A reading companion to Thomas Kuhn’s 1962 book The Structure of Scientific Revolutions: what it argues, how it is built, and the discovery buried under “paradigm shift”.
Everyone knows what The Structure of Scientific Revolutions is about. It is about paradigm shifts: long stretches of settled science, then a sudden overthrow — Copernicus, Lavoisier, Einstein — after which the old way of seeing is gone. The phrase escaped the book within a decade and has been applied to almost everything since. It is one of the few pieces of philosophy of science a general reader has heard of.
It is also not the book’s discovery, and it has hidden the one that is.
Kuhn was a physicist who became a historian and philosopher of science, and the book grew out of teaching the history of science to non-scientists and finding that the past looked nothing like the textbooks. What he found in the archive was not, first, a theory of revolutions. It was an observation about how scientists know what they know. They run on a competence they cannot state. They agree about what counts as a good problem and a good solution, they teach that agreement to students, and they cannot write it down. When they try, they disagree, and the disagreement doesn’t matter. When they have to write it down — and there are such moments — something has already gone wrong.
That observation is this essay’s subject. The revolutions are the drama; the mechanism is in a word Kuhn used so variously that one reader counted twenty-one senses of it,1 and which the public kept in precisely the sense he came to think least important. The word is paradigm, and what it originally meant was an example.
This is a companion, not a substitute. Structure is short but not to be read in one sitting, and not every chapter earns equal attention. What follows is a floor plan, the ideas to watch for, and a suggestion about where the book stops a step short of what it found.
The shape of the book
Thirteen short chapters and, from 1970, a Postscript that matters as much as any of them.2 Four movements:
§I–§IV set up settled science. §I is the manifesto: history, properly read, could change our image of science. §II describes how a field acquires a paradigm — the passage from competing schools to a shared achievement. §III–§IV describe life inside one, which Kuhn calls normal science, and characterise its work as puzzle-solving.
§V stands alone and is the hinge. Its title, “The Priority of Paradigms”, is its claim: shared examples come first, and any rules are abstracted from them afterward, incompletely, and only when needed.
§VI–§IX describe how settled science comes apart: anomaly (§VI), crisis (§VII), the community’s response (§VIII), and why resolution must be a revolution rather than an adjustment (§IX).
§X–§XIII are the famous and contested part. §X argues that a revolution changes the scientist’s world, not just the theory. §XI observes that revolutions are then made invisible by textbooks. §XII asks how revolutions are settled — not by proof. §XIII asks whether science progresses — yes, but not toward anything.
The Postscript answers seven years of criticism, and above all splits the overworked word paradigm in two and says which half is deeper.
If you read four chapters, read §II, §V, §VI, and the Postscript. A fifth: §XI, short and sly.
Ideas worth carrying away
1. “Paradigm” means three things, and the one that got lost is the one that matters
The Greek paradeigma is a model — the word grammarians used for the sample noun you decline and then imitate. You are not given the rule for the second declension; you are given dominus, domini, domino and told to do the rest like this. That instruction — do it like this — is the one I have taken for this essay’s title, because on Kuhn’s account it is how a science is taught and how it holds together. The book pulls the word two other ways first.
The first sense a reader meets (§II) is paradigm as achievement: work striking enough to draw a lasting group away from rival approaches, and open-ended enough to leave them plenty to do — Ptolemy’s Almagest, Newton’s Principia, Lavoisier’s Chemistry. Both conditions are required; a brilliant result that closed its subject would be a finished question, not a paradigm.3
The second, which the book slides into unannounced, is paradigm as world-view: everything a mature science shares — equations, models of what things are made of, values about what a good theory is, and the examples that show how all of it is applied. The Postscript names this the disciplinary matrix and concedes that most of what the first edition called a paradigm was this.
The third is the one Kuhn came to think was the point. Inside the matrix sit the exemplars: concrete worked problem-solutions — the pendulum, the inclined plane, the two-body orbit — that students solve in problem sets and that working scientists use, silently, to recognise a new problem as one they know how to handle. This is paradeigma in the grammarian’s sense, and the Postscript calls it the deeper of the two.4 Exemplar is Kuhn’s word; in what follows I use it and the plainer example interchangeably.
