Revolution is the process by which one paradigm replaces another. The essence of Kuhn's argument is that choosing between paradigms is necessarily an unscientific process. Metascience can't be science. In order to make this argument stick, he must discredit an "accumulation" theory of revolution. According to this metatheory, a scientific theory is developed that explains the data available at a given time. New phenomena are discovered that are inconsistent with the old theory, and then a new theory appears that is objectively superior--it resolves the new problems while continuing to solve all the old problems. A proponent of this idea might point to relativistic mechanics, which reduces to Newtonian mechanics under everyday conditions, and so continues to solve whatever problems Newtonian mechanics used to solve.
Kuhn argues that history does not support such a cumulative theory. Revolutions are destructive as well as constructive. Doubt is cast on previous results. Data must be reinterpreted. Things which were previously explained or thought to be explainable are now left unexplained. "After the pre-paradigm period . . . cumulative acquisition of unanticipated novelties proves to be an almost non-existent exception to the rule of scientific development."
Rather than being clear-cut scientific judgments, paradigm debates are typically circular. "To the extent . . . that two scientific schools disagree about what is a problem and what a solution, they will inevitably talk through each other when debating the relative merits of their respective paradigms." He denies the Einstein/Newton example explicitly. He argues that the equations of the two systems may look the same, but what the variables refer to are completely different concepts in one theory as opposed to the other.
During a revolution, it is not just that new facts are joined by old facts and reinterpreted. The "facts" themselves actually change, because the language in which the facts are stated is itself an embodiment of the paradigm. "Surveying the rich experimental literature from which these examples are drawn makes one suspect that something like a paradigm is prerequisite to perception itself. What a man sees depends both upon what he looks at and also upon what his previous visual-conceptual experience has taught him to see."
Kuhn talks a bit about perception experiments and compares the reaction that subjects have w/r/t anomalous playing cards (red spades, for instance) with the reactions that scientists have to unexpected results: They don't see it, and then suddenly they see it everywhere. Uranus, for example, was spotted many times and interpreted as a star or a comet. Then, suddenly, after it was seen to be a planet, all kinds of new objects (asteroids, I think) were discovered by astronomers. "Normal science ultimately leads only to the recognition of anomalies and to crises. And these are terminated, not by deliberation and interpretation, but by a relatively sudden and unstructured event like the gestalt switch." Kuhn imagines a "neutral observation language, perhaps one designed to conform to the retinal imprints that mediate what the scientist sees. Only in one of these ways can we hope to retrieve a realm in which experience is again stable once and for all." But he sees little hope of developing such a language. In the absence of such a thing, the observing scientist sees only what his perceptual assumptions allow him to see. "The alternative is not some `fixed' vision, but vision through another paradigm."
Each generation rewrites the textbooks to make it appear as if all of science has been leading directly to the current paradigm, disguising the fact that "Earlier generations pursued their own problems with their own instruments and their own canons of solutions." Kuhn sums up his theory of scientific progress: "Theories do not evolve piecemeal to fit facts that were there all the time. Rather they emerge together with the facts they fit from a revolutionary reformulation of the preceding scientific tradition."
"What is the process by which a new candidate for paradigm replaces its predecessor?" Kuhn rejects Nagel's probabilistic verification and Popper's falsification. He claims that a paradigm is never judged by itself, but only in comparison to some competing paradigm. He proposes a two step process: anomalies evoke new competing paradigms, and then some verification process chooses between them. He sees four kinds of evidence that might be used for a new paradigm.
Kuhn claims that (4) is actually the most important in the beginning, since the first researchers who pursue a new paradigm necessarily do so in the absence of (1-3).