Title Page
 Table of Contents
 1. Introduction and list of...
 2. Technical prescreening
 3. The classification of on farm...
 4. Issues in the design and management...
 5. OFE with animals
 6. The analysis, interpretation...
 7. OFR: Linkages with component...
 8. Other issues
 List of participants
 Opening speech

Group Title: Networking workshops report - International Maize and Wheat Improvement Center (CIMMYT) ; no. 5
Title: Report on a networkshop on issues in on farm experimentation
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00080069/00001
 Material Information
Title: Report on a networkshop on issues in on farm experimentation Lilongwe, Malawi, May 24-28, 1984
Series Title: Networking workshops report
Physical Description: iii, 53 p. : ill. ; 30 cm.
Language: English
Creator: CIMMYT Eastern and Southern African Economics Programme
International Maize and Wheat Improvement Center
Conference: Networkshop on Issues in on Farm Experimentation, (1984
Publisher: CIMMYT Eastern and Southern Africa Economics Programme
Place of Publication: Nairobi Kenya
Publication Date: 1985
Subject: Agriculture -- Research -- On-farm -- Congresses   ( lcsh )
Agriculture -- Research -- Congresses -- Africa, Eastern   ( lcsh )
Agriculture -- Research -- Congresses -- Africa, Southern   ( lcsh )
Genre: conference publication   ( marcgt )
non-fiction   ( marcgt )
Spatial Coverage: Kenya
General Note: "This report presents a summary of the group and plenary discussions at the CIMMYT networkshop on Issues in On Farm Experimentation in May 1984"--P. 1.
General Note: "August, 1985."
 Record Information
Bibliographic ID: UF00080069
Volume ID: VID00001
Source Institution: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 55958822

Table of Contents
    Title Page
        Title Page
    Table of Contents
        Table of Contents
    1. Introduction and list of issues
        Page 1
        Page 2
        Page 3
        Page 4
    2. Technical prescreening
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
    3. The classification of on farm experiments (OFE)
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
    4. Issues in the design and management of OFE
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
    5. OFE with animals
        Page 26
        Page 27
        Page 28
    6. The analysis, interpretation and use of OFE results
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
    7. OFR: Linkages with component research and with the extension services
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
    8. Other issues
        Page 41
        Page 42
        Page 43
        Page 44
    List of participants
        Page 45
        Page 46
        Page 47
    Opening speech
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
Full Text

CIMMYT 7,-24




Networking Workshops: Report No. 5

CIMMYT Eastern and Southern Africa Economics Programme,
P.O. Box 25171,

August, 1.985



1. Introduction and List of Issues : : : : : : : : : : : : : : : : 1

2. Technical Prescreening : : : : : : : : : : : : : : : : : : : : : 5

3. The Classification of On Farm Experiments (OFE) : : : : : : : 11

4. Issues in the Design and Management of OFE : : : : : : : : : : : 18

4.1 The issues and their importance : : : : : : : : : : : : : : 18

4.2 Farmers as hosts for OFE : : : : : : : : : : : : : : : : : 18

4.3 Sites, replications and plot size by type of OFE : : : : :20

4.4 Levels of non-treatment variables in OFE : : : : : : : : 21

4.5 Researcher and Farmer Participation : : : : : : : : : 23

5. OFE with animals : : : : : : : : : : : : : : : : : : : : : : : : 26

6. The Analysis, interpretation and use of OFE results : : : : : : 29

6.1 The issue and its importance : : : : : : : : : : : : : : : 29

6.2 What criteria are used in the evaluation of OFE : : : : : : 29

6.3 Analytical requirements for different types of OFE : : : : 30

6.4 The use of OFE results : : : : : : : : : : : : : : : : : : 34

7. OFR: Linkages with Component Research and with

the Extension Services : : : : : : : : : : : : : : : : : : : : : 36

8. Other Issues : : : : : : : : : : : : : : : i : : : : : : : : : 41


1 List of participants : : : : : : : : : : : : : : : : : : : : 45

2 Ministers Opening speech : : : : : : : : : : : : : : : : : : 48

3 Programme : : : : : : : : : : : : : : : : : : : : : : : : : : 52

The smallholder agricultural sector holds the key to economic development and should
be afforded the latest knowhow in order to boost its contribution to the economy,
agricultural researchers were told yesterday. The Minister of Local Government
Mr. Kapichila Banda, made the remarks when he opened a four-day workshop of
researchers on behalf of His Excellency the Life President, who is the Minister of
Agriculture. Mr Kapichila Banda called for a greater efforts to apply new ideas
directly on smallholder farms to see how they worked.
"We are convinced that this approach will yield good dividends, not only for the
nations represented here, but also for their respective smallhoder farmers", said
the Minister.

Agricultural research scientists attending a four-day workshop at the Lilongwe Hotel,
on Tuesday visited maize and groundnut fields at Thiwi/Lifidzi Rural Development
Project in Dedza, which is under the Lilongwe Agricultural Development Division. The
visit was to acquaint the scientists with the on-farm adaptive research trials in the
project. Above, the scientists in the maize fields.


The report presents a summary of the group and plenary discussions at
the CIMMYT networkshop on Issues in On Farm Experimentation in May 1984.
The workshop, held at the Lilongwe Hotel, Lilongwe, Malawi from the
21-24 May, was attended by 30 agronomists from 10 countries of the
Eastern and Southern African Regions.

1.1 Report layout

The Farming Systems Newsletter circulated quarterly by CIMMYT Regional
Office in Nairobi requested issues in experimentation from field agronomists
using a systems oriented On Farm Research approach in the region. These
issues were consolidated to form an agenda for the workshop. The list of
the issues is presented at the end of this introduction. Each issue is
treated in a section of the report. Some sections cover individual issues,
other a group of issues around a common theme. Section 8 covers issues
brought up by participants during the workshop. Finally, there are three
Annexes giving the list of participants, the Ministers opening speech,
and the Programme for the workshop. A draft report was circulated to
participants in May 1985 and this final report incorporates comments
received by August 1st.

1.2 Networking in On Farm Research with a Farming Systems Perspective

Networking is seen as valuable, especially in a new field of activity
such as OFRIFSP for three reasons:

(1) Practitioners are few and scattered.

(2) Formal linkages, for example through specialised journals, do not

(3) The development of sound methods will be speeded by cross-
fertilisation of ideas and experiences.

CIMMYT Economics helps organise and fund two technical networkshops each
year in addition to one networkshop for research administrators. These
technical networkshops follow three thrusts:

(1) Networkshops organised around a research problem identified as common
to several countries across the region.

(2) Networkshops which review an On Farm Research Programme in the field.

(3) Networkshops which discuss issues in the methodology of OFR/FSP.

The networkshop reported here falls into this final category of issues in
methodology, in this case in On Farm Experimentation.

1.3 An outline of the issues for discussion

(1) Technical Pre-screening, How far can we go?

Agronomists working in OFR must make judgements on which technical relation-
ships are relevant to local farm improvement and whether technical relation-
ships, established elsewhere, often under experimental conditions on research
stations, will hold locally in farmers fields under farmers management
conditions. Which types of classical experimental results can readily be
transferred to local situations with expectations of similar performance,
which types almost certainly need a new set of experiments, and what
conditions should such new experiments be conducted under?

(2) What is the best basis for the classification of On Farm Experiments?

The type of experiment required will, to large extent, be dictated by the
agronomists judgement discussed in issue one above. Can this be a basis
for the classification of OFE or are there better bases?

(3) Issues in the design and management of On Farm Experiments

a) How do we select farmers as hosts for On Farm Experiments?
What criteria should be used to select farmers as hosts for OFE?
Should these farmers remain hosts, or how often should we change?
Can we maintain interest on the part of farmers over a period of time?

b) How is the number of cooperators or sites, replicates at sites, and
plot size dictated by the type of experiment?

c) How does the type of experiment dictate the balance between researcher
and farmer involvement? What are the advantages and limitations of
researcher and of farmer management in OFE? How should risks of loss
be split in each type of experiment?

d) Is it fair to say that non-experimental variables should always be
at farmer level in OFE within a system oriented OFR. Are there
exceptions to this; are there certain types of experimental content
which require control over non-experimental variables.

e) Farmers and agronomists management criteria often differ. Is it
possible for agronomists to absorb farmer criteria and management
strategies and use these more frequently in OFE work?

f) Are there ways of designing and managing OFE to obtain and facilitate
more and better farmer participation in their management and evaluation?

g) How can the timing of operations and the synchronisation of management
best be organised in OFE?

(4) On Farm Experiments with Animals

Which research areas with animals can, and which cannot, be handled readily
on small farms. What are the design, management, and analytical problems
in each of these difficult areas? How far have methods been developed and
used to handle these?

(5) The Analysis, Interpretation and use of results on On Farm Experiments

a) Statistics, economics and farmers' assessment have their places in the
evaluation and interpretation of OFE. What is their relative importance
for different types of trial? How should results from each method of
evaluation used be weighed for different types of trial?

b) The most obvious use of results is for planning next year's work and
making recommendations; what should prompt the OFR team to make a
recommendation? How can OFE results be used to re-define Recommendation
Domains or better understand the Farming System being researched?

c) What are the appropriate statistical tools and standards for the analysis
of OFE in which control over non-experimental variables is low?

(6) Operational linkages

a) Links with 'Traditional' Agricultural Researchers.
What scares 'traditional' researchers about OFR? How can traditional
researchers be drawn into the OFR process and persuaded of the

opportunities and benefits it can bring to their research? How can
On Farm Research gain experience and confidence without alienating
the traditional research establishment?

b) Links with extension:
What benefits can OFR bring to extension? What roles can extension
staff play in diagnosis, experimental planning, farmer selection, trial
management and the evaluation of experiments? Can extension do the
job instead of research.

The report follows this ordering of issues moving through the OFR
sequence of planning, managing and evaluating experiments. Final chapters
cover issues in linkage, institutional and questions raised during the
meeting by participants.


