The Man on the Ship
The legend: the fifth century BC, some ship on the Mediterranean. Hippasus, a member of the Pythagorean sect, was aboard — and so were his fellow sectarians, who had learned what he had said. The sect held a creed later condensed into "all is number": reality is constituted of the integers and their ratios; number is the origin of all things. And Hippasus discovered (or let out) that the diagonal of a square is incommensurable with its side — no common unit, however small, can be found that measures both segments at once. In today's language: √2 is not a ratio of integers. By the account that came down, the price of letting the fact out was execution — he was thrown into the sea by his fellows. The legend cannot be verified, and its details vary from one ancient source to another; what can be verified is that the discovery itself is true, and that the sect once strove hard to keep it. What deserves recording first is the shape of the story: some determinacies were once guarded with lives. The first crisis of mathematics, priced in one human life — the bill humanity drew when the foundations of mathematics first gave way.
This chapter joins three such collapses into one profile. The first two are given briefly — each would need a monograph to be told through, and only the skeleton that bears directly on this book's mechanism is taken here; the weight falls on the third, because the place where it cracked is deepest, and Chapter Six has already surveyed that site.
The First Crisis: Number and Magnitude Part Ways
The discovery itself needs only a few lines of elementary argument: suppose the diagonal and the side commensurable; write their ratio as a ratio of integers in lowest terms; carry the equalities through, and one obtains a number that is both even and odd — absurd; therefore, incommensurable. The argument is harmless; the whole of its destructive force falls upon the creed it strikes: if "all is number" and number means only the integers and their ratios, then the length of the diagonal has no registered residence in existence. A crack ran through the sect's world-picture.
The response took a generation or two to take shape, and it came as Eudoxus's theory of proportion: not asking "what a ratio is", but fixing when relations between ratios hold — that two ratios are equal, that one is greater, each given a strict criterion. The theory was later incorporated into book five of the Elements and has been honored as one of the most profound constructions of ancient mathematics — it goes around the question "what is an irrational number" and rebuilds, upon relations of proportion, all the operations upon irrational magnitudes in geometry. In RC's vocabulary, this is re-convergence of the textbook kind: the old observational consensus (all is number) was abandoned, and the new rule framework (the theory of proportions of magnitudes) caught the collapse. But the re-convergence had its bill: Greek mathematics stood thereafter under the sign of magnitude, number swallowed by geometry — the development of arithmetic and algebra slowed in Greece itself for nearly two thousand years, and "number" did not reopen as an independent household until the modern era. After the foundation settled, the building was not rebuilt upon the original site; it took another road — re-convergence does not promise a return to the original site, only that the collapse is caught.
The Second Crisis: The Ghosts of Departed Quantities
The sound of the second collapse waited two thousand years to be heard — and this time, the building upon the foundation had already been carried up into the sky.
In the later seventeenth century Newton and Leibniz each invented the calculus. Its working power was terrifying: tangents, areas, extrema, planetary orbits, the tides — nearly the whole of the problems of mechanics and astronomy fell in batches before the new calculus. But the foundation of the calculus was a something that no one could say: Newton's fluxions let the "ultimate ratio" be treated, in mid-course of the derivation, as zero and not zero; Leibniz's dx was an existence neither zero nor smaller than any positive number. The calculus hit a hundred times out of a hundred, and no one could say what held it up — a tower carried up into the sky, its tip in the clouds, its base hanging in the air.
In 1734 the criticism came, from the side of theology: Berkeley published The Analyst, its subtitle "A Discourse Addressed to an Infidel Mathematician" — the same year, he was made Bishop of Cloyne. His argument was sharp and waspish: you — who compute the orbits of the heavenly bodies with infinitesimals, and on that ground look down upon miracles — your calculus rests upon a logical monster. A quantity that first exists, takes part in the operations, and then vanishes from the result he named the "ghosts of departed quantities". You can digest second fluxions, and yet find absurd the doctrine of a Trinity in three persons? Berkeley's theological purpose takes nothing from the accuracy of the criticism: the calculus did work, and working has never automatically meant that a foundation exists — that gap was what he pressed.
