Proceedings of the Second NAHWOA Workshop

[Previous]    [Index]    [Next]

Agricultural systems research

Lockeretz, W.¹ and Boehncke, E.²

¹School of Nutrition Science and Policy, Tufts University, Medford, MA 02155, USA; Voice: 1 617 627 5264; Fax: 1 617 627 3887;

²Division of Organic Animal Husbandry, University of Kassel, Nordbahnhofstrasse 1a, D-37213, Germany.


Just what do we mean by "systems", and how can we understand them?

The term "agricultural system" has become commonplace in research circles these days, especially in "alternative" agriculture circles. The term is used with widely varying senses, without the differences always being recognised. Thus there is a real risk that efforts to understand agricultural systems will be led astray by confusion over language. Even worse, there is a real risk that, because the term is so fashionable, it will be overused to the point of becoming a trivial vogue word instead of an exciting challenge for agricultural scientists. We would do well, therefore, always to consider the range of possibilities covered by the term, and to state clearly what kind of system we are working with.

At the very least, a system, agricultural or otherwise, must have more than one part. Unfortunately, the term "systems approach" sometimes means little more than studying several things at once. But if we have any kind of real system, presumably those parts interact with each other. How strongly they interact determines what kind of system we are talking about, and how we should study it. This involves two points: how many factors must be included (both those imposed on purpose by the farmer, and those imposed by the external environment), and how much of a difference it makes when these factors work together, compared to simply adding up what each does by itself.

The components of an animal enterprise, such as feeding, veterinary care and housing, might affect each other only weakly. That is, there might be little difference between how the individual components behave in isolation, versus how they function together. Studies of such systems are common throughout agricultural research. The system might be represented by a mathematical model that relates one or several outcome variables to a few independent or driving variables, which mainly enter the model separately. The model may also have interaction terms combining two or more independent variables; the important point, though, is that in a weakly interacting system, the model’s predictions do not change much when these interactions are included. Such models are commonly used with regression analysis, analysis of variance, and many other familiar techniques for analysing empirical data.

Such models often are interesting and informative, but here the "systems approach" does not involve anything fundamentally different. More challenging are systems where the interactions among the components are so strong that it is pointless to talk about the effects of individual variables. In their mathematical form, such models may look like the kind described above. However, a key difference is that they are much more difficult to interpret because the important dependence is not on the individual variables, but on complicated functions of several variables at once.

Still, such systems at least are amenable to mathematical modelling. But the systems approach really starts to get interesting when the components interact so strongly that new phenomena emerge. The system’s components may be affected so strongly by being part of the system that they change their nature. But a cow is a cow, one might say. Not so. A cow on a dairy farm is a different animal than she would have been if – somehow – she had grown up separate from her herd mates, the farmer, the milking facilities, and so forth. Another way that new phenomena emerge is when a system has a great many interacting components. Even if the interactions among them are only weak, the system’s behaviour may be entirely different from how the components interact a few at a time. To deal with such behaviour, an entirely new scientific discipline must be developed.

Familiar examples abound outside agriculture. No matter how thoroughly we study individual nerve cells, we will not understand the enormous number of cells that make up a nervous system. Hence neurology, not just cell biology. No matter how thoroughly we study the nervous system, we will not understand the behaviour of a human being. Hence psychology, not just neurology. And no matter how thoroughly we study the behaviour of individual human beings, we will not understand how they behave as members of human societies. Indeed, a human being who had never been part of a human society would hardly be a human being in the usual sense of the term. Hence sociology and anthropology, not just psychology.

What about in agriculture? Although we use the term casually, we do not really know what level of system behaviour is exhibited by a production enterprise, farm, or agricultural landscape. We would do well to learn, because then we would know how to study the system appropriately. Otherwise, we may be either overestimating or underestimating the task. Suppose the components of a farm interact only weakly. We would be wasting time if we assumed that nothing less than a thoroughly "holistic" approach was needed (to use one of the overblown terms that often come up in this connection). Rather, the best approach would be to start with a simple model – even at the danger of being dismissed as "reductionist". We then could gradually introduce as much complexity as is needed. On the other hand, if a farm truly is like an organism – this is the origin of the term "organic farming" – then the approach just described would be missing too much. An organism can never be understood just by accumulating more information on its components, whether its cells, tissues and organs, or its animals, crops, soils, production processes, and people.


Why is systems research particularly suited to organic agriculture?

A systems approach to agricultural research is discussed particularly often regarding organic farming. The connection is sometimes asserted as a statement of fact – organic farming research does take a systems approach - and sometimes as a statement of what should be. Neither statement has been proven, however, only asserted.

