Farming systems research in Phalombe Project, Malawi

Material Information

Farming systems research in Phalombe Project, Malawi another approach to smallholder research and development
Hansen, Art
Mwango, Emmanuel N
Phiri, Benson
Place of Publication:
Gainesville Fla
Center for Tropical Agriculture, International Programs, Institute of Food and Agricultural Sciences, University of Florida
Publication Date:
Physical Description:
34 leaves : ; 28 cm.


Subjects / Keywords:
Agricultural extension work -- Malawi ( lcsh )
Agricultural development projects -- Malawi ( lcsh )
Agriculture -- Research -- Developing countries ( lcsh )
bibliography ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaf 34).
General Note:
"In cooperation with Farming Systems Analysis Section, Department of Agricultural Research, Ministry of Agriculture, Government of Malawi."
Electronic resources created as part of a prototype UF Institutional Repository and Faculty Papers project by the University of Florida.
Statement of Responsibility:
by Art Hansen, Emmanuel N. Mwango, and Benson S.C. Phiri.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact Digital Services ( with any additional information they can provide.
Resource Identifier:
70962355 ( OCLC )

Full Text
Center for Tropical Agriculture
International Programs
Institute of Food and Agricultural Sciences
University of Florida
Gainesville, Florida 32611
in cooperation with

By Art Hansen,
Emmanuel N. Mwango,
and Benson S.C. Phiri
Farming Systems Analysis Section Department of Agricultural Research
Ministry of Agriculture
Government of Malawi

Smallholders constitute the majority of Malawi's people, and their agricultural production is the foundation of the national economy. The Government of Malawi has consistently
emphasized the important role that smallholders play in national development and has established many institutions and programs to encourage increases in smallholder production, yields, and standards of living. Malawi has achieved a qualified success in its efforts and is one of the few African countries to achieve positive rates of growth in food and export crop production. Yet the success is qualified because .smallholder yields remain low, and smallholders have not adopted technological innovations as rapidly as many planners and development experts had
This paper describes a new smallholder-oriented research program that was established in Malawi in 1981. The farming systems research program is based in the Department of Agricultural Research of the Ministry of Agriculture as one component of the total research effort. The program also links research with the Department of Agricultural Development and the Division of Planning and Evaluation within the same Ministry. We will discuss the principles and methodology of the farming systems approach, its application tn one area of Malawi, and some preliminary conclusions.

The term "farming systems research" is used in different ways. Some people refer to multiple cropping research, while we and many others mean a more holistic approach that integrates production and socioeconomic scientists (Technical Advisory Committee 1978; Gilbert,,. 190; Collinson 1960; Hansen 1981). The farming systems research program in Malawi is specifically designed to focus on smallholder problems, priorities, and opportunities and to help the Ministry of Agriculture identify high priority problems confronting smallholders in different localities, understand the systemic constraints and opportunities in existing farming practices, and make farm-tested recommendations that are appropriate for smallholder conditions and are acceptable to and desired by smallholders. The basic elements in the farming systems approach areas
-pulling together the various research disciplines and extension in a cooperative series of research and development activities;
-understanding the complexities and interdependencies of localized smallholder economies and.ecosystems;
-involving the smallholders themselves in diagnosing local problems and constraints, planning alternative technologies, and then testing and evaluating the alternatives;
-proposing gradual modifications to existing farming systems rather than radical new directions; and
-testing the advantages and costs of proposed innovations under actual smallholder conditions and management.

The central concept'in this approach is the farming system. Hansen,, (1961) define this as "a complicated interwoven mesh, of resources and factors (agronomic, economic, social, cultural, physical, etc.) which are managed to a greater or lesser extent by a farmer." This concept includes off-farm resources and factors as well as on-farm ones. The farmer (an individual or family unit) utilizes some of the technology that is available in an attempt to increase the farmer's or farm household's utility within a given context of accepted preferences, aspirations, and socioeconomic conditions. Utility refers to a broad range of satisfactions. For Malawi's
smallholders, for example, utility obviously includes the production of food, both staples and side dishes, for home consumption, as well as the-generation of cash income.
The farming system concept reflects the empirical complexity of the conditions surrounding smallholder agriculture and the complexity of the decisions that smallholders have to make. Most of their decisions are compromises in which the farmers balance what they want to do against their limited time and other resources. Other compromises occur because the farmers' goals conflict: trying to achieve higher incomes versus lower risks, for instance. Consequently, farmers often fail to practice what they know to be "improved" practic&-. (in terms of improving yield) because they are simultaneously trying to maintain a number of enterprises (some of which may be off the farm) to satisfy a range of desires or necessities.
The farming systems approach considers more factors and relationships than single commodity (maize, groundnut, livestock, etc.) or single discipline (agronomy, pathology,

4 4
etc.) programs. This does not mean that a farming systems
program replaces the more narrowly focused programs. Both are needed to complement each other. The single commodity and discipline programs research in depth specific relationships, while the farming systems program is more concerned with adapting the body of existing research knowledge to smallholder conditions and alerting other research programs to high priority
areas where adaptive research is needed.
Although the government of Malawi initiated its farming
systems program in 1979 as part of a major upgrading of agricultural research, the first demonstration of the farming systems diagnostic survey (the first step in the four step method) took place in early 1980 when Dr. Mike Collinson of the International Maize and Wheat Improvement Center (CIMMYT) conducted a survey of Ntcheu in Lilongwe Agricultural Development Division (Lilongwe ADD). Dr. Collinson s
involvement reflects the interest shown by various international centers and donors in the potential of this approach which is also being established in the neighboring countries of Zambia$
Zimbabwe, Kenya, and Tanzania.
The Farming Systems Analysis Section in the Department of
Agricultural Research began operations in early 1981 with the arrival of Dr. Art Hansen. During 1981 t"e section conducted
diagnostic surveys in three areas of Malawi:
- the Lilongwe Project of Lilongwe ADD,
- the Phalombe Project of Blantyre ADD, and
- the Bulambia plain of Karonga ADD,
and assisted in surveying the Balaka area of Liwonde ADD. Research trials were designed for two of those areas (Lilongwe and