The response dragged on for more than a century, and was completed by Cauchy and Weierstrass. The point: the infinitesimal no longer needed to be "said what it is", because it was expelled from the language. The limit received a strict definition — to say that aₙ tends to L is to say: for every ε there exists an N such that, n greater than N, |aₙ−L| is less than ε — a definition using only inequalities among finite quantities, in which no ghost has room to stand. The derivative is no longer "the quotient of zero by zero" but the limit of the difference quotient; continuity, convergence, the integral were rewritten, one by one, in the same language. This is the ε-δ rigorization: not a residence permit issued to the ghost, but a rewriting of the language such that the ghost is no longer needed. Later, in the 1960s, infinitesimals were given a residence permit again by model theory (nonstandard analysis), proving that the expelled road can also be walked strictly — but that was already a contract of another kind: the ghost was readmitted on a complete set of immigration papers, not by amnesty.
Note the isomorphism of this re-convergence with the one before it: a smoothly operating layer of foundation (the fluent calculus of infinitesimals) was pierced by observation ("the infinitesimal" had never been defined), and "it has always been so" stood out as a locking never re-audited; and the repair did not recover the old foundation but rebuilt the language. The directions of the two repairs form, moreover, a duality worth pondering: the first crisis handed number over to geometry (magnitude took in number); the second expelled geometric intuition from analysis (inequalities took in continuity) — the grammar of the foundations swings between two extremes, and every swing answers one and the same question: which words deserve to appear in the sentences of mathematics. One further destination is to be recorded: rigorization reduced analysis layer by layer — limits rest upon the real numbers, the real numbers upon the rationals, the rationals upon the natural numbers — and the end of that chain of reduction is exactly where Frege took over in Chapter Five, and exactly the detonation point of the third crisis.
The Third Crisis: The Ground Cracks under Everyone's Feet
The third collapse Chapter Six has already surveyed on site: June 1902, Russell's letter of a few lines to Frege; the reflexive dead knot in "the totality of all sets that do not contain themselves"; the load-bearing wall of the Basic Laws of Arithmetic cracking, with Frege himself acknowledging that the foundations had given way at the moment of the building's completion. Beside the first two, this one differs in order of magnitude: the first cracked a sect's world-picture, the second cracked the terminology of one calculus, and the third cracked — by the expectations of the day — the common foundation of all mathematics. And the collapse came at mathematics' most ambitious moment: the rigorization of analysis seemed to have reduced everything to arithmetic; Frege and Russell were on the point of reducing arithmetic to logic; victory was in sight. The ground chose that moment to crack under everyone's feet.
Then broke out the famous battle of the foundations in the history of mathematics. Three stances answered the call; let each lay its cards on the table.
Hilbert's formalism. Chapter Seven told of his program: the object layer formalizes the whole of mathematics thoroughly — axioms written as signs, inference become the transformation of signs; the meta layer uses the most conservative finite reasoning to prove that this system will never derive a contradiction. What must be added here is the philosophical stance behind the program: axioms are not self-evident truths but conventions — the axioms of geometry are not a priori truths about space (non-Euclidean geometry had already demolished that position), nor are the axioms of arithmetic; they are the opening rules of a game of signs. Whether an axiomatic system is legitimate asks one question only: whether it is consistent. Consistency is legitimacy — he even expressed the sense that consistency is existence: so long as a system derives no contradiction, the objects it speaks of have a registered residence in mathematics. Chapter Seven has already told the ending this program met in 1931: the guarantee cannot be issued by the very party that is guaranteed — what Gödel's second theorem struck was precisely the unspoken premise inside "consistency is legitimacy": a certificate of legitimacy must always be signed by someone, and no system can sign its own.