Still, it is plausible for a systems approach to be particularly suited to organic farming. Why? The answer is that organic farming typically intervenes less heavy-handedly to alter the plant’s or animal’s growing conditions. By definition, all agricultural production involves some intervention, so that agroecosystems necessarily differ from natural ecosystems. But various degrees of intervention are possible. In organic farming, the idea is to intervene just enough to get the job done, whereas a conventional system may create an entirely artificial environment that bears no relationship whatever to nature. Granted, even a traditional pig production facility is unnatural. But it is not nearly as unnatural as a total confinement building that closely regulates environmental conditions such as temperature, ventilation, and light, and severely restricts the animals’ movements.

Technology also enables different parts of a production system to be separated, whereas in organic and many traditional kinds of agriculture they occur on the same farm. For example, inorganic fertilisation enables feed production to occur far from the animals who consume that feed and whose manure will fertilise the next feed crop. Producing hogs, and only hogs, in a confined building is a great simplification compared to a farm that has hogs, ruminants, feed crops, pastures, and a barnyard.

In organic farming, therefore, more variables come into play. This is true both within an individual enterprise, such as hog production, and in how that enterprise is connected to the rest of the farm. This makes it even less realistic to analyse the effects of one aspect of one enterprise on an organic farm, such as the composition of the hog ration. There is much more going on, and some level of systems approach is called for. Indeed, that statement is already a platitude in organic farming circles. The more challenging task is to learn what level of systems analysis is needed.


How do we do it? A way to start

How to do systems research is a very difficult question. Much more has been written about why we should use a systems approach in organic farming research than about how to do it. Nor do we have convincing examples from which we might infer an answer; far more research papers claim to offer a systems-level analysis than actually do.

Later we discuss the long-range answer to this question. A good starting strategy, though, would be to do appropriate studies of a variety of actual farms that have attempted to operate as true systems. This, by itself, would not tell us generally how to do systems research: the fact that the producer conceives a farm as a system doesn’t mean we automatically know how to study it as a system. Still, such farms seem like the best place to start.

We offer as an example an organic farm whose production system has been developed over some ten years. It functions very well, and has several notable features:

This outstanding performance invites several broad questions:

How can we go about answering these questions, and who should do it? Besides a team leader responsible for integrating the separate pieces of this effort, we would probably want an animal scientist, an ethologist, a veterinarian, an animal nutritionist, a soil scientist, and a marketing specialist. But it is not just a question of assembling the right disciplinary backgrounds. Very important, too, is the kind of minds the members possess. They must not just contribute their own disciplines’ perspectives to the larger effort. They also must recognise that their contribution is just one piece of something bigger, and they must be able to step outside their disciplines when a disciplinary background becomes a hindrance rather an asset.

Their first task is to observe and record the pigs’ behaviour under the farm’s free-range conditions. We already know that the pigs show a complete range of behaviour, in contrast to confined conditions. Specific behavioural and animal health questions to be pursued might include:

Several questions about the soil and land also arise:

Finally, there are questions regarding the consumer aspects of the system:

These questions should first be studied with the farm operated exactly as before. Eventually, though, one might change individual factors in the system to see what happens.


Achieving true systems research: a long-range goal

Studies of the kind just described are just the raw materials out of which a systems approach to research can eventually emerge. The real goal is much more ambitious. We will still need to have system-level understanding routinely permeate all agricultural research. We will need to incorporate it even into more specific studies, not just the exceptionally intensive and far-ranging kind just described.

How can we achieve such a goal?

It is much easier to say how not to achieve it. It will not be achieved by rejecting out of hand the considerable amount of information and experience that has accumulated about conventional systems using conventional research methods; that such research is "reductionist" does not make it invalid. It may be flattering to think that one is working in a whole new domain – or "paradigm". But at best it is wasteful, and at worst stupid, to be so obsessed with seeing the "big picture" that one doesn’t capitalise on what is already known, even granting its limitations.

As for a positive answer to the question: we don’t have one yet. The challenge to do systems research has been offered and accepted, but still not met. Someday, researchers should no longer have to say, "How does one do systems research?" Rather, their question should be, "How can we best put that concept to work in this particular research?"