2.1. The issue and its importance

Agronomists working in On Farm Research (OFR) must make two particularly
important decisions in identifying an experimental programme appropriate
to a local farming situation: First: What is the range of technical
interventions relevant to solving local farmers priority problems?
Second: Will the technical relationships inherent in each of these inter-
ventions, often established under different circumstances elsewhere, hold
in local farmers fields and under their management conditions? We learn
from the Farming Systems Perspective (FSP) that a significant economics
input, markets and costs being two considerations, is needed to identify
appropriate technical interventions. While economics does often dictate
desirable characteristics of technical interventions, only technical
expertise can make the precise technical specification; set out the
husbandry requirements and the expected technical relationships between
physical inputs and outputs, and judge their replicability under local

The two decisions are particularly crucial because they create a focus for
the start of local development activities which will engage many scarce
professional resources over a period of several years. The OFR team,
typically of three professionals with support staff, will work around this
focus for two or three years. Beyond this the local extension services
and local branches of other service institutions for credit, marketing
and processing must gear their local infrastructure and services
development to the needs of introduced technologies over the medium term.
Thus these are two vital decisions to be taken by the OFR technical
scientists. They deserve attention in the interests of the agricultural
development process.

2.2 Discussion at the Workshop

Participants concentrated their discussions on two themes. The first was
how to widen the range of technical possibilities brought to bear on the
problems and opportunities highlighted by the Farm System diagnosis. The
second was on the sources of uncertainty in evaluating the transferability
of technical experience. Finally participants put forward means of

improving the ability of OFR technical scientists to make these two
crucial decisions; what range of relevant technical interventions are
available and will observed technical relationships hold locally?

2.2.1 Widening the range of technical interventions for solving farmer
problems or developing their opportunities.

Discussions on widening the range of technical interventions centred
around two ideas. First, that wider approaches to problem solving are
needed, second that wider sources can be tapped in a search for solutions.

(1) Widening approaches to problems solving

Participants agreed that traditional research findings had been blindly
prescriptive rather than problem solving. The example of maize weeding
was used in the discussions. Traditional research would, as a result
of experimentation, prescribe an optimal number of weedings and best
time for each in order to maximise maize yields. As a direct,
generalised prescription this ignores factors causing poor weeding, which
may differ in each local situation. Such blind prescriptive solutions
result in exhortation by the extension services with no reference, nor
often indeed relevance, to the cause of the problem. Prescription without
diagnosis jeopardises the relationships between extension staff and the
local community. Workshop discussions centred around two diagnostic
approaches to the identification of relevant technical interventions:
the'indirect approach' and 'the systems' approach, both requiring
knowledge of the causes of technically inferior husbandry in local
specific situations and therefore only possible in an OFR context.
Returning to the maize weeding case, the indirect approach searches
for complementary management practices in the maize enterprise where
inferior husbandry is contributing to poor weed control and improved
husbandry would enhance weed control; better seed bed preparation,
higher stand density and changed rotation are some examples. Identified
husbandry improvements would then be evaluated for system compatability
on technical grounds disease control for example and on economic
grounds, in particular the resource allocations implied by the improved
practices. The systems approach uses an understanding of the inter-
actions between various activities of the farming system. Returning
again to the maize weeding case in a scarce seasonal labour situation;
it is understood that the farmer allocates scarce labour over the maize

weeding period to other crops, perhaps for example planting groundnuts
or weeding cotton. Interventions which change the timing or intensity
of labour use on these other crops, perhaps the use of berbicides on
cotton, or the later planting of a shorter season groundnut, will free
up labour in the system over the critical maize weeding period. Once
labour is freed up the system the direct prescription from a conventional
research approach provides useful guidance on the application of that
labour to maize weeding. On the other hand, participants pointed out
that it may be more profitable for farmers to use the released labour
in other ways. Economic pre-screening should help to identify these.

(2) Wider sources of solutions

Participants agreed that research stations offered a major source of
interventions for OFR. However, it was felt to be important that
technical scientists moving into OFR try to lose their conventional
straight jackets and not be overly dependent on local research station
results for progress in OFR. Relevant results from regional and
international centres and from research stations in other countries
should be tapped in order to solve local problems. The value of
Indigenous Technical Knowledge (ITK) was highlighted. Even within most
local communities some farmers had themselves usually devised solutions
to their most pressing farming problems. Other communities, in country
or elsewhere, have often been through a similar evolutionary process.
Farmers reactions and solutions, locally and elsewhere are a valuable
source of interventions. The example of the sophisticated draft animal
management methods in India and Egypt were advanced as a source of guidance
on interventions in draft animal feeding and training for countries in
the region beginning to experience a build up of population pressure on
grazing land. Finally it was pointed out that the use of their skills
by the agronomist and the animal production scientists, bringing training
and experience to bear in identifying and filling gaps in the exploitation
of biological potential, was a further important source of intervention.
Adding an intercrop or relay crop without penalising the yields of the
main crop was one example. To sum up four sources of interventions were
identified by the workshop.

a) Local research stations.
b) Research stations elsewhere, often in other countries, located in
a similar agro-ecological environment.

c) ITK, both in the local community under study, and in communities
with similar farming systems which have preceded the study community
down a similar evolutionary path.
d) Biological logic; the technical scientists exercising their skills
where they see exploitable gaps in current practice.

2.2.2 The Transfer of Technical Experiences: Sources of Uncertainty

Several of the workshop discussion groups dwelt on the question of how the
agronomist and animal production scientist judge whether a set of technical
relationships, established by research on stations or experienced by farmers
elsewhere, will be replicable in the farming situation under study. It was
emphasised that technical scientists will often be uncertain about the
transferability of technical relationships, and the level of uncertainty
will decide the type of experiment which is appropriate. Confidence in
transfer allows immediate comparison with current farmer practice, strong
uncertainty demands researchers managed trials to establish the technical
relationships under local conditions.

It was suggested that certainty of transfer is strongest with genetic
characteristics least vulnerable to genetic/environment interactions. The
maturity period of a crop variety was given as an example, but was disputed
as being vulnerable to temperature or radiation changes. It was agreed
that these effects are proportional across materials and often relatively
small. Also that they are fairly readily anticipated by 'biological logic'.
The discussion highlighted three points:

(1) Vulnerability to environmental interactions is a crucial source
of uncertainty on the transferability of technical relationships.

(2) Climate and soil are major environmental modifiers, supplemented
by pest and disease pressure where plant material and certain
agronomic practices are concerned.

(3) The 'management background' against which technical relationships
are measured, non-treatment management in the case of formal
experiments, is the other major source of uncertainty.

This third point received some elaboration. In reductionist research the
non-treatment management variables form a context within which technical
relationships are observed and measured. In such classical experimentation

non-treatment variables are usually held at none limiting levels to avoid
inhibiting the expression of responses from the treatments, in
acknowledgement that interactions between treatment and non-treatment
variables can be strong. Technical relationships which have been measured
within a context of optimised, non-inhibiting non-treatment variables
may be radically altered when interpolated into the very different manage-
ment background of many farming systems.

For effective evaluation of the transferability of technical relationships
OFR technical scientists need information on the environmental and
management circumstances under which the relationships were established,
as well as on the local circumstances into which they will be introduced
as interventions. Once apparently appropriate interventions have been
identified in the course of an Informal Diagnostic Survey relevant local
environmental and management circumstances will be defined and will be
the focus for further Informal Survey and perhaps content for the Formal
Survey Questionnaire.

2.2.3 Means to wider approaches to problem solving and to improving
judgement on transferability

All the means brought forward at the workshop, to widen approaches to
problem solving and to improve judgement on the transferability of technical
relationships, can be summarised as more frequent professional interaction
and organised information exchange. Workshop groups expressed the view
that close linkages with their own country component researchers provided
both ideas for intervention and guidance on judgement as to transferability.
Some felt it was very difficult for limited numbers of component researchers
to service the needs of several OFR/FSP teams across the country.
Others felt that part of the component/OFR interaction can be built into
annual research planning meetings as a routine when the OFR teams and
component researchers meet centrally. If each component researcher is
requested for interaction out in the field by say three OFR teams it is
unlikely to take more than two working weeks of his time to meet their
needs. They emphasised the importance of the decisions to be made,
creating a focus of local activity for years to come for significant
numbers of scarce professionals in the rural areas. Participants felt
that a mass of relevant research results were available but not readily
accessible, at country, regional and international levels. They saw two needs:

(1) A common information framework
An initial requirement was a common information framework with
consistent terminology and a keywords system allowing identification
of at least four aspects of each piece of research:

(a) A listing of farming system problems to which the research is
(b) Specification of the technical relationships established by
the research.
(c) The environmental circumstances within which the relationships
were established.
(d) The management background against which the relationships were

It was felt important that such an information framework is, as far as
possible, common to both On Farm and to component researchers. On Farm
Researchers should incorporate findings on problem related ITK from their
diagnostic work in local situations into their reports, using the same

The framework would be aimed towards widening the range of potentially
transferable solutions available to OFR teams working locally, by
establishing sets of analagous circumstances across the region. Sets would
be analagous with respect to environmental circumstances as a base level
and key farming system descriptors (population pressure and power source
would be two) at a secondary level.

(2) An information network hierachy

Participants felt that the information framework should be the heart
of an information network hierachy, at least region wide. The network
might operate at local, country and regional levels with field researchers,
both commodity and systems oriented, being the source and target of
information exchange. Commodity or disciplinary related national coordina-
tors are a traditional feature of research establishments in the region,
national coordinators for OFR/FSP are increasing. These professionals,
preferably tied to some sort of documentation facility, could be the link
at the national level. At the regional level the Southern African Centre
for coordination in Agriculture Research, (SACCAR) within the SADCC and
based in Botswana, is a clear candidate for regional coordination across the
SADCC area. Eastern Africa is not so obvious.

In addition to taking the initiative in developing the mechanics of
information exchange, the regional coordination bodies would play three
other vital roles:

(a) In sensitising the national coordinators to analagous situations
across the region.
(b) In promoting cross country visits by field scientists.
(c) In organising regional meetings of researchers working in analagous


3.1 The Issue and its Importance

There are two benefits from an agreed classification of On Farm Experiments
(OFE). First, it serves as a reference point for detailed discussion,
minimising terminological confusion. Second, it is a valuable training aid
allowing newcomers a systematic foundation for understanding on farm
experimental activities in OFR/FSP.

3.2 Discussion at the Workshop
Participants agreed that different objectives in On Farm Experimentation are
the heart of any typology but the implication of the discussion was that,
because one type of experiment may meet more than one objective, they are
not the best basis for classification.

3.2.1 An inventory of objectives in OFE

In Section 2 above the varying objectives of On Farm Experiments were seen
to spring from the level of confidence of the OFR technical scientists that
identified solutions to diagnosed problems will work under local farming
conditions. Objectives for On Farm Experiments can be summarised as:

(1) To establish the cause of a technical problem.
(2) To establish the technical relationships between treatment inputs and
(3) To verify that technical relationships, established elsewhere, are
replicable in local farming conditions.
(4) To measure performance compared to current farmer practice and record
farmer assessment of treatments.
(5) To identify recommendations for dissemination among the local farming
Objectives range from resolving complete technical uncertainty to final confidence
that the new treatments are attractive and acceptable to local farmers.

