A birthday nobody can prove: Confucius, David Lewis and the logic of convention

Nobody knows what day Confucius was born. Today is nonetheless his birthday, kept as such in Qufu, in Taipei, in Seoul and in Hong Kong, and this is not a paradox but a piece of administrative history. The year, 551 BC, rests on the Shiji of Sima Qian and, with a discrepancy of a year, on the Gongyang and Guliang commentaries to the Spring and Autumn Annals, all written centuries after the event. The day is later still: a traditional reckoning of the twenty-seventh of the eighth lunar month, which Britannica notes "has been questioned by historians", and which the government of the Republic of China had, in the 1930s, converted to 27 August for the purposes of a Teachers' Day. In 1952, in Taipei, the calculation was done again, and the day came out at 28 September. It has been 28 September ever since. Taiwan made it a public holiday in 1963, abolished the day off in 2000, and restored it in 2025. The mainland keeps Teachers' Day on 10 September, though proposals to move it to Confucius's birthday have circulated since at least 2004.

David Kellogg Lewis was born on 28 September 1941 in Oberlin, Ohio, where his father taught government and his mother was a medieval historian. He died in Princeton on 14 October 2001, at sixty, of complications from diabetes. His first book, Convention, appeared in 1969, and it is the reason the story above is philosophy rather than trivia. Lewis's question was how a regularity can hold a whole population without anyone having agreed to it, and his answer explains, better than anything written since, why a birthday nobody can document goes on being celebrated by hundreds of millions of people who know it cannot be documented.

Immanuel Wallerstein, born in Manhattan on 28 September 1930, spent his career attacking one such regularity inside his own discipline, the assumption that the nation is the natural unit of social analysis. Herman Melville died in New York on this day in 1891, and Louis Pasteur near Paris in 1895; each has a claim on the page that runs through the same theme. The thread is convention: how it holds, what it costs to refuse, and where Lewis's account, for all its elegance, runs out.

What Lewis explained, and what the 1952 committee knew#

Convention began as a Harvard dissertation under Quine and was, as Lewis said in its opening pages, an attempt to answer Quine's scepticism about the conventionality of language. The question inherited from Hume and from Rousseau was whether language, promising and the rest could rest on agreement when no agreement had ever been made and none could have been, since making it would have required the language it was supposed to found. Lewis took the apparatus of Thomas Schelling's coordination games and answered that no agreement is needed. A convention, on his definition, is a regularity in behaviour to which nearly everyone conforms, which nearly everyone expects nearly everyone else to conform to, and which nearly everyone prefers to conform to on condition that the others do, where at least one alternative regularity would have served as well. The last clause is the one that bites. Driving on the left is a convention because driving on the right would have done; the point is not which side but that everyone picks the same one.

A birthday for Confucius meets every clause. It does not matter, for the purpose of a public holiday or a Qufu ceremony or a school assembly in Kaohsiung, whether the man was born on 28 September or 27 August or in the spring; it matters that the ceremonies fall on one day, and the 1952 recalculation supplied a salient one. Lewis's account also predicts something the historians of the date confirm, which is that once the equilibrium is reached the evidence stops mattering. Nobody in Taipei proposes to move the day when a new paper on the Guliang chronology appears. The convention has become, in Lewis's phrase, self-perpetuating, and the reasons that first selected it have gone quiet.

The friction is that Confucius himself would not have accepted this picture of ritual, and neither did the reading of him that made him respectable in Anglophone philosophy. Herbert Fingarette's Confucius: The Secular as Sacred (1972) is the book the Stanford Encyclopedia still singles out, and its argument, as the encyclopedia summarises it, is that the ceremonies of li, "when done correctly and sincerely", carry a "magical" quality on which the efficacy of a ruler's virtue depends. Fingarette was borrowing from Austin's theory of performatives, and his point was that a bow, properly made, does not coordinate two people so much as constitute a relation between them. A Lewisian convention is indifferent to its content; either side of the road will do. A Confucian rite is not, since the Analects insist that a rite performed without the inner disposition is not a rite at all. "I am a transmitter and not a maker," Confucius says at Analects 7.1, and the sentence is not the modesty it looks like. It claims that the forms are given and binding, not selected among equals. Lewis has the better theory of birthdays. Confucius has the better theory of why anyone bothers to bow.

Hegel, for what it is worth, read the Analects in translation and found nothing beyond "commonplace" moral teaching; he thought Cicero's De Officiis did the job better and that Confucius's reputation would have fared better had the works never been translated. It took two centuries and Fingarette's slim book to reverse the judgement.

Lewis's other worlds, and the stare#

Convention is not what Lewis is famous for. The fame belongs to a thesis he first set out in Counterfactuals (1973) and defended at length in On the Plurality of Worlds (1986): that possible worlds are as real as this one, spatiotemporally isolated from us and from each other, and that "actual" is an indexical, meaning simply the world the speaker is in, as "here" means the place. The book's first sentence is characteristically flat: "The world we live in is a very inclusive thing." The argument for the rest is that a systematic philosophy that helps itself to possible worlds in analysing necessity, counterfactuals, propositions and properties should pay for them honestly, and that the alternative accounts, in which worlds are sets of sentences or abstract states of affairs, either fail or quietly presuppose the real thing. The objection Lewis heard most often had no premises. He named it himself: the incredulous stare. He conceded that it was a legitimate reaction, that the price of his theory was a serious offence against common sense, and he argued that the theoretical benefits were worth it. Few have agreed. Almost everyone has used his framework.

