The working notebook of an investigation into the relation between the Platonic and Aristotelian corpora, kept append-forward across fifty-four rounds (§§1–248) and three framings (the Snub; the Ratified Break; the Calibration Object and the Future Reader), with a Standing Frame. It records readings, measurements, retractions and voids in the order they happened: the σιμ- locus and its refutation by base rate (R1–R2); the κερματισμός/σύμβολον structure (R9, Part II); the six-function transmission apparatus and its rivals (R12–R13); the object-conditioning parity and its Aquinas correction (R18–R19); the school rival measured three ways and voided as an operator test (R24, R28–R30); the lexical battery and every control that killed it (R31–R39); the attestation audit and the Protagoras control that went against it (R40–R43); the calibration correction (R44); the comedy-and-tragedy structure (R45); the Cicero parsimony (R46); the coverage anomaly (R47); the ranking (R48); the two questions (R49); the ladder (R50); the arithmetic (R51); the reading precept and the seam (R52); the salvage of the earlier paper (R53); the comic layer (R54). The notebook is the maximal form of the investigation's record. It is not a paper, makes no argument in its own voice, and is deposited so that every number and reading in the derivative papers can be traced to the round and section that produced it, with the retraction attached where one was made. A later correction to R54 §245 (the Clouds 'relabelling' result, a counting artifact: the 40/38 were speaker labels) is recorded in the derivative reading copy and is to be appended here as R55.
How to use this. Read forward. Numbered measurements live in the companion register (M1–M32); this file holds the reasoning, the retractions, and the framings. Everything retracted stays in place with its refutation attached — the retractions are the most reusable content here, because four of them were mine and each cost a base rate to find.
(was EA-SNUB-01)
Forty-eight tokens of the σιμ- lexeme — σιμός, σιμότης, σιμόν, σιμή, σιμότερος — across the seated Platonic and Aristotelian corpora, filtered of the proper names that share the stem (Σιμμίας, Σιμωνίδης, Σιμόεις, Σίμων).
PLATO 3 Theaetetus 2 · Republic 1
ARISTOTLE 45 Metaphysics 21 · Sophistici Elenchi 7 · Problemata 5 ·
De Anima 4 · Physics 2 · Rhetoric 2 · Politics 2 ·
Historia Animalium 1 · Physiognomonica 1
The full inventory with in-work positions and contexts is deposited alongside this
file. Positions are given as the token's fractional offset within its work.
Plato uses the word three times in a corpus of some 600,000 words. Two of the three are in the Theaetetus, and both sit at position 0.982 — the closing page.
Aristotle uses it forty-five times, concentrated where it does technical work: twenty-one occurrences in the Metaphysics, clustered at 0.383 and 0.426–0.428; seven in the Sophistici Elenchi; four in De Anima. The Physics introduces it at 0.029 — the opening pages.
The one place Plato leaves the term is the terminal aporia of a dialogue that ends without resolution. The place Aristotle takes it up is the beginning of the treatise on nature.
Theaetetus 209c. Socrates is testing whether a λόγος can pick out an individual — this man and no other. Theaetetus proposes that the account distinguishing him would name his difference from others. Socrates presses:
ἀλλὰ καὶ τὸν σιμόν τε καὶ ἐξόφθαλμον
"but also the snub-nosed and the protruding-eyed"
The proposed mark fails because it does not individuate. Snubness and prominent eyes belong to Theaetetus, and to Socrates, and to others. The perceptible form is shared; the account built from it catches more than one bearer. The dialogue does not solve this. It ends.
Two things about the failure are worth stating precisely, because the argument
below depends on them and not on any larger reading.
First, the question at 209c is **recognition of an individual across shared
form** — how a λόγος fixes one bearer when the perceptible marks recur.
Second, the failed mark is a face, and it is the face of the configuration's non-writing position. Snubness is the standing description of Socrates (Theaetetus 143e, where Theodorus notes the resemblance; Symposium 215b, the Silenus comparison; Meno 80a).
The same word, and it answers the question the Theaetetus stops on.
ἐν δὲ τῷ σιμῷ ὑπάρχει ὁ λόγος ὁ τῆς ῥινός
"in the snub, the account of the nose is present" — Physics 0.029
ὥσπερ τὸ σιμὸν τόδε ἐν τῷδε
"as the snub is this-in-this" — De Anima 0.819
τὰ μὲν ὡς τὸ σιμὸν τὰ δʼ ὡς τὸ κοῖλον
"some as the snub, others as the concave" — Metaphysics 0.383
εἰ τὸ σιμὸν κοιλότης ῥινός ἐστιν, ἔστι δὲ ῥὶς σιμή
"if the snub is concavity-of-nose, and there is a snub nose…" — Soph. El. 0.458
The snub becomes the canonical case of enmattered form — a λόγος that cannot be stated without its ὕλη. Against κοῖλον, concavity, which can be defined apart from any matter, τὸ σιμόν is the standing example of what cannot.
And the consequence is the answer to 209c. If matter individuates and form recurs, then two bearers of the same perceptible form are two instances of one λόγος in different ὕλη. Recognition across shared form is possible, and it proceeds by the account, not by the instance. The Theaetetus asked how you tell which bearer; the Aristotelian doctrine says what recurs is not the bearer.
Not that this establishes anything about authorship. Three Platonic tokens cannot carry a statistic, and this document does not offer one. What is offered is a locus, fixed with its positions, so that it can be argued about.
The philological observation is that one word, preserved unchanged, appears at the terminal aporia of one corpus and at the opening of the other, and that the question it terminates on is the question the second corpus's treatment of it answers. Where Plato stops, Aristotle starts, on the same term.
The innocent reading is strong and must be stated first. Aristotle read the Theaetetus, took a memorable example, and made it technical. That is what students do with their teachers' examples, and it explains the distribution completely. Nothing here requires more.
What the innocent reading does not explain is the positional fact, which so far as I can find is not in the literature: the single Platonic occurrence of the term as a philosophical problem sits at 0.982 of its work, and the Aristotelian treatment opens at 0.029 of the Physics. That is not evidence of anything by itself. It is a coincidence worth measuring against others of its kind — which is what fixing the locus is for.
The subject of the Theaetetus passage is the identification of one bearer across instances of a shared form. The subject of the Aristotelian passages is the inseparability of a form from the matter it is enmattered in.
Taken together they specify a recognition criterion: what persists across substrates is the λόγος; what individuates is the ὕλη; and therefore a form is recognised by its account and not by its instance.