Phalombe) and assistance given to Liwonde ADD in designing its own trials. During this last growing season (1981/82) we have been monitoring these trials, and we are now in process of analyzing the results. This paper discusses our work in Phalombe as an example of how the farming systems approach works to generate smallholder-oriented recommendations. After a brief description of the area and its people, we will trace our involvement, the steps we follow in our approach, and our tentative conclusions at this time.
Phalombe Project is located in southeastern Malawi between Mount Mulanje to the south and Lake Chilwa to the north. The project's eastern boundary is the international border with Mozambique. Much of the area is a colluvial plain at an altitude of 600-700 meters, but the flatness of the plain is broken by a number of steep sided rocky hills and mountains, and the southern half.. of the project is dominated by the towering bulk of Mount Mulanje which rises to 3000 meters.
More fertile and well drained pediment soils surround the mountains and hills, while the plain is more variable in drainage and soil texture (coarse sands to heavy clays). To the north and northwest the plain slopes down to yike Chilwa and the Phalombe River which drains into the lake. The heavy clay soils in these lower areas are seasonably or permanently waterlogged (Land Husbandry Unit, Blantyre ADD).
Crops are much more important than livestock to the smallholders in Phalombe. Maize is the most important crop. More than 75% of cultivated land is devoted to maize, usually grown as

the major crop in a field intercropped with pulses, groundnuts, millets/sorghum, and/or cassava (Table 1). Maize, cassava, millets/sorghum, and rice are staple food crops, usually prepared as nsima, while the pulses (primarily cowpeas and pigeonpeas) and groundnuts are grown as relishes (ndiwo), i.e.$ dishes that accompany nsima. Almost a quarter (23%) of Phalombe rural
households own no livestock at all. Although chickens are the most common animal and are owned by almost three fourths (71%). of the households, many (28%) of these households only own one chicken each. Fewer than one sixth (14%) of Phalombe households own cattle, one fifth (20%) goats, and one 4n ten (11%) pigs.
Unreliable rainfall is a major constraint to agricultural production and stability. Rainfall varies from place to place within the project with higher levels (1000-1300 mm annually) east of Mount Mulanje where agriculture is more secure, lower levels in the central section and along the western flanks of the mountain (averaging 800-900 mm), and dropping off towards the north and west to less than 800 mm a year. The low rainfall in the central,. western, and north,*rn sections is compounded by erratic distribution during the rainy season, especially the prevalence of February dry spells when the maize cobs are forming. An analysis of rainy pentades (five day units) for four rain stations in the central and wester', sections shows that rainfall distribution is adequate for good maize production only one in every four years (Land Husbandry Unit, Blantyre ADD).
Another important constraint for the majority of smallholders is a scarcity of land to cultivate. The population density in
Phalombe (121 people per square kilometer) is more than twice the national average, and- Malawi is one of Africa's most densely

populated countries (Table 2), Although the average holding size (cultivated land per household) in Phalombe is approximately one hectare (2.5 acres), Table 3 shows that almost a third of the households are cultivating less th-an half a hectare', and more than 60% cultivate less than a hectare. Land scarcity and the
drought-prone climate have induced many men and entire households to emigrate in search of land and employment elsewhere. This emigration explains the low average annual growth rate of the Phalombe population (1.6%) and the high ratio of women to men (Table 2).
The scarcity of adult men is also shown in the high percentage of rural households that are headed by women (37% in Phalombe as- compared with 28% for Malawi as a whole). Within Phalombe the sections with the poorest agricultural potential have the highest percentages of women-headed households (Evaluation Unit, Blantyre ADD). Two thirds of these women are unmarried singley divorced, or widowed). Married women are considered the heads of their households when their husbands return home less than once a month. Labor is an inherent constraint for households headed by women because they usually contain only one adult worker, while most households headed by men also contain another adult workers the wife.

1968/69 % of % Inter- 1980/81 % of % InterAcres Total Cropped Hectares Total Cropped
Maize 25800 81% 96% 29800 76% 92%
Pulses 23600(b) 74%(b) 100% 5100(c) 13%(d) 100%(c)
Millet/Sorgh 18000(b) 56%(b) 92% 5400(c)' 14%(d) 63%(c)
Groundnuts 10100(b) 32%(b) 97% 2100(c) 5%(d) 76%(c)
Cassava 6000(b) 19%(b) 77% 3000(c) 8%(d) 30%(c)
Rice 1200(c) 3%(d)
TOTAL 31900 100% 86% 39100 100% 75%(c)
a) The National Sample Survey of Agriculture 1968/69 statistics refer to a smaller area than the Evaluation Unit Working Papers, Blantyre ADD 1982, which refer to the present Phalombe Project dimensions. This is why the 1968/69 acreage statistics are so much smaller than the later hectarage statistics. -The earlier figures are given as a guide to the extent of multiple cropping ,and the overlap of land in the various crops.
b) These statistics include many acres which these crops share with maize and sometimes other crops as well, so they sum to more
than 31900 acres and 100%.
c) Intercropping and the extent of other crops than maize in 1980/81 are underestimated because intercropped fields are
recorded in two categories:
-mixed stand., i.e., more than one major crop in a field, and
-scattered plantings in a field with only one major crop.
Scattered plantings are then joined with monocropped fields to form a "pure stand" category that is opposed to "mixed stand". d)Only the maize statistics have been disaggregated in this table so
that monocropped maize (8% of all land in maize), mixed stands of maize and other crops (43%), and maize fields with scattered
plantings of other crops (49%) me-y be distinguished.

