DEEP READ · 57
Thomas S. Kuhn · 1962
Science is not a building going up brick by brick, with each generation standing a little closer to the truth. It looks more like long stretches of stability punctuated by a few violent changes of regime. Nearly all the time, scientists work inside a shared framework — Kuhn's "paradigm" — doing something much closer to a crossword than to a test: not asking whether the framework is right, but assuming it is and filling the world into it square by square. Only when the things that will not fit pile up, and the best people in the field keep failing to make them fit, does the discipline go into crisis and then swap the whole framework out in a remarkably short time. And the startling part is what the two sides look like afterwards: they are not "right and wrong about the same question." They no longer agree on what counts as a question, what counts as an answer, or even what they are looking at. Which is how Kuhn arrived at the least popular conclusion in the book — science genuinely progresses, but that progress is movement away from, not movement toward.
Thomas Kuhn (1922–1996) was a trained physicist — a Harvard PhD in theoretical physics in 1949 — who got waylaid by the history of science while still a graduate student and never went back. His 1962 book carries a lovely irony: it appeared as a volume in the International Encyclopedia of Unified Science, a series edited by the logical positivists (the movement holding that scientific knowledge reduces to observation plus logic) — and it turned out to be the volume that pulled their foundations out. It has sold well over a million copies, is among the most-cited academic works of the twentieth century, and is the birthplace of "paradigm shift," a phrase since flogged to death in conference keynotes.
Everything in the book circles three claims:
The word has been used so loosely that Kuhn's own meaning is buried under it. The philosopher Margaret Masterman famously counted twenty-one distinct senses of "paradigm" in the book; in his 1969 Postscript Kuhn conceded the fault and narrowed it to two layers.
The first is the disciplinary matrix: everything the practitioners of a mature field hold in common — symbolic formulas (F = ma), metaphysical commitments about what the world is made of, shared standards for what makes an explanation good, and a whole repertoire of instruments and routines. The second — the one Kuhn thought more fundamental — is the exemplar: a concrete problem already solved.
That second layer is the key to him. Think about how you learned physics. The lecture on Newton's laws did not make you able to do the problems; what did was working through the twenty problems at the back of the chapter — inclined plane, pendulum, collision. What you acquired was not a rule you could apply but an ability to see resemblance: faced with a new problem, you recognize that it "goes like the inclined-plane one." A scientific community coheres, on this view, not because its members memorized the same definitions but because they did the same problem sets. Which explains something otherwise odd: scientists agree strikingly well about whether a piece of work is any good while being unable to state the criteria they used. The criteria were never written down. They live in the exemplars.
How it changes what you see: when a field agrees confidently but can't articulate its standard, that's not irrationality. Tacit agreement usually rests on shared worked examples rather than stated rules — and the corollary is practical: to actually enter a field, reading its principles won't do it; you have to work its canonical cases yourself.
Kuhn chose an unflattering and exact phrase: mopping-up operations are what engage most scientists throughout their careers. Normal science does not aim at novelty and — in his cool formulation — when it succeeds, it finds none. Its work is to cash the paradigm's promises: measure the constants more precisely, extend the theory to new materials, solve the equations for harder cases.
He calls this puzzle-solving, and the metaphor rewards being taken literally. A jigsaw has two properties most problems lack: a solution is guaranteed, and the rules are given in advance. Nobody opening a jigsaw box doubts that the pieces fit; if they don't go together, you blame yourself, not the manufacturer. Normal science works the same way: when an experiment fails, the person convicted is the scientist, not the paradigm — "I couldn't measure it" is not a paper, it is a failure.