To be able to provide the answer will require a cumulative effort. No single researcher will do it. Progress can come from two sources. First, some people should make it their primary goal to answer the question, by thinking long and hard about all the research that has already attempted to take a systems approach, in the hope of synthesising it into a higher level of wisdom and understanding. Second, researchers who have particular research questions they must answer – the majority – also can contribute. Granted, they cannot afford to devote themselves primarily to methodological or theoretical issues. Still, they can do their specific research with an eye toward also contributing to the development of systems research methods. Admittedly this requires a trade-off between one’s immediate research goals and a longer-range and vaguer goal, but there are ways to deal with this.

One way is to record more experimental information than might seem relevant at the moment. Generally, we do not know the boundaries of the system we are studying, and cannot be sure in advance which variables matter. This doesn’t mean that we should indiscriminately throw everything into the mix. It does mean, however, that we should seriously think about what variables are likely to be important at the next level of system analysis. We then should measure them so that we, or other researchers, can return to the data when we have a better understanding of the system we were studying. Including additional variables does not necessarily mean including them as "treatment" variables, i.e. as additional variables that are varied in a controlled way to see their effects. It could also mean simply recording more experimental conditions than are routinely recorded for that kind of experiment. Granted, either of these takes effort, and may mean that fewer (but more thorough) experiments can get done with the available resources. But haven’t we all at one time or another said, "If only I had to thought to look at …"?

An example is offered by a recent study of young laying hens in different housing systems (El-lethey, 1999). Compared with a group that had long straw available, the group with no straw engaged in significantly more feather picking, a serious behavioural disorder. At the same time, another researcher measured several physiological variables in both groups. The heterophil:lymphocyte ratio showed that the group that had long straw available was under significantly less stress. Also, this group showed significantly higher antibody titres after sheep red blood cells were used as antigens to measure the immune response. This is a very important result, because it shows that feather pecking was related to stress and immune parameters. It is very fortunate that these measurements were made; otherwise we would only have known that feather picking was related to housing, without knowing anything about the underlying mechanisms.

A second way to let one’s work contribute towards the development of a systems approach to research is to remember that scientific progress lies not in what we already understand, but in what we do not understand. The usual way to present results suggests that many researchers overlook this point. Typically, one announces the "percentage of variance explained by the model", as in a regression analysis, for example. Typically, too, an experiment is regarded as successful if one’s model "explains" at least something. Thus we flag statistically significant regression coefficients, say, even though statistical significance only means that that coefficient probably is not 0, which usually is a trivial finding. Instead of being so proud of how big R2 is, researchers should focus on 1-R2, the part they can’t explain. Too often, the unexplained part of the results is written off as the "random error" caused by factors external to the experiment. But is there really such a clear separation between "relevant" causal factors and "irrelevant" random errors? Whether something is relevant depends on where one draws the boundaries of the system. We will never make progress towards a systems approach if we declare the things we don’t understand to be outside the system. The goal should be to drive 1-R2 down to 0, or at least close!

Finally, one must remember that an individual research project is not self-contained. Its value lies not simply in what the project itself has learned, but also in what it contributes to our cumulative understanding of some phenomenon. This is especially true for developing a true systems approach in agricultural research. An understanding of agricultural systems will emerge, if at all, only from many studies that in some ways repeat but in other ways change previous experimental conditions and procedures. The need to change some things from one experiment to the next is obvious – otherwise, why do it the next time? Less obvious is that experiments must still have enough in common that larger patterns eventually will emerge, not just a confusion of separate, unconnected points.

This means that in designing an experiment, one should not simply ask, "How can this give me the best answer to my specific question under my specific circumstances?" Rather, one should make a point of duplicating some procedures and experimental conditions of earlier experiments, even thought they may be less than optimal for one’s own experiment. That makes it possible to relate one’s results to earlier ones, and thereby greatly increase their meaning.

The notion of studying an agricultural system as a system is very challenging and intriguing. It opens great possibilities for understanding agriculture on a higher level, not just as an immense array of isolated, unrelated specifics. But its great potential is still only potential because the task is extraordinarily difficult. If it were easy, someone would have done it already. Making sense out of agriculture is so difficult because so many things are going on at once, and because they have this annoying way of varying so much from place to place and from year to year.

Yet this is exactly what makes it so important to take a systems approach. We need to be able to see beyond the details of one time and one place. Only then can we take what we have learned under one set of circumstances and apply it to another. Recognising the need is a start, but only a start. The challenge will require a cumulative, long-term attack that engages the best minds from every branch of agricultural science.



El-lethey, H.S. (1999) Stress and feather pecking in laying hens in relation to housing conditions. Master’s thesis, University of   Cairo, Faculty of Veterinary Medicine.


[Previous]    [Index]    [Next]