RURAL POPULATION GROWTH IN PHALOMBE AND MALAWI 1966-1977 (a) 1966-77 Mean Annual 1977 Women
1977 PopuLation Increase Growth Rate Population per 100
Density (b) Men
Phalombe 168,500 19% 1.6% 121 113
Malawi 5,547,500 37% 2.9% 59 107
a) Malawi Population Census 1977: Final Report. b) This refers to people per square kilometer (121 per sq. km.
equals 315 per sq. mile).
HOLDING SIZE IN PHALOMBE 1968/69 AND 1980/81 (a)
1968/69 (b) 1980/81
31% less than 0.5 hectares 41% less than 0.8 hectares 62% less than 1.0 hectares
81% less than 1.6 hectares 83% less than 1.5 hectares
a) National Sample Survey of Agriculture 1968/69 and Evaluation Unit Working Papers, Blantyre ADD 192. The earlier survey
covered only part of the present project. b) These were originally expressed as acreage: less than 2 and 4 acres, respectively.

OUR INVOLVEMENT IN PHALOMBE 1981-1982: THE FIRST STEP Step 1. Review and diagnostic survey May 1981.
Step 2. Plan alternative technologies May-November 1981.
Step 3. Test alternatives November1981-July 1982.
Step 4. Evaluate, recommend, replan Now in process.
The first step in the four step methodology involved reviewing secondary data as well as interviewing local smallholders and observing their fields. Evaluation economists at Blantyre ADD collected and prepared background data to brief the survey team. Project management and extension staff provided information about local cropping patterns they had observed.
The rapid reconnaissance survey (Collinson 1979) or what Hildebrand (1979) calls the sondeo works best when applied to a fairly homogeneous area. Project management noted the high priority of the central section (EPAs 3, 6, and 7), an area of medium rainfall levels where half of Phalombe's population lived, so that was chosen as the focus for the survey.
Ten people, divided into three teams, participated in the survey three days of interviewing and observing followed by a day of discussion. The group included research agronomists, evaluation economists, crops, land husbandry, and credit staff, an extension supervisor, and the firming systems analyst. Team members were -'structed to address smallholders as local experts, people who had survived because they knew their ecosystem. Using a CIMMYT questionnaire (Collinson 1979) as a guide, the survey covered a wide range of topics: cropping combinations and calendars, food preferences, periods of food and labor shortages, local strategies for coping with problems, role of livestock in the system, importance of off.-farm

employment and non-agricultural enterprises, common pests and plagues, what smallholders believed to be their major problems, and what they thought the project could do to help them.
The purpose of the rapid diagnostic survey is to form a rapid synthesis of the most important relationships among resources, constraints, and enterprises and the highest priority targets for action by research and development staff. Sometimes a survey team will discover previously unknown facts about the local agricultural and economic relationships. Often,
especially in areas such as projects in Malawi where a lot of background data are already available, the survey does not
unearth any new data but the survey team becomes more aware and able to appreciate the significance of facts that were already known.
This was what happened in Phalombe. Our conclusions about the basic parameters of smallholder agriculture agreed with the available background information that was presented earlier in this paper, but our recommendations were innovative because we based them on the actual problems and constraints of the majority of Phalombe smallholders. We pointed out that the major hazard confronting local farmers was low and variable rainfall, compounded by a shortage of land to cultivate. Farmers' ,i'ihest priority was ensuring their staple food supply, and they also needed to acquire relish foods and some cash incor~e. Capital was scarce, and credit was feared because they had no security to cushion a bad cropping season and, therefore, feared the consequences of defaulting. Some households sold maize immediately after harvest because they needed money (taxes, etc.), even though they knew they would run out of food

later. Many farmers and their families worked for others during the growing season instead of on their own holdings because they needed food to eat. Some more fortunate smallholders had more land, more capital, and more security, but we emphasized the conditions faced by the majority.
Smallholders were. trying to cope with their problems and constraints by devoting almost all of their land to staple crops interplanted with relish and some market-oriented crops. Insurance staple crops such as sorghum and cassava were grown., often interplanted with maize, to sustain the family if the maize crop failed. Many men and women worked off their farms as laborers or in various kinds of enterprises (fishing and the fish trade, for instance) to supplement their insecure agricultural incomes.
The production packages that were the basis for extension and credit in Phalombe were for acre (or 0.4 hectare) units, were primarily for cash crops, and were integrated packages of seed and several bags of fertilizer that required a lot of capital or credit. All of ihe research and extension recommendations included monocropping as one aspect. Our recommendations for action by research and the project stressed the need to work with units smaller than an acre and the difficulty for smallholders of taking production packages that requis-ed them to assume a lot of credit risk or put in more labor, particularly during the December-January period that was already a labor bottleneck. Since capital and labor were scarce and there was a feeling of insecurity with credit, the best set of innovations would permit a step by step progression with each step not requiring a great jump in resource commitment or risk.

