Contributions of decision-tree methodology to a farming systems program

MISSING IMAGE

Material Information

Title:
Contributions of decision-tree methodology to a farming systems program the case of ICTA
Physical Description:
27, 3, 4 p. : ill. ; 28 cm.
Language:
English
Creator:
Gladwin, Christina H
Publication Date:

Subjects

Subjects / Keywords:
Agricultural systems -- Decision making   ( lcsh )
Agricultural extension work   ( lcsh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Bibliography:
Includes bibliographical references.
Statement of Responsibility:
by Christina H. Gladwin.
General Note:
Caption title.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
oclc - 687734765
ocn687734765
Classification:
lcc - S494.5.S95 G53 1980z
System ID:
AA00007161:00001


This item is only available as the following downloads:


Full Text
(`T, r f


Contributions of Decision-Tree Methodology to a ru8
Farming Systems Program: The Case of ICTA

by Christina H. Gladwin*
University of Fl'orida

During the past decade, anthropologists in several cultures have

used natural process or hierarchical "decision-tree" models to predict,

with a surprisingly high degree of accuracy, the actual historical choices

of individuals. Decision trees have predicted selling decisions made

by Ghanaian fish sellers (H. Gladwin, 1971: C. Gladwin, 1975; Quinn, 1978),

farmers' adoption decisions i~n Puebla, Mexico (C. Gladwin, 1976, 1979a

1979b), farmers' land use patterns in Costa Rica (Barlett, 1977), and

farm families' choice of treatment for illness in Pichatero, Mexico

(Young, 1980). In each case where the methodology has been used, the

predictability of these models has been a high 85 to 95 percent of the

actual choice data used to test the model. These success rates are only

remarkable, however, because most studies of economic decision making do

not test the model against actual choice data collected from individuals

(Anderson, 1974; Benito, 1976; Moscardi and de~anvry, 1977).

More recently, decision-tree methodology has been shown to be genera-

Tizable to a wider geo raphical region than one village or town so that some

agricultural production decision rules are shared by farmers who live in

different agroclimatic, socio-economic zones (C. GladwJin, 1979c). Moreover,

it has been shown that a consumer decision-process model based on in-depth

interviews with decision makers in one region (car-buyers in Orange County,

California, USA) can be tested, with 70% success, against choices made by

individuals selected at random in a national survey (H. Gladwin, 1980;

Murtaugh and H. Gladwin, 1980). These same results can then be used to

predict aggregate market behavior, in this case, national market shares
of cars in particular size categories (H. Gladwin, 1980).







The form of a hierarchical decision-tree model, seen in figure 1,

is amazingly (even disturbingly) simple, with decision criteria or factors

or dimensions at the nodes or diamonds of the tree, and decision outcomes

or choices at the endpoints of the branches. The decision criteria can

either be orderings of alternatives on some aspect or dimension or factor

of the alternatives (e.g., Is Profitability of potatoes )Profitability of

corn?), or they can be constraints that must be passed or satisfied (e.g.

"Do you know how to plant potatoes?" or "Do you have the capital or credit

to buy the inputs (seed, fertilizer, insecticides) for potatoes?"). In

either case, the criteria or constraints are discrete rather than continuous,

that is, the alternative "potatoes" either passes the criteria or constraints

or it does not. A decision tree is thus a sequence or series of discrete

decision criteria, all of which have to be passed along a path to a parti-

cular outcome or choice. For example, figure 1, read from top to bottom,

is a hypothetical model of a farmer's decision of whether or not to plant

potatoes. In this example, "potatoes" must pass profitability, knowledge,

and capital constraints in order for the farmer to choose the outcome,

"Plant potatoes". If "potatoes" fails any one of these three criteria, the

model predicts that the farmer will not plant potatoes.

How to build a decision model

Given a form of decision model, the researcher must then select the

decision criteria or constraints that are used in the model. He must

decide which information is actually considered by farmers when they

make a particula-rsdeeisi on. He then must throw away or discard information

which might be interesting to gather but which farmers don't seem to use

when they make the particular decision. In-depth interviews with the people

who actually make the decisions are absolutely necessary in order to build

the model, because only the decision makers themselves are the experts on








Figure 1: a hypothetical example of a decision tree


-3-


) Profitability
corn(+ beans)


nor

'Don't plant
potatoes


Profitability
potatoes


'yes


SDo youP otte
know how t
plant /


yes /

Do you have
capital or cr
buy the inputs
potatoes?


no


~edit potatoes ;
f o r Dntpa


\~the


yes


Plant
potatoes


Don't plant
potatoes







how they make their decisions. Since only they process the information that

the researcher wants to put or represent in his model, in-depth interviews

using ethnoscientific eliciting techniques are a natural starting-point on
the road to building the model. Eliciting techniques and ethnoscience

ethnographies are a uniquely anthropological invention (Spradley, 1979;

Werner et al., 19'79). Although some anthropologists question the reli-
ability of decision criteria which are elicited from the decision maker

(Barlett, 1977; Chibnik, 1980; Cancian, 1972, 1980; De Walt, 1979); most

anthropologists accept the need for ethnographically-valid, inductively-

built, testable decision models. To test an ~inducti velyi-based decision-

tree model, one must, of course, collect actual choice data from a second,

independent, separate sample of decision makers, whose data were not used
to build the model. When such tests have been taken, decision-tree models

have been found to be surprisingly predictive, give-n their simplicity.

These studies, although highly successful and relatively easy to

produce, have not yet been applied to regionally-important policy questions,
not used regularly by national or international (agricultural) research

centers. The reason for this neglect is that the decision-tree tool

presupposes a farming-systems research and extension (FSR/E) program, in
which the farmer as decision maker is at the focus or center, "directing"

the program. Decision-tree research fits naturally into such a program, even

though it may be a luxury good to an agricultural program or project which
talks about but has little interaction with the farmer. To demonstrate

this point, section I summr~arizes the purpose and uniqueness of a farming

systems program, and describes the methodology of one such program, that
of the Guatemalan Institute of Agricultural Science and Technology (ICTA).

Section II then shows how one decision-tree model was developed and used

as part of ICTA's program to generate appropriate technology in the Western







Highlands (Altiplano) of Guatemala; while section III concludes by

summarizing the policy recommendations implied by and taken from the

results of testing the model. By showing the FSR/E team how farmers are

making their agricultural decisions, decision-tree models can recommiiend

effect program policies, so that the program is --- in reality as well

as in theory --- centered on the farmer.



1. A Farming Systems Program: Focusing on the Farmer as Decision-Maker

A farming systems program treats each farm as a unique farming system

of interrelated activities, e.g., crops, livestock, forest, pasture, off-

farm labor, etc. Within that system, each farmer uses a set of resources

(land, capital, time, labor, energy) and faces a certain physical and

socio-cultural-economic environment, which imposes certain constraints on

his or her farming operation.1_ Given these constraints imposed on the farmers

from environmental forces, the farmer makes management decisions, thus

integrating the set of available resources and the environment (Hansen

et al., 1980:p.1). The farming systems approach thus starts with the

constraints and conditions facing the farmers as given, and develops,

through experimentation on farmers' fields and with knowledge of their

situation, recommendations designed to improve their standard of living.

