4.3 Experimentation

After discussing observation as a scientific method, let us now turn to experimentation.

The difference between the two methods can best be explained by an example: Suppose that you want to know as much as possible about the growth of maize. You decide to find out by watching a maize crop. You have a few ideas about what might be important for good maize growth and what might be not so important. So you keep an eye on the things you think are important. This is an example of observation. But suppose that you have a very strong hunch that maize would yield more if it got more water than it usually gets during dry season. How can you find out? You will probably choose a maize field and, during dry season, you will water it, but not the whole field. You will water only certain rows. This will allow you to see immediately whether your initial belief was right because you can compare growth in the rows left dry with growth in the watered rows. Here, then, you have not left the maize field undisturbed. You have introduced a change in part of it. After this you have observed carefully what happened. This is an example of an experiment. Experimentation always introduces a precisely controlled change into one part of the area under study and leaves the other part undisturbed. Conclusions about the effect of the experimental change are reached by comparing what happens in the two parts of the experimental set-up. The bulk of technical innovations have been discovered through careful and systematic experimentation.

An experiment is a way of testing beliefs or assumptions about facts. It is a planned comparison of two or more objects of the same type under different conditions. An experiment involves a number of steps which are all equally important:

- planning;

- setting up (e.g. getting the two or more sets of objects into place and subjecting them to different conditions);

- observing (an experiment goes on for some time, i.e. the sets of objects always remain subjected to the experimental treatment for some time so that the treatment, the different conditions, can take effect. During this time frequent observations are necessary);

- evaluating (at the end of an experiment, the final comparison must be made: How did the two or more sets of objects do? How did they respond to the differences in conditions? ).Experiments can be very short and might be completed in a period of 60 minutes, like many physical or chemical experiments. Or they might cover a period of a week, e.g. germination experiments. But farming experiments will last much longer, usually stretching over a whole farming season, from planting to harvesting. And while setting up the experiment often attracts considerable interest, at harvesting time this interest has declined. It is important, however, to arouse interest again, and lead pupils through a carefully planned evaluation of the experiment. Unless this is done, the whole exercise has little point as far as teaching/learning is concerned.

As we said earlier, experimentation always actively introduces a precisely controlled change into one part of an area under study and leaves the other part undisturbed. The change introduced in part of the experimental setting is called the experimental treatment. Whether the experimental treatment had any effect is assessed by checking whether it made any difference: we compare the part that received the experimental treatment with the part that was left without the experimental treatment.

This, by the way, leads towards the scientific concept of causation. If the objects which have undergone the experimental treatment are now different from the objects which had no treatment, we will claim that the treatment is the cause of the difference. This claim to a causal relation will be accepted unless it can be proven false, usually by more experiments.

This discussion might appear to be too general and abstract. We shall presently try to make things clear through a number of examples.

Let us imagine you believe that during dry season, maize would grow better if it was' watered than if it was left to grow on water provided by rain and dew. This belief is an assumption or an hypothesis. You want to test this hypothesis. Since you don't want to waste anything, you plant two maize seeds at a fair distance from each other. After they have germinated, you water one of them regularly. This is your experimental treatment. The other one is left alone. At harvest time you compare the two maize cobs, each plant having yielded one. You find that the cob from the watered maize plant is much bigger than the one from the other plant. Besides, even before harvest you have noticed that the watered plant grew faster and taller, remained green for a longer time and showed every signs of being a very healthy plant. Since you believe that other conditions import-ant for maize growth were very much the same for the two plants, you relate the difference in yield to the experimental treatment, watering, and conclude that watering during dry season was the cause of the better yield.

