Understanding a Simple Experimental Design (SIN 401)

Look down each column of a trials table. Most columns repeat the same number trial after trial: same volume, same pH, same time. One column changes, stepping through a range of values. That changing column is the variable being tested; the frozen columns are controls, held fixed so they cannot muddy the result. The design is boring on purpose. If two things changed at once, a difference in outcome could be blamed on either one, and the experiment would answer nothing. Boring and uninterpretable is the trade a careless design makes; boring and clean is what a good one looks like.

More trials does not fix a design that changes two things. A test with twelve trials that quietly swaps growth medium halfway through is weaker than a test with four trials that only ever changes light intensity, because the twelve-trial version has two candidate causes for any result, and no way to tell them apart. Trial count measures effort, not logic. The question to ask is always: how many columns moved, not how many rows exist.

Repeating a trial is a different move entirely. Running each condition once gives a single number with no sense of how trustworthy it is. Running each condition several times and averaging gives both a mean and a sense of scatter, how much a repeated measurement wobbles on its own. That scatter is the yardstick against which a real difference between conditions gets judged. This sits right next to the idea of comparing two experiments: one axis is whether the design isolates one variable, the other is whether it has been repeated enough to trust the numbers it reports.