The goal of a farming systems program thus becomes one of designing

recommendations or strategies to enable the farmers to deal with the con-

straints on their specific farming systems. In most farming systems

programs, four stages of research are distinguished to accomplish this aim:






(1) The descriptive or diagnostic stage in which the actual
farming system is examined in the context of the "total"
environment--to identify constraints farmers face and to
ascertain the potential flexibility in the farming system
in terms of timing, slack resources, etc. An effort is
also made to understand goals and motivation of farmers
that may affect their efforts to improve the farming
system.

(2) The design stage in which a range of strategies are iden-
tified that are thought to be relevant in dealing with the
constraints delineated in the descriptive or diagnostic
state.

(3) The testing stage in which a few promising strategies
arising from the design stage are examined and evaluated
under farm conditions, to ascertain their suitability
for producing desirable and acceptable changes in the
existing farming system. This stage consists of two
parts: initial trials at the farm level with joint
researcher and farmer participation, then farmer's
testing with total control by farmers themselves.

(4) The extension stage in which .the strategies that were
identified and screened during the design and testing
stages are implemented. (Gilbert et al, 1980: p..11)

In ICTA, research and extension methodologies were developed to guide

the program in each stage. Because these are described in detail elsewhere,

a brief summary of ICTA's stages should suffice here (Fumagalli and Waugh,

1977; Hildebrand, 1977, 1978, 1981; ICTA, 1977; Ortiz, 1979).

(1) The survey (sondeo) stage in which a multidisciplinary team of

plant breeders, pathologists, agronomists, anthropologists, agricultural

economists and sociologists reconnoiter an area where technology generation

and promotion is being initiated (Hildebrand, 1979). -Information about

farmers' cropping systems and their socio-economic conditions is gathered,

while at the same time the team that will work on technology generation

in the future is forced to talk to the farmer, the potential adopter of the

technology, to ascertain the problems he or she faces.





(2) The generation-of-technology stage in which the technical team

experiments to better the traditional technology, mostly on farmers' fields

(ensayos de finca). Some experiments of the commodity programs (maize,

beans, swine, etc) are highly controlled trials on the regional experiment

stations. The majority of experiments, however, are conducted on farmers'

fields by technicians assigned to a region. Typically with these experiments,

the technicians perform the work, holding constant the levels of inputs

and cultural practices traditionally used by the farmer, who has loaned

the land for the experiment, and varying only the experimental variable,

for example, the crop variety or plant population (ICTA, 1977:p.3). Farm

budget data are also collected at this stage, to add to ICTA's knowledge

about traditional technologies.


(3) The test stage (parcelas de prueba) in which the farmers them-

selves test ICTA's recommended technology which was generated in stage 2.

At this stage, volunteer farmers perform the work on their own fields, pay

all input costs and typically plant half of one field to ICTA's technology

and the other half to their own technology, in a race to see which does

better (0rtiz, 1979:p.15). Data concerning the time and capital requirements

of TICTA's technology vs. the traditional technology are also gathered.

Clearly, the farmers' tests start the process of technology diffusion and

transference.


(4) The evaluation (evaluacibn de la aceptabilidad) stage in which the
socio-economic team returns to the farmer one year after he or she has tested

ICTA's technology, to see if he or she is still using (i.e., adopting) the

new technology. If farmers are not adopting, the ma*in factors -limiting

adoption are elicited. An "index of acceptability" is calculated for each






CT ---I. .-: ~_;-::01a and Hildebrand, 1979). If this is low,

ne tc anical teal wril reconsider the benefits of the "improved technology"



$1, Th transf~erene or extension stage in which ICTA promotes the

s- of~- ihe acceptable new technology, in collaboration with DIGESA (Dir acci'on

general de Servicios Agricolas), the extension service. At this stage, ICTA

and DIGE5A technicians work with farmer groups, cooperatives, and para-

professionals working with farmer groups, e.g., promoters of Escuela Extra-

Escolar (Adult Education) or World Neighbors (ICTA, 1977:p.5; Fumagalli and

Waugh, 1977: pp.19-20).

The success of ICTA's program in the Altiplano of Guatemala can be seen

by the successful diffusion of several technologies there: a) 'San Marcefio,'

an improved corn variety, adopted in many parts of Quezaltenang'o (especially

the subregion of Llanos de Pinal); b) "Chivito," an improved wheat variety,

adopted in parts of Quezaltenango and Totonicapin; c) urea as a second

application of fertilizer for both corn and wheat; and d) the introduction

of vegetable production in irrigated, terraced parts of San Marcos. Although

agricultural production in. the Altiplano still has a long way to go to

keep up with population increases, the methodology to generate production

increases is clearly there.


Focusing on the Farmer as Decision Maker

Given that farm trials and farmers' tests are on farmers' fields, and

the farmer is consulted both during the diagnostic (sondeo) and evaluation

stages, it is obvious that the farmer is at the center or focus of a farming

systems program, like that of ICTA. This attention to the problems and needs







of the farmer is aptly captured by figure 2, reprinted with the courtesy

of CIMMYT (The CIMMYT Economics Program, 1980:p.12). It shows both internal

economic conditions (such as family income, food preferences, risk aversion)

and external economic conditions (markets and institutions), as well as

natural physical conditions (such as climate, rainfall, frosts, pests,

diseases, weeds, soils, slope) impinging as constraits or decision criteria

on the farmer's decision processes to grow and rotate crops, provide for the

family's consumption needs, hire labor, and make crop production decisions

(such as the timing,method, and levels of inputs used for specific crops).

With this focus on the farmer who, after all, can make or break a farming

systems project by adopting or rejecting the new technology, the program
should know:

(1) what decisions is the farmer making?

(2) what alternatives is he or she choosing between?

(3) why is the farmer choosing the particular outcomes chosen? Why

are farmers doing what they "traditionally" do? What logic, reasoning is the

farmer using? For example, why is he or she planting the crops that are

planted? If the new technology is not adopted, why not?

(4) when, at what stage should a farming systems program do decision

research to answer the above questions?

The answer to the first three questions must come from the particular,

task-specific decision context studied, as will be shown in the next section.

The answer to the last question is obvious: decision modeling is most

appropriate and useful to a farming systems program at first, the diagnostic

stage in which farmers' reasoning behind "the traditional way" must be

understood by the multidisciplinary, team before they attempt to improve on it,











Figure 2.1 Various Circumstances Affecting Farmers' Choice of a Crop Technology


INTERNAL

Farmers' Goals-Ilncome,
food preferences, risk.
Resource Constraints
--Land, labor, capital.


EXTERNAL


1


NATURAL CIRCUMSTANCES

--- wCircumstances which are often major sources of uncertainty for decision-making.


ECONOMIC CIRCUMSTANCES


I
I
C


FARMER DECISIONS


Farming System

Cropping Pattern
Rotations
Food Supply
Labor Hiring,
etc.


Technology for
Target Crop


System Time, Method,
interactions Amount for
Various Practices.