Strange though it might look, this example has all the elements of a true and reliable experiment. Let us analyse it in some detail: - Experimentation grows from previous knowledge. The starting point of our example was a belief, a hypothesis about the effect of watering on maize yields. Such a hypothesis could be derived from clever observation, it could be got from books, or it could be part and parcel of the traditional knowledge of a group of people. For instance there is a belief that maize will only do well on ridges. But according to the International Institute of Tropical Agriculture in Nigeria maize can be farmed successfully on the flat. Three reactions are possible: one is the conservative attitude which would dismiss such a claim as not valid in this particular area. The second reaction is the credulous attitude: since an internationally renowned authority made the statement, one ought to believe it and act accordingly. The third reaction is the truly scientific one: Here are two contradictory beliefs. One should try to find out which one is correct. This leads to the formulation of a testable hypothesis which might say: "Maize grown on the flat will do at least as well as maize grown on ridges". This can be tested by a simple experiment, after which it will be possible to decide whether under local conditions maize can be grown on the flat (which would save a lot of work) or whether it is safer to continue with the traditional method of ridging.

The next step is to work out an experimental design. The design of an experiment consists of all the details of the arrangement which will be used to test the hypothesis. It states:

- the objects on which the experiment will be performed (in our case maize plants, but experiments could be done on any other crop, depending on the problem at hand),

- the experimental treatment,

- the method by which other influences are controlled,

- the way in which the effect of the treatment will be assessed.

Let us turn to the parts of an experimental design one by one.

a) Objects

Usually there is a whole set of objects (e.g. maize plants) that will undergo the experimental treatment, and another set that will be used for comparison. What are the reasons for this? It is simply that one plant might be exposed to the influence of chance, good or bad luck, and that such chance influences might be more effective than the experimental treatment. For example:

By chance the plant which is not watered might be growing on an exceptionally fertile spot. By chance the plant which is not watered might develop exceptionally long roots and get water from a depth that the average maize plant cannot reach. By chance the watered plant might be attacked by the stem borer and therefore do badly. By chance, one of the two plants might be destroyed altogether, thus leading to the collapse of the whole experiment. With a bit of imagination you will find more chance factors that can arbitrarily and unpredictably influence an experiment and thus defeat its purpose. The very fact that these chance influences operate in an unpredictable way helps us, however.

If we subject a whole set of objects (say 100 maize plants in 4 rows) to the experimental treatment and take another set of 100 plants without the experimental treatment, then we have the following situation: The experimental treatment will systematically influence all the plants in one set and will be absent from all the plants in the other set. The chance factors can now operate on all the plants used for the experiment, in our case 200. Since they operate unpredictably, unsystematically by definition, we can assume that they will affect the plants on the treated group as well as the plants of the untreated group. In this case, they would have the same total effect on the two groups and therefore cancel each other out.

This is so because we are not comparing individual plant yields but rather the total yield of the treated group with the total yield of the untreated group.

Result of treated group = effect of treatment + combined effect of all chance factors on treated group.

Result of untreated group = combined effect of all chance factors on untreated group.

Result of treatment group - Result of untreated group = effect of treatment

Since we have shown that we can assume the combined effect of an chance factors on the treated group to be the same as the combined effect of all chance factors on the untreated group, the difference between the two results is in fact the effect of the treatment.

To assume that the chance factors influence the two groups in the same way is the more justified the larger the two groups are. Therefore, school farm experiments should not be done on too small a scale.

b) Treatments

Most of the time experiments are identified and named according to the experimental treatment applied. There are very many possible treatments to experiment with, and we shall list a few of them as examples.

- Cocoyams were grown under shade and compared with cocoyams grown under normal conditions, i.e. exposed to full daylight.

- Pineapples were grown with short planting distances in double rows as compared to the usual wide spacing method used on local farms.

- Maize was systematically thinned down to one plant per stand as compared to local farming practice where up to four plants may grow in a stand.

- Yams were grown using chemical fertilizer. Different planting times for maize were experimented with.

- Two methods of weed control (weeding and mulching) were used on maize and compared to no weed control at all.

Any teacher would find it easy to think up similar experiments of his own. This is all the easier since there is a sharp contrast between the "scientific way of farming", prescribed by many syllabuses, and African ways of farming. Just think of all the possible experiments contrasting mixed cropping with single cropping. A few of them have been listed in the volume II. We could end our discussion of treatments here, but there is a danger - the danger of making treatments very complicated. Let us distinguish experiments according to the number of treatments applied at the same time, and the structure of the treatment. The simplest experiments are those where only one treatment is applied, and where we simply contrast the effect of this one treatment with what happens if that treatment is not given. The first four examples above are of this kind.