This looks like a slight on scientists and is in fact Kuhn's deepest move. It is precisely because the rules are treated as immovable that a tiny deviation can be taken seriously. If everyone held the fundamentals open, every anomaly could be shrugged off onto "well, maybe the theory was never quite right," and none of them would ever cost anything. Dogma is what buys precision, and precision is what detects anomaly.
| Normal science | Revolutionary science | |
|---|---|---|
| The work | Cashing the paradigm's promises: precision, extension, cleanup | Replacing the framework itself |
| The rules | Given in advance, not up for debate | The rules are what's being argued about |
| Who's blamed on failure | The scientist (poor technique) | The paradigm (the framework is sick) |
| What success looks like | Expected, only more accurate | Unexpected; the world looks different |
| What persuades | Argument and data | Conversion, and generational turnover |
An anomaly is an observation that should have come out one way and came out another. Kuhn's neglected point is that an anomaly almost never brings down a paradigm by itself. Every paradigm at every moment drags along a heap of unresolved trouble and everyone carries on regardless — because, as he puts it, the scientist who pauses to examine every anomaly he notes will seldom get significant work done.
So when does an anomaly escalate into a crisis? Two episodes from astronomy answer it, and they are nearly the same story with opposite endings. In the early nineteenth century, Uranus refused to follow the orbit Newtonian mechanics predicted. Le Verrier did not doubt Newton; he posited an unseen planet doing the pulling — and in 1846 Neptune was found where he said it would be. Identical treatment of an anomaly, and the result was one of the great triumphs in the history of Newtonian physics. The same Le Verrier then applied the same move to a different discrepancy, the extra 43 arcseconds per century in the precession of Mercury's perihelion, and predicted a planet he called Vulcan. Vulcan was never found. The discrepancy simply sat there until Einstein's general relativity computed it in 1915. Same method, same confidence: once it drove the paradigm deeper in, once it became a nail in its coffin.
The difference is not the size of the anomaly — 43 arcseconds is pitiful — but that it survived the community's best repeated attacks and sat at the paradigm's core rather than its edge. When enough such anomalies accumulate, the symptoms of crisis appear: a proliferation of competing patches, fundamental questions reopened, and public quarrelling about methodology. A field only starts debating its own method when it is ill.
How it changes what you see: this transfers intact to any organization or belief system. A system does not collapse because counterexamples appear; it collapses when the ablest insiders start working on a counterexample and keep failing. Next time you're judging whether a theory, a company, or a practice is really in trouble, don't count the counterexamples — watch whether its senior people have begun arguing about first principles.
For what a revolution feels like from the inside, Kuhn borrows the psychologists' gestalt switch — gestalt meaning the configuration of a whole. The stock example is the duck-rabbit: the same lines can be seen as a duck or as a rabbit; nothing on the page changes, what you see changes completely, you cannot see both at once, and looking harder will not settle which is correct. After Copernicus, an astronomer watching dawn no longer sees the sun coming up; they see the horizon rolling down.
He also cites a real experiment to show how physical this is. Bruner and Postman flashed playing cards at subjects for very short intervals, slipping in anomalous cards — a red six of spades, a black four of hearts. At short exposures nearly everyone identified the red spade without hesitation as a heart or an ordinary spade: they were not misreading it, their existing categories silently corrected it into what should have been there. With longer exposure came hesitation and real discomfort; some subjects reported they could not make out the suit at all — that it did not even look like a card. Longer still, and most suddenly saw it, with no difficulty thereafter. That curve, from invisible through distressing to obvious, is Kuhn's revolution in miniature.
It also explains why the arguments get ugly. The competition between paradigms is not the sort of battle that can be settled by proofs. Kuhn closes with Planck's famous line: a new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it. Usually quoted as cynicism, it is cited here as mechanism: a change of paradigm is, in large part, a change of generation.
The term comes from Greek geometry: the side and diagonal of a square are incommensurable — there is no common unit that measures both a whole number of times (√2 is irrational). Note what that means: no common measure, not "no comparison." Applied to theories it bundles three things.
How it changes what you see: incommensurability is a diagnostic tool. When two people have argued long and moved nowhere, stop asking whose evidence is better and ask whether their key terms mean the same thing, and whether they agree the question is worth answering at all. If those fail, no quantity of data will end it: they are not comparing readings on one ruler, they are fighting over which ruler to use.
Kuhn asks a fine question. If revolutions are this frequent, why does science look so smoothly cumulative — not only to outsiders but to scientists themselves? Because a scientist's main exposure to the history of the field is the textbook, and textbooks are rewritten after every revolution.