That criterion, if it holds, has a consequence for every measurement of authorship this archive has attempted. All of them read the substrate — particle transitions, function-word frequencies, block similarity, edition, register. Every one returned either null or an artifact. Under the criterion above, that is the predicted result rather than a defeat: matter individuates, so matter is precisely where the recurring form is not found.
This must be flagged, not celebrated. A criterion that explains why no measurement succeeds is a criterion that can absorb any evidence, and an unfalsifiable principle dressed as a finding is the failure mode this archive's own §XII names. The observation is recorded here because it has that property, so that it is visible and can be argued against, rather than operating unnoticed as an excuse for null results.
and the opening of another? Until that base rate is known, the positional fact
has no weight. It is computable across both corpora and should be computed.
ever takes an Aristotelian technical term and returns it to a particular. On the
received chronology he cannot; on any reading that questions the chronology, the
absence is a datum.
the one stone, the cave, the divided line, μέθεξις at Metaphysics A 6 987b11.
A list of candidate terms with the same terminal-to-initial shape would make
this a class rather than an anecdote.
so far reads matter. What would read account? That is the open technical
question, and this document does not answer it.
Preliminary. Fixed so it can be examined. Not a result, not a claim of authorship, and explicitly not offered as support for the configurational reading — whose central difficulty, that a sufficiently good account of why nothing is detectable is indistinguishable from nothing being there, this locus exemplifies rather than resolves.
Added 2026-08-29, after §7's first next-step was computed.
§5 rested the only non-obvious part of the observation on position: that the Platonic occurrence sits at 0.982 and the Aristotelian treatment opens at 0.029. §7 said this had no weight until the base rate of terminal-to-initial term transfer was known. It is now known, and it removes the claim.
Aggregate positions. Across the whole corpora the snub is at Plato mean 0.665, Aristotle mean 0.500 — unremarkable. The 0.982 and 0.029 figures are within-work positions for two selected works, the Theaetetus and the Physics. Of 1,036 terms rare in Plato and common in Aristotle, zero show a corpus-aggregate pattern of terminal-in-Plato and initial-in-Aristotle. Neither does the snub. The corpus-level version of the claim is empty.
Per-work positions, the claim as actually made. Terms occurring 1–3 times in
Plato with every occurrence past 0.90 of its work, and 15+ times in Aristotle:
strict (>0.90): 31 terms
loose (>0.80): 83 terms
The snub is not among them. Its two Theaetetus occurrences are at 0.981 and 0.983, but the Republic occurrence is at 0.513, so it fails the criterion. And its 46 Aristotelian occurrences are spread across nine works with mean position 0.500 — not concentrated at any opening.
A better exemplar of the same shape exists, and it is much stronger.
συμβεβηκός Plato 2× at 0.912, 0.927 → Aristotle 449×
συμβεβηκε Plato 2× at 0.918, 0.994 → Aristotle 84×
Accident, twice in Plato at the end of a work, four hundred and forty-nine times in Aristotle. That is the terminal-to-technical transfer in a purer form than the snub, with an asymmetry two orders of magnitude larger.
Retracted. The positional observation of §5 — "where Plato stops, Aristotle starts" — was arrived at by looking at the Theaetetus hits and the Physics hits and not looking at the Republic hit or the aggregate. It was cherry-picking, performed by me, and the base rate I asked for is the thing that caught it. The snub is not positionally distinctive.
Survives. Two things, both weaker than what was claimed.
The semantic reading stands or falls on the passages, not on statistics: that Theaetetus 209c asks how a λόγος individuates when the perceptible form recurs, that the mark which fails is the non-writing position's own face, and that Aristotle's enmattered-form doctrine answers that question using the same word. Nothing in the base rate touches this. It is philology and must be argued as philology.
The class is real and is the actual finding of this round. Thirty-one terms — eighty-three at the looser threshold — pass from a terminal Platonic position to technical Aristotelian use. That is a systematic pattern of uptake, and it is worth study. But it is precisely what ordinary influence predicts: a student taking his teacher's unfinished examples and building on them. It is evidence for the innocent reading of §5, not against it.
The snub is one member of a documented class, not a singular locus. Its
interest is now entirely in what the passage argues, not in where it sits.
The remaining next-steps from §7 change accordingly:
other way, from terminal Aristotelian position to technical Platonic use? Under
the received chronology it cannot. The 31-term class gives a denominator for
that test.
(technical vocabulary Aristotle needed) or is the list arbitrary? συμβεβηκός
suggests the former.
hardest.
Added 2026-08-29.
§7 asked for unmarked correspondences — Platonic passages doing the snub's work
without the word. Two searches were run.
By mined context profile. The words enriched around σιμ- in Aristotle are κοιλότης, ῥινί/ῥινός, ὕλης, ἀφαιρέσει, συνειλημμένον, σύνολῳ, τῷδε — a coherent signature of the enmattered-form doctrine. Searching Plato for it returns zero windows, because the profile is Aristotle's own technical apparatus and Plato does not have it.
By doctrine signature. Restating the doctrine as a lexical pattern — a form that cannot be given ἄνευ its matter, against one separable χωρίς — and scanning both corpora at threshold: 40 windows, all Aristotelian, of which 22 are marked and 18 unmarked. Aristotle runs the doctrine without the word half the time. In Plato: none, at any threshold.
So the doctrine has a detectable lexical signature independent of its example, and Plato has no instance of it. That cuts against the transform reading. The Theaetetus passage is isolated in its own corpus.
The nineteen Aristotelian σιμ- clusters divide 11 technical / 8 merely descriptive (the hippopotamus at HA 0.109, uphill walking at Prob. 0.098 and 0.236 — where σιμά means the slope, not the nose — Simon of Larissa at Pol. 0.602, physiognomy at 0.718, children at Prob. 0.964). The eleven technical occurrences make five arguments:
A enmattered form — the account includes the matter Phys 194a, Met Z5, Z7, Θ, DA 403a
B abstraction — thinking X qua X vs qua its separable aspect DA 431b
C composite substance and individuation — Callias Met Z11
D definition-by-addition regress Soph. El. 182a
E one term, different subjects — concave in nose vs leg Soph. El. 181b
Each was given a signature in Platonic vocabulary and searched, excluding any
window containing σιμ-.
