A Book Copied for Two Thousand Years
The year 888, Constantinople. A scribe named Stephanos was at work by lamplight, copying a book. His employer was Arethas, a Byzantine scholar renowned for his learning — in later years he would become archbishop of Caesarea — and the colophon of the manuscript still carries the traces of the transaction: for the work, Stephanos was paid fourteen gold pieces. What he was copying was Euclid's Elements. The manuscript still exists today, held in the Bodleian Library at Oxford under the shelfmark "D'Orville 301": it is the earliest complete Greek text of the Elements bearing a scribal colophon that dates the copying.
What deserves a moment's pause beside that lamp is the manuscript's position in time. By the time Stephanos set his pen to work, the book was nearly one thousand two hundred years old — it had been born in the city of Alexandria around 300 BC. And it still had traveling ahead of it: Greek manuscripts were replicated generation after generation in the studies of Byzantium; centuries earlier it had already been translated into Syriac, and in Baghdad's translation movement into Arabic, read and reread by the commentators — among them Omar Khayyam, who had put the problem of parallels on record; in the twelfth century it passed from Arabic back into the Latin world; and in 1482 the Venetian printer Erhard Ratdolt set it in movable type, among the very first mathematical textbooks ever printed. One text, copied for nearly two thousand years, all but without revision.
One must first weigh what "copying" meant in that age. Printing would not enter Europe for another six hundred years; the reproduction of a book had a single road: find a literate man, pen and paper, a lamp, and write the whole through again from beginning to end. This was the only practicable technology of replication — and the most fragile technology of transmission: one stroke copied wrong, and the error travels on with the manuscript. For precisely this reason, the deformation of texts in transmission was the norm: the classics of antiquity almost without exception carry strata of alterations, accretions, and confusions. Yet under just such conditions of transmission the Elements kept its shape: when later ages recovered Saccheri (one of the protagonists of the next chapter) from the old papers, what they collated were the same propositions of the same book; what the Byzantine scribe and the Renaissance compositor set out was one and the same building. Copying brought slips of the pen, the intrusion of scholia, divergences of detail between lines of transmission — but across two thousand years, no one felt the need to "revise" its substance: the propositions remained those propositions, and the proofs those proofs.
"All but without revision" — these few words are the true object of this chapter. Through those same long centuries, no other great edifice of ancient knowledge kept its original form: Galen's medicine was rewritten before human dissection, Ptolemy's astronomy before new observations, Aristotle's physics before the criticisms of the Middle Ages and the Renaissance. Euclid alone, the more it was copied, the more it came to resemble the eternal. By what right does a book go two thousand years unaltered? The standard answer: because it is the truth. The answer this chapter has to give is different: because it changed its interlocutor. A book that no longer converses with the world cannot be revised by the world — and this was done by Euclid's own hand, on the very first page of the book; the method is called the axiomatic method.
Part Three opens here. The question left at the end of Chapter Seven was: at the cost of which cracks does a self-enclosed freedom of expansion come. This chapter tells how this building was raised; Chapter Nine tells of its first crack; Chapters Ten and Eleven, of the full spectrum of the cracks and of the architects' responses.
The Boundaries of Alexandria
Of Euclid's own life almost nothing survives. Only a few facts stand: he was active in Alexandria around 300 BC, in the city where the Ptolemaic dynasty had just built the greatest scholarly institution of the Mediterranean — the Museum and its Library; he flourished under Ptolemy I; and the Elements is the work he completed there (more precisely: compiled — a great part of its material came from predecessors). Of his temper only two anecdotes remain. One is what he said to Ptolemy: there is no royal road to geometry — in this science there is no shortcut reserved for kings — recorded by Proclus. The other is what he said to a student who, having just finished the first propositions, asked what learning them would bring in: give him three obols, since he must make profit from what he learns — recorded by Stobaeus. The credibility of both anecdotes is limited, but the very fact of their transmission fits posterity's understanding of this book: it grants no favors, asks after no uses, and acknowledges argument alone.