Brouwer's intuitionism. This Dutch topologist pulled the firewood out from under the pot: mathematics is not a system of signs at all. Mathematics is the mind's activity of construction in time — prior to language, prior to logic; language is only an imperfect record of this constructive activity, and logic only the ordering of linguistic habit. A mathematical object exists exactly when it has been constructed; to say that an object exists and give no construction is to write a check against no deposit. Brouwer placed the origin of mathematics in the most primitive construction of all: the two-oneness of time — the falling apart of this moment and the moment next, from which comes "one more", and from that the infinite sequence. All of mathematics should be traceable back to this introspective beginning, one any subject can replay. Hence intuitionism rejects the universal validity of the law of the excluded middle: for finite collections a proposition is true or false, with no third possibility; but for infinite totalities — all the natural numbers, all the reals — the excluded middle presupposes that "the infinite can be reviewed from the beginning as a finished thing", and the mind has never once completed an infinite construction. Thus some proofs of classical mathematics are, in intuitionist eyes, no proofs at all: non-constructive existence proofs, and proofs by reductio (assume the negation, derive a contradiction, therefore the assertion) — the force of the latter is pledged precisely upon the excluded middle. Hilbert's counterattack was the famous sentence, to this effect: to take the law of the excluded middle away from the mathematician is like forbidding the astronomer his telescope, or the boxer the use of his fists. Their polemic of the 1920s tore through the mathematical communities of Göttingen and Holland with an intensity unmatched in the history of mathematics — a war over "what counts as mathematics" wounds deeper than any quarrel over "which mathematical conclusions are right".
Platonism. The oldest stance: mathematical objects exist independently. Numbers, sets, structures inhabit an immaterial kingdom that does not shift with human cognition; the mathematician is not an inventor but a discoverer; axioms are not conventions but (fallible) descriptions of that kingdom. Platonism can be traced back to Plato and the Pythagoreans, and its weightiest modern adherent was precisely Gödel: he believed the universe of sets as real as the physical universe, and believed that the continuum hypothesis — to make its entrance later in this chapter — has a determinate truth or falsity, we having simply not yet found the right angle from which to see it. The two-thousand-year misreading of Chapter Eight — taking postulates for truths — received from Platonism a positive rehabilitation: postulates look like truths because they are truths; the self-evidence felt across two thousand years is not the interest on the account but the echo of the principal.
The Common Form of the Three Crises
Now for this chapter's first transcription: lay the three collapses one upon another. They stand two thousand years apart, in different locations; the common form is nonetheless one: the "foundation" at some level of convergence is pierced by observation into the next level down, and the old "it has always been so" stands out as a locking never re-audited.
The first: the piercing came from above downward — the world delivered magnitudes beyond number. The convergence level of everyday surveying had locked "all lengths are commensurable" into default; Hippasus's argument pushed observation beneath that default and found the floorboards empty underneath. The second: the piercing came not from new facts but from an audit of an old concept — the infinitesimal held up the whole of the calculus, and "it has always been so" ran for a century; Berkeley's observation merely asked "what is it", and the floor showed through. The third: the piercing force came from the system itself — the core conclusion of Chapter Six: once a system has gained sufficient capacity for self-observation, the observer falls into his own range of observation; the more successful set theory became, and the more freely "totalities of an arbitrary property" were used, the nearer the reflexive seam drew.
The direction of the three piercings is one: downward, toward the deeper stratum. Each foundation, at the time it was laid, was not wrong — the theory of proportion, the calculus of infinitesimals, naive set theory each carried great mathematics on its back; they had simply never been questioned. The crises are not the disease history of mathematics but its geology: every collapse is the discovery that a "floor" was in fact "the ceiling of the story below". And the three repairs are isomorphic as well: not to recover the old foundation but to rebuild the language — proportion, limit, axiomatic set theory. Through all three collapses mathematics never stopped work — a point to which the end of this chapter returns.
The Three Stances = Three Strategies of Re-convergence
The second transcription: the three foundational stances, on RC's reading, are not three theories about mathematics but three strategies of re-convergence — three construction plans for one and the same site of collapse.
Formalism: admit whatever path is self-consistent. Its re-convergent act binds legitimacy to an internal property of the locking procedure — it does not ask which truth a starting point corresponds to, only whether the road from the starting point can be walked. Once the lessons of the battle of foundations (self-evidence untrustworthy, truth undecidable) had been fully absorbed, only one legitimate foundation remained: freedom from contradiction. In RC's vocabulary this is a lucid contraction: it demotes "truth", a question that crosses snapshots, to "self-consistency", a question internal to the snapshot — and Chapter Seven has already proved the limit of that demotion: self-consistency cannot certify itself; legitimacy is in the end conferred across levels.