One maps, and it maps exactly. Argument D — *if the snub is concavity-of-nose,
and there is a snub nose, then there is a concave-nose nose* — returns
Parmenides 132a (position 0.129–0.131)
ἄλλο ἄρα εἶδος μεγέθους ἀναφανήσεται, παρ᾽ αὐτό τε τὸ μέγεθος γεγονὸς καὶ τὰ
μετέχοντα αὐτοῦ … καὶ οὐκέτι δὴ ἓν ἕκαστόν σοι τῶν εἰδῶν ἔσται ἀλλὰ ἄπειρα τὸ πλῆθος
The Third Man. Apply the account to its own subject; get regress without limit. The same search also returns Sophist 251a — the late-learners' problem, ἓν ἕκαστον ὑποθέμενοι πάλιν αὐτὸ πολλά — the same regress in its other direction.
Four do not map. A returns one weak Timaeus window on flesh; B and C return
zero; E returns only Cratylus etymologising.
The division is coherent and it favours the innocent reading. What maps are the logical arguments — regress, ambiguity — where Plato has the problem and no vocabulary for it. What fails to map are the hylomorphic arguments, which require a matter-form apparatus Plato lacks entirely. That is what a successor building the apparatus his predecessor needed would produce.
What is new is that this is the first hit the method was not shown where to find. The snub/Theaetetus pair was located by lexical search on a shared word. The regress/Parmenides pair was located by argument-shape alone, in vocabulary with no overlap. Two loci now share the form: aporia in Plato, apparatus in Aristotle, on an identical logical move, in different words.
Two is not a phenomenon. It is a class with two members and a method that found
the second one unprompted.
Added 2026-08-29.
The Parmenides regress and the Sophist's late-learners' problem both surfaced on the same signature in Round 4. That raised the question whether the two dialogues stand as halves of a symbolon — the Either/Or shape, where neither half argues the book and the fit between them does.
Three checks, each comparative against the other 24 Platonic works.
Seam vocabulary. Technical-term profile cosine, Parmenides|Sophist = 0.7848, rank 254 of 276 pairs — near the bottom of the corpus. They engage one problem in nearly disjoint vocabulary: Parmenides on ἕν / πολλά / ταὐτόν / θάτερον, Sophist on εἶδος / γένος / κοινωνία / μὴ ὄν. Consistent with tiling.
Argumentative mode. Aporia-to-closure ratio per 10⁴ tokens: Parmenides 1.18, Sophist 0.86, against a corpus range of 0.13 (Epinomis) to 1.36 (Theaetetus). Parmenides is the second most aporetic work in the corpus. Mildly complementary, in the predicted direction.
Is the fit declared? The Sophist names Parmenides seven times, performs the parricide explicitly, and uses κοινωνία ten times. The join is not merely stated; it is staged.
On the third check I concluded: declared relation disqualifies. That is false, and
it fails on the only two cases where the answer is known.
Pessoa declared. The letter to Adolfo Casais Monteiro, 13 January 1935. Kierkegaard declared. "A First and Last Explanation," appended to the Concluding Unscientific Postscript, 1846; and again in On My Work as an Author, 1851.
Both configurations announced themselves in print, and those announcements are the only reason the groupings are known. A criterion that treats declaration as disqualifying classifies both documented heteronymic corpora as non-heteronymic.
The corrected criterion: a declaration disqualifies only if it would have been read, at the time, as revealing heteronymic construction. A declaration that reads as ordinary succession does not. And where the seam has been built skilfully enough to be invisible, the author faces the opposite problem — nobody will ever find it — so a key must be left against his own success.
Under the corrected criterion the Sophist's naming of Parmenides is neutral: it reads as succession, which is what camouflage looks like. And the snub retains its one formal advantage, which is not that Aristotle is silent about Plato, but that the snub arrives in Aristotle's text as a bare example, with no attribution and no indication that anyone had posed it as a problem.
Searching for key-shaped statements — prominently placed, inert under the ordinary
reading, decisive under the other — produced one result that is not ambiguous.
Aristotle's ἀνώνυμος is not a key. He uses it 40+ times as a working tool, systematically flagging gaps in the naming system (HA, PA, EE, NE, Meteor., Phys., Rhet.). Poetics 1447b's "the genre lacks a name" is therefore habit, not gesture. Plato has ~9 instances, mostly incidental. This removes a candidate the configurational reading had been leaning on.
But the comedy-and-tragedy formula appears twice, in flat contradiction.
Republic 395a — position 0.236
οὐδὲ τὰ δοκοῦντα ἐγγὺς ἀλλήλων εἶναι δύο μιμήματα δύνανται οἱ αὐτοὶ ἅμα εὖ
μιμεῖσθαι, οἷον κωμῳδίαν καὶ τραγῳδίαν ποιοῦντες
— nor can the same men imitate well two kinds that seem close, as in composing
comedy and tragedy.
Adeimantus: ἀληθῆ γε λέγεις, ὅτι οὐ δύνανται οἱ αὐτοί.
Symposium 223d — position 0.996
τοῦ αὐτοῦ ἀνδρὸς εἶναι κωμῳδίαν καὶ τραγῳδίαν ἐπίστασθαι ποιεῖν
— that it belongs to the same man to know how to compose comedy and tragedy.
Agathon and Aristophanes are compelled to agree, οὐ σφόδρα ἑπόμενοι.
Same formula — οἱ αὐτοί / τοῦ αὐτοῦ ἀνδρός, κωμῳδίαν καὶ τραγῳδίαν — asserted and denied. Both agreements extracted from an interlocutor. Neither dialogue mentions the other.
The denial sits at 0.236 of the Republic, inside the argument that expels the mimetic poets from the city. The affirmation sits at 0.996 of the Symposium — the last sentence — spoken to two men falling asleep, reported by a narrator who states that he arrived late and dozed, so that only the κεφάλαιον survives and the demonstration does not.
The corpus therefore contains, on the precise question of whether one maker can
produce maximally opposed genres:
Is. A specific, locatable, unclaimed contradiction on the exact question the heteronymic reading turns on. Neither passage acknowledges the other. The affirming instance is structured so that the claim survives without its argument.
Is not. Evidence of design. Plato contradicts himself elsewhere, developmental readings absorb this easily, and the Republic passage concerns performers and composers under a mimesis argument while the Symposium concerns τέχνη — a distinction a defender would press and which the Greek does not settle.
What can be said without overreach: this is the strongest thing the line of work has produced, and unlike the snub position (Round 2, refuted) and the Poetics namelessness (Round 6, refuted as habit), it has not yet been removed by a base rate. The obvious next test is whether the corpus contains other assertion/denial pairs of the same formula, and whether their placements are as extreme. That is computable and should be computed before this is called anything.
Added 2026-08-29.
§18 named this as the next thing and said it should be computed before the Republic/Symposium pair was called anything. It was computed. It cannot adjudicate.