The Elements has thirteen books. The first six are plane geometry — from the congruence of triangles, through the theory of proportion, to circles and areas; books seven through nine are, unexpectedly, number theory — divisibility of integers, the proof that the primes are infinite, the characterization of the even perfect numbers; book ten treats incommensurable magnitudes, in nearly a quarter of the whole; books eleven through thirteen are solid geometry, closing with the construction of the five regular solids — the components of the heavenly bodies in Plato's cosmology. The standard account is that the whole proceeds from twenty-three definitions, five postulates, and five axioms to more than four hundred propositions. The division of labor between postulates and axioms, on the standard account, is that the postulates govern the geometric objects proper (such as "a straight line may be drawn from any point to any other point"), while the axioms govern what holds for all quantities alike (such as "equals added to equals are equal"); through two thousand years of relayed transmission the commentators have never ceased to dispute the details of this division — but the boundary it drew has never been disputed.
The boundary is the book's true invention. At the opening, all the starting points are laid upon the table — definitions, postulates, axioms, a few dozen lines in all — and then the rule is set down: every conclusion from here on may be derived only from these starting points, by explicit steps. What the second book states today as algebraic identities is written as the fitting together of areas; the Pythagorean theorem appears at the end of book one in the form of an equality of areas; and to the very last book, the five regular solids remain achievements led down one road from the same starting points. Beyond those starting points, nothing on the page is any longer "obvious".
The severity of this discipline deserves separate statement: besides the starting points laid upon the table, the reasoning is permitted to smuggle in nothing — not how a figure looks, not "it is obvious", not the naked eye. Euclid himself may not have entirely achieved this. In proving the congruence of triangles in book one he used the move of "superposition" — picking one figure up and setting it upon another — and "moving a figure" is something the postulates nowhere license; in many arguments he also took for granted the order of points along a line, an order of relation that the definitions and postulates equally fail to supply. These smugglings were so small that for two thousand years almost no one perceived them; only at the end of the nineteenth century, when Pasch and Hilbert re-audited the whole foundation, were they itemized one by one — that is the foreshadowing of Chapter Eleven. For now let the lesson of the matter be recorded: half of the "perfection" of the perfect building was designed; the other half came of two thousand years without anyone looking through a magnifying glass. The seams of the building are not the fifth postulate alone; it is only that the other seams lie buried deeper — and the fifth postulate is the one seam that stood in plain sight from the first day.
The Axiomatic Method: The Explicit Publication of the Starting Points
Now to characterize this opening move according to this book's reading.
RC's paper, in its axiomatic system, discusses observational locking (Axiom A3): determinacy is locked only when observation occurs. Chapter Five gave the transcription of this into formal systems: a formal system is the symbolization of an already-locked structure of convergence — its axioms and rules are achievements of convergence explicitly locked. Euclid is the prototype of this, two thousand years early. What he locked was the starting point of the convergence paths of geometric argument: from then on, whether a geometric proposition counts as established would no longer depend on whether the figure looked trustworthy, on the prestige of the teacher, or on whether the conclusion agreed with the naked eye — only on whether it could be walked through, by explicit steps, from those few dozen lines laid upon the table. The granularity differs (Euclid locked the starting points; Frege locked the transformation rules of every step); the action is one and the same: to take "what counts as established" out of the hands of rhetoric, authority, and intuition, and hand it over to an explicit path that anyone can walk again.
RC's paper, in its axiomatic system, discusses consensus reinforcement (Axiom A6): when the results of observation across subjects and levels verify one another, they form a positive reinforcement loop, converging into a stable objective reality. Chapter Five said that the modern symbolic language is an artificial accelerator of consensus reinforcement; this chapter must trace the credit one step further back: the axiomatic method is the oldest specimen of the artificialization of A6. Its principle of acceleration is plain to the point of a single sentence — once the starting points are published, verification no longer needs to re-walk the road of discovery; it needs only to check the road that is given. The students of Alexandria, the commentators of Baghdad, the medieval translators, and the Byzantine scribes checked one and the same set of starting points, one and the same path; they did not know one another, did not share a language, stood a thousand years apart — and were verifying one and the same thing. Every stroke beneath Stephanos's lamp was a re-emphasis of a locking performed one thousand two hundred years before.
Chapter Three said that mathematics' interlocutor is its own axioms and its constructive practice. This identity was not mathematics' birthright; it was the covenant Euclid signed: the conversation with the world — surveying, counting, the calendar: mathematics' mother tongue, whose origins the introduction has told — was shown out of the book, and geometry from then on asked its own questions and answered them itself. "There is no royal road to geometry" says the same thing from the other side: the rules of the boundary hold alike for kings and commoners — once the legitimacy of the starting points is explicitly published, it admits of no exemption by rank. This is the sharpest and the most dangerous stroke of the axiomatic method, and its sharpness and its danger are one and the same place: to publish the starting points is to admit that the starting points were chosen.
