There is a version of the scientific method that appears in almost every secondary school curriculum in the world. It goes something like this: observe something, form a hypothesis, design an experiment, collect data, draw a conclusion. Clean. Sequential. Reassuring.

It is also, in important ways, a fiction.

Not a harmful fiction exactly — it is a useful scaffold for teaching controlled thinking to twelve-year-olds. But somewhere between the classroom and public discourse, that scaffold became the building. People came to believe that science is a procedure rather than a practice, that it produces proof rather than probability, and that disagreement within it signals failure rather than function. These misunderstandings are not trivial. They have consequences for how policy is made, how medical decisions are reached, and how readily people dismiss inconvenient findings as "just theory."

The textbook model presents science as linear. In practice, it rarely is.

Research begins in many places — sometimes with an anomaly no one can explain, sometimes with data collected for an entirely different purpose, sometimes with a hunch that would embarrass a grant committee if written plainly. Hypotheses are revised mid-experiment. Control conditions are added after the fact when confounds emerge. Papers are written in a sequence that does not reflect the actual order in which the thinking happened. This is not fraud. It is science as it is actually practised.

The philosopher of science Paul Feyerabend argued, controversially but not without evidence, that virtually every significant advance in science has involved some departure from the official methodology of its time. Galileo used rhetoric. Darwin delayed publication to amass converging evidence from multiple non-experimental sources. The discovery of the structure of DNA involved data that had not been shared with its collectors' full consent.

Science's strength is not its adherence to a fixed procedure. It is the self-correcting structure that sits beneath any procedure — the requirement that findings be falsifiable, that methods be transparent enough to be scrutinised, and that conclusions remain provisional until the weight of independent evidence supports them.

Karl Popper's criterion of falsifiability is probably the most cited and least understood concept in the philosophy of science. The common interpretation runs: a claim is scientific if it can, in principle, be proven wrong by evidence.

What people often miss is the "in principle" clause. Falsifiability is a property of the claim, not the experiment. The claim that a drug reduces inflammation is falsifiable. The claim that it will definitely work in every patient is not — and that distinction matters enormously when interpreting clinical trial results.

Falsifiability is also frequently weaponised. Climate change denial, for instance, sometimes proceeds by demanding a single falsifying observation — "if the models were right, we wouldn't have had that cold winter" — while ignoring that probabilistic predictions about complex systems are not falsified by individual data points. This is not scepticism in any philosophically coherent sense. It is the misapplication of a criterion that was designed to separate science from metaphysics, repurposed to manufacture doubt.

In the past fifteen years, systematic attempts to reproduce published findings across psychology, medicine, and economics have returned sobering results. Landmark studies have failed to replicate. Effect sizes have shrunk dramatically upon replication. Some findings that shaped clinical guidelines have not held up under independent scrutiny.

This is frequently characterised as a crisis of credibility. The correct framing is more complicated.

The fact that the replication crisis was identified at all is a function of science working correctly. Researchers designed replication studies, published unflattering results, and the community updated its priors. What was exposed was not that science is broken — it was that a specific set of incentive structures (publication bias toward novel positive results, underpowered studies, and pressure to produce) had distorted the literature in ways that the methodology itself eventually caught.

The error would be to conflate a published study with established scientific knowledge. A single peer-reviewed paper is a contribution to an ongoing conversation, not a verdict. The misunderstanding — the expectation that science speaks in final, authoritative pronouncements — is precisely what makes retracted or unreplicated findings feel like betrayals rather than corrections.

No single linguistic collision has done more damage to public understanding of science than the shared word "theory."

In ordinary usage, a theory is a guess. An idea without much behind it. When someone says they have a theory about why the train was late, they mean a conjecture. When scientists talk about evolutionary theory or germ theory or quantum field theory, they mean something categorically different: a well-tested, internally consistent explanatory framework supported by extensive, independent, converging evidence.

A scientific theory is not a hypothesis that hasn't been proven yet. It is a hypothesis that has been tested so rigorously, from so many directions, that it now forms the scaffolding on which further work is built. Evolution is called a theory not because its status is uncertain, but because it is the most comprehensive available explanation for a vast body of biological evidence. Saying evolution is "just a theory" is a category error — it mistakes the technical vocabulary of science for the casual vocabulary of everyday language.

The confusion is not entirely the public's fault. Scientists have not always been careful to flag the distinction, and science communicators have sometimes compounded the problem by overclaiming certainty in ways that make any subsequent qualification sound like a retreat.

Another common misunderstanding inverts the problem. Where some dismiss science by demanding certainty it cannot offer, others treat scientific consensus as the end of inquiry rather than its current state.

Consensus is meaningful. When virtually every relevant expert, working from independent data, arrives at similar conclusions, that is strong evidence. It is the closest science gets to a reliable signal — and on questions like anthropogenic climate change, vaccine safety, or the age of the universe, the consensus is both clear and well-founded.

But consensus is not immutable, and treating it as such misrepresents how science works. The consensus that ulcers were caused by stress held for decades before Barry Marshall and Robin Warren demonstrated the role of Helicobacter pylori — a finding so contrary to established belief that Marshall famously infected himself to generate the evidence he needed to be taken seriously. The consensus was wrong. Science corrected it.

The lesson is not that consensus should be distrusted. It is that consensus is a probabilistic statement about the current state of evidence, not a claim about permanent truth. Healthy scientific culture holds both of these things simultaneously: defer to the weight of evidence as it currently stands, while remaining genuinely open to revision.

These are not purely academic concerns. The public misunderstanding of how science works has measurable real-world effects.

When people believe science produces certainty, they interpret uncertainty as incompetence — and become vulnerable to actors who offer false certainty instead. When people believe a single study is definitive, they can be swung by any outlier finding given enough media attention. When people treat consensus as ideology rather than evidence, they lose the ability to distinguish legitimate minority scientific positions from manufactured doubt.

The scientific method — properly understood — is a set of commitments designed to manage human fallibility. It accounts for the fact that researchers are biased, that individual experiments are imperfect, that initial findings are often wrong, and that knowledge accumulates through failure as much as success. It is not a guarantee of truth. It is the least bad mechanism humans have developed for getting closer to it.

That distinction is worth knowing. And it is not, currently, what most people are taught.