In the cocoyam experiment the researchers divided their cocoyam farm into two halves with the same number of plants each. One half was shaded with palm fronds much like a coffee nursery or a vegetable nursery, the other was left exposed to the full daylight. We can say that the treatment has only one level. Such experiments with one treatment at only one level are the simplest ones yet embody all the logic of scientific experimentation. Therefore they are particularly suitable for use in school.

The idea of treatment level will become clearer if we look at an experiment with one treatment at multiple levels: The experiment involving different planting times is a case in point. People generally believe that the best time for planting maize is immediately after the first heavy rain in March. Earlier planting will result in poor germination, later planting in loss in yield. In order to test this, planting was started early in March and continued at intervals of one week for six weeks. Here the treatment is difference in planting time, but instead of contrasting late or early planting with planting at the traditionally best time, six different planting times were used, each leading to a block of several rows all planted at the same time. These different planting times are like many treatment levels with the one closest to the "correct" time being the "untreated" one. One could also say that the same treatment is being given at different intensities. In a thinning experiment, one might start by planting four grains in each hole, then have one area where all the plants are allowed to grow, one area where a maximum of three plants are allowed, another area where only two plants at a stand may grow, and a last one where thinning to one plant per stand is done. This would give us a three-level treatment with one untreated group. The fertilizer experiment on yam could be carried out with different amounts of fertilizer applied per hectare.

The advantage of a one-treatment-multiple-level experiment is that it not only tells us whether a given treatment has an effect, but what the most effective form of that treatment under local conditions might be. What would be the best quantity of fertilizer to use on yams? Is it more profitable to thin down to one plant per stand than to two plants? How long can we delay maize planting before the loss in yield or the delay in maturity or both become alarming? This takes us one step further towards finding out experimentally the most advantageous ways of farming.

In order to understand experiments with several treatments at a time we just have to remember that during farming we take not only one but several different measures to ensure good yields. The different operations in farming - tilling, planting, weeding, mulching, staking etc. - can be thought of as different treatments. Each of these farm jobs is thought to be important. Leaving any one out or doing any one poorly might result in a poor harvest. Yet farmers will find it difficult to say which of them is the most important. Can poor weeding offset the effect of careful tilling and ridging? Can intensive watering make up for late planting of a dry season crop? Questions such as these can in principle be answered by experiments with several treatments at the same time. To take a simple example from IPAR-Buea's experimental plot:In traditional farming there are two major methods of weed control. One is weeding itself, the other is heavy mulching. Maize is a crop that is very sensitive to competition from weeds. The effect of weeding is known. The effect of mulch on weeds is that by forming a thick covering over the soil, it prevents or at least delays weed growth at a time when maize is most vulnerable to it, during and shortly after germination.

The questions now are:

- Deciding whether mulching or weeding is the more effective in protecting the maize from competition.

- If a woman weeds her maize properly, would it make an appreciable difference if she also did mulching or could she leave it without losing much?

In order to answer these questions, an experiment was carried out with mulching and weeding as combined treatments. Both treatments were one-level treatments: proper weeding (i.e. two weedings) versus no weeding at all, and heavy mulching versus no mulching. The following blocks of four rows each were laid out:

Block I: No mulching no weeding Block II: No mulching only weeding Block III: only mulching no weeding Block IV: mulching and weeding

Block I is the no-treatment block, block II and III are one-treatment blocks, and block IV has both treatments. You will realise that the blocks show all possible combinations of the two treatments. This is necessary for experiments with more than one treatment, and it makes them very complicated and unwieldy. The number of experimental blocks increases very fast. With one one-level treatment we have 2 blocks, with two one-level treatment we have 2 x 2 = 4 blocks, with three one-level treatments we have 2 x 2 x 2 = 8 blocks, with four one-level treatments we go up to 2 x 2 x 2 x 2 = 24 = 16 blocks, and so on. The number of experimental blocks increases even faster if the treatments have multiple levels. What is more, evaluating the results of the experiments becomes very complicated and tedious, and therefore primary schools are well advised to stick to experiments with two treatments.

