There is a comforting picture of science that most of us absorb at school. Knowledge, on this view, grows like a coral reef: patient researchers add small, solid deposits of fact, and the structure rises steadily towards the light. Each generation stands on the shoulders of the last, and nothing of value is ever thrown away. It is a tidy story, and in 1962 the American physicist-turned-historian Thomas Kuhn set about dismantling it. His book, The Structure of Scientific Revolutions, became one of the most cited works of the twentieth century, and its central term escaped the seminar room so thoroughly that it now turns up in business reports and advertising copy.
Kuhn's argument begins with what he called 'normal science'. Most of the time, he observed, scientists are not questioning the foundations of their field; they are working within them. A paradigm (a shared set of theories, methods, standard examples and assumptions about what counts as a good question) tells researchers which problems are worth solving and what a solution should look like. Normal science is therefore a kind of puzzle-solving: a puzzle is interesting precisely because the rules guarantee that an answer exists. This is not a criticism. Without such an agreed framework, he argued, researchers would spend their lives arguing about first principles and never get round to the detailed, painstaking work that makes a mature science productive.
Trouble arrives in the form of anomalies: results that refuse to fit. At first they are usually treated as the scientist's failure rather than the theory's. Kuhn's crucial insight was that a paradigm is rarely abandoned simply because it has been contradicted; every theory, at any moment, lives alongside unsolved problems. Only when anomalies accumulate, resist the efforts of the ablest people and begin to touch the heart of the discipline does a field slide into what he called a crisis. Even then, a paradigm is not given up until there is a rival ready to replace it. Scientists, like the rest of us, would rather have a flawed map than none at all.
The history of chemistry offers a classic example. For much of the eighteenth century, burning was explained by 'phlogiston', a substance supposedly released by anything that burned. The theory was not foolish; it accounted for a good deal. But careful weighing showed that some metals gained weight when heated, which suggested that something was being taken in rather than given off. Defenders of phlogiston proposed ingenious repairs, including the suggestion that it might have negative weight. Antoine Lavoisier, working in Paris in the 1770s and 1780s, reinterpreted the same experiments in terms of a gas taken from the air, which he named oxygen. What changed was not merely a single belief but the whole way of seeing the evidence.
This is where Kuhn becomes most provocative. A revolution, he claimed, resembles a gestalt switch, like the famous drawing that can be seen as either a duck or a rabbit, but never as both at once. Scientists on either side of the divide may use the same words, such as 'mass', 'element' or 'species', while meaning subtly different things. Kuhn described rival paradigms as incommensurable: there is no perfectly neutral standard, he suggested, by which their merits can be measured. Hence his unsettling use of a remark by the physicist Max Planck, that a new truth often triumphs not by convincing its opponents but because those opponents eventually die and a new generation grows up familiar with it.
Critics were quick to pounce. If theory choice cannot be settled by evidence and logic alone, they asked, is science merely a matter of fashion, or of mob psychology, as one philosopher scornfully put it? Kuhn spent much of his later career insisting that he had been misread. He never claimed that scientists choose theories at random. Accuracy, consistency, breadth, simplicity and fruitfulness all matter, he argued; they simply do not work like an algorithm that produces the same verdict for every competent judge. Reasonable people can weigh them differently, which is why revolutions take time and why the losing side is not necessarily stupid.
Nor does the theory fit every case neatly. The acceptance of plate tectonics in the 1960s looks gratifyingly Kuhnian: Alfred Wegener's idea of drifting continents, proposed in 1912, was long dismissed, partly because nobody could explain what force might move them, until new evidence from the ocean floor made the old picture untenable. Yet many fields seem to change through a steady series of smaller upheavals rather than one dramatic overthrow, and historians since Kuhn have tended to stress continuity: the instruments, techniques and habits of mind that survive even when grand theories fall.
What remains of Kuhn's account is less a law of history than a change of sensibility. He taught us to see science as something done by communities of people, with training, loyalties and blind spots, rather than by disembodied reasoning machines. Paradoxically, this makes its achievements more impressive, not less. That fallible human beings, attached to their favourite ideas, have nonetheless repeatedly conceded that nature was telling them something else is one of the more remarkable facts about our species. The coral reef was always a flattering image. The truth is stranger, messier and, in its way, more hopeful.