The popular reception kept the second sense and lost the third. A “paradigm shift” is a change of world-view — dramatic, total, faintly mystical. What the phrase omits is how a world-view is held: by shared examples nobody has to state. The drama survived transmission; the mechanism did not — an irony I will come back to.
2. Competence is carried by examples, not rules — and scientists recognise a solution long before they can say what makes it one
§V is the chapter to underline. Scientists in a mature field agree remarkably about which problems are worth attacking, which methods are legitimate, and what a solution looks like. Ask for the rules behind that agreement and something odd happens: they cannot state them, or they state different ones — and the disagreement makes no difference to the work. Newton’s followers divided bitterly over whether gravity was a real force acting across empty space or a placeholder for some mechanism yet to be found. Everyone computed the orbits the same way.
Kuhn concludes that rules are not what scientists share. They share examples. Rules, where they exist, are abstracted from the examples — later, partially, and only when someone has reason to do the abstracting. Textbooks write the examples down, but what the student learns from them, how to see the next problem as like this one, is on no page; it comes from the problem sets. He invokes Wittgenstein’s observation that “games” share no single defining feature, only overlapping resemblances, and Polanyi’s tacit knowledge, the competence of the expert who can do what she cannot explain.5 The chapter’s sharpest line: scientists can agree that someone has produced “an apparently permanent solution to a group of outstanding problems” while disagreeing — sometimes, Kuhn adds, without being aware of it — about what makes it permanent. Recognition outruns explanation, and normal science runs on the recognition.
Newton is the purest case. By the standards of the science it displaced, the Principia was a scandal: no mechanism for gravity, and an author who refused to supply one. The mechanical philosophers, for whom explaining meant showing what pushed, called it a relapse into occult qualities. The community took it up anyway, because it delivered the tides, the comets, and the Moon. The achievement was recognised in the numbers; the account of what a force is was left to others and never converged.
Kuhn’s own cases are mostly older, but a later one makes the point as well. Heisenberg built quantum mechanics from arrays of numbers without knowing they were matrices; Schrödinger built a rival with a wholly different picture; within months the two were shown equivalent; and the community adopted the machinery while the argument about what it meant was demoted to philosophy. It is unresolved today. A working physicist need hold no view on it, and most don’t. What they hold are the hydrogen atom, the harmonic oscillator, the scattering problem: exemplars.6
3. A science is a community, and it polices by competence, not creed
Kuhn’s test for whether a field has a paradigm is sociological. Look at the literature. Before a paradigm, practitioners write books, addressed to anyone, because there is no shared foundation to presuppose. After one, they write articles — short, addressed to colleagues, presupposing everything — and become, as Kuhn notes without regret, unintelligible to the layman.7 The four-page paper works only because the examples carry everything else. A scientific paper is mostly what it leaves out.
This gives the community a peculiar shape. You join by passing the problem sets, not by professing anything. Inside, you may hold whatever interpretation you like, provided your work resembles the exemplars in the ways that count. What expels you is not heresy but incompetence: work that does not look like physics is not argued with; it is dismissed as not physics.
Readers have found this uncomfortable — Popper called normal science a danger, and the word cult is never far off — and Kuhn half-agreed. His reply was that a degree of dogmatism is the precondition of discovery, for reasons we will meet in §VI.8 What distinguishes a paradigm from a cult is not the strength of commitment but that the commitment is tested by work rather than belief, and the work answers to something outside the community, which keeps producing results the community did not want. A cult that ran experiments precise enough to embarrass its own doctrine would not stay a cult.
4. Normal science is puzzle-solving, and when the puzzle fails it is the scientist’s fault
§IV gives normal science its least flattering and most accurate description. Most of what a scientist does is solve puzzles: problems whose solution is assured in kind, so the only question is how, and who first. The paradigm supplies the terms and the criteria. It also supplies the boundary: a problem too far from the shared examples is not a puzzle, because nobody can see what an answer would look like, and it is not attempted. A paradigm can thereby insulate a community from problems that matter to everyone else.