c) Controlling for Other Influences One of the control methods has been mentioned already:

- having a large number of objects in the experiment. Some other methods are:
- repeating the same experiment at one or two different places on the school farm,
- repeating it over several years. If the results of repeated experiments remain roughly the same, then observed differences between treatment and nontreatment group, blocks or sets can be assumed to be due to the treatment.

d) Assessing the Results of an Experiment

The first problem here is to know what the results are we are looking for. In school farm experiments it will almost invariably be the harvest. Most of the time it will be the quantity harvested, but even quantity could be expressed in different terms: the weight of the crop or the number of cobs, fruit or tubers. But sometimes we are more interested in quality: the resistance of a crop to disease, the storage quality of the harvest, the taste, the content of certain substances (e.g. protein in maize, poisonous matter in yams or cassava) etc. Whatever the case may be, we have to make sure what effects we are looking for in an experiment, and define exactly how we are going to measure these effects. It does not make sense, for example, if we set up an experiment to determine the effect of sunshine on the sugar content in pineapples, and have no way of measuring the sugar content. Or, to put it at a more elementary level, it does not make sense to set up an experiment to find out the effect of compost manure on the weight of yams if the school does not possess any equipment for weighing. Most farming experiments go on for a long time, covering at least two terms. The effect of an experimental treatment does not appear only and for the first time at the harvest. The effect of watering during dry season should be evident very soon after the treatment starts. In the case of the experiment on weeding and mulching, the bad effects of weeds on the crop became apparent soon after the first weeding of the treatment blocks. Continuous observation of plant development 'on experimental plots yields rich insights into the biology of the plants grown because it sees them under different conditions.

· Experiments grows from previous knowledge, ideas, beliefs (= hypotheses) which have to be tested.

· Experimentations and observations lead towards a scientific concept of causation. Traditional African farming methods can be improved by using such an approach.

· Try by all means to check the influence of chance in your experiments. You will get clearer results in comparing not individual plants but a large number of plants. Another method of control is to rep eat the experiment next year.

· Experiments can be made with more than one variable. The number of experimental factors influences the analysis: it will become more complicated. The simplest experiments are those where only one treatment is applied. Therefore they are particularly suitable for use in schools.

How to Present Experiments to Pupils

It might be difficult to use the scientific vocabulary and reasoning we have used so far when introducing experimental work to pupils. Here is an example of how a farm master tried to tackle the problem. We have merely elaborated on his idea. Here is how the experiment on weedings and mulching might be introduced to a class:

"There are four farmers (or four women) who started a contest to find out who would get the best maize harvest. They chose four equal pieces of land with the same kind of soil, in fact, they got one large plot and divided it into four equal parts. One of them said she would spare no effort in order to win and decided not only to weed at the right time but also to mulch very heavily between the maize. Two others said they had no time to do so much work. So, one decided to use heavy mulch. This would stop the weeds from growing, and therefore, there was no need for weeding. The other one did not believe this and decided to weed but leaves out mulching. And the fourth one - well, she had bad luck and fell ill immediately after doing the planting. Nobody was prepared to do any of the work for her, and so her plot was neither mulched nor weeded. IPAR-Buea has tried to show what the four plots would look like so that anyone who wants to know can find out for himself which of the four ways of caring for one's maize would produce the best harvest".

In Part III there is a sub-unit about an experiment with pineapple farming. It shows how such an experiment can be evaluated.

In the case of pineapple-farming, there is another experiment which works well: the different types of suckers are said to differ in terms of the size of the fruit they produce and the time after planting when they first produce a fruit. The class should collect an equal number of

- stumps or ratoons,
-stem suckers,
- slips,

crowns. These should be planted separately so that there is a number of rows with stumps only, a number of rows with crowns only, etc.