We recommended that research examine improving both the yield and the stability of staple food production within a context of multiple cropping. The primary staple to test would be maize, followed by sorghum and cassava. Legumes (principally cowpeas and pigeon peas) would be an essential intercrop to provide for the relish as well as the staple; additional
intercrops could be market-oriented crops such as sunflowers, grams, chickpeas, etc. Trials would be run on local farms under smallholder management to see how they fared under real smallholder conditons.
We recommended that the project promote credit mini-packages of less than acre size; the packages would not be optimal in terms of highest yield per acre but would represent. improvements over present practices that were affordable to the majority. For the more fortunate smallholders the present packages were suitable, but the mini-packages would permit many more smallholders to participate in the project's credit program. Another recommendation was that the project examine its extension coverage, to see whether women smallholders were receiving enough extenson advice. Women provide most of the labor on smallholder crops anywhere in Malawi (Clark 1975), but they are even more important in Phalombe. Chickens were an important source of protein and cash income to many smallholders and were unprotected against Newcastle disease, so we recommended that the project provide vaccine.
The second step in the farming systems process was to plan alternative technologies based on our survey recommendations and

on joint planning with maize and legume agronomists at Chitedze Research Station and Bunda College and the B1antyre ADD management and technical staff. Because we wanted to run the trials under smallholder management to analyze their systemic adaptation, we did not use standard agronomic research designs such as randomized blocks with multiple replications. Smallholders needed to be able to handle the trials and understand the alternatives, so we used a simpler design in which each smallholder had a single replication of each treatment, and we could use the smallholders as blocks for analysis.
The planning step also included smallholders. After agreement had been reached among research and ADD staff, the farming systems staff called meetings in November in the two Phalombe villages where trials were to be conducted, and we consulted the villagers about their ideas. They contributed specific information about planting densities and dates for intercrops and made several criticisms. As a result, our plans were modified.
The final plans were for a simple 2x2 factorial test of two maize varieties under two levels of fertilizer, all treatments to be the same intercropping mix of maize, cowpeas and sunflowers. The maizes were to be the "local" type and CCA, an improved composite. Although hybrid maizes are the highest yielding, their "dentness" (soft starch) leads smallholders to reject them as a food crop because of their pounding (food processing) and storage characteristics. Hybrids are usually only grown by smallholders as a cash crop for immediate sale after harvest. The improved composites, however, are "flint"

(hard starch) enough to be utilized for home processing and consumption as are the "local" types.
Fertilizer levels were to be: I) without any fertilizer and 2) at the level recommended for composites one bag of 20:20:0 and two of S/A per acre (or 2.5 and 5 per hectare). Each of the two fertilizers is 20-21% nitrogen (N), so the recommended level is 30 kilos of nitrogen per acre. Maize and cowpeas were to be planted at the same time, three maize seeds per station three feet (0.9m) apart on ridges three feet apart, while cowpeas were very dispersed, following villager recommendations. Sunflower was to be delayed in planting until after the maize was well established* this reflected smallholder desires to promote the maize crop and treat other crops as bonuses that were not to endanger the staple. Each of the four treatments was to be eight ridges wide and ten maize stations long. This was much larger than usual research plots, but we wanted the smallholders to be able to appreciate any differences in yield, labor costs, etc,
The trials were designed to test how "local" mai:,e and an improved variety fared under identical conditions. We were trying to find out whether fertilizer by itself or improved seed by itself would improve maize yields and stability, which was the highest priority issue noted by tFe survey. The
intercropping pattern was fairly common in Phalombe. Most
smallholders mixed maize and cowpea seeds before planting and then planted them in the same stations. Sunflower was the most commonly grown cash crop and was frequently intercropped with maize. The cowpeas would provide relish; the sunflowers a source of cash; and the plot would therefore provide a complete

mix of nsima (staple food), diwo (relish food), and ndaiama (cash).
In addition to these research trials which were planned as a consequence of the survey, the ADD decided to seriously
consider revising its credit and extension policies to include mini-packages and to better address women smallholders. These decisions were based on recommendations by ADD evaluation and project extension staff (Evaluation Unit Working Papers, Blantyre ADD 1961; Evans 1981).
The third step was establishing trials. The first phase of this involved the selection of farmers and fields, education of the smallholders in our research design, distribution of seeds and fertilizers, and establishing monitoring procedures. Then the farmers planted, fertilized, and cultivated the plots on their own fields, while we monitored the progress of the trials and continued interviewing the trial farmers and their neighbors to learn their reactions to the alternative!,. Whereas most agricultural research focuses on the production, we were interested also in the perceptions .and responses of the farmers. Our goal was to provide farm-tested recommendations that smallholders could and would use, so we needed tz know
socioeconomic as well as agronomic information.
After explaining our purpose and plans to the villagers in public meetings in both villages (project/extension staff always accompanied us in village meetings), we asked each village headman to select eight volunteers to host the trials on their own fields. We requested that the smallholders be those with

less land because they were the ones who needed to intercrop, and we also asked that approximately half be women. As it turned out, the smallholders who were the hosts for the trials were split about half and half between smaller and larger, women and men, and the women between married and unmarried. The total trial thus consisted of 16 farmers in two separate villages.
Each farmer was asked to donate a section of a field, 16 ridges wide by 20 maize stations long. Research and project/extension staff visited each field with the farmer, marked the plots, and took soil samples. In several cases we had to change fields from the ones originally selected by the smallholders because the fields were inappropriate for controlled research.
In two different public meetings we then explained carefully to the smallholders and their neighbors the design of the trial: two plots with each type of maize with a crosscutting application of fertilizers to one plot of each maize. We drew diagrams in the dirt and answered many questions until people were satisfied that they understood what was going to happen and why. As an additional guide the signs given to each farmer to mark' each plot were in abbreviated Chichewa (MAK standing for chimanga cha makolo or "local" maize; CCA for CCA; and -F for the plots with "feteliza" or fertilizer) instead .f the usual scientific shorthand.
When we distributed the seed and fertilizer, along with a regulation cup for applying the fertilizer, we again explained the research design. In addition to giving the farmers the inputs, we also discussed the prices each cost at the government outlet (ADMARC), the units in which they were sold, and the

