I. Stages in the History of a Science
1. Pre-paradigm period. In this period one starts from ground zero and attempts to build a science from scratch. Because there is no paradigm to organize the data, all facts seem equally relevant. Science consists of simple data collection with no real organizing principle.
The acquisition of a paradigm transforms pre-paradigm science into:
2. Normal science. Normal science consists primarily of puzzle solving: of extending our knowledge of facts highlighted as important by the paradigm, increasing the match between the facts and the paradigm's predictions, and further development and articulation of the paradigm. Scientists doing normal science do not work to refute or overthrow a paradigm, or even to find out whether it is true: they presuppose that it is true, and work on that assumption.
The emergence of anomalies pushes normal science into:
3. Period of Crisis. A few anomalies -- cases in which the observational facts do not match up with what our paradigm has led us to expect -- can always be explained away. (The experiment was badly performed, the beakers weren't washed well enough, there must be another planet we haven't found yet, . . . ) But as they accumulate, a sense grows that something is fundamentally wrong. But nothing much can be done about this until . . .
The emergence of a new paradigm makes possible:
4. Revolution. One paradigm is replaced by another believed to me more adequate. The change from one paradigm to another is not dictated by the observational data in any straightforward way. Both paradigms will have ways of accommodating the data, and proponents of the different paradigms may have different interpretations of the criteria for theory choice, so that theory A looks simpler (or more coherent with existing theory, etc.) to proponents of theory A, while theory B looks simpler to proponents of theory B. Moreover, to some extent proponents of differing paradigms have difficulty even communicating with each other, because they will use the same terms to mean different things.
Kuhn stresses a notion he calls incommensurability. This gets applied, I believe, in a number of different areas but with a common pattern. Here are some of them:
1. No neutral language. This is the most basic sense in which Kuhn uses the notion of incommensurability. The idea is that different paradigms, even if they use the same vocabulary, will use it in different ways, so that scientists committed to the differing paradigms will tend to "talk through" each other. The theoretical justification here seems to be that any aspect of a theory can affect the meanings of its terms -- there is no distinction between "analytic" and "synthetic" sentences, between sentences which merely give the meanings of terms and sentences which state facts about the world. So there is no way to give neutral definitions of words shared by different theories, definitions both theories can accept. And so it is extremely difficult (Kuhn doesn't actually say "impossible") for proponents of one paradigm to even figure out what proponents of another are really trying to say.
2. No neutral observations. Observation is "theory-laden": what we observe depends to some extent on our theoretical commitments. Our theories provide the categories in terms of which we classify our observations, and thus to some extent affect what we see. The positivist ideal of theory choice was a situation in which two competing theories made conflicting observational predictions, a "crucial experiment" was performed, and one theory won the day while the other was refuted. On Kuhn's view, things are rarely this simple; often different theories will handle different sets of observations, and even where in some sense they overlap they may not agree in their interpretation of what is observed. (The duck-rabbit drawing is helpful in getting a feel for what Kuhn has in mind here: two people can look at exactly the same drawing and still in some sense see entirely different things.)
3. No neutral criteria for theory choice. In The Structure of Scientific Revolutions, especially chapter IX, Kuhn appears to suggest that each "paradigm" carries with it a set of evaluative criteria on which it scores well, so that there are no neutral criteria that will decide which theory is best. "In learning a paradigm the scientist acquires theory, methods, and standards together, usually in an inextricable mixture. . . . each paradigm will be shown to satisfy more or less the criteria that it dictates for itself and to fall short of those dictated by its opponent" (pp. 109-110). In later writing, notably "Values, Objectivity, and Theory Choice," Kuhn takes what seems to be a more moderate view (though he claims that this is what he meant all along), holding that there are general criteria for theory choice on which nearly everyone can agree -- things like simplicity, scope, coherence with existing theory, etc. But he also argues that proponents of different theories may well interpret these criteria differently.
4. No neutral world. This is the most radical of the claims Kuhn makes. He suggests that scientists committed to different paradigms in a certain sense "live in different worlds." His view here is a nuanced one; he does not deny that there is a real world which is not changed by changes in our theories or paradigms, but nevertheless insists that the world we experience and live in is changed when our theories change. (For instance, he argues that until the medieval period, there were no pendulums, but only swinging objects.)
C. Kuhn's metaphysical views
The fourth claim listed above, that there is no neutral world, appears to involve a commitment to some sort of metaphysical anti-realism about the empirical world, combined with an acknowledgment that there is also a real world that is not changed by our changing theories of it. Kuhn's view here is really very Kantian (except for the view that there can be different worlds for different paradigms; for Kant there is only one human "paradigm" and so only one empirical world). He shares with Kant the idea that the really real, independently existing world (for Kant, the "thing-in-itself") is completely unknowable, and that the empirical world, which is knowable, is partly constructed by our categories and concepts. (It is hard to maintain together the view that there is a thing-in-itself and the view that we cannot know anything at all about it. Just as the philosophers who followed Kant tended either to be realists who argued that we can know the real nature of things, or idealists who rejected the idea that there is a thing-in-itself, so post-Kuhnian philosophers of science tend to be either straightforward realists who think that science gives us real knowledge of the world, or anti-realists such as social constructionists who seem to reject the idea that there is a mind-independent world at all.)