Diagramium
🔬 Narrated diagram

The scientific method

From a stubborn question to a result other people can trust — hypotheses, controls, statistics, peer review and the loop that never really closes.

14 steps3 min readNarrated with a studio voice
The scientific methodnoyestry againThe trigger that begins the processA question you cannot shakeA step — work that gets doneFind out who asked it beforeA step — work that gets doneForm a testable hypothesisA step — work that gets doneName the variablesA step — work that gets doneDesign a fair testA step — work that gets doneRun it, and run it againInformation going in or coming out of a stepRecord every measurementA step — work that gets doneAnalyze the dataA branch — the path splits on a conditionDoes the evidence support it?A step — work that gets doneRevise and design a new testA step — work that gets doneState a careful conclusionA form, record or printed outputWrite and publish the methodA step — work that gets donePeer review and replicationA stadium shape that finishes the processKnowledge that holds up
The scientific method — the complete diagram. Press Present to watch it build itself.
Step 1 of 14

A question you cannot shake

Science starts the way a good argument does, with somebody refusing to accept the obvious. Ignaz Semmelweis noticed that mothers died far more often in the doctors' ward than in the midwives' ward, and would not stop asking why.

Step 2 of 14

Find out who asked it before

Before you test anything, find out who has already tested it. Something like three million papers are published every year, and half an hour of reading can save you six months of walking down somebody else's dead end.

Step 3 of 14

Form a testable hypothesis

A hypothesis is not a guess, it is a promise. It has to say what you expect and, more importantly, what result would prove you wrong. If no possible outcome could sink your idea, it is not science yet.

Step 4 of 14

Name the variables

Now name three things. The independent variable is what you deliberately change, the dependent variable is what you measure, and the controlled variables are everything you stubbornly keep identical, from the room temperature to the time of day.

Step 5 of 14

Design a fair test

A fair test needs a control group that gets nothing, so you have something to compare against. Medicine goes further and hides which group is which from both patient and doctor, because expectation alone can shift a result by as much as a third.

Step 6 of 14

Run it, and run it again

Run it more than once, because a single trial can simply be luck. Decide the number of runs before you start, not after. That is exactly why every clinical trial must now register its full plan in public before the first patient joins.

Step 7 of 14

Record every measurement

Write down everything, including the runs that went wrong. Lab notebooks are dated and kept for years, because the messy little detail you skipped is usually the exact reason nobody else can reproduce your result.

Step 8 of 14

Analyze the data

Now the statistics. You are really asking whether the difference you see is bigger than the noise. The usual bar is a one in twenty chance that luck alone could throw up a result this strong, which is a far weaker guarantee than most people assume.

Step 9 of 14

Does the evidence support it?

Here is the fork, and it is where the ego gets tested. Does the evidence support what you predicted, or are you quietly rewriting the prediction to fit the numbers you happened to get? Physicists distrust themselves so much that they often hide the real answer from their own team until the analysis is locked.

Step 10 of 14

Revise and design a new test

If the answer is no, that is not failure, it is information. Penicillin, the microwave oven and the echo of the Big Bang were all found by people whose experiments went sideways and who paid attention anyway.

Step 11 of 14

State a careful conclusion

If the evidence holds, you may state a conclusion, but only about the thing you actually tested. Twenty mice in one laboratory is not proof about humans, and an honest scientist says exactly that inside the paper.

Step 12 of 14

Write and publish the method

Writing it up means giving away your method in enough detail that a stranger on another continent could rebuild the experiment. That recipe, not the headline finding, is the real gift a paper makes to everyone else.

Step 13 of 14

Peer review and replication

Then other scientists try to knock it down. Reviewers pick at the method and rivals attempt to repeat it. When one large project retested one hundred psychology studies, fewer than half came out the same way.

Step 14 of 14

Knowledge that holds up

Nothing here is ever finished. A scientific theory is simply an idea that has survived every serious attempt to kill it so far, and that survival, rather than certainty, is what makes the whole thing worth trusting.

Watch it explain itself

Every step above is narrated aloud. Play it, or open it in the editor and make it yours — no account needed.