correct timing for applying fertilizers to maize. When someone commented on the unusual color of the CCA seeds, we explained the reason why they had been coated with a chemical. This turned out to be important because later on, when planting was delayed in one village, one elderly woman ate the "local" maize because she was hungry. She then told us she would have eaten the CCA also if she had not learned from us about the poisonous coating.
The farmers provided the actual labor and management for the trial. Each week each of them was visited by a local research or extension agent, and each month we visit d from Lilongwe. The weekly visits noted the dates of significant operations (plantings, weedings, fertilizer applications) and natural occurrences (rain, army worms, etc.) and the condition of the treatments. The agents also offered advice concerning the trial plots.
Farming is a risky business. The farmers and we suffered through an. early drought that caused all of the farmers in one o+ the villages to have to replant their maize in late December (a month later than the usual planting time), a damaging blow in an area where the rains often stop too early. When the rains came, so did erosion in one village where people were cultivating hillside slopes. Then army wora.: attacked in m-iy areas of Phalombe (and other areas in Malawi), and there were always stalkborers and termites to combat. We had originally intended not to use any pesticides and to expose our plots to the same uncontrolled environment of pests that regular fields confronted, but we changed our minds when the plots were really attacked and distributed insecticides for the farmers to apply.

This is one reason why our final yields are much higher than the
local averages.
Our Monthly visits combined inspecting the trial plots and interviewing the -trial farmers about their reactions with continuing investigations into Phalombe smallholder farming. During each visit the farming systems staff met with the project management to talk about the progress of the trials; this
continual interaction between research and extension is necessary to ensure that extension staff understand the trials and can successfully extend their lessons to local smallholders. In February we conducted a special multiple cropping survey for three days to supplement the data collected in the original farming systems survey and by evaluation. Although the 1981/62 trial had a standard intercropping pattern in all four treatments, intercropping is a variable that needs to be investigated under Phalombe conditions, and the next research trial will probably consider various intercropping mixtures.
In May we harvested the maize (the sunflower only being harvested in JLtne and July). In each village the trial farmers helped us harvest each others' plots so all would have the opportunity to see how the treatments responded in various ecological niches. Since smallholders are more accustomed to
evaluating or measuring volumes than weights, we measured the yield from each treatment both ways, using a scale "or weight and a standard calibrated tin for volume. After harvesting all of the plots we called another meeting in each village to
discuss our perceptions of the different treatments.
At both meetings we began by reminding everyone of the original purpose of our presence and the trials, and then we

began to compare our evaluations. Here was the beginning of the fourth step in our four step process evaluating the trial and deciding whether we could make any recommendations and/or whether we needed to plan again another set of trials.
The yields that were harvested from the plots of the eight farmers in one village and the six in the other (two farmers were omitted due to problems with erosion and poor germination) are presented in Table 4. The statistics represent metric tons per hectare (M.T./ha.) of usable grain. Usable grain was
defined by the smallholders themselves as they worked with us to shell and weigh the harvest. They eliminated all rotten grain and that which was very badly eaten away by weevils. As has been noted by the Crop Storage Research Section (Annual Report 1979/80), smallholder criteria for defining usable and unusable grain differ from the criteria used by laboratory technicians, who discard all grain that has been attacked at all by insects.
We asked the trial farmers to share with us their ideas about which variety grew best and about the value of fertilizer. After they gave us their impressions, we reciprocated by telling them our preliminary assessmer, based on c comparison of the means for the treatments and the differences among smallholders. For clarity, we will start here with a review of the actual yields we recorded and our interpretaion of those yields before discussing what the smallholders said in the meetings.

Usable Grain in Metric Tons per Hectare (M.T./ha.)
1 2 3 4 5 6 7 8 Mean Local Maize (LM) 2.2 2.2 1.9 1.2 1.3 0.9 1.0 0.5 1.4 Fert. Local (LM-F) 3.6 3.7 4.3 3.2 2.3 2.3 3.1 2.8 3.2
CCA Maize (CCA) 3.5 2.0 2.9 0.4 0.6 0.5 0.6 0.3 1.3
Fert. CCA (CCA-F) 5.0 4.7 4.3 3.5 2.4 1.7 3.0 2.8 3.4
Mean for Farmer 3.6 3.2 3.3 2.1 1.7 1.3 1.9 1.6 2.3
1 2 3 4 5 6 Mean Local Maize (LM) 1.8 1.1 1.6 1.0 1.6 0.6 1.3 Fert. Local (LM-F) 3.2 2.5 2.9 1.2 1.9 0.8 2.1 CCA Maize 2.2 0.7 0.9 0.3 1.1 0.3 0.9
Fert. CCA (CCA-F) 2.9 2.5 2.1 1.1 0.8 0.4 1.6
Mean for Farmer 2.5 1.7 1.9 0.9 1.4 0.5 1.5