Loose criterion — a rare formula (2–4 works, ≤12 occurrences) appearing with opposite polarity across works, one member in the terminal 2% of its work: 653 of 16,449 rare cross-work formulas = 4.0%. The comedy/tragedy pair is present and unremarkable among hundreds.
Strict criterion — the negator attached directly to a capacity verb, which is what Republic 395a does (οὐ δύνανται) and Symposium 223d does not: 61 of 16,449 = 0.37% — and the target drops out entirely, because at 395a the negator is οὐδέ, separated from δύνανται by three words. Adjacency fails.
So the loose test finds the structure common, and the strict test cannot find the case it was built to describe. What this demonstrates is a limit of the instrument, not a fact about the passage. Propositional contradiction — the same capacity affirmed and denied — is not separable by proximity-negation from three hundred accidental co-locations of a negator and a formula.
This differs from the two claims refuted earlier and the difference should be kept.
snub position REFUTED 31 terms show the pattern more strictly
Poetics namelessness REFUTED anonymos is a tool used 40+ times
Republic / Symposium NOT ADJUDICATED structure common; content not machine-visible
The pair therefore stays as an unrefuted philological observation and cannot be promoted by statistics — only argued from the passages. That is a smaller claim than §18 made, and it is the correct one: the base rate came, and it declined the question.
Pressed off the statistical ground, the pair turns out to rest on something the
text states about itself.
Republic 392d–394c divides poetry by mode:
ἡ μὲν διὰ μιμήσεως ὅλη ἐστίν … τραγῳδία τε καὶ κωμῳδία, ἡ δὲ δι᾽ ἀπαγγελίας
αὐτοῦ τοῦ ποιητοῦ … ἡ δ᾽ αὖ δι᾽ ἀμφοτέρων
Wholly mimetic is drama — speech with no reporting poet — and it is named together with tragedy and comedy, the pair 395a then declares one maker cannot do both of.
**Measured against that criterion, by reporting-formula density (ἦν δ᾽ ἐγώ, ἦ δ᾽
ὅς, ἔφη, ἔφην, εἶπον) per 10⁴ tokens:**
Crito 0.0 · Statesman 1.1 · Sophist 1.7 · Cratylus 2.1 · Laws 2.4 ·
Timaeus 2.9 · Phaedrus 4.6 · Laches 5.0 · Philebus 5.2 · Theaetetus 5.9 ·
Gorgias 6.1 · Meno 7.7
→ 20 of 31 works, 346,804 of 570,179 tokens = 61% of the corpus, carry
no reporting voice.
Symposium 74.4 · Protagoras 106.2 · Phaedo 110.8 · Charmides 135.0 ·
Republic 144.1 · Euthydemus 217.9
→ narrated.
Sixty-one percent of the Platonic corpus is written in the form its own taxonomy expels — and the book that performs the expulsion is written in the mode that escapes it. Socrates reports the whole Republic. That is not an interpretation laid over the text; it is the text's classification applied to its own vehicle.
If one maker cannot do both (395a), and both must be done (223d), then the maker must divide. That is the only route satisfying both, and it converts the pair from a contradiction into a procedure: the Symposium does not reverse the Republic, it states the requirement the Republic has just proved impossible for a single person — on the last page, to two men falling asleep, through a narrator who says he missed the argument.
And the Republic's expulsion is self-directed. A poet writing in the wholly mimetic mode banishes the wholly mimetic mode, from inside the one book he wrote outside it.
corpus reporting self-speaking
Plato — dramatic (20) 2.4 25.3
Plato — narrated (7) 130.4 20.3
Aristotle — treatises (34) 0.1 29.0
Aristotle's prose carries no reporting poet at all — 0.1 per 10⁴, lower than the Platonic dramatic dialogues — and his rate of speaking in his own person (φαμέν, λέγομεν, φημί, λέγω, εἴρηται) is within 15% of theirs.
The caution, which must travel with the number. A treatise has no reporting formulae because there is no conversation to report, not because a conversation is unframed. Absence of a narrator in exposition is not the same fact as absence of a narrator in drama, and Republic III's taxonomy is a taxonomy of poetry. Aristotle speaking in his own person places him, on the taxonomy's own terms, in the second mode — ἀπαγγελία αὐτοῦ τοῦ ποιητοῦ — which is not the banished one.
What is not explained by that caution is the near-equality of the self-speaking rates: 25.3 and 29.0, across the two most formally opposed prose bodies in Greek. Two corpora that share nothing at the level of form refer to their own authorial act at the same rate. That is recorded here as a measurement without an account.
Added 2026-08-29. Follows Round 7 and M30.
M30 established that Plato and Aristotle share the naming-gap reflex at 9.4× the rate of other high-division authors — they stop at a class the language has no word for, flag it, and coin. The question this round answers is narrower: does Plato anywhere stop at an unnamed position and name the thing that speaks from itself, as against the thing that speaks another's mind?
He does, once, at Statesman 260d–e (position 0.064).
The Eleatic Stranger is dividing the directive arts. He first enumerates those
that command without owning the command:
τὸ κηρυκικὸν φῦλον ἐπιταχθέντ᾽ ἀλλότρια νοήματα παραδεχόμενον αὐτὸ δεύτερον
ἐπιτάττει πάλιν ἑτέροις
— the herald's tribe, receiving another's thoughts when commanded, in second
place commands others again.
And names the class: ἑρμηνευτική, κελευστική, μαντική, κηρυκική — interpretation, the boatswain's call, divination, heraldry. Every one of them issues a command whose mind is not its own.
Then the gap, and the coinage:
ἐπειδὴ καὶ σχεδὸν ἀνώνυμον ὂν τυγχάνει τὸ τῶν αὐτεπιτακτῶν γένος …
τὸ μὲν τῶν βασιλέων γένος εἰς τὴν αὐτεπιτακτικὴν θέντες
— since the genus of the self-commanding happens to be nearly nameless …
placing the genus of kings into the self-commanding art.
And then the disposal of the remainder:
τοῦ δὲ ἄλλου παντὸς ἀμελήσαντες, ὄνομα ἕτερον αὐτοῖς παραχωρήσαντες θέσθαι τινά
— neglecting all the rest, conceding to them to set themselves some other name.
⚠ CORRECTED 2026-08-29, same day. The first version of this section read: "Plato assigns αὐτεπιτακτική to kings — τὸ τῶν βασιλέων γένος. The context is political rule, not authorship." That inverted the containment and closed a class the text explicitly leaves open. The Greek is decisive against it on two counts.