Soils/Topography
Soils Type
Slope


(Reprinted from the CIMMYT Economics
Program, 1980, courtesy of CIMMYT)


Figure 2:


Focusing on the Farmer as Decision Maker


Institutions
Land Tenure Poliey
Credit IniCJences


Markets
Product
Inputs


Climate
Rainfall
Frosts


Biological -
Pest
Diseases
Weeds




SII


and fourth, the evaluation stage in which the team must know why farmers

are or are not adopting the new, improved technology. A decision study is

necessary at both of these stages because, although factual, quantitative

data are impressive, easily arranged in a table, and seemingly-rigorous,

they do not tell program planners why farmers are doing what they're doing.

The latter information can only be provided by a decision study which discovers

what strategies or plans the farmer has traditionally used, to cope with

his or her environment, constraints, and problems. An example of a decision

study used by ICTA at the diagnostic stage is presented in the next section.

Although use of the decision-modelling tool is most appropriate at the

initial diagnostic stage, it is also necessary at the evaluation stage, to

elicit from the farmers themselves the reasons for their non-adoption (Ortiz,

1979:p.15). At the evaluation stage it is not enough to interview farmers

to measure adoption accurately and put together an "index of acceptability,"

although this step is essential (Chinchilla and Hildebrand, 1970). Adoption

decision trees must also be formulated and tested at this stage, to pinpoint

the main factors Timiting adoption of each recommendation in each agro-

climatic, socio-economic zone. Fortunately, adoption decision trees are the

simplist and easiest kind of tree to formulate (Gladwin, 1976, 1977, 1979a,

1979b, 1980), and the factors limiting adoption are relatively easy to

elicit from the farmer.

Finally, wh:1e adoption decision models are most common in the evaluation

stage, in an international research center it is sometimes useful to evaluate

experimental technology prior to transfer to a national research/extension

center. The objective of decision research in this case would be to help the

international center "select, from'an immense range of possible experimental




-IZ-


treatments, those to be recommended for testing at the farm-level in col-

laboration with the ... national research institution" (Ashby, personal

communication). An excellent example of this way to use decision models is

given in the paper by Ashby and de Jong, which uses a model of farmers' choice

of tillage methods (manual, oxen, vs. tractor) to evaluate fertilizer appli-

cation technology now being developed by the IFDC/CIAT Phosphorus Project

(Ashby and de done, 1980).



II. Farmers' Cropping.Decisions in the Altiplano of Guate~mala

The farmer's cropping decision is a two-stage choice process in which

the farmer first narrows down the complete set of possible crops to plant

in the Altiplano to a feasible subset that satisfies certain minimal con-

ditions. For example, given eight to ten possible crops to plant, a farmer

may eliminate (rapidly, often unconsciously) vegetables because of a lack

of irrigation. He or she might not consider planting potatoes because of a

lack of knowledge of how to plant them or apply pesticides. Alternatively,

the farmer might not even _thi of growing coffee because the land is at too

high an altitude to grow coffee. With the smaller subset of feasible crops

that emerges from this "elimination-by-aspects" stage (Tversky, 1972), the

farmer proceeds to stage 2, the "hard-core" part of the decision process,

the theory of which is described elsewhere (Gladwin, 1980). Stage 2 of the

cropping decision allocates the farmer's available land to the crops which

pass stage-1 constraints. If the farmer has a lot of land, stage 2 is a

simple decision process: the farmer will plant all the crops that pass

stage-1 constraints. If, however, the farmer does not own or operate much

land, the crops 'that pass stage-1 constraints compete for the little land

there is, and the decision process and model become more complicated. In






the most general terms, stage 2 of the model proposes that farmers in the

Altiplano give first priority to crops or systems of crops that are at least
two times as profitable as corn, the main consumption crop. Second, they

plant as much corn as is necessary to fulfill the family's consumption require-
ments between harvests. Third, if farmers still have more land, they then

plant a crop or system of crops that is not twice as profitable as corn,
but may be equally as or a little more or less profitable than corn (+
beans.

Figure 3 represents the choice process of stage 1, in which a farmer
eliminates, unconsciously or "preattentively" (H. G~ladwJin and M~urtaugh,

1980), some of the possible crops seen in the set at the top of the tree:

corn in association with two kinds of beans (corn + frijol + haba), wheat,

potatoes, vegetables, fruit trees, monoctop of beans (frijol), a monocrop
of bush beans (haba), and coffee in.association with avacado. (To shorten

interviewing time with the farmers, only those crops that had some possi-

bility of passing the stage-1 constraints in the Altiplano are considered

to be alternatives and put in this set.) Each possible crop (e.g., potatoes)

of each individual farmer is then "put down" this tree; .i.e., each farmer

is.asked a series of six questions about each crop in the possible set,

whether or not it is actually planted. To avoid elimination at this stage,

the crop must pass or satisfy six conditions or constraints: a consumption

or market constraint (there must be a demand for the crop); altitude, soil,

and water constraints (the crop must respond on the farmer's 1.and at his or

her altitude with his or her soils and available water); a knowledge constraint

(the farmer must know how to plant the crop or have someone to help); a time
or labor constraint (the farmer must have either the time or (family or hired)

labor available to help plant the crop; and a capital or credit constraint:







trestant~s. t_~r ar top ~e. r mended~ for ies~ting~ at the farm-level in col-

aboratc :ita ... ?I- anal research1 institution" (Ashby, personal

communications neclent eg:-:al of this way to use decision models,s

giveln -in ;i scan-~ L' :-uc~: and ;e cng, which uses a model of farmers' choice
of tillage~ ,et.::d: 1Era:1:-l, oxen, vs. tractor) to evaluate fert}iz~er appli-

cation techni'c-~ ~ being developed by the IFDC/CIAT Phosphdrus Project

(Ashby and de Jonei 1980).


II. Farmers' Cropping Deci ions in the Altiplano of Guad emala

The farmer's cropping decision is a two-s a~ge choice process in which

the farmer first narrows do n the complete Xet of possible crops to plant

in the Altiplano to a feasible subset t,Wat satisfies certain minimal con-

ditions. For example, given ei~f~ tB ten possible crops to plant, a farmer

may eliminate (rapidly, often un ~hciously) vegetables because of a lack

of irrigation. He or she mig~ not c nsider planting potatoes because of a

lack of knowledge of how to plant them i apply pesticides. Alternatively,
the farmer might not eve hik f roin coffee because the land is at too

high an altitude to g owcoffee. With the s aller subset of feasible crops

that emerges from s"el imi na tio n- by-a aspect s stage (Tversky, 1972), the

farmer proceeds ,to Stage 2, the "hard-core" part qf the decision process,

the theory of which is described elsewhere (Gladwin\ 1980). Stage 2 of the

cropping decision allocates the farmer's available la ~d to the crops which

pass stage-1 constraints. If the farmer has a lot of la d, stage 2 is a

simp1,e decision process: the farmer will plant all the crot that pass

stage-1 constraints. If, however, the farmer does not .own o operate much

Stand, the crops that pass stage-1 constraints compete for the 1'ttle land

there is, and the decision process and model become more complica ted. In





the farmer must have the capital or credit to obtain the necessary inputs

(seed, fertilizer, insecticides, labor) to plant the crop.

For a monocrop or stand of tree crops such as fruit trees, coffee,

avacado, there is an additional "investment" constraint: the farmer must

own land for the crop, and be able to wait five years for a product on the

land. If a farmer passes all six or seven constraints with a crop, then the

model in stage 1 "sends" the farmer to stage 2 with that crop. If a farmer

answers no to one question or constraint for a crop, then that crop is

eliminated from the set of feasible crops that are sent on to the more

detailed decision process of stage 2, represented by figure 4.