Now watch where blame goes when a puzzle resists. It goes to the scientist. Kuhn makes the point with a proverb: “it is a poor carpenter who blames his tools.”9 The paradigm is not on trial; the practitioner is.
This tells you the exemplars are not mere habits. A habit that fails is a surprise; a standard that fails is a violation, and someone is answerable. Kuhn’s word for what the paradigm lays down is, tellingly, not “rules” but commitments — a layered network of them, from the local (this instrument is reliable) to the nearly universal (the world can be brought under ever more precise order at all).10 Commitments are things one is held to. The community holds its members to them without ever having to say what they are. Notice, though, where Kuhn puts the word. He calls the stated layer commitments; when he describes the examples themselves, he speaks of seeing and recognising, never of being bound. Hold onto that: it is where his account stops.
5. Anomaly is a product of precision, not its absence
If examples are what a science shares, how does it learn what it did not expect? §VI inverts the ordinary picture. We think the open mind discovers and the committed mind resists. Kuhn shows, with oxygen, X-rays, and the Leyden jar, that discovery requires commitment first. An anomaly is a violated expectation; you cannot be surprised unless you had a definite view of what should happen, and you cannot be sure you are surprised unless your instruments are precise enough that the discrepancy is not noise. The paradigm supplies both. The more exact a science’s expectations, the more sensitive an indicator of anomaly it becomes. This is the dogmatism of idea 3 earning its keep. Kuhn’s motto for it is Bacon’s: truth emerges more readily from error than from confusion. The confused inquirer has nothing to be wrong about.11
Two details reward attention. The resistance: novelty emerges only with difficulty; the first response to an anomalous result is to distrust the result. Kelvin’s first verdict on X-rays was that they were a hoax. Kuhn cites an experiment in which subjects shown a red six of spades reported a six of hearts and grew distressed before they saw the card.12 The commitment shapes what is seen at all.
And the assimilation: detecting an anomaly is only the beginning of a discovery. Priestley isolated oxygen and did not discover it, because he took it for ordinary air with the phlogiston removed. Lavoisier discovered it by adjusting the categories until the gas was expected rather than anomalous — at the cost of the whole phlogiston theory. This is why Kuhn insists that fact and theory cannot be separated except inside a settled paradigm: a new fact cannot be taken on board without the theory moving, and when it moves, the old facts are reclassified.13
6. Rule-talk erupts where paradigms crack
Here is the book’s central claim about rules, and this essay turns on it.
Normal science, Kuhn argues in §V, can proceed without stated rules only while the community accepts its examples without question. It follows — he presents this as a prediction — that rules should become important, and the usual unconcern about them should vanish, whenever a paradigm is felt to be insecure. He then observes that debates about method, problems, and standards cluster in exactly the periods his account picks out: before a field has a paradigm, and in the run-up to a revolution. In between, they go quiet.14
§VIII fills in the crisis side with a symptom list: competing articulations proliferate; researchers will try anything; discontent is open; and there is recourse to philosophy and debate over fundamentals. Scientists, Kuhn says, have not usually needed or wanted to be philosophers; it is “particularly in periods of acknowledged crisis” that they turn to philosophical analysis.15
Consider what this means. The rules of a science are not its foundation, laid down first and built on. They are what a practice writes down about itself under duress — when the examples have stopped licensing agreement and the community must argue about what it was doing all along. Explicit method is a symptom.
