What is science? How the scientific method works

Science is a way of finding out how the natural world works by testing ideas against evidence. A scientist proposes an explanation, works out what it predicts, and checks the prediction by observation or experiment. Ideas that survive many tests are kept; ideas that fail are changed or dropped. What science produces is not a list of fixed facts but the best-tested explanations so far, held with confidence that matches the evidence.

In short

• Science is a method more than a body of facts: propose an idea, say what it predicts, test it, and let the result decide.
• A scientific idea must be testable, meaning some possible observation could show it wrong.
• Science corrects itself because other people check, repeat and challenge each result, though that can take years.

What makes something science?

Science covers subjects as different as the structure of atoms, the spread of disease and the evolution of life, but the subjects have one habit in common: claims are judged by evidence that anyone could, in principle, check. A claim earns its place by what it predicts and how well those predictions hold up, not by who made it.

The philosopher Karl Popper put the key test sharply in the 1930s. A theory is scientific only if some possible observation could prove it wrong. As the Stanford Encyclopedia of Philosophy summarizes him, Popper was struck that Einstein's theory had testable implications which, if false, would have falsified it, while the theories he set against it, such as Freud's and Adler's, could be confirmed by almost any outcome. Einstein's theory of gravity made a risky prediction: starlight passing the Sun would bend by a set amount. Eddington's measurements during the 1919 solar eclipse matched it. Had they not, the theory would have been in trouble. Philosophers have since argued over the details of his rule, but its core, that a scientific claim must stick its neck out, is widely shared.

How does the scientific method work?

The scientific method is usually taught as a sequence of steps. One real case shows them better than a list does.

In the 1840s, the Vienna General Hospital had two maternity clinics. In the First Clinic, staffed by doctors and medical students, 11.4% of mothers died in 1846, most of childbed fever. In the Second Clinic, staffed by midwives, 2.8% died. The figures come from the records the doctor Ignaz Semmelweis later published, tabulated year by year.

Step	What it means	What Semmelweis did
Observe	Notice something that needs explaining	Mothers died about four times as often in one clinic as the other
Ask	Turn it into a precise question	What does the First Clinic do that the Second doesn't?
Hypothesize	Propose a testable explanation	After a colleague died of a fever-like illness from a cut during an autopsy, he guessed that doctors carried "cadaverous particles" from autopsies to mothers on their hands
Predict	Say what should happen if it's true	If doctors cleaned their hands in a chlorine solution before seeing patients, deaths in the First Clinic should fall to the Second's level
Test	Check the prediction	In May 1847 he required hand-washing with chlorinated lime
Analyze	Compare the result with the prediction	First Clinic deaths fell to 1.3% in 1848, the same as the Second Clinic
Share	Publish so others can check	Colleagues spread the result in lectures and letters; his own book came only in 1861, and many doctors rejected the idea

The last row matters as much as the others. Semmelweis's idea was resisted for years, partly because he had no mechanism to offer and partly because it implied that doctors were killing their patients. The mechanism arrived later, with the germ theory of disease developed by Louis Pasteur, Robert Koch and others. A result is not settled when one person finds it, but when others can explain it and repeat it.

Real research is messier than any table. The University of California, Berkeley's Understanding Science project calls the real process "complex, iterative" and able to take "many different paths": surprises send work in new directions, and questions are revisited at deeper levels for decades.

Is a scientific theory just a guess?

No, and this is the most common misunderstanding about science. In everyday speech, "theory" means a hunch. In science it means the opposite: a broad explanation, tested many times, that ties many facts together.

Term	What it is	Example
Hypothesis	A proposed, testable explanation for something specific	Hand-washing will reduce deaths from childbed fever
Law	A description of a regular pattern, often as an equation, without saying why	Newton's law of gravity: the pull between two bodies falls with the square of the distance
Theory	A well-tested explanation of why a wide range of things happen	The germ theory of disease; the theory of evolution; general relativity

A law doesn't grow up into a theory; they do different jobs. Evolution by natural selection and the theory of relativity are called theories not because they are uncertain, but because they explain.

How does science correct itself?

No scientist is trusted to check their own work alone. Before a paper is published in a journal, other experts in the field review it in a process called peer review, looking for mistakes, gaps and claims the data don't support. After publication, other teams try to repeat the result, which is called replication. A finding that only one lab can produce is treated with caution.

This is slow and imperfect. In 2015, a group of 270 researchers called the Open Science Collaboration tried to repeat 100 published psychology studies. Of the 97 that had reported a significant effect, only about 36% did so again, and the effects found were on average about half as large, as the Wikipedia account of the replication crisis sets out. The response has been part of the correction: more fields now ask researchers to register their plans before collecting data and to share their data afterwards.

Still argued

How deep the replication problem runs is still debated. Some researchers read the 2015 results as a sign that much of the published literature in some fields is unreliable, driven by pressure to publish surprising results. Others argue that a failed replication can reflect differences in method or setting, and that a single repeat attempt is weak evidence either way. Both sides agree that the problem was found by scientists checking each other, and that results differ a great deal between fields.

What does science cover?

The natural sciences are usually grouped into the physical sciences, such as physics and chemistry, and the life sciences, such as biology. Physics runs from quantum mechanics, the rules for very small things, to gravity and the cosmos. Chemistry organizes matter through the periodic table. Biology explains living things down to DNA. Even an everyday question like why the sky is blue has an answer built the same way: an idea, a prediction and a measurement that matched. The social sciences, such as economics and psychology, use the same methods on human behavior, where controlled experiments are harder.

Science's answers come with stated uncertainty, and good scientists say how sure they are. That is a strength, not a weakness. Albert Einstein overturned parts of Newton's physics, yet Newton's laws still send spacecraft to other planets, because a better theory tends to include the old one where the old one worked.

Check yourself

• Without looking back, walk through Semmelweis's case using the steps of the scientific method.
• What would make a claim impossible to test, and why does that put it outside science?
• What is the difference between a scientific law and a scientific theory?

Sources

• The real process of science — Understanding Science, University of California Museum of Paleontology
• Scientific Method — Stanford Encyclopedia of Philosophy
• Karl Popper — Stephen Thornton, Stanford Encyclopedia of Philosophy
• Historical mortality rates of puerperal fever — Wikipedia, from Semmelweis, 1861
• Replication crisis — Wikipedia