Tabls 4 shows obvious differences: 1) between the means for villages, 2) among the means for farmers in each village, and 3) between the means for the two fertilizer levels in the first village. The difference between the two villages is easily explainable by reference to rainfall and the time the maize was planted. Both villages planted in late November, but the second village did not receive enough rain to sustain the maize and had to replant in late December after they received their first real planting rains. The first village received enough rain in late November and early December, so their maize had a month
headstart. The second village also suffered severe attacks by army worms in January, but this was less of a factor than the rain and time of planting. Yields in the first village, therefore, show how the treatments responded to better conditions. Smallholders anticipate planting in late November, s prhaps the first village conditions mipht be considered more normal. Yields in the second village reflect the adverse conditions that continually threaten Phalombe agriculture.
We were testing two varieties under two levels of fertilizer. What is very clear in Table 4 is that both maizes responded strongly to fertilizer. Both 'local" and CCA more than doubled their yields in the first village under better conditions, although the effer of fertilizer was not as marked in the second village with its generally lower performance. What is also apparent is that there was little difference between the two maizes at either level of fertilizer in the first village under better conditions, and "local" performed somewhat better in the second village under poorer conditions.

These conclusions were supported when we later ran analyses of variance on the yield data. The analyses apportioned the total variance among the treatments, the blocks (farmers), and the individual values (random/error): the higher ,the mean square the greater the variance attributed to that factor. In the
analysis of the combined villages we used the villages as blocks and the farmers as replications within each block. The F ratio expresses the difference between the mean squares of that factor and of error: the higher the ratio the more unlike are the two populations. The significance statistic expresses the
probability that such a difference was not caused by chance.
Table 5 shows that the fertilizer effect was highly significant (>99%) in the first and in the combined villages, and it was significant in the second village. The variable of maize type was never significant. The differences among farmers were significant in each village, and the difference between the two villages (expl-nnnd :o,' by reference to rainfall and time of planting) was highly significant. Differences among farmers will be addressed after first discussing the maize type versus
-ertilizer issue.
The importance of fertilizer to maize yields is obviously supported by our data, but the insignificant relationship between maize type and yield needs to be examined more closely. This set of relationship. may also be studied in the yield data collected by the Evaluation Unit of Blantyre ADD during the 1980/81 National Sample Survey of Agriculture. They sampled many smallholder maize plots in Phalombe during that cropping

Farmers (2) 7 3.947 2.41 94%
Fertilizer (2) 1 25.740 15.74 >99%.
Maize Type (2) 1 0.428 0.26 Insignificant
Fort x Maize 1 Insignificant
Error 21 1.635
Farmers (6), 5 2.049 3.53 97%
Fertilizer (2) 1 3.450 5.94 97%
Maize Type (2) 1 1.000 1.72 Insignificant
Fert x Maize 1 0.010 Insignificant
Error 15 0.581
Villages (2) 1 11.550 10.23 >99%
Fertilizer (2) 1 25.515 275.93 >99%.
Maize Type (2) 1 0.026 Insignificant
Vill x Fort 1 3.676 3.45 93%
Vill x Maize 1 1.403 1.3-2 Insignificant
Fert x Maize 1 Insignificant
Error .. 49 1.066
a) In the analyses for the individual villages,the smallholders are used as blocks, and there is only one replication of each treatment per block. In the analysis of the combined villages, the villages are used as blocks, and there are eight replications (farmers) in the first village/block and six in the second.

season, although there were very few fields (only 12 in their sample) with anything other than "local" maize. All of those 12 were UCA, another composite similar to CCA. Their data show a significant difference in mean vield between "local" and UCA, but that difference appears to be largely a matter of
differential fertilizer application rates (Table 6).
(kilos/ha.) (M.T./ha.) (M.T./ha.)
None 0.8 431 1.8 for 12 plots
1-49 1.6 5 throughout the
50-99 1.3 21 Phalombe Project
100--149 1.3 22
150-199 1.4 19 1.5 when plots
200)-249 1.9 6 in EPA 8 (area
250-299 2.0 4 with highest
300-349 1.9 6 rainfall) are
350+ 1.2 6 excluded
a) Data provided by Evaluation Unit, Blantyre ADD.
When comparisons are made between the two maize types at similar levels of fertilizer, the differences in yield are
greatly diminished. Recommended levels of fertilizer for composites (UCA and CCA) are three bags (150 kilos) per acre or 7.5 bags (375 kilos) per hectare. "Local" maize yields equal the mean for UCA once 200 kilos are applied (four bags per

hectare or 1.6 per acre).
Ano other important aspect that we were studying was yield stability. Farmers want higher yields of the staple, but they also want a more secure or stable production Stability of the two maize types may be measured to some extent from our data (Table 4) by looking at the range of yields, i.e., more stability may be defined as reduced- variability. The coeffficients "+ variation (c v.) in Table 7 measure the extent to which the individual yield values deviate from the mean. This statistic :is corrected frr the magnitude'of the different means (c.v.= standard deviation divided by mean), so all of the Q.v. values in Table 7 are directly comparable: the higher the c.'v. the more variable and unstable. As may be seen, 'local' maize is more stable than CCA in each village and under both fertilizer conditions. The inverse of stability may be responsiveness; CCA is more responsive to its environment, whether adverse or favorable. Both mazes are more stable when .f. ertilized (except f+r "local in the second village), a feature most noticeable in the first village under better growing condi tions.
Treatment c.v. Treatment c.v. Treatment c.v.
Local Maize 45 Local Maize .35 Local Maize 39
:ert. Local 22 Fert. Local 45 Fert. Local 35
CCA Maize 98 CCA Maize 78 CCA Maize 88
Fert. CCA 34 Fert. CCA 3 Fert. CCA 52