Kuhn touches the 1905 case in §VII; following Einstein’s career further turns this observation into biography. In 1905 electrodynamics was in exactly §VIII’s condition: the ether experiments had failed, Lorentz had a patch, Poincaré was writing about conventions. Special relativity was not a new calculation — Lorentz’s equations already gave the numbers — but an act of articulation: Einstein found an assumption nobody had stated, that simultaneity is absolute, stated it, and dropped it. That is what crisis calls for, and it made him. Thirty years later he brought the same method to quantum mechanics and was politely ignored, because quantum mechanics was not in crisis but in the most productive normal science physics has known, and philosophical analysis was not wanted.16
7. A crisis ends in a new example, not a rulebook — and the rulebook is forgotten
What happens to the explicit argument once the crisis is over? On Kuhn’s account it does not settle anything and is not what settles things. Revolutions end (§XII) by persuasion, conversion, and the old generation dying out. There is no neutral algorithm, because the standards by which a decision could be made are among the things in dispute. What replaces the old paradigm is a new achievement — striking enough to draw adherents, open enough to give them work. The community retrains on it, and the debate over fundamentals, which had felt like the real business, goes quiet.17
The same two physics cases show it. Physicists did not learn relativity from Einstein’s analysis of simultaneity but from Minkowski’s spacetime geometry of 1908, which Einstein at first dismissed as superfluous erudition — because four-vectors can be practised in problem sets and an argument can only be understood. The interpretation debates of the 1920s produced the Copenhagen interpretation, which most working physicists today neither use nor could state; what they inherited was the exemplars, which carry no interpretation. The explicit phase is a phase. Practice returns to fluency at a new location, and the articulations that got it there are left behind like scaffolding.
Kuhn saw this clearly and did not, I think, see what it implied.
8. The revolution is then rewritten out of the record
§XI is one of the shortest chapters in the book and repays every page. Revolutions are invisible, and the invisibility is manufactured by the textbook. Written after a revolution, from inside the winning paradigm, with the job of making students fluent fast, it tells the field’s history as a straight road ending at the present: earlier scientists appear for whatever looks like a contribution to today’s problems, and the rest — including their actual problems — is dropped. Chemistry texts credit Dalton with answering a question about how elements combine; Dalton was a meteorologist asking something else. Newton credited Galileo with the relation between force and acceleration, which Galileo neither stated nor could have.18
Kuhn calls the distortion probably functional — a textbook that told the real history would fail at its job.19 Near the end of the book he goes further, and concedes that the scientist is, like a character in Orwell’s 1984, “the victim of a history rewritten by the powers that be”; the comparison, he adds, is not wholly unfair, because revolutions bring losses as well as gains and scientists tend not to see the losses.20
Notice what the rewriting removes. It keeps the before (as error or anticipation) and the after (as truth). What it drops is the crisis — the period of explicit argument in between. The phase in which a science said out loud what it was doing is exactly the phase the record deletes. The rules that crisis produced vanish twice: when practice goes tacit again, and when history is rewritten to make the new practice look inevitable.
Here is the irony promised earlier. Kuhn’s book was popularised and fixed in the public mind as “paradigm shift” — a reconstruction that kept the drama, dropped the mechanism, and made the drama look like what he had been saying all along. Structure is a case of Structure.
Where Kuhn stopped
§X is the chapter everyone remembers and the one I find least defensible. It argues that after a revolution scientists work in a different world — not that they see the same world differently, a formulation Kuhn considers and rejects as too weak. The evidence is perceptual analogy: the duck-rabbit, inverting lenses, the anomalous cards. Where the pre-Copernican astronomer saw a planet, the post-Copernican saw a satellite, and there was no neutral thing both were seeing.
The chapter has been read as relativism and defended as not-quite-relativism for sixty years. Kuhn reaches the conclusion through the psychology of perception. I want to suggest why, once there, he could not easily get out: his own findings leave him nowhere else to stand. He has shown that a science’s standards live in its examples, that examples are replaced in a revolution, and that the arguments in between settle nothing by neutral rules. With nothing more than that, “the world changes” is hard to resist. There is no vantage above the examples from which to compare them, because the only thing Kuhn has found above the examples is rules, and he has shown rules to be late, partial, and abandoned.
What is missing is a theory of what the explicit phase does. Kuhn treats crisis-era articulation as heat generated by friction, dispersed once the friction stops. He cannot ask whether an articulation was faithful to the practice it articulated, because on his account the practice is trained perception, a kind of skilled seeing — and skilled seeing is not the sort of thing a statement can be true or false to. The vocabulary of obligation, the commitments of §IV, enters his story only at the level of stated rules. The examples themselves are described as something nearer to reflexes. A reflex cannot be violated, only surprised.