For generality, it is assumed here that three systems of crops and corn

have made it to the feasible subset at the top of figure 4. The first

criterion in the flowchart then considers each crop system, independently

of the others, and looks for a "very profitable" crop, i .e., a crop which

is at least twice as profitable as the consumption crop corn. Each alter-

native cropping system is compared with corn since---as the farmers testify

---Maiz es principal or "Corn is first." Since all the feasible crops are

not rank-ordered on profitability, the order in criterion 1 is a partial

but not a full order: e.g., the profitability of wheat is compared to that

of corn, the profitability of potatoes is compared to that of corn, but

wheat and potatoes are not compared or ordered on profitability.

The very profitable crops, which may be three or four or five times as

profitable as corn but must be at least twice as profitable, are then "sent

down" the left-hand branch of the tree. In this flowchart, only crop i ft

considered by the farmer to be twice as profitable and sent down the left-

hand path. Crops j and k and of course corn are not considered very pro-

fitable with respect to corn, and so are sent down the right-hand path to

criterion 3. On the left-hand path, however, the model predicts that the



































































FIGURE 3: ELIMINATION BY, ASPECTS IN STAGE 1


Ceorn + rriio1 + haba-, whea~t, potntoes;, vegetables, fruit trees,'
monocrop of frija monorop of hanba, coffee +e avocado


Do you eat or
can you sell crop
in a nearby market
.to a merchant?


"1-------
or/


Demand


yes


SDoes
SCrop X yield well \ --2 lmn
at your altitude,/o
onl your soils? Co


yes
Do you have.
Eliminate
irrigation or land no Co
moist enough to plant/ L
Crop X?


Altitude '
soils


Water
Requirements


yes
Do `
,i you know how \-----
to plant /
\~Crop X ?,


Knowledge


yes

Do you have the Eiiae
time or labor available )(-- o-
to help you plant
crop X ?


Time
Labor


yes

the capital or
tobuy inputs ro


yles


IEliminate
credit n rp)
for


Capital,
Credit





Ability to invest
in tree crops:


SDo you own land
and have thle capital to'\
waitt 4 or 5 years until
\ he tree produces?
(Sellers, 1977)/

yes

GO TO STAGE 2 WITH CROP


Eliminate
no Crop X


__---- -------j Eliminate
no Crop X


Eliminate
---I Crop X
L.






farmer will plant the very profitable crop first, even though the farmer

has to take some land out of corn, with the result that the farm may not

be able to produce the family's yearly consumption requirements for corn.

If the farmer still operates more land after planting the very-pro-

fitable crop 1, criterion 2 in the model sends him or her to the consumption

criterion 3 on the right-hand branch of the tree, which asks if the farmer

has enough land to plant the not-so-profitable cash cropss, after he or she

has planted enough corn to fulfill the family's consumption requirements.

If there is enough land, the model predicts that corn be planted first before

the decision of whether to plant one or more cash crops, in the subset below

criterion 3.

The latter "diversification" decision, between two or more cash crops,

is simple if the farmer has enough land to plant both crops. If there is

not enough land and the farmer cannot rotate the crops within the year, then

he or she must decide between them, by "trading off" the profitability and

riskiness of the cash crops. For brevity, and because results show that most

farmers manage to squeeze in both cash crops on their land, the model of this

subdecision is presented elsewhere (Gladwin, 1980).


The cash crops) and corn' compete for land

If farmers do not have enough land to be self-sufficient in corn and

plant a cash crop, they are asked the string of questions in figure 5, to

see if there might be extenuating circumstances which would lead them to

take some land out of corn to put into a cash cropss, even though by so

doing, they would then have to buy some corn in the market for home con-

sumption. The decision for farmers now is between a crop-mix of cash crops)

and corn, versus a' crop-mix of just corn. There are four criteria which










system of crops 1,

GGross Returns System of crops X


yes: crop X = system of crops i


Plant crop X = crops 1, even
though the family's consumption
requirements are not fulfilled


Decide between
crops) j and crops) k
in Figure 7


FIGURE 4
STAGE 2 of the
cropping decision
in the AZltiplanno


system j, system k, corn (+ beans)

)(2 times) Gross Returns corn (+beans)?


1


no: crop X= system j, system k, corn (+ beans)3


1











I .


3
If you plant
all the corn you need to
fulfill the family's consumption
requirements, do you still
have land to plant crop X?

yes no

Plant enough corn' (+ beans Go to
+ crop X intercropped) to Figure 8
fulfill the family's corn
needs for one year


c~rop(s) j, crop(s) k)


yes


After you plant
corn, do you have enough
land to plant both crops) j and
crops) k in separate fields, or can
you rotate them in the same field
in one year?


-- -


s) j
Sk


yes


Plant crop(
and crops)




~ IEUL 5 _1 sLI~ UCC~J-CVII LV t~LOllL a ~aJI1 c~v~ airu cvLI~ vL JUDL LVLLI \dllll Ura115/


Plant cas
and corn


cash crops) X,~ corn (and beans)3

S the consumption
~riterion in figure 67


/ Can you \
/interplant crop X1
/and corn on the same
piece of land, or in
surcos doubles,
\a la ICTA?/


yes /
jPlant cash crop X
land corn (and beans)


no
The subdecisio \
to rent land to
plant crop X

yes rent no; ~don't rent land
;h crop X Are there
(and beans) special conditions
-----limiting produce tion
of corn?


yes no
Plant cash crop X i
Sand corn (and beans)

/ Do
// you needJ
\to plant crop X
\to/ have c ash ?


Do you have 'Plant
'the capital or credit, corn
to buy corn in the :(and beans)
marketplace when
\you need to?
yes no
/Can you Pl+bantj
Stake the risk corn
of buying corn (+beans
int0amarket? --

Is it 1xn
proitable to grow\
and sell crop X, and ,--
buy corn in the no ~Plant
marketplace? / /corn
I(+beans)




'lU'


lead farmers to grow a cash crop and corn under these circumstances. First,

they can multicrop or interplant cash crop X and corn, in such a way that

their yields of corn per land area do not diminish substantially. Second,

they can rent land for the cash crop, and devote their own land to corn.

This constraint is really a subdecision: a farmer will rent land if rented

land is available, and they have the time to search for the owner before

planting time, and they have the capital) available to pay for the land, and

they think that renting land proves profitable. Third, they think there are

special conditions which limit the production of corn to only a portion of

their land. For example, farmers may plantcorn only on the fields around

the house, to discourage the theft of green corn in the field by people and

birds. Fourth, farmers need cash, and don't have another source that would

give them enough cash, like full-time, off-farm employment.

Besides processing these criteria, which act to encourage farmers to

grow a cash crop, Altiplano farmers who need cash must also pass three

constraints which act to discourage cash crop production. First, they must

anticipate that they'll have enough cash/capital to be able to buy corn in the

marketplace when the family runs out of corn before harvest time. For farmers

with severe capital constraints, planting and then storing a year's supply

of corn is insurance against a later shortage of capital. Second, they must

be willing to take the risk that there will be corn in the marketplace when

they go to buy it, a not inconsiderable risk. Finally, they must think it

is profitable to plant and sell a cash crop, and then buy corn in the market,

before they will plant some "corn land" to a cash crop. In summary, the first

four criteria in figure 5 are factors which encourage farmers to switch

some (needed) corn land to a cash crop, while the last three criteria act as

a brake on this switch.






