea mays) through cultivar mixtures insemi-arid zones of the Rift Valley in Ethiopia. Ex-perimental Agriculture 31:161-168.

Tripp, R. 1996. Biodiversity and modern crop varie-ties: sharpening the debate. Agriculture and HumanValues 13:48-63.

Trutmann, P. 1996. Participatory diagnosis as an es-sential part of participatory breeding: a plant pro-tection perspective. Pages 37-43 in P. Eyzaguirreand M. Iwanaga, editors. Participatory plant breed-ing. Proceedings of a workshop on participatoryplant breeding, 26-29 July 1995, Wageningen, TheNetherlands. International Plant Genetic ResourcesInstitute, Rome.

Trutmann, P., and M. Pyndji. 1994. Partial replace-ment of local common bean mixtures by highyielding angular leaf spot resistant varieties to con-serve local genetic diversity while increasing yield.Annals of Applied Biology 125:45-52.

Uphoff, N. 1992. Learning from Gal Oya: possibilitiesfor participatory development and post-Newtoniansocial science. Cornell University Press, Ithaca,NY.

van Oosterom, E. J., M. L. Whitaker, and E. Wel-tzien R. 1996. Integrating genotype by environ-ment interaction analysis, characterization ofdrought patterns, and farmer preferences to identifyadaptive plant traits for pearl millet. Pages 383-402in M. Cooper and G. L. Hammer, eds., Plant ad-aptation and crop improvement. CAB Internationalin association with IRRI and ICRISAT, Walling-ford, Oxford, UK.

Walker, T. S. 1989. Yield and household income var-iability in India's semi-arid tropics. Pages 309-319in J. R. Anderson and P. B. R. Hazell, eds., Vari-ability in grain yields: Implications for agriculturalresearch and policy in developing countries. JohnsHopkins University Press, Baltimore, Maryland.

Weltzien, E., and G. Fischbeck. 1990. Performanceand variability of local barley landraces in Near-Eastern environments. Plant Breeding 104:58-67.

Weltzien R., E., M. L. Whitaker, H. F. W. Rattun-de, M. Dhamotharan, and M. M. Anders. 1998.Participatory approaches in pearl millet breeding.Pages 143-170 in J. Witcombe, D. Virk, and J. Far-rington, eds., Seeds of choice. Intermediate Tech-nology Publications, London.

Wilkes, G. 1989. Maize: domestication, racial evolu-tion, and spread. Pages 440-455 in D. R. Harrisand G. C. Hillman, eds., Foraging and farming: the

2000] 393

ECONOMIC BOTANY

evolution of plant exploitation. Unwin Hyman,London.

Witcombe, J. R. 1998. Introduction (Participatory Ap-proaches). Pages 135-141 in J. Witcombe, D. Virk,and J. Farrington, eds., Seeds of choice. Interme-diate Technology Publications, London.

- . 1999. Do farmer-participatory methods applymore to high potential areas than to marginal ones?Outlook on Agriculture 28:43-49.

Witcombe, J. R., A. Joshi, K. D. Joshi, and B. R.Sthapit. 1996. Farmer participatory crop improve-

ment. 1. Varietal selection and breeding methodsand their impact on biodiversity. Experimental Ag-riculture 32:445-460.

Wood, D., and J. M. Lenne. 1997. The conservationof agrobiodiversity on-farm: questioning theemerging paradigm. Biodiversity and Conservation6:109-129.

Zeven, A. C. 1998. Landraces: a review of definitionsand classifications. Euphytica 104:127-139.

Zimmerer, K. S. 1996. Changing fortunes: biodiver-sity and peasant livelihood in the Peruvian Andes.University of California Press. Berkeley, CA.

394 [VOL. 54

COPYRIGHT INFORMATION

TITLE: A biological framework for understanding farmers’ plantbreeding

SOURCE: Economic Botany 54 no3 Jl/S 2000WN: 0018303497011

The magazine publisher is the copyright holder of this article and itis reproduced with permission. Further reproduction of this article inviolation of the copyright is prohibited. To contact the publisher:http://www.nybg.org/.

Copyright 1982-2002 The H.W. Wilson Company. All rights reserved.

A BIOLOGICAL FRAMEWORK FOR UNDERSTANDING FARMERS' PLANT ... · A BIOLOGICAL FRAMEWORK FOR...

Documents

Transcript of A BIOLOGICAL FRAMEWORK FOR UNDERSTANDING FARMERS' PLANT ... · A BIOLOGICAL FRAMEWORK FOR...