What is the scientific method, and how closely is it followed?

The scientific method is the set of practices researchers use to test ideas against evidence in a way that lets other people check the result. Most of us met it as a numbered list on a classroom wall: observe, question, hypothesise, experiment, conclude. That list is a reasonable summary of the logic. It is a poor description of how research actually proceeds.

The cycle, in its usual form

The conventional account runs roughly as follows. You notice something that needs explaining. You turn it into a precise question. You propose a hypothesis — a candidate explanation. From that hypothesis you derive a prediction: if this explanation is correct, then under these specific conditions we should observe this specific thing. You then arrange a test capable of showing the prediction to be false, gather the results, and decide whether the hypothesis survived.

The crucial word there is false. A test that could only ever confirm your idea tells you nothing. The prediction has to be one that reality is genuinely free to contradict, and the experiment has to be arranged so that a contradiction would be visible rather than explained away. This is the part that most often goes missing when the method is taught as a sequence to be recited.

Why falsifiability does the real work

Consider a claim that can absorb any outcome — one where a positive result confirms it and a negative result is attributed to interference, or bad timing, or the wrong conditions. Such a claim may well be true. It is simply not doing scientific work, because no observation could count against it, and so no observation can count for it either.

A useful hypothesis sticks its neck out. It says what should not happen if it is right. That is why control groups exist: not to be fair to the alternative, but to give the result somewhere to fail. If a treatment group and a control group improve identically, the treatment did not do it, and the design forces you to see that rather than allowing you to attribute the improvement to whatever you were hoping for.

Where the tidy version breaks down

Real research is considerably messier. Hypotheses often arrive after the observations rather than before them, suggested by a pattern someone noticed in data collected for another purpose. Whole fields cannot run experiments at all: astronomers, geologists, epidemiologists and economists mostly observe systems they have no ability to manipulate, and must find their controls in nature or in history rather than in a laboratory.

The order of the steps is not what makes such work scientific. What makes it scientific is that the reasoning is exposed — the data, the method, the assumptions and the limits are all published in enough detail that someone unconvinced can go and check. A geologist cannot rerun the formation of a mountain range, but can state precisely what evidence would count against their account of it.

What happens after the experiment

A finished study is not yet knowledge. It goes first to peer review, where other researchers in the field read the manuscript and judge whether the methods support the conclusions. It is worth being clear about what this does and does not involve: reviewers assess reasoning, not raw honesty, and they do not repeat the experiment. Peer review is a filter for obvious error, not a guarantee of truth, and treating publication as proof is one of the more common misreadings of science coverage.

What actually settles a question is replication. When independent teams, with their own equipment, incentives and blind spots, keep arriving at the same result, the finding becomes hard to dismiss. When they do not — and over the past two decades a great many published results in several fields have failed to reproduce — the original claim weakens regardless of where it appeared or how widely it was reported.

This is the practical lesson for reading about research. A single striking study is a reason to pay attention, not a reason to change your mind. The question worth asking of any new finding is not whether it was published, but whether anyone else has since found the same thing.

It is also why the phrase “scientists have proved” should raise an eyebrow. Proof belongs to mathematics, where conclusions follow from premises with certainty. Empirical science deals in evidence that accumulates, and its strongest conclusions are the ones that have survived the most determined attempts to knock them down — held with confidence, but always in principle open to revision if something better-supported comes along.