Popper's picture of science is inspiring: bold theories, risky predictions, honest testing, ruthless rejection of refuted hypotheses. Scientists as intellectual heroes, constantly trying to prove themselves wrong.
Thomas Kuhn (1922–1996) read the history of science and found a rather different story.
Kuhn's The Structure of Scientific Revolutions (1962) is among the most influential books of the 20th century, and not just in philosophy. His central observation: scientists almost never operate in Popperian fashion. When an experiment produces a result that conflicts with the dominant theory, scientists' first response is almost never to abandon the theory. It is to question the experiment, look for errors in the apparatus, hypothesize an undetected confounding factor, or simply file the anomaly away as unexplained. Theories are not abandoned when they face recalcitrant data; they are abandoned when a better theory is available.
Kuhn introduced the concept of the paradigm, a framework of assumptions, methods, exemplary problem-solutions, and values that scientists in a field share. The paradigm defines what counts as a legitimate question, what counts as an acceptable answer, and what counts as good science. Scientists working within a paradigm are doing what Kuhn calls normal science: puzzle-solving within the established framework, extending and applying the paradigm rather than questioning its foundations.
The analogy he uses: normal science is like solving crossword puzzles. You assume the puzzle has a solution; you assume your existing vocabulary is adequate; you work within those constraints to fill in the grid. If you can't solve 14 Across, you don't conclude that the puzzle is wrong, you conclude you haven't found the right word yet.
Anomalies, observations that resist explanation within the paradigm, accumulate over time. As long as the paradigm is productive, scientists tolerate anomalies. When anomalies multiply to the point that the paradigm's credibility is severely strained, and when a rival paradigm emerges that handles the anomalies better, a scientific revolution occurs. The old paradigm is replaced by the new one, not through gradual accumulation but through a relatively rapid gestalt switch in the scientific community. The Copernican revolution, the Newtonian revolution, the Einsteinian revolution, the quantum mechanical revolution, each involved not just new theories but new ways of seeing.
Kuhn's most philosophically provocative claim is that competing paradigms are incommensurable, they are not just different answers to the same question, they are partly different ways of framing the questions. Newtonian mass and Einsteinian mass are not quite the same concept. Pre-revolutionary scientists and post-revolutionary scientists are, in a sense, working in different worlds. This is not irrational, Kuhn insists, but it does mean that paradigm choice is not simply a matter of following the evidence, because the paradigms partly determine what counts as evidence.
Popper found this disturbing and resisted it. If paradigm choice is not purely evidence-driven, is science just sociology? Kuhn replied: no, the transition to a new paradigm requires that it solve problems the old one couldn't, and scientists are responsive to this. But the process is messier, more social, and more historically contingent than Popper's rational reconstruction suggested.