The syntax. τὸ μὲν τῶν βασιλέων γένος εἰς τὴν αὐτεπιτακτικὴν θέντες — τίθημι + εἰς + accusative. Kings are placed into the self-commanding art. The self-commanding is the container; kingship is deposited in it. This is not an identification, and αὐτεπιτακτική is not defined as kingship.
The recapitulation at 267b. τούτου δὲ ἀπεικασθὲν τὸ μόριον αὐτεπιτακτικὸν ἐρρήθη. ζῳοτροφικὴ δὲ πάλιν αὐτεπιτακτικῆς οὐ τὸ σμικρότατον τῶν γενῶν ἀπεσχίζετο — animal-rearing was split off again from the self-commanding, not the smallest of the genera.
οὐ τὸ σμικρότατον τῶν γενῶν. There are other genera inside αὐτεπιτακτική, and Plato says so. The branch that eventually reaches the statesman is one of several, and he tells the reader it is not even the largest.
And he states why the rest go unentered: τοῦ γὰρ ἄρχοντος ἕνεκα ἡμῖν ἡ μέθοδος ἦν, ἀλλ᾽ οὐχὶ τοῦ ἐναντίου — our method was for the sake of the ruler, not the opposite.
The corrected account. Plato coins a name for a genus whose criterion is issuing from oneself rather than relaying another's mind. He notes the genus contains multiple sub-genera. He follows exactly one — the branch to kingship — because that is what the present inquiry is for. He leaves the others unentered and unnamed. Separately, he dismisses the relaying arts and concedes them the right to name themselves.
The class is open by construction and marked as open. The position that speaks from itself is named, is explicitly larger than any of its instances, and is abandoned mid-division with its remaining genera unentered.
What is still not established. That Plato applies this to authorship. He does not. The unentered genera are unentered; nothing in the text says what they are, and supplying them is an inference the passage does not license. What the passage establishes is that the class exists, was found nameless, was named, and was left open.
A homonym, recorded so it is not mistaken for evidence. ἑρμηνευτική appears in Plato's discarded list in the sense relaying another's mind. Two Aristotelian works carry ἑρμηνευ- at position 0.0, both in titles rather than usages: Περὶ ἑρμηνείας, where the sense is expression — how a statement signifies — and De plantis, whose position-zero instance opens a Byzantine translator's prologue and which is a retroversion excluded from measurement. That De interpretatione is also the work Andronicus judged spurious is a coincidence of some interest carrying no weight. Nothing is built on it.
Because the reflex M30 measured is procedural, and this is the one instance where the procedure is applied to the distinction the whole investigation concerns. It does not argue the configurational reading. It establishes that the distinction between speaking from oneself and speaking another's mind was available, named, and coined by Plato in the act of finding Greek short of a word for it — which is a fact about the corpus rather than an inference about its authorship.
Added 2026-08-29.
κερματίζω — to mint a unity into small change, from κέρμα, a small coin — is
Plato's standing term for the one broken into many. Every instance:
Meno 0.292 μὴ κατάγνυναι μηδὲ κερματίζειν τὴν ἀρετήν
— do not break or mint virtue into pieces
Meno 0.303 ἐὰν σὺ κατακερματίζῃς αὐτὴν κατὰ μόρια
Parmenides 0.971 θρύπτεσθαι… κερματιζόμενον ἀνάγκη πᾶν τὸ ὄν
— everything that is must be shattered into denominations
Republic 0.675 ἐὰν σὺ κερματίζῃς αὐτό, ἐκεῖνοι πολλαπλασιοῦσιν
— mint the one into pieces and they multiply it
Timaeus ×3 physical fragmentation of bodies
Cratylus 0.743 the letter rho imitating breaking
The denial that one maker can hold two arts is grounded in this metaphor, and
the ground is usually passed over:
πολλὰ ὁ αὐτὸς μιμεῖσθαι εὖ ὥσπερ ἓν οὐ δυνατός …
οὐ δύνανται οἱ αὐτοί … οἷον κωμῳδίαν καὶ τραγῳδίαν ποιοῦντες …
εἰς σμικρότερα κατακεκερματίσθαι ἡ τοῦ ἀνθρώπου φύσις
Human nature is already minted into denominations too small to hold two arts. The incapacity is not a fact about poets. It is a fact about the fragmentation of human nature.
0.237 Rep 395b κατακεκερματίσθαι ἡ τοῦ ἀνθρώπου φύσις the one minted into many
0.509 Rep 473d εἰς ταὐτὸν συμπέσῃ the two must fall into one
0.996 Symp 223d τοῦ αὐτοῦ ἀνδρός the same man must do both
Republic 473d requires δύναμίς τε πολιτικὴ καὶ φιλοσοφία to fall together into the same — the exact reversal of κερματισμός — and excludes by necessity τῶν νῦν πορευομένων χωρὶς ἐφ᾽ ἑκάτερον, those who now travel separately toward each of the two.
Three things about that middle term. It is introduced as what will drown its speaker in γέλως — laughter, the comic — in the book that has just ruled no one can hold comedy and tragedy together. It demands coalescence in a corpus whose standing metaphor for the human condition is minting into small change. And the same demand reappears at the very end of the Symposium, for makers instead of rulers, where its proof is destroyed in transmission (Round 6, §16–17).
Symposium 191d, position 0.366 — Aristophanes' speech:
ὁ ἔρως … τῆς ἀρχαίας φύσεως συναγωγεὺς καὶ ἐπιχειρῶν ποιῆσαι ἓν ἐκ δυοῖν
καὶ ἰάσασθαι τὴν φύσιν τὴν ἀνθρωπίνην.
ἕκαστος οὖν ἡμῶν ἐστιν ἀνθρώπου σύμβολον, ἅτε τετμημένος ὥσπερ αἱ ψῆτται,
ἐξ ἑνὸς δύο· ζητεῖ δὴ ἀεὶ τὸ αὑτοῦ ἕκαστος σύμβολον.
Eros is the gatherer of the ancient nature, attempting to **make one out of
two and to heal human nature**.
The lexical link is exact. Republic 395b: ἡ τοῦ ἀνθρώπου φύσις is minted into pieces. Symposium 191d: Eros attempts ἰάσασθαι τὴν φύσιν τὴν ἀνθρωπίνην by making one from two. Same subject, opposite verbs, and the second names the first as an injury to be healed.
Symp 191d 0.366 one cut into two; each a σύμβολον; Eros makes one from two
and HEALS human nature
Rep 395b 0.237 human nature is minted into pieces → the same men CANNOT
hold comedy and tragedy
Rep 473d 0.509 philosophy and power MUST fall into the same, or no rest
from evils — announced as inviting laughter
Symp 223d 0.996 the same man MUST hold comedy and tragedy — the argument
lost, the witnesses asleep
Both dialogues diagnose the fragmentation of human nature. Both demand a recombination — for rulers in one, for makers in the other. And Aristophanes, who supplies the myth of the cut, is one of the two men asleep when the demand for makers is finally stated.