But his own evidence needs more. The blame on the carpenter, the resistance to the anomalous card, the refusal to abandon a paradigm without a replacement — none of these make sense unless the examples carry standards one is answerable to before anyone has stated them. Kuhn found the binding force of standards — what philosophers call their normativity — where it becomes visible, in the stated rules, and mistook the place it becomes visible for the place it begins. That is the gap, and §X is its cost.
This is not a scandal. Kuhn found the pattern before anyone had the account of norms that would explain it. What he could not yet say was why crisis, of all things, should demand explicitness — why a community that never needed to state its standards suddenly cannot proceed without doing so.
Why it still reads
The book has been argued with continuously since it appeared. Incommensurability has been softened, by Kuhn among others; historians have shown that instruments, experiments, and theories have partly separate histories and do not shift as one bloc; philosophers have picked apart the gestalt analogy.21 It has survived them.
What has not been argued away is the observation under the drama. A practice can be precise without being explicit. It can hold its members to standards nobody can state. It detects its own failures only because of the commitments those failures violate. And when forced to say what it has been doing, it says it under pressure, incompletely, and stops saying it as soon as it can.
Which leaves a question the book raises and cannot answer. If rule-talk is a phase — if a science becomes articulate in crisis and goes quiet again on a new example — what, if anything, do the articulate episodes leave behind? Kuhn tracks content: which theory replaced which. On that ledger the crisis contributes nothing lasting; its arguments are abandoned and its history erased. But there is another ledger the book gives no way to see: not what a science believes after a crisis, but what it has become able to say — about evidence, about method, about what would count against it — that it could not say before. Kuhn’s own top-level commitment — which he treats as belonging to science as such, whatever the paradigm — hints that this ledger is not empty.
He did not keep it. That is his unfinished business.
Footnotes
Margaret Masterman, “The Nature of a Paradigm”, in Lakatos and Musgrave (eds.), Criticism and the Growth of Knowledge (1970). Masterman counted not less than twenty-one; the Postscript, citing a sympathetic reader, says at least twenty-two.↩︎
Section numbers (§I–§XIII) are Kuhn’s chapter numbers, which are the same in every edition; the Postscript’s numbered sections are cited as such.↩︎
§II. The two-condition definition is early in the chapter; the list of founding works follows.↩︎
Postscript, section 2 (disciplinary matrix) and section 3, “Paradigms as Shared Examples”. Kuhn’s “Second Thoughts on Paradigms” (1974, in The Essential Tension) gives the cleanest statement.↩︎
§V, second half; developed in Postscript section 4, “Tacit Knowledge and Intuition”, including Kuhn’s reply to the objection that the competence must “really” be rules.↩︎
“Shut up and calculate” is the usual name for this attitude; the phrase is David Mermin’s (1989), often misattributed to Feynman.↩︎
§II, on what happens to a field’s literature once a paradigm is in place.↩︎
“The Essential Tension: Tradition and Innovation in Scientific Research” (1959). Popper’s objection is his essay “Normal Science and its Dangers”, in Criticism and the Growth of Knowledge.↩︎
§VIII, in the argument that anomalies are not treated as refutations.↩︎
§IV, the chapter’s last few pages: the network of commitments, its levels, and the top-level commitment without which no one is a scientist.↩︎
Novum Organum, Book II, aphorism 20; Kuhn quotes it in §II and its logic governs §VI.↩︎
§VI. The experiment is Bruner and Postman’s (1949).↩︎
The closing paragraph of §VI.↩︎
§V, near the end; presented as a consequence of the priority of paradigms, then as confirmed by the record.↩︎
§VIII, second half.↩︎
Einstein’s own distinction between “constructive” and “principle” theories (1919) is his name for the kind of work special relativity was.↩︎
§XII for persuasion and generational change; §II for the conditions a new achievement must meet.↩︎
§XI, central pages.↩︎
§XI, near the end, where the depreciation of history is called “probably functionally” ingrained in the profession.↩︎
§XIII, in the discussion of what a revolution loses.↩︎
Peter Galison, Image and Logic (1997), for the partly independent histories of theory, experiment, and instrumentation.↩︎