Edition: University of Chicago Press, 1970 (Buy from Amazon)
Review number: 1101
Of the thousands of books written about science during the twentieth century, The Structure of Scientific Revolutions is the one which has caused the most controversy. It is a famous account of how (in Kuhn's opinion) science actually works, as opposed to the way in which scientists think that they work, and it acts as a summary of Kuhn's ideas; much of the rest of his writing is basically a series of historical case studies which back up and illustrate this summary.
The general view of science is that it proceeds by accretion, new facts gradually allowing theory to become a closer approximation to reality. Kuhn proposed a different model: a generally accepted paradigm determines not just what questions are investigated but how facts are perceived by scientists. However, anomalies appear and grow in importance, until a period of confusion is ended by a revolution which establishes a new paradigm.
The main reason that this is controversial is basically that it seems to make science a matter of opinion rather than fact, especially because of what Kuhn says about the way that the facts change or appear to change when the current paradigm changes. Summaries of Kuhn's arguments tend to make what he says more alarming than it actually is, and I found The Structure of Scientific Revolutions far more convincing than I expected.
Part of the problem with this kind of analysis of science is that the pattern of scientific communication has changed over time, as private letters and public books gave way to refereed journal articles, which in turn are now giving way to the exchange of pre-prints, electronically or otherwise. This is a factor unconsidered by Kuhn - though he was surely aware of it - and in my opinion makes it difficult to compare scientific practice even across relatively short periods as is necessary for this kind of study.
The major idea that Kuhn has which is difficult to agree with is that there is no such thing as scientific progress, that the paradigms are more or less equivalent as explanations of the universe. He explicitly says that there is no rational reason why, in a time of crisis with a current paradigm, one of the competitors to replace it is chosen rather than the others. (The evidence he gives for this is the frequent inability of a new paradigm to predict even some well known facts accurately, as with the century it took for a correct derivation of the moon's orbit from Newton's theory of gravity.) It is, I suppose, an almost inevitable consequence of Kuhn's feeling that a paradigm embodies its own special world view, as then two paradigms would have concepts far enough apart to make them impossible to compare meaningfully. The standard interpretation of issues like the moon's orbit would be that it can take time for a new theory to be properly understood, especially when, like Newton's, it uses unfamiliar mathematics.
However, it seems to me (and to many who have been involved in the practise of scientific research) that there is one very important property of a paradigm which Kuhn minimises: its explanatory power. By pre-supposing that this is unimportant, he anticipates his own conclusions. A scientist's perception of a paradigm is that it should be a model of the phenomena being investigated, and in particular that it should match with investigations already carried out. Exceptions can be made (as in the case of the moon) but they are just that: exceptions. Kuhn objects to this characterisation on the grounds that problems investigated under one paradigm are quite possibly meaningless under another, but I would expect that many would not be; Einstein's theory of general relativity would be useless as a model of the universe under gravitation if it failed to explain the path of a thrown ball on Earth, or the motion of the moon, as well as the Newtonian theory it replaced. (He specifically objects to the common characterisation of the relationship between these two theories that says that Newton's laws approximate to Einstein's when velocities and energies are low, because they make incompatible underlying assumptions, but to say the predictions of the motions of bodies are approximately correct seems to me to be perfectly valid.) Thus, any paradigm which is a candidate to replace one under crisis should explain almost all the well known facts already explained together with some of the more problematic areas. (Another reason for the word "almost", as well as the difficulties in correctly applying new paradigms is that those pushing them may in their enthusiasm make erroneous measurements, as Galileo seems to have done when he reported that pendulum swings take a uniform time even for large displacements even though this is true only when the angle of swing is small.) This more traditional description has the advantage that it preserves the idea of progress towards the goal of describing the universe (not in the more unfortunate cultural sense in which scientific progress is often taken, and which is probably one of the reasons behind Kuhn's rejection of the concept). Another factor which Kuhn doesn't mention which tends to influence the acceptance of a new paradigm (at least when it is mathematical in nature) is the feeling among researchers that it is elegant and beautiful.
Even after forty years, Kuhn's book is thought provoking - more so than reports of his arguments elsewhere. Most readers are unlikely to agree with all his ideas, but it is definitely better to read them here at firsthand.