1. A Mathematician Who Never Existed

At the École Normale Supérieure, a student could become an authority for an afternoon by putting on a disguise. One traditional prank involved an apparently distinguished visitor lecturing to newcomers, who had to notice when solemn mathematics became nonsense.

Around 1923, Raoul Husson played such a professor, complete with false beard and exaggerated accent. His increasingly improbable theorems culminated in one attributed to an unfamiliar name: Bourbaki. André Weil was among the students who heard him.

We know the scene principally through Weil's recollection, published decades later in his memoir, whose English edition appeared in 1992. It is a story told by a participant after almost seventy years, not a contemporary transcript. Husson later studied the human voice rather than pursuing a mathematical career.

The surname was not invented. Charles-Denis Bourbaki was a French general, born in 1816, whose career included the Crimean War, the Italian campaign, and command of the Army of the East during the Franco-Prussian War. That army eventually entered Switzerland and was disarmed in 1871, under his successor Justin Clinchant.

Thus the familiar claim that the mathematicians invented a nonexistent general conflates two figures. The general existed. The mathematician Nicolas Bourbaki did not. The real name supplied a resonance of military adventure and disaster before it acquired mathematical authority.

The name appeared in mathematical print in India in 1931. Weil, then teaching at Aligarh, had recruited the young D. D. Kosambi. Kosambi published a short paper invoking a second theorem of Bourbaki. Weil later explained the fictional attribution as a joke aimed at a difficult senior colleague; its supposed source was an obscure Russian mathematician said to have died during the revolution.

The publication is firmer evidence than all the motives assigned to the prank. Its title survives independently of the anecdote, and the scholarly account of Kosambi's writings identifies its place in this prehistory.

At that point Bourbaki was still a movable joke. A name could be borrowed, given an imaginary biography, and used to puncture someone else's authority. In 1935, a circle of young mathematicians began using it to construct an authority of their own.

2. Lunch

In Strasbourg in 1934, Henri Cartan and Weil were teaching analysis from a curriculum they found unsatisfactory. Édouard Goursat's established treatise remained a standard point of reference, but modern algebra and topology had changed the background against which analysis could be organized.

Cartan repeatedly asked Weil how he handled particular topics. Weil proposed pooling the experience of several friends teaching similar courses elsewhere. Instead of answering the same questions independently, they would write a new account together.

At noon on Monday, December 10, six of them met at the Café Capoulade on Boulevard Saint-Michel in Paris: Cartan, Claude Chevalley, Jean Delsarte, Jean Dieudonné, René de Possel, and Weil. Their working identity was a committee for a treatise on analysis.

The early plan was modest only by comparison with what followed: roughly a thousand pages, a rapidly settled programme, and work expected to proceed with extraordinary speed. Important decisions were to require agreement rather than a simple majority vote.

Further meetings followed during the first half of 1935. The membership expanded, with departures and replacements. Small working groups investigated algebra, integration, differential equations and forms, variational methods, special functions, geometry, and Fourier series. Each group deliberately included someone who was not a specialist in its subject.

That provision matters. The expert would not simply write a chapter and receive polite correction from colleagues. A nonspecialist could ask why the account was arranged that way, whether a definition was necessary, or whether an apparently standard argument had actually been explained.

The project then retreated toward its prerequisites. Analysis required an account of limits and continuity. Those required topology. Topology required a sufficiently explicit language of sets and mappings. Every proposed beginning seemed to depend on an earlier beginning.

The group did not set out to reopen every philosophical dispute over mathematical foundations. Set-theoretic methods had become routine working instruments in large parts of mathematics, even while foundational questions remained unsettled. The task was to establish usable starting points and arrange dependencies.

A textbook intended to repair a course was turning into a reconstruction of the order in which mathematics could be presented.

This is not the same as discovering the order in which mathematics had developed. The order of exposition would deliberately differ from the sequence of problems, guesses, examples, and accidents that had produced the subject.

3. A Week in Besse

The first official congress took place at Besse-en-Chandesse in the Auvergne, July 10–17, 1935. The founding circle comprised Cartan, Chevalley, Jean Coulomb, Delsarte, Dieudonné, Charles Ehresmann, de Possel, Szolem Mandelbrojt, and Weil.

Most were about thirty. Chevalley was twenty-six; Mandelbrojt, thirty-six. Cartan worked in analysis and topology and was the son of Élie Cartan. Weil pursued number theory and algebraic geometry. Coulomb brought mathematical physics and geophysics. Dieudonné would carry an exceptional share of the final writing for the next quarter-century.