Results

The model in figures 3 to 5 was tested against actual cropping-choice

data gathered from 118 farmers in six sub-regions of the Altiplano which are

quite different in terms of altitude, the predominant crop mix, the extent

and type of off-farm labor opportunities available, the languages) spoken,

and the percent of the population which is rural, indigenous, and in agricul-

tu~re. For more detail about these sub-regions, the reader is referred to

a longer report (C. Gla~dwin, 1979c). Even though there are similarities

among farmers in the same subregion, there is also considerable individual

variation in crops grown by farmers within the same subregion, so that the

model tests or processes data from each farmer separately or independently

of the other farmers.3/ Indeed, there can be a separate decision-tree for

each farmer, with different subsets of crops preceding from stage i to stage 2,

and to different branches in stage 2. For brevity, however, the results of

testing figure 4 are summarized for all farmers in figure 6, again assuming

that a subset of crops 1, j, k and corn (+beans) has passed stage-1 constraints.

(The reader is referred elsewhere for the results of testing stage 1 and

figure 5 (C. Gladwin, 1979c: pp.40-45; pp. 50-52).)

The results show that, of the 118 farmers whose data test the model,

only 44 farmers. (37% of the total) have a crop or system of crops which passes

stage 1 and moreover is twice as profitable as corn. Data from these
farmers pass to the left-hand branch of-the tree to the outcome "Plant that

crop even though you may not fulfill the family's consumption needs for

corn." Thus only one-third of the farmers sampled have a cash crop which is

so profitable that it is planted first, or given first priority, before corn
and beans.






The results also show thtol adf "cs c. r o- ra

by farmers, to be so profitat'c~ _:~~ at~cl .,he ; f plantC -.s~t, b~efore ca

(Ibid, table 3). These cash across are mos.; z.* stems or rotastions of croos

which require irrigation or special so': :- 'rate conditions, m~arked~ by

sandy soils and an afternoon cloud cover. ':.8 results shiow that one crop/

year of rainfed vegetables, potatoes, or wheat is ne:j profitablee enough to be

planted first, before corn. Later, the implications of these results for

diversification of crops in the Altiplano will be discussed.

After planting the twice-as-profitable crop, the farmers on the left-

hand path then pass on to criterion 2, to see if they have more land left to

'plant another crop. Of the 44 ~farmers, only two farmers do not have more

land left to plant another crop.


Farmers without ver~y-profitab'le cropping systems

Seventy-four farmers go directly to the right-hand path of the tree in

figure 11, because they do not have a crop which passes sta e 1 and is twice

as profitable as corn. Thus two-thirds of the farmers sampled go directly to

the consumption criterion (3). For them, "Maiz es principal (corn is first):'

i.e., they consider their family's consumption requirements for corn before

their need or desire to plant a cash crop. In total, 97 farmers proceed

to the decision process on the right-hand path of the tree: of those, 23

farmers come from the left-hand path because they have more land left after

planting the twice-as-profitable cash crop, and have two or more crops left

in their feasible subset. At this point, the decision process stops for

two farmers, because corn is the only crop left in the feasible subset.

Of the 95 remaining farmers, 59 farmers (50%/ of 118) "pass" the con-

sumption constraint, i.e., they do have enough land to plant their family's





data from 118 farmers


System of crops 1, system ), system k, corn (+ beans))


Gross Returns
e System of crops X

yes = 44(37%.) farmers
T--

LPlant system X even though '
1 ~error) the family's consumption re-
quirements are not fulfilled


ns ?
corn (+ beans)/


)
2 ( times)


Gross Returr


no = 79-(63)' cases come here directly
-- -- -- + 23(19%) cases fuom left-hand patio
97(82%) cases


a)Only corn is left in this subset, so
decision process stops for 2 farmers.
b): Decision process continues for 95(81.)


cases.


Do `. 2
< you stti
hJave more
\land?~

farmers


SIf you plant 3
,' all the corn you need to
fill the family's consumption'
requirements, do you still
ave land to plant crop X?


no =2(1%) farmers


yes 1 42(36%)


Yes = 59(50%) farmers


Plant enough corn (+ beans
+ crop X intercropped) to
fulfill the family's corn
needs for one year


no 1 36(31%) farmers

Go to
Figure 5


Subset of remaining crops:
Only 1 crop is left in this subset,
so decision process stops for 19(16%.) farmers.
More than 1 crop Is left in subset'
so decision process continues for 23(19%) farmers


a)


Subset of remaining crops:
Only 1 crop is left .in subset,
so decision process stops for 30(25%.) farmers. (1 error)
More than 1 crop is left in this subset,
so decision process continues for 29(25%.) farmers


Stage 2err:

7 Figure 4
2 Figure omitted
4 Figure 5
13 errors
success rate I 105 90%
..118


After you plant
corn, do you have enough
,land to plant both crops) j and
\jcrop(s) k in separate fields, or can
you rotate them in the same field
\ in one year


yes: 26(22%) farmers (5 errors) no: 3(

Plant crops) De
iand crops) k cro De


3%) farmers(2 errors)

ide between
p(s) j and crops) k


* All percentages in this figure are calculated


Figure 6: Staege-2 results
in six zon~es o te ltplno






consumption requirement of corn and one or more cash crops. TheseE carers~

therefore proceed to the outcome or command, "Plant corn (+ho;~n:' :1as-

cash crop that can be interplanted with corn in the same field (t.g.,

potatoes, or peas or fruit trees)." After planting enough .corr o !
their consumption needs between harvests, these farmers then allocate t .2

remaining fields to the cash crops which remain in their feasible subsets.

For 30 of the 59 farmers (25%/ of the total 118), only one cash crop is left

in the feasible subset at this point; the model therefore predicts that they

plant this cash crop and the decision process stops. For 29 of the 59

farmers (25% of the 118), two or more cash crops are still in the feasible

subset at this point, so that the decision process continues on to the diver-

sification criterion 4. Here, the results show that most (26) of the 29

farmers with 2 feasible cash crops (feasible because they pass all stage 1

constraints) manage to squeeze out the land required to grow them both; or

alternatively, the climate is such that they can "fit in" or rotate the two

crops on the same field within the year.