3.2.2 Classifications of On Farm Experiments

Some criticism was made of the CIMMYT classification of OFE which is set
out by Palmer, Violic and Kocher, 1982:

(1) Exploratory Research to identify components likely to contribute
the most significant increases in yield.

(2) Levels trials which seek the optimum levels of those identified
components and the extent of the interactions between them.

(3) Verification experiments which package two or three key components,
at their best levels and compare with farmers' practice.

(4) Experimental production plots and demonstrations.

Participants felt that this sequence was not oriented to problem solving. It
was solely yield oriented in the conventional research tradition and perscriptive
in nature. They felt that there was only late involvement of the other local
actors in the technology development process; the farmers and extension staff.
Palmer, Violic and Kocher (1982) do emphasise however that "it is not a strategy
of transferring findings from the experiment station to farmers' land but rather
a program of research actually conducted under farm conditions by research
personnel in conjuction with extension personnel and farmers". They do see the
involvement of all the actors but because of a somewhat different orientation
on the OFR effort with it seen as a complete research program to design new
technology extension and farmers are only involved in the later stages. The
emphasis in the workshop discussion was in identifying relevant but available
technology, drawing it into, and where necessary adapting it to, the local
situation. Clearly, however an alternative way of organising is to have the
same scientists doing both technical and On Farm Research.
Participants supported the idea that, to reflect the unique orientation
of OFR/FSP, classification should be based on the role of the principle actors
in the experiments. The classification table drawn up by Mr. E. Shumba from
Zimbabwe, was largely accepted by the workshop. It was set out (p.13) with a
definition and some explanation of the three categories.

Ref: Palmer AFE, Violic AD and Kocher F. "Relationship between research and
extension services and the mutuality of their interests in Agricultural
Development'. Paper prepared for the 3rd FAO/SIPA Seminar on Field Food Crops
in Africa and Near East, Nairobi, Kenya, June 6-24, 1982.

Table 1. Classification of O.F.Es based on management considerations

Involvement in implementation


Extension workers
Random variables

Quality of host farmer

Who selects host farmer

Design considerations

Number of treatments

Number of treatment levels

Number of replications

Plots size (m2)

Number of sites

Performance trend expectations


Researcher Farmer
managed managed managed















R, E & F















j _________________ J I

(1) Research Managed (RM) trials

These are aimed at generating information on how a technology performs

under farmers' conditions and not on how farmers respond to it. To
obtain reliable data on the interactions, farmers with representative

physical and ecological conditions must be selected, mainly for research

purposes, and research staff should provide the management.


Such trials are similar to multi-locational trials conducted by component
researchers at the national level. Results can be very useful to the
OFR team during the design phase. However, researcher managed trials
can also be handled by OFR teams in situations where there are no
results from multi-locational trials by component research teams.

(2) Researcher Farmer Managed (R-FM) (Hybrid) Trial

Conducted when the researcher is not very sure of the performance of
the technology under local farmers management even though information
may exist from past research work. The objective is to screen a number
of technical innovations addressed to an identified problem and to
evaluate them within the management regime of target group farmers.
The degree of involvement of the three actors is similar so that each
uses his own criteria to evaluate the technology. Both farm and farmer
characteristics must be considered in the selection of hosts so that
sensitivity of technological options to these particular characteristics
can be evaluated. To facilitate extension worker and farmer participation,
the designs are simpler than for Researchers Managed trials.

(3) Farmer Managed (FM) trials

Designed where the technical relationships of the proposed intervention
are well known from past research work and there is confidence these
will hold locally. The proposed intervention is directly compared
with farmer practice. The objective is to allow the farmer to test,
compare and evaluate the proposed technology against his current
practice. To get a fair evaluation of how the technology fits farmer
circumstances researchers involvement is minimal and farmers
representative of important strata for extension purposes, are involved
in the testing. The extension worker is more involved in the
implementation and in the explanation of the technology to farmers.

The degree of control of random variables decreases from RM to FM trials
while the plot size and number of sites increases. The type of OFE
conducted in any locality depends on the amount of information available
to the researcher.

The idea is to conduct these types of trials sequentially for a set of
technical relationships identified as a possible solution to a farmer
problem. The researchers doubts on the transferability of the set of
relationships will determine the entry point into the final sequence.

Several points were raised on the classification though none challenged
its basis. Perhaps the most important point raised was that the objectives
covered in the description of each type of trial did not include the
investigation of the causes of technical problems. For example, where poor
crop stands are evident and no clear indication can be obtained either by
farmer discussion or by observation of the cause, an exploratory experiment
might be used to identify causes and rank their impact on crop stand.
Possibilities may include, poor seed bed preparation, poor seed, poor planting
method, insects, birds or rodents pre-or post germination. The discussion of
OFE classification was used to develop a diagram of the OFR technical
scientists decision sequence. It is set out below as Diagram 1. The diagram
adds one dimension to the workshop discussion, that of research methods other
than exploratory trials being used to investigate problem causes. Technical
causes can be investigated by detailed surveys; examples are soil deficiencies,
insect, disease or weed pests etc. Resource constraints as the other main
source of researchable problems can be investigated and elaborated either by
detailed survey or case-study techniques.

The original draft diagram was discussed in Farming System Newsletter No. 18.
This version has benefited significantly from comments on that draft. The
choice among possible solutions has been clarified, through ex ante pre-screening
both technical and economic, to arrive at the type of experiments, and in rare
cases, directly to a recommendation.



Diagram 1


SUnknown Cause
SPECIFY CAUSE Technical Res..ourc
(-- ----- -L -- -
Invest iative Detailed Cast
Experiment Survey Study


2. IDETIFY AND System Directly NONE A
SPECIFY SOLUTIONS Interactive Prescriptive AGENDA




Technically transferable?













Economically Viablet

Input Servicing
Avilabil to Targt Group

Resource levels implied
available to Target Group

Syste Compatible
(Complementarity in
resource requirements)

Productivity enhancing
risk reducing



Comments made on the classification in table 1 at the workshop are listed below:

(1) In presenting the classification Mr. Shumba made the point that
Researcher Managed trials, specified with controlled random variables,
can be considered the province of component researchers as they seek
new technical relationships. Zimbabwe has Communal Area Research (CART)
whose job is to adapt technology to local climate and soil. Perhaps
there can be one of three alternatives:

(a) special off station teams of component researchers;

(b) to pass such work back to station component researchers;

(c) the OFR teams do the work themselves (if they believe the potential
is very significant and component researchers are either not available
or cannot give the work priority).

(2) The question was raised whether simulated farmer practice is considered
the same as control of random variables. This would put such RM
experiments more clearly in the On Farm Researchers portfolio.

(3) Participants felt that single replications per site were often adequate,
as long as sites were enough and could serve as replications, particularly
in the case of FM experiments.

The point was re-iterated that this is not a necessary sequence. The type
of trial will depend on the level of confidence of the OFR technical scientist
that he knows the cause of the farmer problem and that the technical relation-
ships of the identified solution are replicable and are appropriate under local

farming conditions. Comments were received on the draft report to the effect
that who does the trial is not the determining factor in final design and
management and not the most appropriate basis for classification. Indeed,
like other aspects of design and management, who does the trial is determined
by its objective. Trial objectives are fundamental to an acceptable


4.1 The issues and their importance

The seven agenda items listed under this general heading were allocated
to Workshop groups for discussion. The items were seen to be of varying
importance, not all were fully covered and there was a good deal of
overlap in comments on different items. In particular the classification
of On Farm Experiments stimulated discussion on replication, plot size
and number of sites which pre-empted a number of the issues listed under
this heading.

4.2 Farmers as hosts for On Farm Experiments

4.2.1 Methods of selection

Participants accepted that different methods of selection were appropriate
for different trial objectives. For researcher managed trials sites
require careful specification and the selection of cooperating farmers
should be done by the researchers, normally in cooperation with extension
staff. For farmer managed trials, normally thought of as the heart of an
OFE programme, community involvement in farmer selection is desirable.
It needs care to avoid selection dominated by community leaders. One way
to reduce this danger is to be very clear on selection criteria, and gain
initial acceptance by leaders that farmers selected should meet these criteria.

4.2.1 Selection Criteria

The key to getting good hosts for OFE is strong community and farmer
participation in the programme. If the diagnosis of the farming system
has been effective, and if the community has participated properly, the
majority of farmers in the Target Group will identify with the importance
of problem being addressed by the experiments, and will see the possible
benefits of the research thrusts underlying the experimental programme.
Apathy or even antagonism from the community almost inevitably means bad
diagnosis or poor participation.

Farmers hosting experiments should generally be from the Target Group
being researched. They should be representative in terms of the resources
at their disposal, the farming system being operated, and the methods
used. The formal survey can include questions which allow an assessment
of farmer representativeness, which test whether farmers identify with

the problems highlighted in the diagnosis, and which test the logic behind
the research thrusts proposed. At the same time a formal survey question
can establish their willingness to host experiments, willing farmers can
then be reviewed for representativeness as potential hosts. Thus the
knowledge of the Target Group provides some criteria for selection, the
type of experiment and the content of the experiment provide others.

It was generally agreed that both types of researcher managed experiments
benefitted from having better farmers as hosts. The precision needed is
easier to obtain from target group farmers who are themselves better managers.
Again survey information can help identify these farmers. For the farmer
managed trials a major aim is to expose the technology to the range of
management abilities and site situations found across the target group and
a representative cross section of the farmers is needed.

Two points were made qualifying these general principles:

(1) Logistics: To help with cost effective implementation of the
trials it may be useful to cluster hosts to allow several sites
to be supervised by a single staff member, and to be visited in
a day by the OFR team. It is always desirable to avoid sites which
are extremes in logistical terms.

(2) It is valuable to know whether other programs are or have been
operating in the area, and under what terms and conditions farmers
participated in these. If rewards are given by other programmes
they may be necessary to gain cooperation in the siting of experiments.

4.2.3 Continuing or changing hosts

Participants were clear that a degree of pragmatism was desirable on the
issue of whether to continue with farmers as hosts, or whether to change
hosts, moving experiments around the community. There are two clear needs
for retaining farmers hosts:

(1) Longer term experiments on soil fertility maintenance, residual
effects, experiments with cattle breeding and trees, all these
clearly need a long term commitment from trial hosts.

