A week of BIO 100.

Do coffee grounds really help plants grow?

BIO 100 · How scientists test an idea · about 690 words

Say a friend tells you that coffee grounds make garden plants grow faster. Her tomatoes looked great last summer. That is a claim, and it might be true. But one good summer cannot tell you whether the coffee did it, or the rain, or the sunny spot, or a better batch of seeds. Science is largely a set of habits for sorting this out.

The first step is turning the claim into a hypothesis: a specific, testable statement that could turn out to be wrong. "Coffee grounds are good for plants" is too fuzzy to test. A better version is this: radish seedlings grown in soil mixed with used coffee grounds will be taller after three weeks than seedlings grown in plain soil. It names what will change, what will be measured, and what result would count against it. The philosopher Karl Popper argued that this openness to being proven wrong is what makes a claim scientific.

Next come the variables. The independent variable is the thing the experimenter changes on purpose: here, coffee grounds or no coffee grounds. The dependent variable is the thing measured to see whether the change mattered: seedling height. The dependent variable depends on the independent one, if the hypothesis is right.

Many other things could affect height: light, water, temperature, pot size, soil type, seed packet. These are controlled variables, and a fair experiment keeps them the same for every plant. If the coffee pots sit by the window and the plain pots sit in the hallway, any difference in height could be about sunlight, and the experiment can no longer answer its own question.

The plants in plain soil form the control group. They get everything the other plants get except the one change being tested. The coffee plants are the experimental group. Without a control group, you might see that the coffee seedlings reached 8 centimeters and have no idea whether that is good. The control gives you something to compare against.

How many plants do you need? One pot of each is not enough. Seeds vary. One might be a dud, and another might be unusually strong. With a single pot per group, a lucky or unlucky seed could decide the whole result. A larger sample size, say twenty pots per group, lets those random differences even out, so a real effect is easier to see and a fake one is less likely to appear by chance.

Scientists also try to assign subjects to groups at random, for example by flipping a coin for each pot. Otherwise you might accidentally give the healthiest seedlings the coffee. Some studies go further and keep the person doing the measuring from knowing which group each plant is in, so their expectations cannot nudge the numbers. This is called blinding.

Suppose the coffee plants do come out taller on average. That supports the hypothesis, but it does not prove it. Maybe the result only holds for radishes, or for this soil, or it was a fluke. A single experiment, even a well-designed one, is one piece of evidence.

That is why the next step is replication: other people repeat the experiment, ideally in different places with different materials, and see whether they get the same result. A finding that keeps showing up earns trust. One that disappears when others try it was probably a coincidence or a flaw in the first design. In 2005 the medical researcher John Ioannidis argued that small studies of surprising ideas are especially likely to report results that do not hold up.

Scientists publish their methods in detail so others can repeat the work and look for problems. Peer review, where other experts read a study before it is published, catches some errors, but it is not a guarantee. Science corrects itself slowly, through many people checking each other.

None of this requires a lab coat. The same habits work anywhere: make the claim specific, change one thing at a time, keep everything else the same, compare against a control, use enough cases, and wait to see whether the result holds up a second time. Next time someone swears a trick always works, you will know what to ask.

Sources: K. Popper, The Logic of Scientific Discovery (1959); J. P. A. Ioannidis, "Why Most Published Research Findings Are False," PLoS Medicine (2005).