The cash crop and corn compete for land

Of the 95 farmers on the right-hand branch of the tree, 36 farmers

(31% of the 118 farmers) "fail" the consumption criterion, i.e., they do not

have enough land to plant all the corn the family consumes in a year and a

cash crop. Data from these farmers are therefore sent to the decision process

in figure 5, which is tested with 48 cash crops of 36 farmers. This model

predicts planting of a cash crop, even though consumption requirements of

corn are not met, in 7 cases in which crops are interplanted or multicropped

with corn; in five cases in which land can be rented for the cash crop; in

15 cases where special conditions limit the production of corn on some

farmers' fields; and in 9 cases in which the farmer needs a cash crop to have







cash, and passes the profitability, capital, and risk constraints associated

with buying corn in the marketplace. On the other hand, the model predicts

that the farmer plant only corn (+beans) in 8 cases in which he or she has

other sources of cash and doesn't need a cash crop; in two cases in which he

or she cannot take the risk of buying corn in the marketplace; and in two

cases in which he or she does not consider it profitable to grow and sell

the cash crop in order to buy corn in the marketplace. The results thus

show that two-thirds of the 36 farmers (20% of the total 118 farmers) plant

a cash crop, even though they then do not have enough land to fulfill their

fami lies' consumption requirements for corn. Only one-third of the 36 farmers

(10% of the total farmers) plant just corn (+beans).


III. _Implications of the Results for ICTA

The results of testing the model in section II have policy implications

for ICTA, in two main directions. The first direction concerns recent debates

in Guatemala over the value of corn production in the Altiplano, expressed
in statements such as:

"Corn is not the right crop for the Altiplano";

"The growing season in the Altiplano is too long for corn";

"There is too little rain for corn";

"Corn is not a profitable crop that will help conditions change".

If one were to suggest, "But the people eat corn--three times a day," the

typical reply would be:

"But farmers should grow and sell higher-valued cash crops (fruits
and vegetables) and buy corn, the consumption crop, in the market-
place."

Since one of ICTA's original aims was to increase foodgrain production

(maize, wheat, rice, sorghum), and ICTA's main commodity program is the




-LI-


maize program, debates over the value of corn production in the Altiplano
hit home.

The counter to the argument, of course, is that what farmers should

do is not always (or often?) what they actually do. As is shown by the

results of testing the decision tree model, in section II, half of the

farmers sampled (50%) plant the family's consumption requirements for corn

first, and a cash crop second. Another 10% of the farmers sampled plant

only corn and beans, because they do not have enough land to be self-
sufficient in corn and plant a cash crop. Thus 60% of the farmers sampled

plant a cash crop only if they can first me-et their consumption needs for
corn. These results suggest that any attempt to diversify farmers' cropping

Patterns in the Alti lano must try to improve corn yields per land area.

When corn yields are improved, farmers can then take some land out of corn

to put into a cash crop. Improving corn yields would thus seem to be the
diversification strategy most capable of reaching the majority of farmers

in the Altiplano.

Other diversification strategies, although of possibly lesser impact and

importance, are also implied by the results. Since a sizeable minority of
the farmers sampled (37%) now plant a very-profitable cash crop first, and

corn second, ICTA might try to introduce such a profitable cash crop into

more subregions of the Altiplano, to increase the percentage of farmers who

successfully plant and market such a profitable crop. However, the results

show that only a handful of systems of crops qualify for the classification

of "twice-as-profitable-as-corn." These systems include: two crops of

potatoes per year; two to three crops of vegetables plus one crop of

potatoes; a rotation of wheat and vegetables (or potatoes); coffee, and a

monocrop of fruit trees. Few farmers perceive one crop of rainfed vegetables







or potatoes or wheat to be twice as profitable as corn, and thus capable of
replacing corn as the "number one", predominant crop. Further, only a few

subregions of the Altiplano have the climate and/or irrigation necessary to

produce these special multiple-cropping systems. Finally, the lack of a
strong market for some of the crops, e.g., vegetables and potatoes, may set
a limit on their future profitability (Smith, personal communication).

Therefore, introduction of a very-profitable cash crop into an area will not

be an effective diversification strategy capable of reaching the majority of

farmers in the Altiplano. Clearly, such a strategy would be effective for

only a few farmers.
The results of figure 6 show other ways to diversify a farmer's crop

mix. Multiple cropping, or planting a cash crop within double rows of corn,

without significantly decreasing corn production, should prove to be the
most effective diversification strategy to reach very-small farmers (i.e.,

those with one-quarter of a hectare or less) who also have family labor

available (Hildebrand, 1976). Unfortunately, knowledge of "relay crops"

or doublee rows" a la ICTA has not yet diffused widely in the subregions of

the Altiplano sampled. Another problem with this diversification strategy is
the shortage of family labor in some parts of the Altiplano (e.g.,

Totonicapan) due to competition from the indigenous weaving (artesania)

industry (Smith, 1978).
The results also show that farmers plant a cash crop when they can rent

land to do so. The increasing scarcity of "rentable" land, however, makes
this-ai-velsti fiction strategy of limited future significance. Surprisingly,

the results also show that the "special conditions" criterion accounts for

relatively more cases of cash crops planted than any other criterion in figure
5. Clearly, many farmers feel that corn is not suitable on some of their






fields, and so plant a cash crop on those fields. Unfortunately, this

particular criterion is not very amenable to policy intervention. (The

exceptional policy recommendation would be to encourage extension agents to

urge farmers to rotate their crops, e.g., wheat and corn.)
Finally, the results suggest a rather Machiavellian diversification

strategy which actually has been employed in the Altiplano, starting from

the time of the Conquest. Since farmers plant cash crops to have cash, one

way to push them into diversifying their crop mix is to increase their
need for cash, by intensifying their involvement in a cash economy, thus

decrea si ng thei r sel f- suffi ci ency The Spanish in Pedro de Alvarado's time

did exactly that in Totonicapan, by introducing the crop of wheat and then

levying taxes on the indigenous population that could only be paid in cash
or wheat. Needless to say, this diversification strategy has adverse

secondary effects, such as an overall decrease in real rural family income,

so that the (social) costs of implementing it are greater than the benefits.

For this reason, its use as a policy instrument is not recommended.

In short, the policy ICTA has been following, since its establish-

ment in 1973, of supporting and extending its maize program in the Altiplano,

is clearly the most effective means of encouraging Antiplano farmers to

diversify their crop mix and thus increase their incomes. The results of

testing the decision-tree model in section II clearly support ICTA's current

policy and emphasis on maize production, and would not support any switch

of emphasis or scarce resources away from corn to vegetable, potato, or

fruit production.


Implications of the Results for the Design of Farm Trials
The second area where policy recommendations taken from this research







apply, for ICTA's goals and objectives, is in the design of farm trials for

cash crops (ensayos de finca). As mentioned in the review of ICTA's stages

of research in section II, before farm trials are planned in a new region,

a multidisciplinary team surveys the region, both to gather information

about the traditional technology and problems farmers perceive, as well as

to make recommendations to the technical team who plan and plant the farm

trials. The recommendations are written up in the survey (sondeo) report,

but in most sondeos, the technical team who plant the farm trials also parti-

cipate in the initial survey or sondeo." The sondeo thus gives the biological

scientists on the technical team a chance to see the area and traditional

technology, and talk to the farmers themselves before they plan the farm

trials. Further, it gives the social scientists on the sondeo an input into

the planning of the farm trials, via the written recommendations and via

the three-way direct communication between biological scientist, social

scientist, and farmer that is set up during the sondeo. The importance of

the three-way communication cannot be over-emphasized because, on the one

hand, the social scientist learns first-hand about the physical constraints

imposed on the farmer by the physical environment surrounding him or her.

On the other hand, the biological scientist hears the farmer talking about

socio-economic constraints, in response to the questions of the social

scientist. While the farmer is educating them both during the sondeo's

interviews, they are given the opportunity to educate each other about their

respective domains of expertise and their solutions to the farmer's problems.