(2) Where experiments start as RM/RI or RM/FM there will be a minimum
of two or three seasons work. Similarly, under variable climatic

situations two or three seasons will be needed, even where experiment
start at the FM/F1 level. One value in carrying the same host through
these sequences is for his progressive assessment of the outcomes.

The argument against continuity was the danger in prolonged contact with a
special research programme distorting the representativeness of the host
farmer. Maintaining farmers' interest over a prolonged period was reported
as a difficulty particularly where results were lagged. Several participants
felt if the problems being addressed were farmers' priorities and both
technical and economic pre-screening were effective, farmers would be naturally
interested, and, if properly drawn into the programme, will understand the
time lags to results. There was perhaps a consensus that effective farmer
and community participation is the key to maintaining interest.

Two things were felt to be important:

(1) That the community, not just the hosts, be kept fully-in the picture,
including visits to sites and group meetings, especially at the FM/FI

(2) That hosts be recruited in the full knowledge of what was expected
of them, included the period their cooperation is sought for.

4.3 Sites, replications and plot size by type of experiment

Attempts to formalise the calculation of numbers of replicates for On Farm
Experiments were felt to be too academic. It was pointed out that the

S(cv )2 (tl+ t2)

even when used for experimental design under station conditions gave
impractical numbers of replications. Participants thought that the rules
of thumb given in the classification of experiments was adequate (Table 1).
As noted in section 2 above the only question raised on these concerned
the number of replications per site for FM/FI experiments. It was felt
that single replications at each site using sites for replications was adequate.

Participants felt that conventional designs and conventional statistical
standards were relevant to RM. types, in which new technical relationships
were being sought. Statistical comparison remains important at the RM/FM
stage in verifying that technical relationships, measured elsewhere, hold
locally. However relaxed statistical standards were acceptable for RM/FM
types facing increased sources of variation from farmer determined non-
treatment management.

It was agreed that the major differences in design are in the Farmer
Managed (FM) experiments, usually a direct comparison of improved
management with farmers current management. At this point farmer assessment
supplemented by economic analysis on farmers criteria are the primary
means of evaluation. The sample of sites needs to be large enough to allow
for losses and still catch the effects of different locations and
different non-treatment management on outcomes, 10 20 sites were judged
adequate. Plot sizes need to be large enough to allow the measurement
of labour and machinery inputs where required, and to permit the farmers
management methods to be readily implemented, over 50 square metres was
the guideline. Most participants agreed that as farmers'involvement
increased,the complexity of the experiment, especially the number of
treatments, and the levels of each,must reduce. This is reflected in
Table 1. which summarises sites, replicates and plot size related to type
of experiment and largely accepted by the meeting. Some expressed the
belief that this was not necessary if there is strong participation by
farmers in the specification of treatments. This minority felt it was
important to place a range of options before farmers, often options designed
to meet different known local circumstances such as seasonal variations in
weather, with farmers fully involved in considering the importance of
these variations and the type of options which might manage them effectively.

4.4. The level of non-treatment variables in On Farm Experiments

Convention, in so far as it is established in OFR, has said that non-treatment
variables should be managed at farmer levels. Discussion at the workshop
noted two approaches to this; first that representative farmer practices
are used as non-treatment variables in experiments a simulation of what
target group farmers do, but imposed by researchers on each host farmer.
Second, that each host farmer be free to manage non-treatment variables as

he sees fit. The two have different implications; the first allows
control of non-treatment variables, the second introduces farmer management
as a source of variation across sites with no control and hence reduced
precision. Participants felt it was clear that RM/RI trials should be
conducted against a standardised background management to keep precision
high and replicate numbers realistic. However at some point, treatments
being developed towards an intervention, need exposing to the variability
in farmer management to test their robustness under conditions in which
the intervention will be evaluated by farmers. It is clearly better that
this be done as part of the technology development process rather than
after recommendation.

Much can be done during pre-screening where knowledge of current farmer
management strategies, including variations in practices and the reasons
for such variations, are brought to bear with a view to identifying potential
interactions affecting expected benefits from proposed treatments. The
pre-screening process should identify the potential difficulties with
farmer management as a context for treatments. Some participants felt that
the RM/FM stage of OFE was a turning point where farmers hosting experiments
provide their own background management. Others felt that this loosening
should not occur until the FM/FI stage, with a proposed intervention being
tested against current farmer practice at a large number of sites. Perhaps
a compromise would be to say the earlier the better, to save research effort
on interventions which are particularly vulnerable to such interactions,
with emphasis on careful pre-screening of possible interactions.

Oneparticipant (T. Maynard, ARPT Zambia) raised the issue that the nature
of the experimental treatments sometimes dictate non-experimental treatment.
He gave examples of the comparison of row and broadcast planting when assess-
ment requires consideration of the weeding implications of the two treatments.
Trial management needs differences in weed control to be expressed
and measured, either by measuring yield differences which does not allow
attribution of yield differences to weed differences, or by increasing
weeding input to treatments requiring it and counting the extra as a cost.
This is an unresolved complication to the general agreement on the issue
at the meeting.

4.5 Researcher and Farmer Participation

The remaining issues under the design and management of on farm experiments
centred on the balance between researcher and farmer participation, especially
in management. These followed fairly easily from the discussion of researcher
and farmer management of the non-treatment variables. Several issues were
brought up in presentations, these and others raised in discussion, are
summarised below:

Types of participation were summarised as observation, implementation
management and evaluation both farmers and researchers may be involved
in each or all of these. The meeting felt that the degree of farmer
participation was dictated by the complexity of the trial, confidence by
the researchers in the outcome, risks of loss and safety risks, and expected
interaction between treatment variables and farmer management. Observation
and evaluation are relatively neutral, although even here farmer participation
is more difficult with complex trials. Farmer participation places limits
on the sophistication in experimental design. Farmer management and implemen-
tation have more straightforward consequences particularly affecting precision.
Farmer and researcher management were recognized to have special attributes:

Farmer: broader basis for inference, confidence in recommendations
and a saving in research effort.

Researcher: Assumption of risk, better control and more precision,
better observation and records.

The assumption of risk emerged as a controversial topic in the discussion,
with participants divided about its treatment. Broadly speaking one
group felt that farmers should be insured from the risks of on farm
experimentation, the other group felt that if the focii for experimentation
were selected properly farmers would readily identify with the work, with
their eyes open to the risks. Much of the discussion of risk centred around
the question of farmer participation, the attitudes of researchers towards
farmers, and researcher confidence in the technology. Where researchers
are starting a programme at the RM/RI end of the experimental sequence -
looking for relationships or confirming they exist where their confidence
in outcomes under local conditions is low, researchers should carry the risks
of loss and costs of experimentation.

Where researchers are confident they have an intervention which is transfer-
rable into local conditions. Farmers should be partner in risk bearing, and
perhaps after a successful round of experimentation confirming clear benefits,
should share in, or carry the costs. This confidence does not obviate risks.
The natural environment is full of risks and even the most profitable technology
may fail to deliver the goods under extraordinary seasonal conditions. What
is clear is that farmers appreciate this, perhaps often better than researchers.

The pre-screening process should alert researchers to the sources of
uncertainty in local conditions which could jeopardise the performance of
selected treatments. They should gain some idea of the expected frequency
of such events from farmers or from secondary data. Researchers will be
looking for host farmers who identify with the priority problem targeted by
the diagnosis, who recognize the possibilities of the solutions) being
researched and who express an awareness that, should extraordinary events
arise in the experimental season, the solutions will fail. Farmers know
that plants die without water, and given rapporte, are not likely to blame
that on the research team. The risks faced should be placed before farmers
in advance and freely discussed. Nevertheless, again there is a turning
point at which farmers must bear the full costs and risks. It may be that
this is usually at the first round of farmer managed verification trials
where farmer management has been the context for RM/FM experimentation.
It may be at the second round of FM/FI trials where RM/FM has had researcher
imposed non-treatment variables, or where the OFR team has decided to go
immediately to FM/FI trials due to a high level of confidence that the
treatment benefits will readily transfer into the local situation.

A logical follow-on from the discussion of risk bearing was the issue that
the way in which researchers manage often increases risks. They have a time-
table to follow and cannot be on the spot, as can the farmer, to react on a
day to day basis to the vagaries of the season. This often leads to differences
in the management perspective of the researcher and the farmer, and gives
rise to the question of how far trial management can be reactive to the state
of the crop. A heavily fertilised treatment which looks good may, in
the farmers eyes, warrant and even require an extra weeding. On the
otherhand, from the same farmers perspective, a poor stand in an establishment
trial may simply not justify the blanket, non-treatment fertilizer application.

Comparisons may be distorted by interactions between non-treatment management
being differentially applied to treatments. Should such differentiation be
discouraged, or encouraged and where possible be observed, recorded and
measured? Similarly farmers judge timing by local conditions, for example
soil moisture and the likelihood of rain over the next few days as a guide
to planting. Tied to a pre-planned programme the researcher may plant when
the fields are too dry in the farmers eyes, or weed when the land is too wet,
or the weed burden does not justify it. Such inflexibility often results
in the research plots looking poor in comparison with adjacent farmers crops.
How can more flexibility be built into programming to allow trials to benefit
from the farmer, as a manager, being on the spot?

The final issue of discussion in this overall topic of the design and
management of on farm experiments was in some ways seeking answers to
questions just posed: How to design and manage to facilitate farmer partici-
pation. Overall the major comment on this was the need to educate both
researchers and extensionists to the contribution farmers can make to the
effectiveness of the research.

farmers know their own objectives and local conditions far
better than researchers or extensionists.

farmers usually do their own experimentation and thus know what
research is.

On farm research programmes need to exploit both of these farmer advantages.
In many countries it requires a major change of attitude on the part of
agricultural officials to see the farmer as a partner in research and
development efforts, not an ignorant pupil. Collaboration with village
based extension staff may allow much more flexibility in trial
implementation. Like the farmer, he is a man on the spot.