The smal1 hc-l ders are concerned about both yi el d and stability, and both improved with fertilizer-. Unfortunately, fertilizer is a costly input, and the lack of capital and fear o.f credit inhi:ibit people from acquiring fertilizer. For this reason we carefully examined the costs and benefits of applying fertilizer in our trial.
Fertilizer c-ost-s per hectare total K61.25 (61.25 Malawian Kwacha) at the recommended rate of one bag (50 kg) of 20:20:0 and two bags (100 kg) of S/A per 0.4 hectare (one acre), using the 1981/82 ADMARC selling prices of K8.50 per bag of 20:20:0 and K8.00 per bag of S/A. Seed costs per hectare for CCA are K6.25 at the recommended rate of one 10 kg bag per 0.4 hectares and the 1981/82 selling price of K2.50 per bag. ADMARC is bIDyiOg maize in 1982 at KO. 11 (11 tambala) per kg or KI1 per 0.1 metric tons (M. T.). Thus, a yield increase of 0.6 M.T. is. more than enough to offset the cost of fertilizer (0.1 M.T. will pay for CCA seed costs for a hectare).
'Irable 8 shows that fertilizer is profitable to apply to both types of maize in both villages, although there is very little average profit in the second village. Examining individual cases, each of the eight farmers in the first village and 50% of those in the second would have made a profit by fertilizing and selling the extra yield. Unfortunately, 50% in Ohe second village would have lost money.
This brings us back again to the variability among smallholders, which was significant in both villages. Graph 1 illustrates this by plotting the mean yields (combining both
varieties) of all fourteen farmers.

PROFITABILITY OF FERTILIZER APPLICATION TO MAIZE (M.T./ha.) Each Metric Ton is Worth 110 Malawi Kwacha at 1982 Prices
Local 1.8 + 1.2 0.8 +0.2
CCA 2.1 +1.5 0.7 0. 1
The Graphed Values Represent the Means for Combined Maize Types
First Village 1-8 Second Village A-F
4.5 1,3
4.0 2
3.5 4
Yield -. 0 1 7,8,9 A
in 2.5 3 5 B,C
M.T./ha. 2.0 2 A 6
1.5 C,E E
1.0 4,5,7 B D
0.5 6,8 D,F F
Three smallholders in the first village (numbers 1-3) achieved yields of 3.5 M.T./ha. or better on all of their
fert:i I i zed plo .ts (and >4.5 on two: plots) and generally demonstrated the advantage of CCA over "l local" in high-yielding

situations. The other five farmers in the same village (4-8) harvested 0.6 M.T./ha. or less on their plots of unfertilized CCA, severely depressing the mean for that treatment, and their plots showed no differences between maizes or, with the lowest yields, advantage of "local" stability.
Why the gap between 1-3 and 4-8 ? We are unable to answer that question satisfactorily because our research design focused on the treatments rather than the farmers (blocks). Probably part of the answer lies in differential residual fertility of the trial fields. Fertilized tobacco had been planted the previous year in the fi-elds of two farmers (1,2), but this relationship has not yet been tested.
Another part of the answer is associated with the relationship among tobacco growers, larger holdings, and higher yields. Four farmers (1-3,8) in the first village are tobacco growers (none of them women), and tobacco is only grown in this area by those with enough land to alienate some from food crop production. Of the other four farmers in the village, one was sick a lot; another experienced serious domestic problems (influencing labor availability); and a third did not weed on schedule.
The range of variability in the second village was not as great, perhaps because the delayed planting depressed the multiplier effects of residual fertility and differential management. Farmer A may have achieved his higher yields because of his slightly different ecological niche (more rainfall ), but the more interesting question raised in this village by Graph I is why two farmers (E.F.) showed no response to fertilizer, and another only a minimal response.

Smallholders A-C showed impressive yield increases with the application of fertilizer (as did 1-8 in the first village), but D-F did not, an anomaly for which we do not as yet have an an swer.
The pre::eding is our evaluation o. the maize yield results. Another important contribution was made by the smallholders themselves in our summary meetings after the harvest. In some ways their* comments supported our own thoughts, but :in other ways they differ-ed.
Smallholders in the .first village agreed that fertilizer was an important input, but they all complained about the cost and the difficulty of getting credit for fertilizer. One of the farmers with less land complained that she wanted to grow a cash crop or a crop on credit, but all of the packages were for an acre. An acre was all the land she had, and putting it into the credit crop meant taking it away from staple food production, which she cculd not afford to do. When we mentioned the
po.s: lb ili.:ty of sharing a credit package with a frimi.nd (one way to lower the cost) the uni versal, rep 1 y was that "You cannot
trust a friend." They all wanted to be individually responsible
-for what they received on credit, and not be jointly responsible with someone else.
The .: arme's in the .first village surprised us, however, with their comments about CCA maize and ADMARC. When we asked them to compare the two maizes, we expected them to say they saw little difference. The most successful farmers (1-3) replied fi rst, and they said that CCA was better than "local" at both .levels. of fertilizer. In fact, as was mentioned earlier,. CCA