Intuitionism: acknowledge only constructions the subject can replay in real time. This is the plan that pushes Axiom A3 of RC's paper (observational locking: determinacy is locked only when observation occurs) to the extreme — whatever locking cannot actually be re-enacted by the subject of this moment is taken as never having occurred. The infinite totality has not finished being constructed — then it does not exist; the law of the excluded middle fails for it, because the observation "review the whole" never took place. Intuitionism thus makes the presence of observation the criterion of citizenship in mathematics — the one school of the three that carries "locking must be present" through with full strictness, at the price of ruling most of the territory of classical mathematics a building site.
Platonism: extrapolate the results of convergence into an independent kingdom. Its move is the opposite of the other two: since all subjects, all ages, all civilizations out of communication with one another converge upon one and the same structure (the two thousand years without failure of Chapter Eight are the thickest evidence), let the structure be declared to exist prior to all subjects — the stability of convergence is not an achievement but the developing image of an ontology. On RC's reading, this reads the output of Axiom A6 as its premise: the thick account of consensus reinforcement, read backward into a proof that the object of the account preceded the account.
RC's General Outline says, discussing ideas and consensus in Section 2.4, that different ideas lock in different possible snapshots. The three foundational stances are footnotes to this principle: one and the same question — what is mathematics — shows three different snapshots under three modes of observation. In the formalist snapshot, mathematics is a family of symbol systems; in the intuitionist snapshot, an activity of mental construction; in the Platonist snapshot, a kingdom of eternal objects. On RC's reading, none of the three is simply wrong — each snapshot truly converged upon one face of mathematics (formal systems, constructive practice, the cross-subject stability of structure) — what is wrong is each taking its own snapshot for the whole Ground. This also explains why the war had no victor: no one lost, because no one ever stepped outside the boundary of his own snapshot; no one won, because adjudicating the battle of foundations would require a standpoint outside all three snapshots, and no such standpoint exists.
The working mathematicians' actual practice became the standing joke, to this effect: on weekdays be a Platonist — believe the structures you study really exist, or you cannot work; on weekends be a formalist — so long as the proof is written and checks out, pass it, or you cannot publish. Behind the joke lies real wisdom: mathematical practice has always been the rotation of three snapshots in use — present while constructing (intuitionism), absent while writing (formalism), in the king's employ while imagining (Platonism). Chapter Six said that on "where the boundary was crossed" the community sees clearly, while on "where the boundary should run" it can only bargain — the battle of foundations extended this configuration from set theory to all of mathematics.
The Building Never Stopped
A final inventory of the war's legacy turns up a fact that surprised every combatant: the thirty years in which the battle of foundations raged hardest (1900 through the 1930s) were precisely the thirty years of the most extravagant growth of buildings in the history of mathematics — measure theory, functional analysis, point-set topology, abstract algebra: half the skeleton of modern mathematics took shape in that period. No one waited for the philosophical verdict before building.
Each school's legacy was shelved on its own account. Hilbert's metamathematics grew into proof theory; intuitionism brought forth constructive mathematics and computability theory — Chapter Five said that the dream of handing all checking over to machines had to wait until 1936 for the proof of its exact boundary, and the whole computational age has since lived on the rigorization of "construction first"; Platonism continues, as working faith, to govern most mathematicians' daily practice, and its hardest problem to swallow — whether the continuum hypothesis is true or false — will be rewritten in the next chapter by Gödel and Cohen in a way no one anticipated. Chapter One said that re-convergence is local surgery: the philosophical ledger of the foundations was reopened three times, and the tenants upstairs scarcely noticed — theorems went on being born, textbooks went on issuing new editions, and after 1902 not one already-proved theorem was recalled.
But the war did rewrite something, and what it rewrote is exactly what Part Three tracks. The three crises and the three stances together forced onto the table a question that two thousand years had never asked head-on — since the starting points are chosen, the language is exchanged, and there are further foundations beneath the foundations, what, in the end, does mathematics study? The next chapter tells the answer mathematics has given in the modern era: a quiet revolution, the objects withdrawing, the relations taking the stage.