5.1 The importance of the issue

There is a widespread integration of crops and livestock enterprises in
the small farm systems of the Eastern and Southern African region. The
complex interactions between crop and animal enterprises highlights the
importance of on farm experimentation with animals as well as crops.
The use of draught animals and its influence on the timeliness of crop
planting and the quality of crop seedbeds is a common example. On farm
experimentation with animals is clearly a more complex process than with
annual crops, particularly handling the long term reproductive and growth
cycle. It has something in common with other long term research with
perennial crops and soil fertility, but, in that animals are mobile and
asynchronous, it has special problems in on farm experiments. Pastoral
systems pose a further unique set of problems. There is a traditional
gap between crop and animal research institutions in countries of the
region, yet at the farmer level the two types of enterprise are often
closely integrated. There is a need for full participation by animal
scientists in OFR teams working in systems with strong livestock components.
They need guidance on what can and what cannot be handled on farms and
on what to do about what cannot.

5.2 Discussions at the workshop

In presenting some ideas on on farm experimentation with animals
Dr. Frank Anderson from ILCA laid emphasis on farmers multiple objectives
in keeping livestock: for food, for sale and to provide intermediate
goods in the form of draft power, manure and security. The same animals
may meet several of these objectives at the same time, weighting
these is difficult and the interactions may run throughout the whole
farming system. The importance of the identification and specification
of costs and benefits across the system, when considering an intervention
in the livestock enterprise, cannot be over emphasised.

Participants agreed that several areas of animal management posed to
particular problems. Increasing quantity or quality of animal feed,
whether by improved gross or legume forage, or by the better handling
of by-products is analogous to experimentation on annual crops. Pasture
improvement with communal grazing was recognized to pose special problems

and like many soil and water management interventions, requires more effort
at the group or community level. When the animals themselves are brought
into experimentation then problems do arise, these are exacerbated when
results must await the completion of the reproductive cycle for responses
to be measured.

Where work is required on animal nutrition; the responses to changes in
feed and water management on calf rearing, milk production or weight gain,
are relatively short term. The main problems posed to on farm experimentation
in animal nutrition are mainly in obtaining statistical precision from
experimental design. Small farms tend to have small numbers of animals
often of different ages and, in the case of adult females, in various conditions.
Where results depend on the completion of the reproductive cycle the time
period involved compounds these design difficulties and makes heavy demands
on host farmers, often considerably disrupting their own management routines.

Three possible ways forward were suggested to ease the problems of carrying
out experiments on animals in farm or herd situations.

(1) Modelling: modelling of the growth of individual animals of herds and
of farm systems with strong animal components was suggested as an
alternative means to some field experiments, especially those subject
to the long term reproductive cycle. It was also seen as a means
to aid the diagnosis of interactions in the farm system and to anticipate
the effects of proposed interventions on those interactions. It was
emphasised that modelling may range from a qualitative description of the
relationships of the system, to a full scale quantitative effort to
measure both expected and realized effects of responses to change.

(2) Modified experimental sequence: It was suggested that systems based
livestock research where animals themselves are involved in experimentation
may require a different sequence, omitting the RM/FM stage. It was
tentatively suggested that Diagnosis should guide both RM/RI and FM/FI
experimentation. RM/RI experimentation should be on stations though
with appropriate local animals, with FM/FI rather less formalised on farms
observed and monitored by the on farm research team.

(3) Farmer assessment: the view was put forward that in farming communities
with a close knowledge of their animals, farmers would observe the
beneficial effects of improved feeding, breeding or general management
on their animals, without waiting to see for example if, in fact,
calving interval was reduced. This would lead to adoption and detailed
measurement of the precise response would be superfluous. To a significant
degree farmer participation and assessment was seen as an alternative
to formal monitoring and evaluation.

It was widely recognized by the workshop that experiences in OFE with animals
were limited and that every effort was needed to draw crops and animal
researchers together at the on farm level.


6.1 The issue and its importance

Analysis, interpretation and use of the experimental results is the product
of the whole diagnostic and experimental sequence. At this stage the purpose
of OFR to identify and, where necessary adapt, technologies relevant to the
development of local farmers comes to fruition. This interpretation and
use of results cannot realistically be divorced from the planning and design
of experiments at which point the pattern is set for analysis and interpretation.
Much of the discussion wandered back and forth between these interrelated
stages in the process.

Three major issues were presented and discussed under this general head:

(1) What criteria are used in the evaluation of on farm experiments?
(2) What are the analytical requirements of different types of on farm

(3) What are the uses of OFE results.

These three are discussed separately here:

6.2 What criteria are used in the evaluation of OFE

The workshop agreed that the traditional criteria of classical agricultural
research, physical output per unit area, would rarely be adequate in on
farm experiments. Participants emphasised that the relevant particular
criteria would depend on the objective of the experiment and the local
circumstances of target group farmers. In all cases the understanding of
the local situation by the diagnostic survey work would be the initial
basis for a decision on evaluation criteria for experimentation. This
would include investigative experiments aimed at further diagnosis and
understanding. The need for these and therefore the criteria by which they
are evaluated, would arise out of the initial diagnostic survey. The aim
in diagnosis would be to identify the criteria that local farmers will use
in the evaluation of experiments; small farmers are variously concerned
with one or several of the following criteria: value of returns to land,
levels of costs, value of returns to labour used at particular times of the
season, value of returns to cash outlay, and with risks of losses. Qualitative
aspects of new technology may feature strongly in farmer assessments;
processing, taste and storage characteristics of food, and the work habit
implications of new techniques are two examples. Clearly, effective
evaluation of experiment requires congruency between criteria used by

researchers and farmers. Clearly, also, where a number of criteria are
involved, their weighting is subjective to particular farmers and is
very difficult to reflect in formalised analysis. This is one foundation
of the need for farmer participation and assessment. With multiple criteria
it is extremely difficult to second guess farmers by formal analysis. The
surest way to ensure effective evaluation is to ensure farmer participation
and to create opportunities for farmer assessments of experimental plans and
experimental results. Often evaluation criteria are not dependent on
results farmers may judge a treatment for method of planting on the labour
required for it and the plant stand resulting from it. This may be anticipated
and a farmer assessment arranged at the appropriate time, however, not all
criteria will be anticipated and continuous interaction with host and
community farmers at experimental sites is perhaps the best way to catch
unexpected reactions.

Placing options in front of farmers was another topic brought up in discussion
and related to the idea that their complex of evaluation criteria are often
difficult to reflect in formal planning and analyses. Related but different
ways of solving a problem different treatments can offer different balances
of various evaluation criteria. Explicit options may also be offered which
cope with the problem under different seasonal conditions, farmers may choose
to take up the option for one set of conditions which he finds most problematic.
One reason for emphasising the idea of options was the problem of correctly
anticipating farmers evaluation criteria. Another reason was that options
enhanced the ability to manage circumstances as they arise, creating
flexibility. It was noted by the meeting that classical recommendations -
one best way to manage were rigid and frequently increased farmers'
vulnerability to vagaries in their natural and economic circumstances, whereas
increasing their flexibility is an important objective, particularly in
uncertain environments.

6.3 Analytical requirements of different types of OFE

In presenting this issue Mr. E. Shumba (R&SS, Zimbabwe) distinguished
hierarchies of analytical strategies and analytical tools by types of trial.
There was little criticism of his presentation and it is extracted here.

Table 2 summarises different analytical strategies for the three types of
Table 2. Analytical strategies for OFEs

Analyses Types of Trial


Statistical ***** *
Economic ** ***
Farmer assessment **
Evaluation unit Individual Individual Individual
site and across and across
Degree of variation Low Medium High

6.3.1 Researcher managed trial

Since such trials are aimed at assessing biological performance among
technical alternatives statistical analysis should be the most important
type of evaluation. Economic analysis could be done to identify gross
tendencies with respect ot additional benefits and costs that are relevant
for defining future direction. Farmer assessment in these trials is of
minor importance.

6.3.2 Researcher Farmer managed trials

Statistical analysis is also a very important evaluation criteria in these
trials. However, technical components are also analysed for economic
viability and compared to farmer practice. The analyses of such trials coupled
with farmer and extension worker assessment lead to selection of 'best bets'
which can be advanced into farmer managed trials.

6.3.3 Farmer managed trials

Emphasis is on across site analysis based on farmer assessment and economic
evaluation. Statistical analysis is of minor importance because of the high
coefficients of variation arising from environmental and management differences
between farms. An important consideration is whether or not farmers start to
adopt the trial technology on other parts of their farm. This is important
because _ince the farmer views a technology within the context of his whole

system he might adopt the technology for reasons other than biological or economic
performance. To ascertain the other reasons for adoption a supplementary question-
naire or regular meetings with farmers and extension workers are necessary. The
specific analytical tools which may be used in the different strategies are
presented in Table 3.

Table 3 Analytical tools for the different types of On-Farm Experiments

Trials managed by:

Possible Tools Researcher Researcher/Farmer Farmer

1. Statistical analysis

Simple mean comparisons *** ** **
Analysis of variance *** ** *
Regression and factor analysis *** ** **
Biological risk determination *** **

2. Economic analysis

Partial budgeting ** ** *

Programme planning *** ***
Linear programming ***
Economic risk determination (sensitivity) ** ** ***

3. Social analysis

Household organisation (decision making,division
of labour) ** ***

Community organisation **

Social cost/benefit analysis

I I _ I _

There were few comments on the analytical tools suggested. The sophisticated
data requirements and therefore high cost of data collection for Linear
Programming were highlighted. It was agreed that these costs were not so
serious if data was collected in parallel with experimentation, when staff had
to be in the field to supervise the experiments. On strategies one point made
was that farmer assessment of RM/RI trials may indeed be useful. There might
be highly acceptable treatments immediately obvious to farmers, or treatments
immediately and obviously unacceptable, despite experimental design complexity.
It is clearly advantageous to pick these up as early as possible, in the one
case to exploit the treatment as quickly as possible, in the other to drop it
as quickly as possible. Even at this early stage it was felt that farmer
assessment can help.

There was a good deal of controversy on the need for statistical analysis
of FM/FI type trials. Participants accepted that mean comparisons would
be an adequate tool, some participants felt that lower levels of significance
would be acceptable, others that sites should be expanded to seek conventional
levels of significance. A third line to this argument emphasised that
importance of seeing this issue in the context of the OFE sequence. If it
is a sequence it must be progressive, confidence in the robustness of treat-
ments must be increasing, each stage cutting away some doubts. When RM/FM
experiments are conducted with a background of individual host farmer
management of non-treatment practices, and treatments show through success-
fully across sites perhaps subsequent FM/FI experiments need no further
statistical analysis as no new sources of variation are introduced. When
RM/FM experiments are conducted with a background of simulated, and therefore
standard farmer management, statistical analysis will remain important at
the FM/FI stage as individual host farmer management of non-treatment practices
will be a new source of variation. Clearly judgement is inescapable, but it
is best done in the light of strong, ex ante evaluation of expected interactions
between treatments, non treatment management and variability in seasonal

The problem of statistical precision in OFE in general was seen as a formidable
one due to difficulties of control and management and to multiple sources of
variation. The counter argument was that if our diagnosis has focused us on
the problems which when solved will afford a major impact on the system and if
the appropriate evaluation criteria are being used, differences between identified

technologies and current practices should be large. Participants also felt
it important to do any statistical analysis on the appropriate evaluation
criteria (which of course may not be yields per unit area) as this is where
the large differences would express themselves.