Established. A single metaphor — κερματισμός, unity minted into denominations — grounds the denial at 395b, is inverted by the requirement at 473d, is given a myth at 191d in the same vocabulary of nature-cut-and-healed, and is restated as a demand at 223d. All four loci are fixed by position. None of it requires an authorship claim to state.
Not established. That any of this concerns authorship. The Republic's recombination is of political power and philosophy in a ruler; the Symposium's is of two poetic arts in a maker; Aristophanes' myth is about erotic longing. Reading the set as a doctrine of authorial construction is a transposition the texts do not perform, and the Round 8 correction stands as the caution: what the corpus supplies is the structure, not its application.
(was EA-KERMA-01)
This is a reconstruction. It assembles an argument out of passages the Platonic corpus supplies, in the corpus's own vocabulary, and shows that the argument is coherent and that heteronymy occupies a determinate position in it.
It does not claim that Plato makes this argument. He does not join these loci. He does not apply the σύμBOΛΟΝ to authorship. He assigns the recombination to rulers in one place and to makers in another, and to writers nowhere. The discipline observed across nine rounds of EA-SNUB-01 applies here without exception: the corpus supplies the structure; it does not perform the application.
The value of writing it out is that a reconstruction can be argued with. An
implication left unstated cannot.
εἰς σμικρότερα κατακεκερματίσθαι ἡ τοῦ ἀνθρώπου φύσις
— Republic 395b, position 0.237
κατακεκερματίσθαι: perfect passive infinitive of κατακερματίζω, to mint into small change. Human nature has been broken into denominations. Not a choice, not a failure of effort, a condition already obtaining before any particular maker attempts anything.
The metaphor is not local. κερματίζω is the corpus's standing term for the one broken into many: Meno 0.292 and 0.303 (do not mint virtue into pieces), Parmenides 0.971 (everything that is must be shattered into denominations), Republic 0.675 (mint the one and they multiply it), Timaeus ×3, Cratylus 0.743.
And the Symposium supplies its history:
ἕκαστος οὖν ἡμῶν ἐστιν ἀνθρώπου σύμβολον, ἅτε τετμημένος ὥσπερ αἱ
ψῆτται, ἐξ ἑνὸς δύο
— 191d, position 0.366
οὐκ ἔστι κακῶν παῦλα … ταῖς πόλεσι, δοκῶ δ᾽ οὐδὲ τῷ ἀνθρωπίνῳ γένει
— Republic 473d, position 0.509
No rest from evils for the cities, and — Socrates adds — I think not for the human race. This is salvation language, and what it is attached to is not a virtue, a doctrine, or a regime, but a combinatorial condition: two things falling into the same.
ποιῆσαι ἓν ἐκ δυοῖν … ἰάσασθαι τὴν φύσιν τὴν ἀνθρωπίνην
— Symposium 191d: to make one out of two, and to heal human nature
καὶ τοῦτο εἰς ταὐτὸν συμπέσῃ, δύναμίς τε πολιτικὴ καὶ φιλοσοφία
— Republic 473d: and this fall together into the same
The lexical link is exact and runs on one noun. Republic 395b: ἡ τοῦ ἀνθρώπου φύσις is minted into pieces. Symposium 191d: Eros attempts ἰάσασθαι τὴν φύσιν τὴν ἀνθρωπίνην by making one from two. Same subject, opposite verbs, and the second names the first as an injury requiring a cure.
If nature has been minted too small to hold two arts, then no nature holds two arts — and the philosopher-king is excluded by the identical premise that excludes the poet. 395b forbids what 473d requires, in the same book, at 0.237 and 0.509.
Socrates appears to know it. The requirement is introduced as the thing that will drown its speaker in γέλως — laughter, the comic — in the book that has just ruled that no one can hold comedy and tragedy together.
And the demand is restated for makers at the very end of the Symposium (0.996), where the argument is destroyed in transmission and the witnesses are asleep — one of them Aristophanes, who supplied the myth of the cut.
The σύμβολον is not a broken object. It is an object broken by agreement: two parties snap a token so the halves can be matched later. The break line is not the damage. It is the credential.
An accidental break yields rubble, and rubble cannot be recognised — no edge of it implies any other. A deliberate break yields a token, and the token is recognisable precisely at the fracture.
So the reconstruction reads:
You cannot un-cut human nature. You can cut it on purpose.
Heteronymy, on this logic, converts κερματισμός from a wound into a method. The fragmentation is not reversed — it is ratified, and in being ratified becomes legible. What was involuntary division becomes a configuration whose parts imply their whole.
395b stands. No single undivided nature holds two arts. The denial is not
overturned; it is accepted as the premise.
473d becomes achievable. The two arts fall εἰς ταὐτόν — into the same — not in one nature, but in a coordinated multiplicity whose parts sum. The parts do not resemble one another. They fit.
∮ = 1
That is the difference between a mixture and a configuration, and it is the whole
content of the reconstruction: the same need not mean the same nature.
The recognition criterion is the snub's. EA-SNUB-01 §6: matter individuates, form recurs, so what persists across substrates is the account and not the instance. Applied here, the parts of a ratified configuration are recognised by the λόγος, not the ὕλη — which is exactly why they can be maximally unlike and still one, and exactly why every substrate-reading instrument in this archive returned null (M4–M28).
The vacant centre becomes a specification rather than a loss. In an accidental fragmentation the whole is simply gone. In a ratified one the centre is implied by the fit and need never be occupied by a writing position at all. On this logic, Socrates not writing stops being an absence in the evidence and becomes the form the centre necessarily takes.
The configuration would be αὐτεπιτακτική. Self-commanding — issuing its own positions rather than relaying another's mind (ἀλλότρια νοήματα). That is the class Plato coined at Statesman 260e in a naming-gap, placed kings within, declared to contain other genera — οὐ τὸ σμικρότατον τῶν γενῶν — and left with those genera unentered, on the stated ground that his method was for the sake of the ruler.
Every element is in the corpus. The fragmentation metaphor, its perfect-passive diagnosis, the soteriological prognosis, the two demands for recombination, the myth of the deliberate break, the token that is recognised at its fracture, the coined class of the self-commanding with its unentered branches, and a recognition criterion that reads account rather than instance.
