Conducting Experiments on the Farm

By Brenda Frick, Ph.D.

Farmers conduct a wide range of experiments on their farms every year. Most of these are done for the farmer’s own information and don’t require the “blessings of science”. However, a few simple techniques from scientific methodology can give the results of an experiment more credibility, making it easier to share with scientists, academics and even skeptical neighbours.

The first step of a scientific investigation is to clearly form the research question. Although we often think of science as very objective, this first stage depends on the researcher’s (in this case the farmer’s) point of view. A more specific question is often easier to answer. It is always tempting to ask complex questions, but experiments can quickly become overwhelming unless a serious effort is made to keep it simple. Science is an on-going process. No experiment can include everything. A simple one-comparison test can be very useful.

A control is pivotal in being able to make comparisons. If a field produced a bumper crop when the farmer harrowed, for instance, this does not necessarily mean that harrowing was beneficial. Perhaps it was just that sort of season. To scientifically test the effect of harrowing, the farmer would need to have a portion of the field that was harrowed, and a portion of the field that was not. The portion that was not harrowed is the control. The effect of harrowing is seen by comparing the harrowed area to the control area.

A meaningful control has to be as similar to the treatment as possible in every way except for the treatment. For instance, if the treatment is applied to the centre of the field and the control to the headlands, or the treatment on the knolls and the control in the valleys, the difference between treatment and control might be due to position as well as to the treatment. Landscape position and field history are particularly important to consider when deciding where to place a test.

If something happens in the same way several times, it is more likely to be meaningful. This is the idea behind replication – repeating both the treatment(s) and control several times. A minimum of 3 replicates is needed to do a statistical analysis. The more replicates there are, the easier it is to find statistical differences between treatment and control. Of course, the more replicates there are, the more work is involved. Many scientists use 4 replicates in their field trials. Some recommend 6 replicates for on farm trials.

For each replicate, it is important that treatments are in random order. This is to help avoid unintentionally favouring one treatment over the other. For instance, if one treatment is always downwind, or uphill, or nearer the shelterbelt than the other, this might unfairly influence the comparison. Randomization is simply a matter of making sure that each treatment has an equal chance of being in each field position. One way to randomly assign treatments is to put a piece of paper with the names of the treatments in a hat and draw them out as many times as there are reps.

Scientific experiments on the farm can be a way of testing new ideas or demonstrating interesting innovations. They can be the focus of farm field days. By doing these experiments themselves, farmers can make sure that the research they want is done in their environment and with their constraints. By making sure that each test follows a scientific methodology, with at least a control and some replication, these experiments can form the basis of a credible network of farm based information.

Expanded version of this article (PDF)

Brenda Frick, Ph.D., P.Ag., is the Prairie Coordinator for OACC (the Organic Agriculture Centre of Canada) at the College of Agriculture, University of Saskatchewan. She welcomes your comments at 306-966-4975 or via email at brenda.frick@usask.ca.

This article first appeared in The Western Producer, and is published here on the OACC website with permission.

Posted on the OACC website, June 2006