A Building Becomes the Template of Knowledge
The radiance of this building soon passed beyond geometry. Archimedes and Apollonius built on with its methods — the one computing areas and buoyancy, the other studying the conic sections; the astronomy, optics, and mechanics of Alexandria took its style of argument for their paradigm. After the Renaissance it became the very template of "what knowledge ought to look like". Spinoza's Ethics carries the subtitle "demonstrated in geometric order", setting God, mind, and the emotions into the format of definitions, axioms, and propositions; Newton's Mathematical Principles of Natural Philosophy is written in the order of definitions, axioms of motion, propositions, and proofs — structurally an astronomical edition of the Elements; and still in the eighteenth century, when philosophers asked how there can be knowledge that is necessarily true and yet extends our knowledge of what exists, the exemplar lying on the desk was geometry — Kant set geometrical judgments up as the prototype of synthetic a priori judgments precisely because, in his eyes, they were the specimen in which necessity and efficacy upon reality were most perfectly joined. A building that stands two thousand years without falling — and everyone who wants to raise a building comes to study its drawings.
RC's General Outline, discussing ideas and consensus in Section 2.4, says that when a particular interpretive framework shows predictive power and practical value, individuals and groups will form a consensus of ideas through comparison, and the established consensus will in turn act back upon the subjects' perception, closing the loop. The establishment of the template is precisely this loop running at its grandest scale in the history of knowledge: the successes of the axiomatic method (one irrefutable derivation after another) made it the very definition of "rigor"; once it became the definition, it prescribed in return how posterity would see "knowledge" — any discipline that did not look like geometry was taken for work unfinished. There is nothing mysterious in this loop; it is the account and the re-emphasis that Chapter One told of, magnified from a single desk to a whole civilization. And the cost of the loop stands in the same place: once a method becomes "rigor" itself, its starting points are locked away in the vault of the "self-evident" — to ask "why are the starting points thus", one must first have someone who doubts that the method itself could be otherwise.
The Fifth Postulate Was Off from the First Day
Of the five postulates Euclid laid upon the table, the first four take a line apiece: a straight line may be drawn from any point to any other point; a finite straight line may be extended continuously; a circle may be described with any center and any radius; all right angles are equal to one another. The fifth runs thus: if a straight line falling upon two straight lines makes the interior angles on one side together less than two right angles, then the two straight lines, produced indefinitely, meet on that side on which the angles are less than the two right angles.
Set in the ranks of the postulates, it does not look like its colleagues. The first four say what may be done — draw a line, extend a line, describe a circle; the fifth says what must come to pass — two lines meeting at a distance too far to be drawn. Its statement bends around a conditional clause, and its conclusion falls at a remove that the page cannot check. In the later phrase: it reads not like a starting point but like a theorem — a theorem that happens not yet to have been proved. This unnaturalness was seen from the first day. Eight hundred years after Euclid, the fifth-century Neoplatonist Proclus recorded the doubt in his Commentary on the First Book of Euclid's Elements: this proposition ought to be struck out of the postulates altogether — it is not an unprovable starting point but a theorem awaiting proof. Proclus recorded Ptolemy's attempted proof and gave one of his own; seen in retrospect, all those proofs smuggled in an assertion equivalent to the fifth postulate — proving the thing by means of the thing to be proved. He recorded the earlier futilities as well: already in the Hellenistic period attempts had been made to derive it from the remaining axioms. From then on this seam was never forgotten: the mathematicians of the Arabic Middle Ages opened case files for it, and the commentators of the Renaissance resumed the audit. The doubt itself became a tradition — for two thousand years, upon the wall of geometry's perfect building there stood a seam that everyone could see, no one could explain, and no one could erase.
It is worth setting this picture against the law of Chapter One: the more complete the convergence, the more it looks as if there had never been convergence. The fifth postulate is the law's inverse specimen — it is the one brick in the building not thoroughly covered by two thousand years of re-verification, and therefore the only brick of which one can still see that "it was laid on". Everyone saw the seam; everyone's explanation was that it belonged there of itself; as to why it belonged there, the world would wait another two thousand years.