Nine young people proposing to reorganize university mathematics can sound like a generational coup. Interwar France provided an unusual setting. The First World War had killed a substantial part of the generation that might have become their immediate senior teachers and colleagues. Estimates for particular École Normale cohorts vary greatly with the population counted; the demographic loss is clearer than a single universal percentage.

This absence did not create Bourbaki mechanically. It helped explain why peers could see both an opportunity and an obligation to rebuild a curriculum they considered outdated.

The organization was horizontal in an important sense: colleagues of similar age, rather than a master with a school of disciples. It had no permanent chair directing the congresses. Individual influence certainly existed, but it did not take the simple form of a founder assigning everyone a permanent intellectual territory.

During that week, the temporary committee became a continuing enterprise. The fictional author took a durable form. The members also accepted that the familiar division into separate volumes for algebra, calculus, and geometry would not support their intended arrangement.

In 1935 a note under Bourbaki's name reached the Comptes rendus through Élie Cartan. The invented mathematician was becoming a published author. The often-repeated story that Eveline de Possel supplied the first name Nicolas comes through later reminiscence and deserves a lighter evidentiary weight than the publication record.

The title Éléments de mathématique was settled in 1938. Its echo of Euclid was deliberate. So was the unusual singular mathématique: the subject was to appear as a connected whole, not merely a collection of branches.

The first booklet, published in 1939, summarized set-theoretic results without their proofs. The first two fully developed chapters of set theory arrived only in 1954. The intended logical order was already diverging from the chronology of actual publication.

Bourbaki's finished books can seem inevitable in their sequence. Their beginnings show choices, delayed commitments, changing plans, and unfinished foundations underneath the appearance of inevitability.

4. Meetings Without a Chair

The working rhythm eventually involved three congresses a year: two of about a week and a longer summer gathering. Sessions could occupy roughly seven hours a day.

A draft was read aloud line by line. Anyone could interrupt a word, demand a proof, challenge a choice of generality, or insist that an entire section be rewritten. When Armand Borel first attended in 1949, the simultaneous arguments could sound like several monologues colliding.

Participants and visitors later described the meetings with affectionate horror. Weil defended their cultivated disorder: premature assignment of territories, a presiding authority, and routine voting might produce an ordinary committee document rather than the desired work.

Disorder was therefore combined with a demanding constraint. Publication required unanimity. In practice every member could prevent a text from becoming Bourbaki's text.

The first draft often came from someone expert in the topic. Later drafts might pass to other hands. Six or more versions were not unusual. A complete revision could be rejected at the next congress; a discarded earlier approach might return.

Borel's account of the chapter on root systems captures the inflation. A draft of around seventy pages became one of about 130. Someone objected that too much effort was being spent on so small a topic. The eventual printed chapter ran to 288 pages.

Unanimity also generated large amounts of unpublished work. A member might spend months writing a draft that was condemned after a few pages had been read. Others went into the collective “refrigerator,” a storage category offering no promise of revival.

Borel recalls more than 260 pages by Weil on manifolds and Lie groups, followed by about 150 pages by Roger Godement, that did not appear in the projected form. The labour was real even where the publication count was zero.

This system could be wasteful, frustrating, and extraordinarily productive. It prevented a chapter from being simply the property of its first author. It also made completion dependent on a community's ability to turn objection into a shared alternative.

At moments when an argument produced an unexpectedly satisfactory idea, the group described a spirit as having passed through the meeting. The phrase acknowledged something their formal publications concealed: a text assembled through rules still depended on the unpredictable arrival of an idea.

Borel's participant account explicitly identifies its own subjectivity. It is evidence of working practice, not an omniscient view of every congress or member.

5. The Same Shape

Traditional divisions follow the apparent nature of their objects: numbers, figures, functions. Bourbaki increasingly organized material by recurring patterns of relations and operations.

Consider integer addition, rotations about a fixed point in a plane, and permutations of a deck of cards. These are different kinds of things. Yet each offers a composition operation, associativity, an identity, and inverses.

They are groups. A theorem proved using only the group axioms applies to each example and to other groups not yet encountered. The shared framework does not assert that every group is identical. It identifies precisely which conclusions can travel between different settings.

The permutations in Galois's work were an earlier instance. The subsequent abstraction made the common pattern itself a systematic object of study.

Bourbaki called such organizing patterns structures. Algebraic structures govern operations; order structures organize comparisons; topological structures govern notions of neighbourhood and continuity. More complicated objects combine these patterns.

The framework changed where a result belonged. Instead of being filed under numbers or shapes, it could be located according to the assumptions actually used in its proof. A fact about a vector space might apply equally to geometric arrows, polynomials, or functions because its proof depends on the same linear structure.