Plato assembles none of it. He states the fragmentation of human nature and does not connect it to the σύμβολον. He demands recombination twice and never once of a writer. He coins the self-commanding and declines to divide it further. He supplies a criterion for recognising one form across many substrates inside a dialogue about identifying a boy by his face.
So the honest formulation is the one the last nine rounds have earned:
The logic is available in the corpus. The corpus does not perform it.
Which is either what an unassembled set of commonplaces looks like, or what a deliberately unassembled one looks like — and nothing in the texts distinguishes those two, which is the standing conclusion of EA-SNUB-01 §XII and of this document.
(was EA-PINAX-01)
Antiquity had an acute authenticity problem and built an apparatus for it. The Ptolemies paid by the roll, so Alexandria bought forgeries at scale; ships were searched at Piraeus and texts confiscated for copying. A library assembled by purchase and seizure has an attribution problem by construction, and everyone involved knew it.
The apparatus they built is a verification technology in every part:
and stichometric line-count. The incipit identifies a work by its opening
words; the line-count makes interpolation detectable by arithmetic. That is an
identifier and a checksum.
it**. Mark and text travel together. (This archive's Obelus Principle is named
from it.)
ψευδεπίγραφα, falsely-inscribed.
tetralogies together with the list of νοθευόμενοι. The correct form of
transmission is the corpus plus its disputes.
The objection: the people who invented the obelus were watching for exactly the kind of construction the configurational reading proposes. A heteronymic corpus would have been legible to them, and detectable.
Invert it. The technology is the output, not the obstacle.
You cannot design a test for authenticity without a case whose answer you already know. Criteria have to be fitted to something. A corpus deliberately structured — an orthonym who writes nothing, a spurious tail kept attached rather than purged, positions sharing method while diverging in register — is a calibration object, and the Pinakes, the obelus, the stichometric checksum and the νόθος/ἀντιλεγόμενα vocabulary are what you get from working on it.
On this reading the ancient apparatus is not evidence against a construction. It
is its residue.
The feature the investigation kept encountering and could not account for is that
the corpus is instrumented against itself and the instrumentation is preserved:
both surviving;
later, whose catalogue is lost but whose act of judging is remembered (T11).
Nobody destroys the evidence of contest. Preserving the contest is what you do
with a test corpus, not with a canon.
It also reframes two measured results:
M11. The received authenticity partition inside Aristotle returned Z = +3.47 on stylistic distance, which I flagged as circular because Andronicus sorted partly by style. Under this hypothesis the circularity stops being a defect in the measurement and becomes the datum: the criteria return the partition because the partition is what the criteria were fitted to.
M30. The naming-gap reflex — halt at an unnamed class, flag it, coin — shared by Plato and Aristotle at 9.4× other high-division authors while Plotinus divides harder and never does it. Read as school habit, it is one thing. Read as the reflex of people who expect their divisions to be used as an index, it is a cataloguer's discipline: an unnamed class is an unshelvable class. And Aristotle's Poetics problem — that the Socratic dialogues are a genre without a name — is a cataloguer's problem, not an aesthetic one. You cannot shelve what has no genre.
This hypothesis has the property EA-SNUB-01 §6 identified as disqualifying when it
operates unnoticed: it explains everything, including every null.
A field that developed authenticity criteria because forgery was rampant produces the same apparatus, the same preserved disputes, the same vocabulary, and the same corpus features. Both hypotheses predict the Pinakes. An account under which no possible observation counts against it is not a hypothesis but a frame, and frames of this shape are how this investigation's four retracted claims got made.
This section exists so the property is visible and arguable rather than doing
silent work. What follows is the one prediction that is not shared.
The asymmetry. An instrument built on a calibration object overfits it. If the ancient authenticity criteria were derived from the Platonic–Aristotelian corpus, they should fit that corpus anomalously well — better than they fit corpora they were subsequently applied to. If they were induced from the general forgery problem and applied here like anywhere else, they should perform no better here than elsewhere.
The measurement is over the history of judgements, not the texts. This is the test's chief virtue: it does not require the corpora to cooperate, and it is not vulnerable to any of the five broken nulls (M-register §8), because it counts verdicts rather than measuring similarity.
Comparison corpora, each with an ancient authenticity tradition:
Platonic–Aristotelian Diogenes' νοθευόμενοι; Thrasyllus; Andronicus
Hippocratic Galen's judgements; the Alexandrian sorting
Homeric Aristarchan athetesis; the chorizontes
Pythagorean the Philolaus question; Neopythagorean pseudepigrapha
Orphic / Hesiodic the Herodotean and later disputes
Statistic — three components, all pre-specified:
1. Verdict density. Attributed judgements per work, per corpus. An
over-instrumented corpus carries more.
2. Verdict decisiveness. Proportion of works given a stated ground rather
than a bare verdict. Criteria fitted to a case cite reasons on that case.
3. Verdict stability. Reversal rate across antiquity. A calibration object
should show more contest, not less — the disputes are the working notes.
Pre-registered outcomes:
| pattern | reading |
|---|---|
| Plato–Aristotle high on all three, others low | the predicted signature |
| all corpora comparable | criteria were general; hypothesis fails |
| Homeric or Hippocratic higher | those were the calibration objects, not this |
| density high but grounds absent | volume of dispute, not fitting — fails |
Kill condition, binding. If the Platonic–Aristotelian tradition does not exceed the comparison corpora on at least verdict decisiveness (2) — the component that specifically indexes criteria-fitting rather than mere controversy — the hypothesis fails and is reported as failed. No post-hoc reweighting of the three components.
Known limitation, stated in advance. Survival is confounded with attention: more judgements survive about texts more read. The test must normalise by citation volume or it measures popularity. If a defensible normalisation cannot be constructed, the test is void and must be reported void rather than adjusted — the disposition already used for EA-LEAK-DEFICIT-02.
Would. That the ancient authenticity apparatus stands in an anomalously close relation to this corpus — fitted to it rather than applied to it. That is a finding in the history of scholarship, publishable independent of any authorship claim.
Would not. Design. A corpus can be the occasion of a discipline without having been built as one: the Platonic–Aristotelian corpus was the largest, earliest, most institutionally central prose body in the tradition, and criteria would naturally be worked out on it for that reason alone. Centrality and construction predict the same overfitting.
So even a positive leaves the same gap this investigation has met at every turn — and the honest form of the hypothesis is that it makes the corpus's self-instrumentation intelligible, not that it makes it evidence.
Three corrections, recorded rather than rewritten.