6.4 The use of OFE results

The meeting identified three uses for OFE results in the OFR process. It
recognized that other uses for planning and policy were outside the scope
of the workshop. The three uses were:

(1) Re-definition of recommendation domains and improved understanding of
the farming system.

(2) Planning next years On Farm Experiments.

(3) Making recommendations to farmers.

Comment on the draft report added a further important use for results,
communication with appropriate station based researchers.
Discussions centred around item (1) listed above, with some comment on item (3).

6.4.1 Re-definition of Recommendation Domains (RD) and better system

Participants recognized RD's as stratifying device for mapping agro-ecological
and socio-economic variation that determines present farming systems. They
also recognized stratification as an iterative process. An initial definition
of domains may be attempted from secondary data and refined during the
diagnostic surveys. Definition may be further refined during experimentation.
For example fertilizer trials may show different responses on different soil
types dominating different parts of the present domains, justifying a sub-
division into discrete domains. Where such variations are hypothesised from
the diagnosis the initial on farm experiments can be distributed to allow
groupings of sites to test the hypotheses. Participants were doubtful whether
the initial cycles of OFE should aim at seeking out such differences even if
not suspected from the diagnostic survey work. It would involve putting
a network of sites down, enough to allow ex poste groupings where differences
emerge. It would require very detailed recording of environmental parameters
at all sites to permit grouping. Consensus suggested this was only practical
where differences were hypothesised in the course of the diagnostic process
giving guidance to site locations and recording needs. Similarly, it was
felt that investigative or diagnostic trials to improve understanding of the
biology of the system were valuable when focused by the diagnsotic process.

The implication of the consensus was that without reasons for focussing on
specific leads, the task was too broad and resource consuming to undertake
across the board. In both cases, both tightening the specification of RD's
and better understanding the system, it was acknowledged that leads might
arise out of early cycles of experiments as well as out of the diagnostic
survey. Some facets important to the development of the system are unlikely
to be fully captured in the initial survey, and indeed are unsuitable to
diagnosis by survey. In such cases ex-poste grouping of sites could verify
differences and improve the planning of future experiments.

6.4.2 Recommendations to farmers

The discussions emphasised that the objective of On Farm Experimentation was
to make appropriate recommendations to farmers. The point was stressed that
farmers might pick out treatments at any stage of experimentation and begin
(input supplies permitting) to use and perhaps modify them in their own
fields. Monitoring this farmer adoption and adaption, where innovations
are picked out early in the experimental sequence, is an alternative to
pursuing the sequence through. The danger of farmers' enthusiasm being lost
due to poor linkages with extension and input supplies was emphasised. There
is great value to subsequent dissemination if local extension staff learn
the in's and outs of emerging technologies by joining the OFR team and
farmers in assessing the ongoing experiments. A local mechanism for reviewing
OFE results, making recommendations, initiating and coordinating complementary
input and credit supplies, is vital. The local extension staff participating
in the OFR process form a core of trainers for extension staff throughout
the domain, to speed the diffusion of technology once recommended.


7.1 The issue and its importance

Participants agreed that one justification for OFR is its role as a linkage
between the three sets of actors in the technology development and diffusion
process; researchers, extensionists and farmers. Boldly stated a crisis
has arisen from a traditional process in which researchers 'pass down' blind
prescriptions to extension staff to 'pass out' to farmers. Field level
extension staff are caught in a squeeze between these 'downward flowing'
words of wisdom and the real needs of their farmer clients. They suffer a
'morale crisis' turning to 'special cases' of resource rich farmers who can
use some of the prescriptions in return for ready access to credit and
information. In bringing research, farmer and extension perspectives together
in the development of appropriate local specific technologies, OFR links
the interests of the three groups.

However, introducing OFR into the traditional research and extension
hierarchy and building a capacity for its operation is a long process.

7.2 Discussion at the Workshop

Participants focused on two needs: Clear understanding on the part of
existing, traditional research and extension establishment of the role of
OFR. Secondly, clear procedures for OFR operation, including the mechanics
of cooperation between component researchers and OFR's, and OFR's and
extension staff.

(1) There was consensus on the roles of OFR: To mobilise technical research
results, by identifying, and where necessary adapting, findings from stations
research appropriate to the priority problems of local specific target groups
of farmers.

(2) To pass technical problems, identified as important to the development
of target groups of farmers, back as agenda for technical research on stations.

Discussion centred on the failures in communication which inhibited others
understanding these roles, and the operational procedures needed for their

effective execution.

7.3 Linkages between component and On Farm Research

Participants acknowledged that the distinction between roles of component
and On Farm Researchers; and the methodological consequences of this
distinction, needed careful exposition to component workers. They felt
that component researchers evaluated OFR using their own criteria of
precision and therefore validity. There was insufficient appreciation
that the control necessary for experiments aimed at establishing basic
technical relationships invoked a serious distortion of the real world.
This is removed by OFR. Adaptation, by exposing basic relationships to
the wide sources of variation experienced by farmers, has its own methods,
standards of precision and criteria for evaluation. This distinction must
be part of a dialogue with component researchers during the introduction
of OFR, which continues throughout the OFR process. Participants stressed
that interaction with specialists can make an important contribution at
all stages of the OFR process:

(1) Diagnosis: To help specify problems a weed spectrum, insect or
disease pathogens, are good examples in the field. To help identify
technical causes of problems; for example rotation management, seedbed
preparation,weeding method any of these may be the basis for a particular
weed spectrum.

(2) Planning: To help identify and specify solutions and assess their
transferability to conditions in local situations.

(3) Experimentation: To help with design and records required to explain
performance, to assess unexpected phenomina in treatments and.help with the
biological interpretation of experimental results.

While recognizing that the scarity of specialists precluded their
involvement in every local situation participants acknowledged the value
of OFR in determining which specialist contributions are most needed in
any particular situation. The meeting recognized that OFR could enhance
the relevance of specialist work by information on farmer needs and manage-
ment conditions. For example information on how farmers assess varieties
and the kind of management varieties would be subject to provides guidance
for breeders in making up and managing crossing blocks and in selecting

from them. Reactions were mixed on the idea of helping specialist researchers
with their off station programmes. While it was seen as a favourable public
relations exercise for example to supervise multi-locational trials, there
is a danger of specialists coopting OFR teams as Trial Officers, pre-occupying
them with extended off station programmes and usurping their OFR activities.
This is particularly so when specialists dominate the decision making in
the research planning process. The idea was offered that all newly recruited
professionals should serve time in OFR, then a proportion be invited to
opt for specialisation. Eventual specialists would gain a grounding in
farmer perspective, seeing their specialisation in a more balanced way.
This 'apprenticeship' would allow them to understand the complementary role
of OFR in passing technical problems important to farmers to appropriate
specialists, and, at the same time, slotting specialist findings into the
local specific situations for which they are appropriate and timely.

It was agreed by the meeting that convincing personalities on the role and
contribution of OFR was not enough. Operational procedures had to be
developed to facilitate and perpetuate the interactions between component
and On Farm researchers. Several devices were noted:

(1) Annual research planning meetings with cross representation of relevant
OFR and component researchers.

(2) Ready access to specialists for on farm researchers

(3) Scope in specialist work plans for a number of field visits, to be
implemented at the request of on farm researchers.

(4) Mechanisms for prioritising technical problems passed to specialists
by on farm researchers and for the incorporation of high priority problems
into specialist work programmes.

(5) Scope in (selected) specialist work programmes for them to be repositories
of relevant technical knowledge in their field, from outside the country as
well as within.

7.4 Linkages between OFR and Extension

It was observed that linkages were needed all the way through the hierarchy
of the agricultural institutions in the country; at national, local and

field levels. Discussion was dominated by the need for coordination
at the field level, between contact extension workers and On Farm Research
teams operating in the same areas. It was recognized however, that this
coordination was dependent on effective linkages through the whole

7.4.1. The mutual roles; Extension workers in OFR and OF researchers
in Extension

Given the bridging role of OFR, with one end of the bridge in technical
research and the other in dissemination of information and diffusion
of innovations, an overall aim is to make the Extension worker part of
the technology development process with a commitment to the recommend-
ations made for local farmers.

Extension workers were seen with several roles in OFR:
Help in the definition of Recommendation Domains
Liason with community leaders
Selection of farmers:- during the Informal Survey and as hosts for
experimental sites
Help in the evaluation of possible solutions to farmer problems
Help with trial layout, supervision and recording
Liason with farmers and community on their evaluation of the experiments
Decisions on recommendations.

Controversy centred on two issues, first the potential biases from
extension workers selecting farmers both for informal survey and as
hosts for experiments. Second, the level of involvement expected of
extension workers in trial management. Several participants expressed
the opinion that because extension workers contacts with the community
are often with 'special case' farmers, allowing them to select farmers
to cooperate with the OFR team would create 'a bias and result in
unrepresentative results. If the extension workers perceives benefits
coming to selected farmers, he will again be tempted to select those
he would like to see benefit, or who may, in turn, benefit him later.
Participants also highlighted 'male' bias, with most extension workers
as males. Some measure of control was seen to be possible. Participants
suggested that if extension workers were clear about objectives their
own self interest would dictate unbiased selection. They pointed out
that for the selection of experiment hosts at least the Formal Verification

survey provides standards which hosts should measure upto. Others
noted that extension workers had to abide by local political leaders,
often higher resource farmers, who may want to be involved. It
is essential to recognize this when it happens. Also, even within a
fairly homogenous target group, there will still be innovators
and laggards. If extension workers can identify innovators who are
representative of the group on the basis of key parameters, this may
be a positive, purposive selection. Results from 'politically selected'
OFE hosts can be omitted from analysis.