The rest of the work is of a piece with this. Lewis's stated project, which he called Humean supervenience, was to show that everything there is supervenes on a spatiotemporal arrangement of local qualities, and that laws, chances, causation and mental states can all be accounted for on that basis, with no fundamental necessary connections. He spent a year at Oxford in 1959 and 1960, tutored by Iris Murdoch and attending Ryle, Grice, Strawson and Austin, and went back to Swarthmore a philosopher rather than the chemist he had meant to be. He taught at UCLA from 1966 and at Princeton from 1970 to his death. He went to Australia nearly every year and took from its philosophers, Jack Smart above all, the materialism about mind that "Mad Pain and Martian Pain" (1980) turned into a functionalism with room for a madman whose pain does the wrong causal work and a Martian whose pain is hydraulic. Smart is in the entry for 16 September.

It should be said plainly that Lewis thought most of what matters, from the laws of nature to the pattern of possible worlds, was not conventional at all. The 1952 recalculation in Taipei decided a matter he would have filed under convention; whether Confucius was born in 551 BC he would have filed under fact, and he would have been exact about the difference.

Wallerstein's refusal#

Wallerstein, who was born on the same day of the calendar eleven years earlier and outlived Lewis by eighteen, dying in Branford, Connecticut, on 31 August 2019, spent his career telling social scientists that one of their conventions was a mistake. He took all three degrees at Columbia, the doctorate in 1959 on Ghana and the Ivory Coast, and began as an Africanist. He was on the Columbia faculty through the occupation of 1968, an event he later said made him what he became, moved to McGill in 1971 and then to Binghamton in 1976, where he ran the Fernand Braudel Center for twenty-three years and published, in 1974, the first volume of The Modern World-System.

The argument of that book is easily flattened into a slogan about core and periphery, and this newspaper has argued elsewhere that the slogan misses the point. The point was methodological. Comparative sociology since Weber had taken the national society as its unit, compared nations with one another and asked why some developed and others did not. Wallerstein said the unit was wrong, that there had been since the long sixteenth century a single European world-economy with an axial division of labour, and that a nation's "development" was a position in that system rather than a property of the nation. On this view the industrial revolution in England and the second serfdom in Poland were not two national histories but two aspects of one process. The convention of the nation as unit was, in Lewis's terms, self-perpetuating: everyone used it because everyone else did, and the data were collected by nations, which made the alternative expensive.

The critics landed real blows. Robert Brenner's 1977 essay in New Left Review accused Wallerstein of a "neo-Smithian" Marxism that located capitalism in exchange rather than in the relations of production, and Theda Skocpol's review the same year in the American Journal of Sociology found the framework too total to be tested. Both objections have force. What neither disputed was that the unit-of-analysis question had been raised, and it has not been put down since. Wallerstein was not a Confucian and would have been startled to appear on this page. He is here because refusing a convention is the one thing Lewis's account cannot easily describe. A convention is stable, on Lewis's definition, because unilateral deviation is worse for the deviator. Wallerstein deviated, unilaterally and for forty-five years, and it was worse for him for a good part of that time.

Melville's refusal, and Pasteur's network#

Herman Melville died at 104 East Twenty-sixth Street on 28 September 1891, aged seventy-two, of what the death certificate called cardiac dilation. He had worked for nineteen years as a customs inspector on the New York docks, most of his books were out of print, and one newspaper's brief notice managed to call his masterpiece "Mobie Dick". Billy Budd lay in manuscript in a breadbox until 1919 and was published in 1924. The revival that followed made him the American novelist philosophers read, and they read one story above all. "Bartleby, the Scrivener" (1853) is about a copyist who answers every instruction with "I would prefer not to", and neither refuses nor complies. Deleuze's essay "Bartleby; or, The Formula" (1993) and Agamben's "Bartleby, or On Contingency", published the same year, both fasten on the grammar of the sentence: it is not a negation, which would still belong to the convention of instruction and response, but a withdrawal from the game in which instructions are given. Agamben reads it through Aristotle's dynamis, the potentiality not to act that is the shadow of every capacity to act. Bartleby is the limiting case of Lewis's theory: the man for whom the conditional preference that holds a convention together, "I will conform if you do", has simply been switched off. He dies in the Tombs, refusing to eat, and the narrator's last line, "Ah Bartleby! Ah humanity!", is the admission of a man who does not know what he has witnessed.

Louis Pasteur died on the same date four years later, on 28 September 1895, at Villeneuve-l'Étang near Marnes-la-Coquette, west of Paris, having been left half-paralysed by a stroke the year before. What earns him a place in this series is not the list of vaccines but the campaign of the early 1860s against Félix Pouchet of Rouen over spontaneous generation, which the Académie des Sciences turned into a prize competition and Pasteur won in 1862 with his swan-necked flasks, in which broth stayed sterile until it was tilted to touch the dust in the neck. Bruno Latour's Les Microbes: guerre et paix (1984), translated as The Pasteurization of France, took this episode and the decade that followed and argued that Pasteur's triumph was the work of a network rather than of a man: hygienists who had been waiting for a cause, farmers who could be shown a field trial at Pouilly-le-Fort, an administrative state that wanted something to enforce. The microbe, on Latour's account, became a fact when enough people had been enrolled to make disagreeing with it costly.

That is a theory of convention too, and Latour knew it, which is why realists have never forgiven him. The objection is the one Lewis would have made: that there is a difference between the day we agree to keep and the germ that is there whether we agree or not, and that a theory which cannot mark it has lost something. Latour's reply, in his later work, was that he had never denied the microbe, only the idea that it did its persuading alone. The flask, the field trial and the ministry are all in the story, and so is the fact. Whether that answer is enough has been argued for forty years; those who want the older realism have Lewis.

The day, and the series#

The last entry in this series to reach the archive was 24 September, on Cardano and Paracelsus. For readers who want to check Confucius's birth for themselves: the Stanford entry gives the year as "551?", question mark included. The ceremonies in Qufu will go ahead this morning regardless, as Lewis would have predicted and Confucius would have required.