As an anthropologist-agricultural economist, I participated in two

sondeos in the Altiplano during my time in Guatemala working with ICTA: a

sondeo initiating technology generation in the potato zone in Quezaltenango,







and another investigating conditions in a zone of fruit trees centered

around Chichicastenango (Hildebrandet al ., 1979; SER/ICTA y DIGESA, 1979).

Based on my previous development of the decision-tree model in section II,

which was built and tested in different subregions than those in the sondeos7

I was able to understand the cropping patterns observed after interviewing

ten to twenty farmers in each sondeo. I was also able to informally chec"

on the decision-tree model in the two new zones during the sondeo inter-

views.

Based on the decision-tree model and the sondeo interviews, the sondeo

report on the potato zone discusses the constraints and factors limiting

the production of potatoes in a peripheral subregion of the potato zone

(Varsovia, Monrovia, etc. (Hildebrand et al., 1979:p.14)).

It also compares the center of the potato zone, conception Chiquirichapa,

where 2 crops of potatoes are grown per year, to the periphery of the zone

(Varsavia, Monrovia), where climate and lack of seed limit the second crop of

potatoes, and farmers plant potatoes and wheat in rotation if they can. If

they can't squeeze in the rotation, they decide between wheat and potatoes

as their cash crop, planting consumption corn on land given first priority

(Ibid, pp.15-16). The report concludes that, although most of the farm

trials should be placed in the center of the potato zone, a few trials

should be planted in the peripheral area, where potato production has future

potenti al In addition, some experiments should rotatee a first planting of

potatoes followed by a precoz variety of wheat, in the peripheral area. Thus

through the sondeo mechanism in ICTA's diagnostic stage, the cropping

decision-tree model of section II provided policy recommendations for the

design of farm trials in the potato program.




-26-


The sondeo in Chichicastenango applied the ideas of the cropping

decision-tree model to explain why some farmers in the zone plant a stand of

monoctop of fruit trees, and other farmers plant fruit trees associated with

corn. The model also predicts which farmers will gradually plant more and

more rows of fruit trees in their corn fields, and thus slowly switch to a

monocrop of fruit trees (SER/ICTA y DIGESA, 1979:pp.8-10). Farmers will

slowly switch to a stand or orchard of fruit trees if:

a) they have more than enough land planted to consumption corn and thus

can safely switch some corn land to fruit trees, if they also pass knowledge

and capital (i.e., stage 1) constraints,

b) they don't have sufficient land to plant all the corn they need

plus an orchard of fruit trees, but they (rightly) perceive fruit trees to

be more than twice as profitable as corn. (These farmers "go down the left-

hand branch" in figure 4 to plant theversuprofitable crop first, before corn.)

Farmers will continue to plant fruit trees associated with corn, and

will not switch to a monocrop of fruit trees, if a) they don't have sufficient

land to plant all the corn they need plus an orchard of fruit trees and b)

they don't know that fruit trees are 5 to 8 times as profitable as corn, due

to a lack of technical assistance and knowledge. Actually, the majority of

farmers in Chichicastenango fall in this latter category. The sondeo

report therefore recommends that more technical assistance be given to farmers

growing fruit trees interplanted with maize, and not just to farmers with
orchards, or farmers with credit from DIGESA. With more technical assistance

and knowledge, farmers will naturally transform land with associations of

fruit trees and corn to orchards of fruit trees, given continuing good prices

for fruit. The decision-tree model of section II applied to fruit trees as




-Ll"


the cash crop thus predicts a slow diffusion pattern for fruit trees as a

monocrop in Chichicastenango, and produces policy recommendations, through

the sondeo, to increase production of fruit trees in the area.

Conclusion

In summary, a decision-tree model, built with the aid of ethnoscience

eliciting techniques, a uniquely-anthropological invention, can be used by

a farming systems program, such as ICTA, to provide policy recommendations

to a technical team designing farm trials. It can also be used to support

current institutional policies, such as ICTA's continued support of a maize

program in Altiplano. It could also be used to recommend changes in existing

policies: a previous study recommended that the Plan Puebla drop one of its
fertilizer recommendations (C. Gladwin, 1979a). Li kewi se de -ision-tree

model could ur e the creation of a new kind of farm trial: data' from

Guatemala suggest that ICTA should experiment with optimal combinations of

chemical and organic fertilizers on corn, potatoes, and vegetables (C.

Gladwin, forthcoming). By showing a farming systems team how farmers are

making their agricultural decisions, decision-tree models can recommend

effecti-ve program policies, so that the program -- in reality as well. as in

theory -;s centered on the farmer.







FOOTNOTES


"The field work for this paper was supported by a Rockefeller post-

"octoral ellowship and the International Fertilizer-Development Center

1173C) at M~uscle Shoals, Alabama. Invaluable help in the Altiplano was

5given by colleagues at the Institute of Agricultural Science and Technnology

(ICTA), and Carol and Ron Smith in Guatemala. Naturally, the views expressed

in this paper are mine and do not necessarily reflect those of IFDC, ICTA,

or the Rockefeller Foundation. I appreciate the advice and support given by

these institutions and colleagues, especially M. Angel Garcia, M. Chinchilla,

K. Byrnes, K. Davidson, E. Guerra, P. Hildebrand, H. Orozco, and R. Ortiz.

Special thanks are due to the farmers in these rural areas who patiently

and graciously taught me how they made their decisions.

To emphasize the fact that female farmers exist and contribute to

agricultural production in the Al~tiplano and all over the world, the present

author will use the feminine as well as masculine pronouns when referring to

farmers.


A system of crops is a set of crops that are interplanted or multi-

cropped (Hildebrand, 1976). In the Altiplano, an area of roughly 22,000 km2

or 8,400 square miles, and here considered to include the departments (or

states) of Chimaltenango, Solol%, Totonicapthn, Quezaltenango, San Marcos,

Huehuetenango and El Quichf, corn is intercropped with beans (frijo and haba),

so is written corn (+beans). For brevity, however, corn (+beans) shall be

hereafter referred to as corn. A system of crops is also defined here as

a set of crops that is harvested on the same field in one year ( e.g., a

first harvest of wheat and a second harvest of peas, or 2 harvest/year of

potatoes or 3 harvests/year of vegetables).




Page 2


Footnotes


In fact~:refer to the test as "putting the farmer down the tree."

The tree then processes the farmer's data and "gives him or her commands"

about which crops to plant at different logical points in the tree, or

eliminates certain crops from the farmer's predicted crop mix.

The validity of this conclusion depends on whether the results taken

from the subregions sampled in this study can be extended to the entire

region of the Altiplano. This in turn depends on the representativeness of

the sampled subregions.

In my judgement, if the sample of zones in this study is biased or

unrepresentative of most parts of the Altoplano, thabias works to make our

results seem weaker than they really are. Indeed, a truly unbiased sample

or census of the Altiplano might show that 70-80% of Altiplano farmers do

not have a very profitable cash crop; they therefore plant the family's

consumption requirement for corn first, and a cash crop second. 'Farmers'

dependence on corn might be under-estimated in this study because the sample

of subregions or zones was chosen to highlight diversity or variation

in cropping patterns. Zones with different crop mixes were selected, in

order to provide as wide a test of the decision model as possible. Therefore,

zones with a very-profitable cash crop might be over-represented, and zones

with not-so-profitable cash crops might be under-represented in this study.