d ibtLe for. those 4:4 arrier s withi th-e h-igch Est yielids. he l ess S.- Lt C: C : U 1 f a l me' r- -B) w whc h acd a c.t U a 11 Y e x pe e i e d n o
di *f f erence bte ric types or" hiad -et t er ucss t
j. o: ia I var-I: al I I Ysu port ed t rie f toc sp ea k a nd n ot ed .ho w mu Ch be t ter- C was. Th--eir o~nly cdissent was-, -to mention the
1 increase Luscepti bility to weev,,il damage cof the composites (IJCA more than CCA)
'The menition of weevil attack.,s star-ted a series c)f comments. V1 Il. Aqers cr 3 tiCi Zed -the compotiites +or- their- vulr nerabi .i ty 'to wevi (il attacks and compi al nedi about tI-i& delay in opening ADMVIRC mi-ar kets where -the smallhol dec-s coul cid sell their mai ze. Maize is hiir as ted in 1- pI-i 1 b bu-t th.-e AE)MARFC markets did not open for several mocre months. By the time the markets had opened, stor-ed C m pos.i tees w ere a I r-e ady badly weevi l edi; and -this inhiibited p c.,opl e gr-owing c ompcs i t es f or sale. The delayed opening of
~;ocrnentbuyngpoits bviusl des act -to -the disadvantage
o i, ccal+far mer s. We noticed many far-mers selling ma .z~e -to Pr 1:i va t-e tri ad er* s at. pr -i cePs toel I be-l ow teADMIARC price. When lUeSti oned about t he sales, the .farmers poiyited out 'that -they need ed moi--ney th en and could niot it.t -for- severa,.-l months. Other comment, a bouLAt PIDMRC re-flIec ted what appears -to the senior au..thor- o-f th i s paper -toc represent a uni versal suspi ci on by f a rmTIe rs o ~f whoever- mark ets their *prOduce. Far-msi -.voied susp i c i onei about be~i rig cheated and shortchanged because they colId niot see the scale where their produce was being weighed. I': a 1C-i ric th e sc:a I e in a place wh-ere it Could be seen seems a
..niIstep, baut itI would r-educe far-mer diseconten.

Basec:i o n the results of this -first year' s research, we may mak e so Cme qual i .I i e d r ecoc:)mmendai.: T 0ns 1 hey must be qualif ied
-ec:ause they onl y represent one c."rop pi ng sea son and part of Phal ombe (t'. o vi Illages in two of: the six EPAs. The vil ages di ffy ered in rainfall, but they are I ocated in the central secti. : o' of Phalombe, not in the wettest nor/ the driest areas, and each cropp i ng season conf ronts smali holders wi th new ri sks.
1. We recommend that the project make credit available for fertilizer to be used on "local" maize.
2. We recommend that the project permit smallholders to get credit in units smaller than 0.4 hectares (one acre).
3. We recommend that research and extension (the project) examine the utility of multiple cropping for smallholders with limited land resources before making blanket recommendations for monocropp ing.
4.' We recommend that ADMARC consider opening its buying markets in Phalombe earlier (perhaps late April) and moving its
scales so. that farmers can see what, their produce weighs.
5. We recommend that on-farm farmier-managed comparisons of co:mpos:ites and locala" maize s be conducted elsewhere to check their performance under other management condi ti ons. The farmers who manage the trials should represent a spectrum of resources and experience.
6. We recommend that research investigate the effects of pest damage on maize yields and the effectiveness of variable
_cst control measures by conducting on-f arm farmer--managed
Other recommendations were made earlier as a result of the initial diagnostic survey.

The 198]./82 trial examined the relationship o.f maize type, +ertilizer-, .a' d yield. Multiple cropping was held constant, and f armer management was not crit i cal l y examined. Next year' s trials need to examine:
1. maize responses to varying levels of fertilizer (many fa.armers only apply one bag per acre/field) since farmers need ir'tcrmati on about the costs and benefits of minimal increments;
2.. maize yielc:Is under conditions of monocropping and
1 it. ercr opp :i ng;
:. various multiple cropping mixtures in terms of the maize yield and provision, of cash income (maize yield being h:ig her pr iority); and
4. farmer management practices and their effects on yield.
The first three may be easily handled using the same format as we used in 1981/82, i.e., on-farm farmer-managed trials with a few tr eatments. The fourth factor will be more difficult to test, and we have not yet determined the best format.
The farming systems approach to small Iholder research
differs from other research programs ii, r-,any ways. This paper describes our involvement in one locality, and those who are familiar with agricultural research will be able to note the differences in technique. We stress a prior understanding of local smallholder systems, an integrated team of production and soc i economic c sc i enti sts and development staff, conti nual invoivemenit of smallholders in all stages of re4earchn. on-farm

tria tha: are managed by small hol ders, and sel acting smalholders who represent the majority to test our proposed al ternat ives.
Clark, Barbara 1975 The Work Done by Rural Women in Malawi. in Easter-n Afric::an Journal of Rural Development 8:2:80-91.
Col1 inson, Michael 1979 Understanding Smali Farmers.
Col linson, Michael 1980 A Farming Systems Contribution to Improved Rele vancy in Agr icul tural Research: Concepts and Procedures and Their Promotion by CIMMYT in Eastern Africa.
Crop Storage Research Section, Department of Agricultural Research 1980 Annual Report 1979/80.
Evaluation Unit, Blantyre Agricultural Development Division 1981 Worl.k in n Papers.
Evan-s, Jan is 1981 Report on Extension-Oriented Research with Women in Phalombe.
Gij1bert,. E.H., 1980 Farming Systems Research: A Critical Survey.
Hansen At 1981 Farmin ng Systems Research: Theory and Pr'"actice in Malawi.
Hansen, Art, 1981 Farming Systems of Alachua County, Flori da: Ahn Overview with Special Attention to Low Resource
Far menu
Hi ldebrand, Peter E. 1979 Summary of the Sondeo Methodology Used by ICTIA.
Malawi Population Census 1977: Final Report.
National Sample Survey of Agriculture 1968/69.
"echni cal Advisory Commi ttee, The Consultative Group on Internati onal Agricultural Resea-ch 1978 Farming Systems Research at the International Agr:icultural Resear.. Centers.