How science actually decides what is true
The strength of a theory is how easily it could have failed and did not.
School science often arrives as a list of things that are true: forces, cells, elements. That is the output. The interesting part is the method that produced it, because the method is usable on things no textbook covers.
A hypothesis has to be able to be wrong
The central requirement, and the one most often missing from claims that sound scientific, is falsifiability: a hypothesis must predict something that could fail to happen.
- Testable: 'a dropped ball accelerates at 9.8 m/s squared' - measure it, and if it comes out at 4 the claim is dead.
- Not testable: 'an undetectable force guides all events' - nothing could ever contradict it, so no experiment can inform it.
The second sounds like a stronger claim because nothing can touch it. That is exactly what makes it empty. A statement compatible with every possible observation tells you nothing about which observations to expect.
Why controls exist
Suppose a new fertiliser is used on a field and the crop grows better. You have learned almost nothing, because the weather, the soil and the season all changed too.
A control is an identical setup without the one thing you are testing. Two neighbouring fields, same soil, same weather, same seeds, fertiliser on one. Now a difference between them can be attributed to the fertiliser, because it is the only thing that differed.
This is the same discipline as taking one attribute at a time in a reasoning puzzle: change one variable, hold the rest still, and the result means something. Change several and the result means nothing in particular.
Expectation is not dishonesty. A doctor who believes in a treatment genuinely records borderline improvements more favourably. Double-blinding removes the effect without anyone having to be trusted, which is why it became standard.
Correlation is not causation, and the reason matters
Two things moving together allows four explanations, and only one is the one people jump to.
- A causes B
- B causes A
- Something else causes both
- Coincidence
Ice cream sales and drowning deaths rise together. Ice cream does not cause drowning; hot weather causes both. That third option - a common cause - is the one most often missed, and it explains a large share of confident, wrong conclusions in the news.
A claim to test yourself on: 'students who eat breakfast get better grades'. Possible: breakfast helps. Also possible: households organised enough to provide breakfast also help with homework. The correlation is real either way; only a trial that changes breakfast alone can separate them.
Theories are not guesses
In ordinary speech a theory is a hunch. In science a theory is an explanatory framework that has survived extensive testing and organises a large body of evidence - the theory of gravity, of evolution, of plate tectonics. A hypothesis is the untested proposal.
So 'it is just a theory' inverts the meaning: reaching the status of theory is the highest confidence the method offers. It is still not certainty, because science does not deal in proof the way mathematics does - it deals in explanations that have repeatedly failed to be knocked down.
Why replication is the real test
One experiment is a result. The same result, in another lab, with different people and equipment, is evidence.
This matters because a single surprising finding is more likely to be an error, a fluke or a subtle flaw in method than a discovery. Whole fields have had to revisit results that looked solid until someone tried to reproduce them and could not. Being unimpressed by a single study is not cynicism; it is the correct default.
The part that transfers
Most people will not run experiments. The habits transfer anyway: asking what would have to be true for a claim to be false, noticing when only one variable changed, distinguishing a result from an explanation, and treating one striking anecdote as the start of a question rather than the end of one.