Two senses of “structure” must remain distinct. Bourbaki also proposed a formal general theory of structures using sets and relations. Historian Leo Corry argues that this technical apparatus played much less of a role in ordinary mathematical research than the broader structural image of mathematics.

That is a historical interpretation, not a theorem showing that nobody ever used the formal definition. Corry's study asks how the formal theory related to the group's actual mathematical practice.

The lasting influence was largely methodological and expository: attend to isomorphism, common axioms, and transferable arguments. Those habits cannot be measured solely by counting uses of a particular foundational definition.

In “The Architecture of Mathematics,” published in French in 1948 and English in 1950, Bourbaki offered a public account of this vision. Usually associated principally with Dieudonné's writing, it presented the axiomatic method as a means of exposing connections beneath apparently separate branches.

An individual drafted an essay signed by an invented individual to explain an organization of knowledge that did not depend on individual authors. The paradox is revealing rather than accidental. Bourbaki's most consequential product was not one universal technical device but a way of arranging mathematical thought.

6. Instructions for Use

The first six books formed a prescribed sequence: set theory, algebra, general topology, functions of a real variable, topological vector spaces, and integration. A result could depend on material already established in the permitted earlier portions. The main exposition largely avoided outside references.

History appeared separately, in notes placed after the mathematical development. Discovery and logical dependence were given different rooms.

The discipline had costs. A later chapter could be delayed by missing prerequisites. Meanwhile, real publication proceeded out of sequence: algebra and topology appeared while the full opening set-theory chapters still awaited completion. The fifteen-year interval after the 1939 summary exposes the practical compromises needed to maintain the imagined linear order.

Within a section, the rhythm typically ran from definition through general proposition, proof, corollary, special case, and exercise. It did not begin by reproducing the reader's probable route from problem to example to failed conjecture.

Borel called the style dogmatic in this specific sense. The books aimed to furnish a cleared foundation for research, not to reenact how research happens.

Each volume nevertheless addressed its reader through directions for use. The promised starting point was elementary in logical dependence, but the required reader already needed familiarity with mathematical reasoning and considerable powers of abstraction.

A dangerous-bend symbol, borrowed from French road signage, marked places where misunderstanding or technical difficulty threatened. This was not the absence of care. It was care for a reader navigating an extensive system of precise references rather than for a beginner seeking an inviting first encounter.

Numbered statements made the internal architecture searchable. One cited a prior theorem by its location instead of continually returning to an author's paper and date. Names and origins, where discussed, emerged in historical notes rather than directing the argument.

Pictures and motivational passages were scarce. According to Borel, an intuitive explanation could enter a draft, be challenged as imprecise, resist immediate repair, and disappear. Even Dieudonné could have a draft rejected as excessively abstract.

The finished austerity was therefore partly a filtering effect. It does not show that the writers lacked imagination or visual intuition. It shows what survived their collective standard for published exposition.

This distinction also qualifies the present series' own argument. Twentieth-century mathematics can become harder to narrate because its expository forms move definitions forward and personal histories backward. That is an interpretation of a powerful tendency, not a claim that all modern mathematics was deliberately designed to exclude people.

Bourbaki pursued one such separation with exceptional consistency. The reader encountered a clean order whose production had been anything but clean. The difficulty of the book and the activity that produced it were related, but they were not the same thing.

7. Everybody Knew Who They Were

“Secret society” is an imperfect description. The current membership list was not publicly issued, but informed mathematicians generally knew who belonged. Borel learned the local identities soon after arriving in Paris and then received an invitation to a congress.

The fictional personality was maintained playfully. When Bourbaki sought individual membership in the American Mathematical Society, the response recognized the collective and suggested institutional membership. When Ralph Boas explained the pseudonym in an encyclopedia, Bourbaki countered with the suggestion that Boas himself might be a collective invention of Mathematical Reviews.

These exchanges sustained a joke, not a successful concealment of human existence.

The more consequential anonymity concerned contributions. Which person first drafted a paragraph? Who insisted on a definition? Whose argument was rejected? Which four hundred pages never left the refrigerator? The printed books did not say.

Dieudonné makes the asymmetry particularly visible. For roughly twenty-five years, final versions, exercises, and preparation for publication fell heavily to him. Borel estimated an output surpassing that of several colleagues combined. His daily routine included piano playing followed by Bourbaki writing.

In September 1940, amid war, an issue of the internal circular La Tribu carried his ambitious plan for twenty-seven books. Much of the supposedly impersonal voice depended on his sustained labour. It was a cultivated Bourbaki voice, not simply the way he wrote under his own name.