There was concern about the level of involvement of extension workers
in trial management. It was agreed that the earlier the extension worker
is involved with a technical package which becomes a recommendation, the
better. However this was not seen as the same as trial management.
Broadly speaking participants agreed that extension workers, and even
research technicians, could not be substitutes for the involvement of
professional researchers. It was observed that Researcher Managed trials
designs may be too complex for extension workers to fully understand,
while Farmer Verification Trials should be easily understood. It was
acknowledged there is a difficult balance to be managed; the extension
worker should not be seen only as a menial, 'working for' the OFR team.
He should be seen as essential to technology development and diffusion.
The roles he is given, while within his capabilities, should reflect
this partnership.

OFR was also seen with several roles in extension:-
- to improve the extension workers understanding of farmers aims,
management logic, and problems.

to expose the extension worker to potential technologies and, by
incorporating him in their development, familarise him with all
management aspects.

to train extension workers in the laying out, recording and evaluation
of plots in farmers field.

All these roles will fit the extension worker to train other extension
staff working with the target group, in the management and demonstration
of new recommendations which emerge.

7.4.2 Institutional procedures for effective OFR extension linkage

There was no detailed discussion of institutional procedures which
promoted effective OFR/extension linkages. Some participants wondered
whether OFR should have a home in extension, rather than in research.
Others expressed doubts on the grounds that Component Research would
remain isolated from small farmers unless OFR was used as a device to
expose technical researchers to farm situations. It was recognized
that OFR would usually be a rolling programme, spending 2-4 years with
two or three target groups then moving on to others while results
diffused through the groups. The workshop recognized the need for a
national policy on OFR, and thus research and extension dialogue about
it at the national, and local as well as field levels. At the national
level guidelines for OFR operational procedures and the division of
responsibilities between research and extension, would be spelled out.
Several linkage actions were seen as important for senior agricultural
staff at the local level:

the selection of priority Target Groups for OFR (on national policy
grounds, or on grounds of poor adoption from present recommendations
or indeed no present recommendations).

the selection of better extension contact workers to cooperate with
OFR teams in these target group areas.

guidance on modified workplans for these selected extension workers.

initiatives on selecting supporting and servicing new recommendations
made for target group farmers.

Similarily -it was felt to be important that the OFR team professionals
and local senior agricultural staff combined to train selected extension
workers. Their collaborative responsibilities, and their future work
as extension staff trainers should be spelled out for them.


The final morning of the workshop was set aside on the programme for the further
discussion of issues, or the discussion of new issues. Six issues were raised
by participants, none were thoroughly discussed due to limited time and some
were not discussed at all due to a dearth of knowledgeable participants.

The issues are listed with the main comments on those discussed summarised
under each issue.

8.1 Experimentation in Agro-forestry

An issue was raised on the feasibility of agro-forestry experimentation.
No participants claimed any experience in agro-forestry experiments and
the issue was not discussed.

8.2 Training for agronomists in On Farm Experimentation

Issues were raised about training for agronomists in On Farm Experimentation
these were:

(1) Could graduate training usefully handle the farming system perspective
and On Farm Experimentation?

(2) What are participants views on:
(a) The International Course for Development Oriented Research in
Agriculture (ICRA), in the Netherlands.

(b) CIMMYT's Regional Training Workshop in association with the
University of Zimbabwe.

There was no reaction on the question of graduate training. The only
comment on the ICRA course was that national programmes had difficulty
finding people with the full qualifications asked by ICRA. There were
several comments on the CIMMYT/University of Zimbabwe Regional Training

More policy makers should attend the workshops in order to be able to
guide their donor projects more effectively.

The field part of the workshop deserves a greater proportion of the total
time. The On Farm Experimentation practical side is underemphasised because
this workshop is during the dry season with nothing in the ground.

There is difficulty for participants in attending both of the two parts
of the workshop in the same year. It should be given as a single five
week unit in the long vacation.

8.3 What publications accept articles in On Farm Experimentation; in
methods and results

Two journals were listed by participants; Experimental Agriculture
and Agricultural Systems. A probable future journal Southern African
Journal of Agricultural Research, to be coordinated through SADCC
(SACCAR) should be approached to cover this field. Tropical Agriculture
is another journal of international repute with an interest in OFR/FSP

8.4 How can experiments be evaluated in terms of returns to labour

This issue, discussed briefly in section 6.2, was raised again. In some
countries of the region cash and labour, especially seasonal peak
labour, are the factors limiting the productivity of small farm
systems, not land. Farmers would evaluate experiments against these
limiting factors. How could researchers best evaluate experiments

against seasonal labour peaks.

It was pointed out that where returns to labour are known to be farmers
evaluation criteria from the diagnostic work an attempt should be
made during the economic pre-screening in planning experiments to
relate expected treatment production differences to the expected
differences in labour required to elicit them, and to put values on
both the production and the labour. The calculation itself is simple.
Obtaining precise data to do the calculation is often difficult. As a
first step rough estimates should be used, together with sensitivity
analysis to see if precision is required in the data. If precision
is needed then efforts should be made to measure the data requirements
for the specific labour operations involved in the comparison. These
can be measured on local farms during the cycle of experiments, or
from secondary sources when the operations can be matched with those
in the experimentation.

8.5 Flexibility in recommendations

A 2nd airing of this issue compared the variability in farmers
circumstances, from farm to farm, field to field, household to household
and season to season, with the rigidity of classic 'best way'

recommendations. It emphasised the point that the 'best way' may
change with all these sources of variation. The question posed was
how can recommendations be made more flexible, with more initiative
left to the farmer and sufficient information to exploit the flex-
ibility. Is flexibility more feasible with some components and not
others? What implications does this have for trials, their design
and evaluation?

Discussion of this issue acknowledged its importance, and acknowledged
that the rigidity of classical recommendations can make farmers more
susceptible to hazards. The example of monocropped hybrid maize
(compared to lower yielding but drought escaping intercropped local
variety) jeopardising food security in a poor year, was offered. No
full solutions were identified but OFR/FSP was seen to offer a way
forward, with three angles on the issue:

(1) Diagnosis should be directed towards an understanding of the
influence of different sources of variation on farmer objectives and
management strategies where variation is identified as important.

(2) Where diagnosis gives these insights solution options should be
identified for different sets of local circumstances.

(3) Solution options should be laid out before farmers and discussed
in terms of their relevance to different household or seasonal

8.6 The evaluation of OFR/FSP programmes

The discussion of this issue raised many questions but few answers.
Important questions were raised on evaluation criteria: should farmer
adoption or progress in institutionalising OFR/FSP be dominant?
Project objectives and strategy were emphasised as the appropriate basis
for evaluation. Important questions were also raised on who does the
evaluation. The danger of evaluation by people with 'funny ideas' about
OFR/FSP was recognized. Particularly of evaluation by academics with
high flown ideas about what OFR/FSP ought to accomplish, but no idea of
the practical problems of institutional change.

Annex 1.


Richard J. Edwards

R.B. Jones

J.D. Siebert

Synai Masikara


Frank Anderson


Mike P. Collinson

A.E. Chabeda

James B.W. Matata

R.A. Kirkby

Robert McColaugh

Ann Stroud

A.M. Marimi Mpuria


P O Lobatse

Agricultural Dev.
P O Box 71

P 0 Box 90

P 0 Box 10275

P O Box 5689
Addis Ababa

P O Box 25171

Min of Agric.
P 0 Box 30028

P O Box 14733

IDRC, P 0 Box
62084, Nairobi

P 0 Box 30261

P 0 Box 62084
Nairobi, Kenya

P O Box 450


Extension Agronomist

(Team Leader)

Research Agronomist

Technical Officer

Team Leader

Farm Economist

Director of

Senior Research
Officer (Biometrics)

Programme officer
Crop and Livestock
Production Programme

Agriculture Advisor


Sorghum, Maize various
Fodders (small stock &
Cattle feeding.)

Millet, Sorghum, Maize
drought oxen.

Sorghum, Maize Cowpweas,

Legumes, Maize Cattle,
tea and coffee fur sheep
sorghum maize & cowpeas

Ruminant Livestock

Farming System

Pasture Agronomist/
Specialist & Livestock
(non-disease) policies.

Assistant both crop and
Livestock scientists in
the design, analysis and
in the interpretation of

Maize, Oilseeds rootcrops,

Ag.Extension & Research

Dryland crops, ox culti-
Consulting Agronomist eating, weld management
FAO, Conservation Village vegetable crops.

Senior Research
Officer (Agronomist)

Maize, beans, potatoes



L.J. Janicki

J.T. Legg

D.R.B. Manda

F.M. Nyirenda


P.K. Sibale


Chitedze Res.Station
P O Box 158, Lilongwe

P 0 Box 30134

P 0 Box 30134

Chitedze Res.
P O Box 158,Lilongwe

P 0 Box 158, Lilongwe

Chitedze Res.Station
P O Box 158,Lilongwe


Deputy Chief
Research Officer

Farming Systems

Research Management

Crops & Animals Research

Maize, G/nuts, Tobbacco

Chief of Party UF/

Breeder Groundnuts


Powell J. Mark

A.H. Suleman

Sampson A-Koi-

Elhag Hassan

Musa Babiker Taha

Themba Masuku

C. Seubert

ILCA, PMB 2248
Kaduna, Nigeria

Kaduna, Nigeria

Min.of Agriculture
JUBA, Sudan

Western Sudan
Agric.Research Project
P O Box 5141,Khartoum

WAD HRS, Eddamer
P 0 Box 31,Sudan


P O Box 4,Malkerns
Research Station

Crop Agronomist
crop livestock

Field Programme

A/Director of

Acting Director
Elobeid Research

Research Scientist
Agronomy section


Farming Systems

Sorghum,millet, maize
G/nuts cattle, goats

Forages, Cattle sheep
and Goats.

Maize,Sorghum, cattle
sheep and few G/nuts
Millet, coffee tea and

Pearl millet, sorghum,
Roselle and Legumes

Faba beans, chicken
pea Lentils and
Fasiolus beans

Maize, fruit, and
Horticultural crops.

Maize, dry beans




T. Stilwell


Robert Hudgens

Tom Maynard

G.A.S. Mitti

C.M. Ndiyoi


Crops & Soils
Michigan State
E.Lansing, MI
4882 USA

Kabwe Regional

Mt. MakuLu R.S.
P/Bag 7,Chilanga

P O Box 510089

P 0 Box 910064,



Adaptive Research
Planning Team

Farming System

F.S. Agronomist

F.S. Agronomist



Maize, sorghum, G/nuts
beans, Sunflower


Maize, Sunflower,
Sorghum, G/nuts, beans,
rice, cotton, F.Millet.