Thus the importance of improving corn yields in the Altiplano, in order to

encourage diversification of farmers' cropping patterns, cannot be over-

emphasized or explained away on the grounds of sampling errors.

This doesn't always occur, e.g., the case of the Zacapa sondeo,

but the biological scientists and technicians who dlo participate in a sondeo

and later plant farm trials give the sondeo high ratings as an educational

instrument for them. However, the reader should not get the mistaken




footnotes rage a



impression that biological and social scientists in ICTA are in agreement

all of the time. Both during and after my stay in Guatemala, serious

disagreements arose regularly, and the status and role of the socio-economic

unit ("socio-eco~nomia rural") was always in question within the institute,

with the biological scientists questioning its role, influence, and budget

within +he institute.




REFERENCES


Anderson, Jack R.

1974 "Risk Efficiency in the Interpretation of Agricultural Production
of Research." Review of Marketing and Agricultural Economics
42(3): pp. 131-184.

Ashby, Jacqueline, and Geshand de Jong

1980 "Farmer Field Preparation and Tillage Practices: Implications
for Fertilizer .Technology Research." Cali, Colombia: International
Center for Tropical Agriculture (CIAT) and International Fertilizer
Development Center (IFDC).

Barlett, Peggy F.

1977 The Structure of Decision Making in Paso. American Ethnologist.
4(2): 285-307.

Benito, Carlos A.

1976 "Peasants' Riesponse to modernization projects in Minifundia
economies." American Journal of Agricultural Economics, Vol.
58, No. 2: 143-151.

Cancian, Frank

1972 Change and Uncertainty in a Peasant Economy: The Mlaya Corn
Farmers of Zinacantan. Stanford: Stanford University Press.

Cancian, Frank

1980 "Risk and Uncertainty in Agricultural Decision Making"
In Agricultural Decision Making, P. Barlett, Ed. New York:
Academic Press.

Chibnik, Michael

1980 Working Out or Working In: The Choice Between Wage Labor and
Cash Cropping in Rural Belize. American Ethnologist F(1):
pp. 86-105.

Chinchilla, Maria, y Peter lIildebrand

1979 "Evaluacion de la Aceptabilidad de la Tecnologia Generada
Para el Cultivo de Maiz en Quezaltenango, 1977-1978." Guatemala:
Informe del ICTA.

CIMMYT Economics Program

1980 "Planning Technologies Appropriate to Farmers; Concepts and
Procedures." El Batan: CIMPMYT.








De Walt, Billie

1979 "Drinking Behavior, Economic Status, and Adaptive Strategies
of Modernization in a Highland Mexican Community," American
Ethnologist 6(3); pp. 510-530.

Fumagalli, Astolfo, and Robert K. Waugh

1977 "Agricultural REsearch in Guatemala." Presentation at the Bellagio
Conference, Bellagio, Italy, Informe del Instituto de Ciencia y
Teenologia agricola, Guatemala, C.A.

Gilbert, Elon H., David W. Norman, and Fred E. Winch

1980 Farming Systems Research: A Critical Appraisal. Michigan State
University, Rural Development Paper No. 6. East Lansing:
Department of Agricultural Economics, Michigan State University.

Gladwin, Christina

1975 "A Model of the Supply of Smoked Fish from Cape Coast to Kumasi"
Itx Formal Methods in Economic Anthropology, S. Plattner,
Ed. Wash, DC A Special Publication of the American Anthropological

Gladwin, Christina

1976 "A View of the Plan Puebla: An Application of Hierarchical
Decision Models." American Journal of Agricultural Economics
58(5); pp 881-887.

Gladwin, Christina

1977 A Model of Farmers' Decisions to Adopt the Recommendations of
Plan Puebla. Ph.D. Thesis Stanford University.

Gladwin, Christina.

1979a "Cognitive Strategies and Adoption Decisions: A Case Study of
Nonadoption of an Agronomic Recommendation." Economic Development
and Cultural Change 28(1): pp. 155-173.

Gladwin, Christina

1979b "Production Functions and Decision Models: Complementary Models."
American Ethnologist 6(4): pp 653-674,

Gladwin, Christina

1979c The Future of Corn Production in the Altiplano of Western Guatemala:
How Do the Farmers Decide? Report to the International Fertilizer
Development Center (IFDC) and the Guatemalan Institute of
Agricultural Science and Technology (ICTA).








Gladwin, Chlristina?


151~ 3 1 T::,r of "sal--iife (3Chr c: Appl~~zicarca to Agricultural
--iI; oi-Ji3_ .main ;: Anthtropological
AYn:::buucive -2u~ral Development, P. Earlett, Ed. New York:


Gladwin, Hugh

1971 Decision :I z110~ :.. r: ape; Coast (Fante) Fishing and Fish
Marketing Systl -. 3. Thesis, Stanford University.

Gladwin, Hugh

1980 "Test of a Hierarchical Model of Auto Choice on Data From the
National Transportation Survey" Mimeo.

Gladwin, Hugh, and Michael M. H2urtaugh

1980. "The At tentive/Pre-attent ive .Dist inct ion in Agr icul tural
Decisions," In Agricultural Decision Making, P. Barlett Ed.
New York: Academic Press, pp. 115-136.

Hansen, Art, Griffith, David, Butler, ;John, Powers, Sandra, Gilbert, Elon,
Lauriault, Robin,` and Masuma Downie.

1981 Farming Systems of Alachua County, Florida: An Overview with
Special Attention to Low Resource Farmers, Gainesville, Florida:
Center for Community and Rural Development, University of Florida

H~ildebrand, Peter E.

1976 "Multiple Cropping Systems are Dollars and 'Sense' Agronomy"
In Multiple Croppings American Society of Agronomy, Inc. Eds:
pp 347-371.

Hildebrand, Peter E.

1977 "Generating Small Farm Technology an Integrated Multidisciplinary
System." Paper presented at the 12th West Indian Agricultural
Economics Conference, Caribbean Agro-Economic Society, Antiqua,
Guatemala, 24-30 April.

Hlildebrand, Peter E.

1978 'iMotivating Small Farmners to Accept Change." Paper presented at
the Conference on Integrated Crop and Animal Production Optimize
Resource Utilization on Small Farms in Developing Countires, The
Rockefeller Foundation Conference Center, Bellagio, Italy,
October 18-23.

Hildebrand, Peter

1979 "Summary of the Sondeo Methodology Used by ICTA". Guatemala:
Informe del Instituto de Cilencia y Tecnologia Agricolars.







Tversky, Amos

1972 "Elimination by Aspects: A Theory of Choice".
Psychological Review 79 (4): pp. 281-299.

Werner, Oswald, and G. Mark Schoepfle

1979 The Handbook of Ethnoscience: Ethnographies and Encyclopedias
Evanston, Illinois: Department of Anthropology, Northwestern
University.

Young, James C..

1980 "A Model of Illness Treatment Decisions in a Tarascan Town."
American Ethnologist 7 (1) : pp 106-131.