Science is often held up as our most successful way of gaining knowledge about the world. But what makes it successful? Philosophers of science have offered competing answers, and understanding them helps explain both why science deserves trust and why individual scientific claims can be wrong.

Induction and its problem

Science seems to generalize from observations: every piece of copper tested conducts electricity, so all copper does. This is induction. David Hume pointed out that no amount of past observation logically guarantees the future. Our expectation that nature will continue as before cannot be proved without assuming the very thing in question. This problem of induction has never been fully solved.

Popper and falsification

Karl Popper (1902 to 1994) offered a striking response in The Logic of Scientific Discovery (first published in German in 1934). Science, he argued, does not prove theories true by piling up confirmations. No number of white swans proves that all swans are white, but one black swan proves it false. Science advances by proposing bold conjectures and trying hard to falsify them.

This gave Popper a test for what counts as scientific: a theory is scientific only if it is falsifiable, meaning it rules out some possible observations. Einstein's general relativity predicted that starlight passing near the sun would bend by a specific amount, a prediction checked during the 1919 solar eclipse. It could have failed. Popper contrasted this with theories he thought could explain any outcome at all, such as some versions of psychoanalysis, which in his view made them unscientific rather than strong.

Criticisms: scientists rarely abandon a theory at the first failed prediction, and often they are right not to. The physicist Pierre Duhem and later the philosopher W.V.O. Quine stressed that theories are always tested together with many auxiliary assumptions (about instruments, conditions and so on), so a failed prediction does not show which part is wrong.

Kuhn and paradigms

Thomas Kuhn (1922 to 1996), in The Structure of Scientific Revolutions (1962), studied how science actually develops. Most of the time, he argued, scientists do normal science: solving puzzles within an accepted paradigm, a shared framework of theories, methods and model examples. Results that do not fit, called anomalies, are often set aside. When anomalies pile up, a crisis may lead to a scientific revolution in which a new paradigm replaces the old, as when the sun-centered model of the solar system replaced the earth-centered one.

Kuhn argued that no simple rule settles the choice between paradigms. Some readers took this to mean science is mere fashion or power. Kuhn rejected that reading, holding that later theories are generally better at solving problems, even if science does not progress in a straight line.

Replication and self-correction

Individual studies can be wrong through chance, error, bias or fraud. Science's strength lies less in any single result than in its social process: publishing methods openly, peer review, independent replication, and competition among researchers trying to prove each other wrong.

That process has had real failures. In the 2010s, a large effort to repeat 100 published psychology studies found that well under half produced significant results again. This replication crisis, which reached other fields too, was traced to small samples, flexible analysis and pressure to publish striking findings. Reforms followed, including registering study plans in advance, sharing data and valuing replications. Many see this as the self-correcting process working, if slowly.

Lessons

Science produces our most reliable knowledge, but it is fallible and provisional. A well-replicated finding backed by many lines of evidence deserves far more confidence than a single striking study.

Worked example: continental drift

In 1912 the German scientist Alfred Wegener proposed that the continents had once been joined and had slowly drifted apart. His evidence was impressive: the coastlines of South America and Africa fit together like puzzle pieces, matching fossils and rock formations appeared on continents now separated by oceans, and traces of ancient glaciers turned up in places that are now tropical.

Yet most geologists rejected the idea for decades. Their main objection was reasonable: Wegener could not explain what force could move whole continents, and the mechanisms he suggested did not work. The existing framework explained a great deal, and the anomalies were set aside or explained in other ways, such as land bridges that later sank.

In the 1950s and 1960s, new evidence arrived, much of it from the ocean floor. Researchers found that rock on either side of mid-ocean ridges carried matching stripes of magnetic orientation, recording reversals of the earth's magnetic field. This showed new seafloor forming at the ridges and spreading outward. The theory of plate tectonics emerged, explaining both Wegener's evidence and the mechanism he lacked, and within about a decade it became the accepted framework of geology.

The story fits Kuhn's picture: anomalies resisted, then a revolution. It also fits Popper's: the new theory made risky predictions about the seafloor that could have failed and did not. And it shows that the earlier skepticism was not foolish; it was a demand for a mechanism. When the evidence came, the field changed its mind.