Maize, Millet, Sorghum

Enos M. Shumba

R & SS, Agrono-
mist Institute
P 0 Box 8100

FSR Agronomist

Annex 2



Mr. Chairman, the District Malawi Congress Party Officials, your Worship the
Mayor of the City of Lilongwe, CIMMYT Representatives, Distinguished Partici-

pants, Ladies and Gentlemen.

1. First and foremost, I wish to express my sincere thanks to His Excellency
the Life President, Ngwazi Dr. H. Kamuzu Banda who is also the Minister for
Agriculture for directing me to come and open this important workshop.

2. I feel greatly honoured to do so and therefore,

3. On behalf of His Excellency the Life President, Ngwazi Dr. H. Kamuzu Banda,
the Government and the people of Malawi and on my own behalf, I wish to extend
a very warm and hearty welcome to all our distinguished participants to the
joint CIMMYT Government of Malawi sponsored On-Farm Experimentation Network-
shop. Please feel free to visit any part of the country you wish to go to and
I hope that you will enjoy your stay with us, and that you will have an oppor-
tunity to see something of the surrounding countryside to enable you to
appreciate the efforts that Malawi and her farmers are making in the area of
agriculture development.

4. Mr. Chairman, Malawi has achieved a creditable degree of economic develop-
ment since Independence, despite the formidable presented by the growing
population, our distance from the sea, and lack of mineral wealth. The economy
of Malawi is primarily oriental to agriculture, and to support this the country
is endowed with relatively fertile soils, a climate generally favourable
to a wide diversity of agricultural production, and good water resources. This
has enabled Malawi to maintain adequate domestic food supplies and the production
of export crops, as well as to provide stimulus to a growing manufacturing base.

5. Agriculture is the most important sector of Malawi's economy, as it engages
87%, of the employed population, accounts for 40% of the Gross Domestic product

and 86% of export earnings. At present Malawi's agricultural production comes
from two subsectors; the smallholder and estate subsectors. The smallholder
subsector contributed in 1983, 80% of total agricultural production providing
almost all of the foodstuffs (maize, cassava, rice, sweet potatoes, beans,
groundnuts, vegetables, livestock products) for domestic consumption and some
of the agriculture raw materials for industry.

6. Mr. Chairman, the main government objectives are to maintain food self-
sufficiency expand agricultural exports and improve rural incomes. The current
agricultural policy emphasizes the sustaining productivity increase per unit
area per unit animal and per unit water input in case of irrigated crops.

7. Earlier agricultural development strategy was based on Rural Development
Projects which targeted on areas with high agricultural potential. It was
however realized, as early as 1978, that although these projects were making
a big contribution to agricultural production, it would take decades to reach
the majority of the population a strategy redesign was undertaken which resulted
in the formulation of the concept of the National Rural Development Programme
(NRDP). In this regard, priority is placed on those investments having an
immediate impact on agricultural production, leaving out costly or less direct
investments. NRDP's main emphasis is on the provision to smallholders of
agricultural inputs, extension services, credit, markets, soil conservation,
watershed protection and on afforestation.

8. Improved growth of the smallholder sector is the key element in the
development strategy. Faster growth is needed to generate employment for
workers not absorbed by the expansion of the non-farm sector, and to provide
inputs for the expanding agro-based and other industries. To maintain and
increase crop yields as fallow periods are shortened and more marginal
land is brought into production will require development and increased
use of higher-yielding varieties, more intensive and efficient use of fertilizers
and pesticides, better crop rotation, irrigation and soil conservation.

9. Under the National Rural Development Programme (NRDP) both technology
development (Research) and technology transfer (Extension) are going to assume
a more important role in the development of smallholder agriculture. Three
conditions must be satisfied if farmers are to increase their use of improved

agricultural technology: (i) the technology must be developed; (ii) the farmers
must know how to use it effectively and (iii) they must have incentives to use
it efficiently. The extent to which these three conditions are satisfied depends
on the institutional structure within which the farmer lives and works. The
more this structure encourages the flow of resources into development of and
dissemination of knowledge about new technologies and strengthens farmers'
incentives to adopt them, the faster will be the place of agricultural development.

10. To this end Mr. Chairman, the Malawi Ministry of Agriculture is
restructuring and re-organizing its Department of Agricultural Research (DAR) so
that it can best serve the interest if the nation and the farmer. Under the NRDP
the country is divided into eight Agricultural Divisions (ADDs). A basic
aspect of the reorganization of DAR is the change from a research-station structure
to a national coordinated and integrated multidisciplinary team approach to
research. Another important change is the formation of eight Adaptive Research
Teams (ART), composed of an agronomist and an agricultural economist, one team
for each ADD.

11. There is growing evidence in developing countries that research and extension
programmes work most useful information when all three components of the total
effort meet and work together. The greatest opportunity for true success exists
when the smallholder farmer, the research scientists and the teacher meet, plan
and work on the farmer's problems in the farmer's garden, To be adapted to the
farmer's situation and must help meet a recognized need.

12. Mr. Chairman, I would like to remind our distinguished participants that
successful research starts with the farmer planning research must explicitly
take into account the circumstances of farmers for whom the technology is intended.
The Adaptive Research teams in the ADD's will work hand-in-hand with extension and
farmers in studying farmer-circumstances, identifying their problems and in
adopting the technology to suit farmers' conditions and meet their needs. This
will be undertaken through On Farm Experimentation.

13. Clearly farmers reject available technologies not because they are conservative
or ignorant but because they rationally weigh the changes in incomes and risks
associated with these given technologies under their natural and economic circum-
stances and decide that for them the technology does not pay. The task of the
scientists, both research and extension, is to show how to incorporate a knowledge
of farmer circumstances into the design of technologies so that they are consistent
with farmer priorities.

14. Through on-farm experimentation, both research and extension personnel
will develop an appreciation for the importance of developing and extending
or adapting the technology that suits farmers' circumstances. Malawi
attaches great importance to on-farm experimentation and to this networkshop.
We are convinced that this approach will yield good dividends not only for
the nations represented here but also for their respective smallholder farmers.
The smallholder farmer is the key to agricultural development and ultimately
to our economic success.

15. The Malawi Government and CIMMYT have embarked on a collaborative effort
aimed at training our agriculturists in the field of on-farm experimentation.
This exercise is already yielding positive results. I am confident, Mr. Chairman,
that the experienced scientists from CIMMYT,from participating countries, and
from Malawi, will use this week for the purpose of developing on-farm
experimentation strategy aimed at increasing smallholder farmer productivity.
I believe that the papers and topics you will consider, and the discussions
that will take place on them, will be directed at meeting practical problems
experienced by our farmers. Our farmers are cautious more than conservative.
We need to learn from their experience; it is an essential input in any
agricultural research. Farmers and scientists working together can advance
agricultural production in Africa.

16. Distinguished participants, Ladies and Gentlemen, I now have much
pleasure to declare this Networkshop open.

Thank you very much.






Issues have been identified from correspondence with agronomists involved with
On Farm Experiments (OFE) in system oriented research programmes. These issues
have been grouped following the sequence of On Farm Research (OFR) programme, to
serve as an agenda for the networkshop.

Each set of issues will be introduced in plenary session by a nominated partici-
pant. He is asked to spell out the operational questions raised by the issues.
Participants will seek clarification where necessary, and have been asked to
identify and be prepared to present examples of decisions taken in their own
programmes relevant to these issues. The workshop will divide into groups to
discuss the issues and the examples raised and seek conclusions how best to
resolve them operationally. Each group will have a chairman and a rapporteur.
Selected rapporteurs will present their group findings, especially answers to
operational questions, to a plenary session for comment by other groups and
for discussion. Thursday morning will be set aside for other issues that
participants feel should be discussed, and for further discussion of unexhausted

Proceedings of the workshop will be a statement of these issues, discussion
of each and conclusions reached on how best to handle them operationally.


Monday 21 May

8.30 a.m Registration of participants
9.00 a.m Opening of the Workshop
9.15 a.m Introduction to the objectives, format and procedure of
9.45 a.m Issue 1 Technical pre-screening how far can we go?
10.15 a.m Coffee
10.30 a.m Group Discussion on Issue L
12.00 noon Presentation of Group Conclusions and Discussions.
1.00 p.m Lunch

2.00 p.m









Tuesday 22 M

8.30 a.m

9.30 a.m

10.15 a.m

10.30 a.m

11.15 a.m

1.00 p.m

2.00 p.m

7.30 p.m

8.00 p.m

Wednesday 23

8.30 a.m

9.30 a.m















- Issue 2. What is the most useful basis for the classification

of the On Farm Experiments?
Group Discussions on Issue 2.

Group Discussion, Issue (contd.)

Presentations of Group Conclusions and Discussion.

Reception for participants.


- Issue 3. The design and management of On Farm Experiment.

- Group Discussions on Issue 3.

- Tea

- Group Discussions (contd.)

- Presentation of Group Conclusions and Discussion.
- Lunch

- OFE in Malawi Field visit

- Issue 4. On Farm Experiments with animals.

- Group Discussions on Issue 4.


- Presentation of Group Conclusions and Discussions on Issue 4.

- Issue 5. The Analysis, Interpretation and use of the results

of On Farm Experiments.
- Coffee

- Group Discussions on Issue 5.

- Presentation of Group Conclusions and Discusssions.
- Lunch

- Issue 6. Operational linkages with Component Research and Extension

- Group Discussions on Issue 6.

- Tea

4.15 p.m Discussions

Thursday 24 May

- Participants: issues needing further discussion, or new issues

for discussion.

- Group discussions on issues raised.

- Coffee

- Group Discussions (contd.)

- Presentation of Group Conclusions and Discussions.

- Closing remarks

- Closing lunch.

8.30 a.m








CIMMYT International Maize and Wheat Improvement Center

The CIMMYT International Maize and Wheat Improvement
Center holds all rights to the source text and shall be
considered the copyright holder for the text and images of
these publications.

The CIMMYT International Maize and Wheat Improvement
Center has made this publication available to the University of
Florida, for purposes of digitization and Internet distribution.

The CIMMYT International Maize and Wheat Improvement
Center reserves all rights to this publication. All uses,
excluding those made under "fair use" provisions of U.S.
Code, Title 17, Section 107 are restricted.

Contact the CIMMYT International Maize and Wheat
Improvement Center for additional information and

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs