A Jesuit's Reductio ad Absurdum
1733, Milan. Giovanni Girolamo Saccheri, professor of mathematics at the University of Pavia and a Jesuit, published a book whose title declared the whole program: Euclides ab omni naevo vindicatus — Euclid vindicated from every flaw. Chapter Eight told of the seam: the fifth postulate reads not like a postulate but like a theorem owed. Saccheri's plan was to pay that debt on Euclid's behalf, and with a weapon tempered in a thousand theological disputations: the reductio ad absurdum. Suppose the fifth postulate false, then derive a contradiction — the moment the contradiction appears, the fifth postulate must be true, the seam closes, and the building stands without flaw.
His instrument was clean and sharp. Take a line segment; erect at its two ends segments of equal length perpendicular to it; join the two new endpoints: the figure is a "Saccheri quadrilateral" — its two base angles are right angles, and the case of the two summit angles is undetermined. Only three possibilities exist for the summit angles: right, obtuse, acute. It is not hard to prove (without any appeal to the fifth postulate): if one such quadrilateral has right summit angles, then all such quadrilaterals do; likewise for obtuse and acute — within each of the three hypotheses, all or none. The hypothesis of the right angle is exactly equivalent to the fifth postulate: accept it, and Euclidean geometry returns intact. So the reductio had only two moves left: prove that the hypothesis of the obtuse angle leads to contradiction, and then that the hypothesis of the acute angle does.
The obtuse hypothesis did indeed run into a wall. Under the standing assumption that a straight line can be extended indefinitely, Saccheri derived a contradiction, clean and quick. Then he entered the acute hypothesis — and entered a world no one before him had ever seen. Theorems were born one after another: the angles of a triangle sum to less than two right angles, the deficit growing as the triangle grows; there exist no triangles different in size but the same in shape — similar triangles must be congruent; through a point outside a line there can be drawn not merely "one" parallel. Each was unthinkable; each was derived without a flaw. Any student of geometry today can recognize it: he was marching through the heartland of hyperbolic geometry — the geometry later called Lobachevskian. Thirty-three propositions, each standing, and the contradiction never came. All that separated Saccheri from one of the greatest mathematical discoveries in human history was the acknowledging of it. He did not acknowledge it. At the end of the book he set down his conclusion: the acute hypothesis is absolutely false — because it is repugnant to the nature of the straight line. This is not a derivation; it is a verdict. The contradiction was not proved; it was proclaimed. That same year Saccheri died; the book slept for nearly a century and a half, until Beltrami dug it out of the old papers in 1889 — by which time it could at last be seen that the man who had proclaimed "absurdity" had in fact built the first city blocks of the new geometry.
The Relay: The More It Was Proved, the More It Looked like a Real Building
Saccheri was not walking the night road alone. In 1763 Klügel's doctoral dissertation took stock of some twenty-eight prior attempts to prove the fifth postulate and concluded that all were suspect — this inventory is itself testimony: by the mid-eighteenth century a considerable skepticism had accumulated inside the professional community. Lambert, around 1766, wrote his Theory of Parallel Lines (published posthumously in 1786); working with a quadrilateral of three right angles, he went further than Saccheri, and more honestly: he derived a whole batch of theorems of the new geometry, noticed that the quadrilateral's area is proportional to the angular deficit, and saw that this forms a perfect counterpart to the formula of spherical excess on the sphere — and he left behind that astonishing remark: that world behaves like the geometry of a sphere of imaginary radius. Yet he too kept looking for the contradiction, did not find it, and did not declare the search closed. Legendre's Éléments de géométrie, from 1794 onward, went edition after edition; across thirty years he gave proofs of the fifth postulate again and again, and was refuted again and again — among them one theorem that genuinely remained in mathematics: without the parallel postulate, the angles of a triangle cannot exceed two right angles. That theorem forced everyone to the foot of the cliff: only two options remained — the angle sum always equal to two right angles (Euclid), or always less (the new world); there was no third road. Legendre himself believed the first to be the truth until his death.
Lay the three trajectories one upon another, and a strange picture appears: the new geometry was built, hammer stroke by hammer stroke, by its opponents. Saccheri sank some thirty-odd piles for it; Lambert found its law of area; Legendre cleared away every middle road. This is the first RC-style lesson of the chapter: re-verification does not distinguish between positions — an opponent's re-verification is also re-verification, and the paths an opponent has trodden are also trodden solid. The ledger of the defeated and the ledger of the victor sit in one and the same account; a community's knowledge is saved up in the mutual checking of the two sides of a disagreement — which is exactly how RC's paper, discussing consensus reinforcement in its axiomatic system (Axiom A6), operates, without praise or blame.
Three Men, One and the Same New World
Then came three men who arrived almost at once.
Gauss's work on the problem of parallels began in his youth — in later life he recalled puzzling over it from 1792 on. By the 1820s he had in mind a fully formed geometry: exchange the fifth postulate, and everything remains self-consistent. In 1824 he wrote to his young friend Taurinus of the possibility that a completely self-consistent geometry exists, and warned him not to let it circulate. He never published. In 1829 he wrote to Bessel and gave the reason: he feared "the clamor of the Boeotians" — the din of fools, by which he meant the attacks he expected to meet. The giant whom posterity calls the Prince of Mathematicians dared to declare war on every battlefield but this one, and here chose silence. He also did one thing no one understood at the time and whose weight only later showed: directing the geodetic survey of the Kingdom of Hanover, he measured the great triangle formed by three mountain peaks — the Brocken, the Hohenhagen, and the Inselsberg — with sides running from seventy kilometers to more than a hundred; the angle sum agreed with two right angles within the margin of error. Whether or not Gauss meant these measurements as a test of the nature of geometry, they remain the earliest specimen of "letting the world adjudicate geometry". Remember this posture; it will return at the end of the chapter.
Lobachevsky did not choose silence. This professor at Kazan University (later its rector) began publishing "On the Principles of Geometry" in the Kazan Messenger in 1829, announcing openly: withdraw the fifth postulate, put in its place the contrary assumption — that through a point outside a line more than one parallel can be drawn — and geometry remains self-consistent all the same. He later called this geometry "imaginary geometry", and in 1840 restated it for the European community of mathematicians in a small German book. In Russia his harvest was mostly silence and mockery; blind, he kept dictating his last work, Pangeometry. He once likened himself to a Copernicus of geometry — a comparison that does not read as excessive today: what the two men rewrote was alike "how the world must be".
Bolyai's story is the shortest, and the most painful. When the Hungarian youth János Bolyai threw himself into the problem of parallels, his father Farkas — a mathematician who had himself burned half a lifetime on this very problem — wrote to warn him: for God's sake, give up the theory of parallels; it will devour everything you have, and rob you of your health, your peace, and your joy. In 1823 the son wrote back the sentence since carved into the history of mathematics: out of nothing I have created a strange new world. In 1832 that new world appeared, as an appendix of twenty-four pages, at the end of the father's textbook. The book was sent to Gauss; Gauss replied: to praise it would be to praise myself — for I had thought all this through years before. The father was relieved; the son was broken — to him it meant that the one thing that was truly his own was his no longer. He published almost no mathematics again. Three men came to one and the same door within one and the same decade: one dared not push it; one pushed and was ignored; one pushed, and was told that people had been standing behind it all along.
Riemann: Turning the Foundation Itself into a Variable
The entrance of the third man lifted the question a full story higher. On June 10, 1854, in Göttingen, the twenty-seven-year-old Bernhard Riemann delivered his habilitation lecture, the topic picked out by Gauss from a list of candidates: "On the Hypotheses Which Lie at the Foundations of Geometry". Published after Riemann's death, edited by Dedekind in 1868, it became the birth certificate of modern geometry.
Riemann did not tarry at the level of "how many parallels can be drawn through a point outside a line". He asked the more fundamental question: what, in the end, is the "space" that geometry speaks of? His answer: the object geometry studies is the manifold — a system of quantities assigned point by point and pieced together point by point; the geometric properties of space are not given a priori but are determined by the metric upon the manifold — the metric prescribes how "distance" is computed in this space, and curvature can vary from point to point. Euclidean geometry, the geometries of Lobachevsky and Bolyai, the sphere-like geometry — all are merely three special cases of constant curvature: curvature zero, negative, positive. The differences among the three are not differences between "true geometry and counterfeit", but differences among three positions of a parameter within one and the same framework — the problem of parallels is not the vital point of geometry at all, merely one striation upon a surface. And Riemann did one more thing that left a door open for physics: he said that the validity of geometric propositions in the infinitesimally small is a question that needs to be tested by experience — the boundary between geometry and physics was redrawn by his own hand.
Note the difference of posture between Riemann and Lobachevsky. Lobachevsky exchanged one brick to see whether the building would still stand; Riemann redrew the whole set of drawings, and made "exchanging the brick" an inconspicuous knob upon them. In the language of Chapter Three: the former verified that the foundation can be replaced; the latter turned "the foundation" itself into a variable — the freedom of expansion rose from the layer of postulates to the layer of the concept of space.
Beltrami: Innocence by Translation
One last formality remained. Lobachevsky said the new geometry was self-consistent; Riemann said it was a special case — but how is "self-consistency" itself to be proved? To say that a system is free of contradiction is to say that no contradiction can ever be derived in it — and an assertion about infinitely many steps of reasoning cannot be cashed by "no contradiction so far". The formality was completed in 1868, by the Italian Eugenio Beltrami. In his "Essay on the Interpretation of Non-Euclidean Geometry" he constructed the pseudosphere — the surface generated by revolving a tractrix about its axis — and proved that the intrinsic geometry of this surface realizes, within a bounded region, the theorems of Lobachevskian geometry one by one. More generally, he established the decisive principle: non-Euclidean geometry is consistent relative to Euclidean geometry — if Euclidean geometry is free of contradiction, then non-Euclidean geometry is free of contradiction.
The mechanism deserves a slowed look, for it is one of the most important inventions of nineteenth-century mathematics: the model, that is, translation. Translate the "points, lines, angles, distances" of Lobachevskian geometry, one by one, into the points, geodesics, angles of intersection, and arc lengths upon a certain surface inside Euclidean geometry; thereupon every theorem of the Lobachevskian geometry becomes the rewriting of a Euclidean theorem. If a contradiction existed in the new geometry, then, translated, it would necessarily exist in Euclidean geometry; believe Euclidean geometry innocent, and the new geometry is innocent automatically. Klein's projective model of 1871 and, after it, Poincaré's disk model followed in quick succession; dictionaries of translation came one after another. By the turn of the 1870s and 1880s non-Euclidean geometry had been accepted by the community as legitimate mathematics — no longer as a monster, but as a member of a family standing on equal footing with Euclidean geometry, each standing surety for the other.
Here hides a fine irony, and this book must enter it in the ledger: the certificate of innocence for the new geometry was issued by the old geometry itself. The new geometry's license of legitimacy was signed by the very geometry it had come to displace. It is the same shape as the inner limit of Chapter Seven: a system's determinacy was never a certificate it issued to itself — the guarantee is always a conferment across levels, issued by another (usually stronger) structure of convergence. Gödel proved that such conferment has no terminus; Beltrami demonstrated its actual operation: translation. The two, one negative and one positive, establish the same rule: legitimacy always stands between systems, never inside one.
The Echo, Sixty Years Later
The last piece of the puzzle came from physics. In 1912, Einstein, having recognized that gravitation must be geometrized, was introduced by his classmate Grossmann to Riemannian geometry and Ricci's tensor calculus; in November 1915 the field equations of general relativity settled into place: gravitation is not a force but the curvature of space-time — matter tells space-time how to curve, and space-time tells matter how to move. The geometry of physical space was the very variable of Riemann's lecture of 1854, assignable point by point. A language that had grown up sixty years before for the sake of purely internal expansion at last received its commission from physics. The eclipse observations of May 29, 1919 (Chapter Three told of them) were the world's reply: the deflection of starlight at the sun's edge chose the curved side. Gauss's measurements between his three peaks were, in principle, the rehearsal of this adjudication — only, at everyday scales the difference between the two geometries is too small for any instrument to resolve, and the verdict could be rendered only in a strong gravitational field. The fact itself deserves a second's pause: a disagreement inside mathematics was finally adjudicated by the world stepping forward — and the answer the world gave fell within the several possibilities mathematics had long since laid ready. Mathematics supplies the grammar of the possible; physics picks out this one frame of the snapshot.
The Core Transcription: Necessity Does Not Cross the Boundary of the Snapshot
Now for this chapter's core transcription — the pivot of Part Three, in three steps.
Step one: the principled character of the two-thousand-year failure. From Proclus to Legendre, the attempts at proof across two thousand years all failed; the standard explanation is "the problem was too hard, and the geniuses not enough". On RC's reading, the cause of the failure lay not in difficulty but in nature: to try to prove the fifth postulate is to try to argue that the starting point of one convergence path is the only possible starting point — to argue a choice into a necessity. This cannot be completed in principle. RC's paper lays down conservation of margin as Axiom A7 of its axiomatic system: in locking in determinacy, observation does not exhaust the Ground; there always remains available margin not yet locked. To lock the fifth postulate is to lock one path within the possibility space of "geometric space"; so long as margin remains — so long as other combinations of postulates do not logically self-destruct — "the only possibility" is forever a proof that will never arrive. Every failure of a reductio, every theorem derived without running into a wall, is the available margin taking shape inside mathematics. Saccheri's thirty-three propositions were not failures; they are the footprints A7 left upon paper.
Step two: the birth of non-Euclidean geometry is the textbook case of "observational divergence, system rebuilt elsewhere, re-convergence". RC's General Outline gives this dynamics in Section 1.4, discussing the unity of subject and object: observational consensus, too, can be questioned or negated by different modes of observation, producing observational divergence; the subjects of the divergence re-intervene in the Ground of Possibility to achieve re-convergence, completing an innovation of observational consensus. Match the items one by one. Observational divergence: two thousand years of futility — Klügel's inventory, Saccheri's verdict, Legendre's repetitions — gradually eroded through the consensus that "the fifth postulate is self-evident", and skepticism accumulated inside the community into a formal divergence. System rebuilt elsewhere: Lobachevsky, Bolyai, and Riemann each re-intervened, exchanged the starting point, laid another foundation — Chapter Three said that the architect can lay another foundation on other ground and by another set of conventions raise a building; the three men built three entrances to one and the same building. Re-convergence: Beltrami's translation supplied a checkable guarantee; Klein and Poincaré completed the dictionaries; around the 1880s the community completed the innovation of consensus — the new consensus was not "Euclidean geometry was wrong", but a wholesale rewriting of the question of the nature of geometric necessity.
Step three, the statement this chapter files away for Chapter Sixteen. The rewritten picture is two-layered. Within the Euclidean snapshot, the angles of a triangle sum to exactly two right angles — that is iron: two thousand years, zero failures, hundreds of millions of entries in the account; every schoolchild today makes further deposits in it; its validity is not diminished by a hair, and no "non-Euclidean" touches so much as a thread of it. But "space must be Euclidean" — that was a usurpation: the necessity of the paths inside a snapshot, extrapolated into legislation binding upon every possible snapshot. The existence of non-Euclidean geometry proves the extrapolation illegal: necessity halts at the boundary of the snapshot. Hence the formulations to be filed, first: mathematical necessity = self-consistency inside the snapshot — "necessary" says exactly the freedom from contradiction of what is derived, by explicit paths, from a chosen starting point, and not one grain more; second: the choice of axioms = the choice of convergence paths — the adopting and discarding of postulates is the selecting of a path, not the pronouncing of a truth. The two formulations together are the complete cashing, in the history of geometry, of "mathematics is the self-closure and expansion of convergence paths" (the definition of Chapter Three): closure gives the iron necessity inside the snapshot; expansion rests precisely upon the exchangeability of the knob called "axiom" — and necessity and exchangeability do not conflict, because necessity never crossed the boundary of the snapshot in the first place.
The Customary Clarifications
The customary clarifications, two of them.
First, this chapter is not saying that from now on, anything goes in geometry. This is the relativist slide refuted in Chapter Seven appearing in its mathematical edition, and it must be blocked here as before. The three geometries are each rigorously self-consistent, not equivalent to one another, their propositions distinct; inside mathematics the choice of axioms is free, but this freedom does not mean that the options are indistinguishable — under the unified framework (Riemann's), the Euclidean and the hyperbolic are two positions of the curvature parameter, comparable, convertible, mutually translatable. Before the world, the choice is still less free: which geometry physical space uses is a question with a right answer and a wrong one, and the photographic plates of 1919 recorded the answer. That a map has a border does not mean all maps are drawn equally true; that geometry is exchangeable does not mean geometry is arbitrary.
Second, this chapter is not saying that self-evidence is worth nothing. The account saved up by two thousand years of re-verification is a real deposit: Euclidean geometry, as a high-precision approximation of physical space, is still exact to a hair at everyday scales — general relativity never repealed it, only marked out its domain of validity (Chapter Three said as much of Newtonian mechanics: the old path continues in office as a limiting case). Self-evidence is the interest on the account, the operating result, in hard fact, of Axiom A6 of RC's paper. The error lies not in trusting the account — every entry in the ledger is true; the error lies in reading the thickness of the account as an ontological guarantee — as though the more often something were verified, the more "necessarily" true its starting point. However thick the interest, it cannot turn into a proof of what belongs to the principal.
The Next Step
The battle over geometry's foundations ended with "the foundation can be exchanged" — an ending more honorable than anyone had foreseen: the old building did not fall, new buildings stood beside it, and sixty years later physics came and rented offices in one of the new ones. But geometry is only one of mathematics' foundations. In the same period, two other foundations were beginning to sound: one called the infinitesimal — it held up the whole of the calculus, and yet no one could ever say what it was; the other called the infinite set — it promised foundations for all of mathematics, and was itself pierced in 1902 by a letter of a few lines. Three crises, three responses, one army building as usual upon foundations in collapse. The next chapter joins the three earthquakes into a single geological profile.