The rewriting follows a pattern: predecessors are presented as people who went partway down the road we are on — their questions swapped for our questions, their answers translated into our language, the wrong turnings deleted entirely. The history of science thereby becomes a straight line relayed by geniuses. Kuhn remarks that this continual revision of the past is reminiscent of Orwell's 1984 (paraphrased). He is not accusing anyone of fraud; he is naming a structural consequence: revolutions are invisible because the first thing a victorious revolution does is erase the fact that it was one.
He was woken by this himself. In 1947, a physics graduate student preparing a science course for humanities undergraduates, he read Aristotle's Physics and grew steadily more baffled: how could a mind that acute say such patently absurd things about motion? Then it turned over. Aristotle was not doing bad Newtonian mechanics. His "motion" meant change of every kind — an acorn becoming an oak, a sick man becoming well — with change of place as one small subspecies. Put back into his own questions, the remarks are not absurd; they are rather good. That was Kuhn's own gestalt switch and the seed of the book: past science is not bad science, it is a different self-consistent science.
The last chapter is the hardest to swallow and the deepest. If paradigms are incommensurable and no neutral evidence adjudicates between them, is science progressing at all?
Kuhn's answer: yes — but you have to change your definition of progress. Drop "closer to the truth," because that phrase presumes there is, at the far end, a paradigm-independent account of how the world really is, available for comparison — a position none of us has ever occupied. His replacement comes from Darwin. What was revolutionary in evolution was not that species change but that the change has no goal: no perfected organism waits up ahead. Nobody concludes from this that evolution isn't progress; life has plainly grown more intricate and more differentiated as it moved away from its beginnings.
Science is the same: an evolution from a primitive beginning, not an evolution toward anything. A successor paradigm is judged by how many puzzles it solves, how precisely, how well it predicts, and how many new problems it breeds — all genuine, comparable achievements, none of which needs the metaphysical guarantee of "closer to the truth" to be real. It is the sentence both camps misread: not "science is merely social consensus," and not "science converges on reality," but — it really does progress, and the progress has no destination.
Strung end to end, the book runs like this.
One: observation is not neutral. To collect facts you must first know which facts are worth collecting, and only a framework already in hand can tell you. So a field before its first paradigm is pre-paradigmatic: rival schools talking past each other, everyone re-arguing fundamentals, progress slow (Kuhn thought most social sciences were still here). The first paradigm wins not mainly by being truer but by solving a few big problems handsomely while leaving plenty of work undone.
Two: with a paradigm, efficiency explodes and vision narrows. Fundamentals stop being discussed and detail gets deep. That is normal science, and it buys extreme precision with the assumption that the rules cannot be wrong. Precision is the radar for anomaly.
Three: anomalies are inevitable and usually harmless. Only when one resists the community's strongest patching over a long period, and sits at the paradigm's core, does crisis arrive: patches multiply, fundamentals reopen, methodology becomes a topic. The paradigm is loosening from inside, not being knocked in from outside.
Four: crises end in replacement, not refutation. Here Kuhn takes Popper head on (Karl Popper, for whom falsifiability is the mark of science): a theory is never abandoned for failing to match the facts; it is only ever displaced by another theory. With no substitute on the bench, a counterexample is merely filed as an open problem.
Five: the switch is a conversion, not a calculation. Since standards, meanings and perceptions have all moved, no neutral algorithm can compute the winner. Choice rests on shared values — Kuhn later listed five: accuracy, consistency, scope, simplicity, fruitfulness — but these conflict (a new theory is often broader and more fruitful while temporarily less accurate; the Copernican system was at first neither simpler nor more accurate than Ptolemy's) and different people weight them differently. Which is why the community splits into conservatives and gamblers — not a malfunction but the configuration it needs: someone must squeeze the old paradigm dry, and someone must bet early on the new one.
Six: after victory the history is rewritten, normal science resumes, and the loop runs again. At no point in that chain is the assumption "closer to the truth" doing any work. Remove it and everything still runs.
Misreading 1: "paradigm shift" = any large change. The word's grim fate — a new CEO or a new CRM now counts. Kuhn's version has strict preconditions: a field mature enough to have exactly one paradigm replacing its framework wholesale under internal crisis. By that standard most domains have no paradigm to shift.
