Science Puzzle
The Fertiliser That Proved Nothing
A gardener tests a new fertiliser. Pot A gets the fertiliser, a large pot, a sunny windowsill, and daily watering. Pot B gets no fertiliser, a small pot, a shady corner, and weekly watering. Plant A grows twice as tall.
The gardener concludes the fertiliser works. Why can he not draw that conclusion, and what is the single most important fix?
The Answer
Four things differ between the pots: fertiliser, pot size, light, and water. Plant A grew taller, but there is no way to know which of the four did it, or whether it was some combination. The experiment is said to be confounded.
The fix is to hold every variable identical except the one being tested. Same pot size, same light, same watering schedule, same soil, same seed variety, same starting height. Then the only difference between the two plants is the fertiliser, and any difference in growth can be attributed to it.
The vocabulary is worth having precisely. The independent variable is the one thing you deliberately change (the fertiliser). The dependent variable is what you measure (the growth). The controlled variables are everything you deliberately keep the same. A confounding variable is one that changed when it should not have, and which offers a rival explanation for your result.
Repeating a confounded experiment more times, or with more pots, does not rescue it. It simply produces the same uninterpretable result with greater precision. A bad design cannot be fixed by more data.
The principle: Confounding variables. If more than one thing differs between the groups, the result cannot be attributed to any one of them. More repetition does not fix a confounded design.