The title page still named Nicolas Bourbaki.

A parallel institution worked differently. The Séminaire Bourbaki, established in 1948, presented recent work by other mathematicians in regular sessions. Speakers included members and outsiders, and published seminar reports bore their actual names. The same network produced anonymous collective books and individually credited expositions.

Nor was Bourbaki the only mathematical pseudonym. Blanche Descartes in Britain and, from 1952, John Rainwater in the United States also became collective names associated with real publications. Bourbaki's distinction lay in duration, mathematical strength, scale, and the authority the name acquired.

Retirement at fifty was intended to prevent founders from holding the organization forever. It did not function flawlessly from the first day. When the earliest members reached the threshold, practice remained unsettled; Weil's explicit retirement in 1956 helped establish a stable convention.

Legal continuity acquired a body in 1952, through the Association des Collaborateurs de Nicolas Bourbaki. Contracts and copyright required actual people and an accountable organization. Members lived principally on salaries from universities and research institutions; book revenue sustained collective work rather than becoming their common wage.

Drafts, counterdrafts, assignments, comments, and La Tribu survived. Jokes, poems, and nicknames coexist with technical decisions. Liliane Beaulieu used these materials to reconstruct the early meetings; digital archival projects later made substantial portions accessible.

The author disappeared from the finished exposition, not from the historical record. The surviving paper trail sets a limit to the title of this episode.

8. What Came Out at the Other End

The members did not conduct all their research in the books' manner. Borel described research and preparation of the Elements as largely distinct activities. Between formal sessions, participants exchanged concrete problems, fresh results, and speculative approaches.

Pierre Deligne recalled being introduced to algebraic geometry through a drawing of a cubic curve and a line through two points: an invitation to beautiful geometry. Such an encounter scarcely resembles opening a Bourbaki volume at its first definition.

The separation was not absolute. Cartan's papers could display a sequence of lemmas that seemed naturally to prepare an eventual theorem; Serre joked about the habits acquired through long Bourbaki service. Editorial practice could shape an individual's mathematical presentation without dictating the process of discovery.

Dieudonné supplied another comparison. Alongside his anonymous collective work, he wrote public explanations under his own name. Beginning in 1960 he developed the nine-volume Éléments d'analyse, completed in 1982. In 1987 he addressed a broader readership in Pour l'honneur de l'esprit humain, a title to which the final episode will return.

Criticism of Bourbaki took several forms. Emil Artin's 1953 review of Algebra is often reduced to its complaint about relentless abstraction. The positive continuation matters: a reader surviving the initial difficulty could gain a deeper understanding. Discomfort was not his whole verdict.

Paul Halmos objected in 1953 that Integration's emphasis on locally compact spaces did not fit the needs of probability theory. This was a substantive criticism of mathematical organization, not merely of prose. Later additions responded to that gap.

Probability, partial differential equations, logic, and combinatorics never occupied positions proportionate to their importance across contemporary research. A structural presentation was still a selection. It did not become neutral or exhaustive by suppressing individual names.

The connection with school-level “New Math” also requires restraint. Dieudonné was an outspoken advocate of structural reform, including his provocative 1959 call against Euclid as the governing model of school geometry. But national reforms had distinct institutions and histories. Bourbaki did not collectively issue a universal school curriculum, and the Elements were not written for children.

The organization itself encountered limits. The original foundations approached completion while research changed. A strict dependency order could postpone new topics indefinitely; increasing specialization made it harder for every member to master everything. Grothendieck's ambitious proposals for homological and categorical foundations were not simply absorbed, and large projects remained unresolved.

By the early 1980s the publication comprised many fascicles and thousands of pages, while the analysis initially promised had never been completed as originally envisaged.

How much should be credited to the collective? Witold Hurewicz suggested that extraordinary mathematicians would have produced extraordinary mathematics without it. Borel agreed that postwar French achievements, including those of Leray and Thom, had other sources. He also recalled Weyl withdrawing earlier criticism by 1952.

Count original theorems and Bourbaki can be overcredited. Count shared language, training, and standards of exposition and it can be undercredited.

The author's final metaphor is a machine of authorship. Individual drafts, temperaments, arguments, and intuitions entered one end. At the other emerged not an obvious compromise among nine people but a text sounding as though one author had always intended it.

That author did not exist. The labour and institutions making the illusion possible did.

An order that appears natural can be the result of revision after revision. In 1976, a proof involving roughly twelve hundred hours of computer calculation would pose a different question about the disappearance of individual oversight.