CHAPTER TWENTY-TWO

How to Run an Honest Trial

You do not have to take my word, or anyone’s, about what works in your own soil. You can find out. A fair garden trial costs almost nothing extra, and it will teach you more than any book. The whole art lies in removing the ways we fool ourselves.

Rule one: compare like with like, at the same time

Never compare this year with last year, or your garden with your neighbor’s. Grow treated and untreated plants in the same season, from the same seed packet, in the same soil, with the same water and the same sun.

Rule two: replicate

One treated plant and one control tells you nothing, because any two plants differ. Use at least six pairs, more if you can. Identical containers filled from one well-mixed batch of soil are ideal, because they remove differences in the ground. If you use beds, divide them into small matched plots.

Rule three: let chance decide

Do not choose which plants get the treatment. You will, without meaning to, pick the better ones or the better spots. For each pair, flip a coin: heads, the left pot is treated; tails, the right. This single step is the difference between an experiment and an anecdote. It was introduced into agriculture in the 1920s at Rothamsted, the very station where electroculture was being tested, and it came too late to be used on most of those trials.

Rule four: use a sham

Give the control plants a dummy that looks and behaves as much like the treatment as possible but lacks the active part. For a copper coil, the control gets an identical dowel wound with string or plastic-coated garden wire. For a galvanic pair, the control gets the same two metal plates buried in the same positions with no wire joining them. Now both groups have a stake to shade them, the same disturbance of the soil, the same metals, and the same attention from you.

Rule five: decide beforehand what you will measure

Before you plant, write down the one thing that will decide the question. Total weight of ripe fruit per plant is a good choice. Do not wait to see which measurement favors the treatment.

Rule six: measure blind if you can

Have someone else label the pots A and B and keep the key until harvest is weighed. If you cannot, at least weigh everything, every time, on a scale, and write it down on the spot. Do not judge by eye.

Rule seven: treat them identically otherwise

Same water, measured. Same feeding. Same pruning. Keep treated and control pots mixed together, not in two separate rows.

Reading the result

At harvest, for each pair, note which plant won. If the treatment does nothing, each pair is a coin toss. With six pairs, the treated plant winning all six would happen by luck only about once in sixty-four tries; that would be worth taking seriously and repeating next year. Five out of six happens by luck about one time in ten and proves little. Four of six is nothing.

Then look at the size of the difference. If treated plants average three percent more, that is noise. Twenty percent across all pairs is interesting.

Whatever you find, write it up honestly, including a result of no difference. A careful null result is worth more to other gardeners than a hundred enthusiastic videos, and it is rarer.