How Eternal Truth Was Saved Up
The question with which this chapter opened can now be answered: by what right does a book go two thousand years unaltered?
On RC's reading, the answer has two layers. The first is the account. Chapter One recalled the consensus reinforcement that RC's General Outline describes in Section 1.2, discussing the emergence of determinacy — observations across subjects verify one another, forming a positive reinforcement loop — and called the deposit saved up behind every stable fact by countless mutual verifications its account; self-evidence is an account grown so thick that people have forgotten it was ever opened. The account of the Elements is the thickest in the ancient world: every student who verified the Pythagorean theorem, every field divided by geometric rule, every voyage reckoned by triangulation, every scribe who re-walked the proofs of book one by lamplight — each was a deposit, or a re-emphasis. Two thousand years, across languages, across civilizations, zero failures. No other ancient text ever saved up such an account.
The second layer is the choice of ground. Zero failures is not the merit of the building but the merit of its site. Medicine's ledger has corpses outside it; astronomy's ledger has the starry sky — both owe the world, and the world comes in time to rewrite. Geometry owes nothing: its interlocutor stands within; a proposition is examined only by the postulates and by logic, never by the shape of the field boundaries after the Nile's flood. RC's General Outline, discussing space, time, and causality in Section 1.3, says that observational consensus lets coherent subjects share a broadly consistent reality, and that within the snapshot there can form possibility convergence paths that are stable and continuously emphasized — causal laws appear to possess apriority; and to push this sentence to its conclusion — this apriority "is only the appearance cast by a convergence path sufficiently stable" — is what RC's paper does in Section 2.3, discussing space, time, and causality. Euclidean geometry is this sentence's most complete realization in the ancient world: its paths were stable to the point of two thousand years with zero failure, and so, seen from inside the snapshot, it could not but look like eternal truth.
Then occurred the misreading this chapter has to diagnose. The thickness of the account was read as an ontological guarantee: the postulates — a set of starting points explicitly published, publishable otherwise at any time — were read as truths beyond question. "Starting points not yet questioned" acquired the countenance of "starting points that cannot be questioned". The words of Chapter One here come due in mathematics: people forgot that the account had been opened, and so took its balance for a deposit made before the world began. The bill for this misreading is the whole content of the next chapter.
The Ledger of the Perfect Building
The customary clarifications, two of them.
First, this chapter is not saying that Euclid was wrong, still less that axiomatization was a swindle. Quite the opposite: axiomatization is among the greatest achievements in the history of mathematics, and its greatness carries a clause often overlooked — it is precisely axiomatization that preserved the interface for replacement. The whole legitimacy of the starting points lies upon the table, and the reverse of this is that to exchange the starting points uses exactly the rules Euclid himself laid down. When, two thousand years later, people exchanged the fifth postulate, they did not overturn this building; they simply laid another foundation beside it, by its own method — this capacity of the architect's was foreshadowed in Chapter Three, and Chapter Nine will follow it through in full detail. The axiomatic method promises nothing about the truth of the starting points; it promises only that every step after them can be checked. This is a far smaller promise than "eternal truth", and for that very reason a promise that has never once been defaulted.
Second, this chapter is not saying that the two thousand years were wasted. The re-verification of two thousand years is a real deposit: within the Euclidean postulates, that the angles of a triangle sum to two right angles has been verified hundreds of millions of times, valid as iron — its validity is not diminished by a hair to this day, nor will it ever be. What is to be rewritten is not any entry in the ledger but the nature of the ledger: it is the most thoroughly verified set of starting points the world has, not the only set of starting points the world could have. Chapter Three said that self-evidence is a choice — the view from inside a snapshot after countless convergences along one and the same path. The story of non-Euclidean geometry is the process of turning that sentence from an assertion into a history.
The question Part Three tracks was laid down in Chapter Three: at the cost of which cracks does a self-enclosed freedom of expansion come. This chapter has taken inventory of the building: the boundary, the starting points, the account of two thousand years without failure, and the seam visible upon the wall from the first day. The next chapter tells of a man fixing his gaze upon that seam — a Jesuit who in 1733 set the long ladder of reductio ad absurdum against this wall, climbed to the door of a world no one had ever seen, and there, at the threshold, stopped.