Misreading 2: Kuhn said science is irrational and truth is just consensus. He denied this all his life, with visible irritation. "No neutral algorithm for theory choice" is not "anything goes"; he maintained that successor theories are better puzzle-solvers and that the scientific community is the best knowledge-producing arrangement we know of. What he attacked was the myth of an automatic decision procedure, not science. His use as cover by creationists and anti-vaccine campaigners — "yours is just another paradigm" — is the most complete abuse of him there is.
Misreading 3: incommensurable means incomparable, or unable to communicate. Kuhn spent decades clarifying that incommensurability is local, confined to a small cluster of interlinked concepts (late in life he called it taxonomic incommensurability). The two sides can still talk; it just takes learning the other language, and some residue never translates.
The objections with real weight:
Reading only Structure both undersells and distorts him. Three additions.
The Copernican Revolution (1957) is the historical foundation the later book stands on, and it does something counterintuitive: it patiently explains why Ptolemaic astronomy was good science. The geocentric model was respectably accurate, predictive, and consistent with both the physics of the day (heavy things fall toward the centre of the universe) and plain experience (we feel no motion). Copernicus's alternative was initially no more accurate and not really simpler — he used some three dozen circles himself. A revolution is not clever people overturning stupid ones; it is one coherent system displacing another, usually while the newcomer is still behind on points.
The Essential Tension (1977) collects the key essay on theory choice (the source of those five values), and the title is the position: science needs both the stubbornness of tradition and the capacity to break it, and the second grows only out of the first.
Black-Body Theory and the Quantum Discontinuity (1978) is often read as an ironic test of his own thesis. Kuhn argues that Planck in 1900 did not intend to propose that energy comes only in discrete packets; quantization was read into his work later, above all by Einstein. Historians of physics still argue about the verdict, but the posture is pure Kuhn: even "in year X, person Y made breakthrough Z" — the most basic unit of scientific storytelling — is a product of retrospective rewriting.
① The normal state of science is not criticism but finishing work inside a framework nobody questions. Normal science does not aim at novelty and, when it succeeds, finds none — not a jibe, a description.
② The deepest layer of a paradigm is not theory but exemplars: problems already solved. A community coheres because its members did the same problem sets, which is why experts agree on quality while being unable to state the rule.
③ Normal science is puzzle-solving: a solution is guaranteed and the rules are given. So when the experiment fails, the one convicted is the scientist, not the paradigm.
④ And that dogma is the engine: only when everyone believes the rules absolutely does a violation get noticed. Dogma buys precision; precision detects anomaly.
⑤ Anomalies rarely kill on their own — the scientist who pauses to examine every anomaly will seldom get significant work done. Uranus's discrepancy crowned Newtonian mechanics; Mercury's 43 arcseconds helped bury it. The difference isn't size, it's whether it survives the community's best patching.
⑥ A theory is never abandoned for failing to fit the facts; it is only displaced by another theory. With no substitute available, a counterexample is filed as an open problem.
⑦ Revolution is a gestalt switch: not one line changes, and what you see changes entirely. In the Bruner–Postman experiment people saw a red spade as a heart — not misperception, but existing categories quietly correcting the world into what it ought to be.
⑧ Incommensurable comes from the side and diagonal of a square: no common measure — which is not the same as no comparison. Newton's mass and Einstein's are both m and are not the same thing; Ptolemy's "planet" includes the sun and excludes the earth. When an argument stalls, check whether the key words still mean the same thing.
⑨ Revolutions are invisible because the first act after victory is rewriting the textbook, recasting predecessors as people who went partway down our road. Kuhn's own awakening was exactly here: Aristotle was not doing bad Newtonian mechanics — his "motion" meant all change, and inside his own questions the remarks are rather good.
⑩ The hardest and most important line: we may have to relinquish the notion that changes of paradigm carry scientists closer and closer to the truth (paraphrased). Science is like evolution — an evolution from a primitive beginning, not a march toward an end. It really does progress, and the progress has no destination.