The Structure of Scientific Revolutions

The bulk of work in any mature field is not testing the framework — it's working out the details the framework already implies. Kuhn calls it 'mop-up work.' The framework is ASSUMED; only the puzzles are solved. Most 'productive work' is elaboration and confirmation, not questioning. You are not testing whether the system is right; you are extending how far it reaches.

A contradiction does not immediately overturn a framework. First it registers as an 'anomaly' — a violation of expectation — and the framework absorbs it: people work around it, ignore it, or reclassify it. Frameworks don't die from one bad result; they die from anomalies that accumulate and keep resisting being made to 'conform.'

A framework doesn't collapse cleanly — it enters 'crisis,' and the tell is behavioral, not a single decisive event. The rules loosen, people will 'try anything,' explicit discontent surfaces, and the field splinters into competing versions of itself. You diagnose crisis by its symptoms: first-principles arguments, contradictory 'fixes,' and open complaint.

The switch from one framework to another is not additive — you don't build the new on top of the old. It's a discontinuous leap: 'far from a cumulative process.' After the switch, the old framework isn't a special case of the new one; it's simply left behind. Change here is conversion, not accumulation — and the difference matters, because a conversion can't be achieved by piling up more of the old.

After you adopt a new framework, the SAME inputs read differently. It's like the gestalt figure — 'the marks on paper that were first seen as a bird are now seen as an antelope.' The change isn't in the data; it's in what the data shows. Afterward, the person 'works in a different world' — so you can't convert someone by adding data, only by getting them to see differently.

After a framework wins, the community rewrites its own history to make the winner look inevitable and the losers look naive. The record-keeping documents 'inevitably disguise not only the role but the very existence of the revolutions that produced them.' The path that was actually contingent and contested gets edited into a straight line of progress.

A mature framework is conservative BY DESIGN: it 'often suppresses fundamental novelties because they are necessarily subversive of its basic commitments.' The resistance isn't stupidity or bad faith — it's the framework doing its job. Novelty threatens the shared order that makes normal work possible, so normal work suppresses it. The people most invested in the current way are, structurally, the last to see the new thing that will replace it.

'Progress' is measured by the framework's own internal standard of good puzzle-solving, not by any external approach to truth. 'We may... have to relinquish the notion... that changes of paradigm carry scientists and those who learn from them closer and closer to the truth.' A framework can be getting objectively better at solving its own problems while drifting from the thing it claims to be about.

Two frameworks can't be fully judged against each other by a shared, neutral standard — because the standard itself lives inside one of them. 'Incommensurable' means the successor is 'not only incompatible but often actually incommensurable with that which has gone before.' You cannot referee the dispute from outside both frameworks, because there is no outside.

A paradigm is not a big idea or a theory; it's a shared example of what a solved problem looks like — an accepted instance of actual practice (including its tools, methods, and standards) that a group uses as the template for its next problem. The mechanism: a group can only do work as a group once it tacitly agrees on what counts as a good piece of work — and that agreement is carried by concrete exemplars, not by stated rules.