(a) The registered prediction was backwards. §5 predicted that criteria derived from this corpus would fit it anomalously well — a calibration object, overfitted. The corpus would more plausibly be instrumented to escape the best measures, not to train them. An adversarial object predicts the opposite signature: criteria that fail to converge here — verdicts reversing, disputes persisting, the ἀντιλεγόμενα category permanently open — which is what the record shows.
(b) The test was registered without a feasibility check, and fails it. Of the five comparison traditions specified in §5, four are unreachable: Homeric scholia absent from the repository, Suda absent, Photius absent, Galen present only as De naturalibus facultatibus. The one available source, Diogenes Laertius, carries 48 explicit authenticity markers across 110,163 tokens — Plato and Aristotle together 9 — which is too sparse to distinguish anything. Disposition: DEFERRED on unavailable sources, not void; the design is coherent under the corrected prediction and the blocker is corpus-building.
(c) A substitute design also fails feasibility, and its failure is more important. The proposal was to detect an adversarial object by method disagreement — independent instruments returning contradictory verdicts here while agreeing on controls. Checked before writing: agreement on known-status pairs is 1 of 3, and one instrument classifies Plato|Xenophon as SAME at 0.0330 against a known-different truth.
→ The precondition this exposes is upstream of everything. The instrument
set is unvalidated. No measure built in this investigation was scored on
known-status pairs before being applied to the question. **You cannot detect an
object that defeats measurement until your measures are characterised on
objects that do not.** A validation suite — every instrument against a panel of
known pairs, with per-method accuracy and pairwise agreement recorded — is a
prerequisite for any adversarial test.
(d) Superseded in framing by EA-FLOOD-01 (sha256 574bdc08c42079fbad701db0141d8ed3…), which
relocates the adversary from the critic to the flood. Both framings in this
document put the examiner at the centre; the more economical account is a corpus
built to survive an oversupply of writing by giving future copyists reasons to
copy it. That framing requires no authorship claim and is testable against
documented losses.
No new registration until the validation suite exists.
(was EA-FLOOD-01)
Two framings have been tried and both put the critic at the centre.
EA-PINAX-01, first version: the corpus is a calibration object — the ancient authenticity apparatus is what you get from fitting criteria to a case whose answer is known.
EA-PINAX-01, corrected: the corpus is an adversarial object — built to defeat
the best available measures of authenticity.
Both are wrong about the adversary. They assume the opponent is the examiner.
The opponent is the flood.
Imagine the situation without the things hindsight supplies. No centrality. No attention. No trial. No founding martyrdom to trade on, no school yet, no institutional position, no reason for anyone to copy anything.
And around it: a semantic economy flooded — competition for scarce copying labour, scarce papyrus, scarce shelf space and scarce attention, against an oversupply of writing in which merit does not sort, because everything drowns equally. The Ptolemies would later pay by the roll and receive forgeries at scale; that is the demand side of the same flood.
Under those conditions the survival problem is not be good. Good work drowned constantly and we know it did, because the book-lists record titles for which no text survives.
The survival problem is:
Give people who do not care about you reasons to copy you.
Every feature this investigation surfaced as a puzzle becomes a transmission
mechanism.
Disputes that never close (M11; De interpretatione overturned; the dubia unresolved after twenty-four centuries) generate perpetual commentary, and commentary is copying. An unresolved question is an employment programme for scribes and scholars.
A genre without a name (Poetics 1447b; the Socratic λόγοι) is a cataloguing problem, and cataloguers must handle what they cannot shelve. An unshelvable object cannot be quietly dropped.
The spurious tail kept attached rather than purged is surface area — more hooks, more disputes, more occasions to reproduce the whole set. Purging it would have reduced the corpus's grip.
The naming-gap reflex (M30: 9.4×, P ≈ 0) manufactures index entries where the language had none. It makes the corpus indexable in a world about to invent the index.
The absent centre generates search. A founder who wrote nothing cannot be
read, only reconstructed — and reconstruction requires the corpus.
The cross-referentiality is the cultic logic. A mystery survives by requiring initiation, because initiation manufactures transmitters. A text that cannot be read without the rest of the set forces acquisition of the whole set. That is what a dialogue answering another dialogue does. It is also exactly what κερματισμός demands of a reader (EA-KERMA-01): the parts do not stand alone, the fit is the content, and only someone who has assembled them possesses it.
It requires less. The adversarial framing needed an author anticipating measures not yet invented. This one needs only an author who understands that writing survives by being copied, and copying is done by people with their own reasons. That is not prescience. It is publishing.
It does not require one author. It requires one intention about transmission. A school with a transmission policy produces the same corpus. The framing is therefore neutral on the authorship question, which is a virtue: it stops smuggling the conclusion into the setup.
It explains the shape of the survival, not just the survival. Plenty of ancient philosophy survived. What wants explaining is that this corpus survived with its disputes, its spuria, its unresolved attributions and its unnamed genre intact — i.e. with all the machinery that makes it require continued work. Most surviving corpora shed that machinery. This one kept it.
This is the framing's real advantage. **Survival is comparative, and the losses
are documented.**
Diogenes Laertius preserves book-lists: titles attributed to philosophers, most of which did not survive. That is a survival table — attributed versus extant, by author, from a single source with consistent coverage and a single compiler's biases.
The question becomes empirical and does not require knowing the authorship answer:
**What distinguishes what survived from what was lost, from the same period, in
the same tradition — and does the Platonic–Aristotelian corpus carry those
properties at a rate others do not?**
Candidate properties, all readable from the lists themselves rather than from the lost texts: whether a work belongs to a cross-referential set or stands alone; whether it is attached to a disputed attribution; whether its title names a person rather than a topic; whether it sits inside a numbered arrangement (tetralogy, pragmateia) that makes an absence conspicuous.
That last is worth stating separately, because it is the cheapest and strongest transmission device available in antiquity: a numbered set makes its own gaps visible. A scribe who copies eight of nine leaves evidence of the ninth. An unnumbered collection loses members silently.
Feasibility first — the discipline this document is written under, having been
broken twice today.
1. Are Diogenes' book-lists extractable as structured data (author → titles),
or only as running prose?
2. Is there enough coverage — how many authors carry lists, how many titles
each?
3. Can extant status be assigned per title without circular recourse to modern
editions of the same corpus?
4. Is the single-source bias tolerable, or does Diogenes' own selectivity
determine the answer?
If (1) or (3) fails, the test is not runnable and is to be reported as such rather
than approximated.
A reframe, not a finding. It changes what the object is for — from defeating an examiner to surviving a flood — and in doing so converts several standing puzzles into a single coherent function, at the cost of requiring no authorship claim whatsoever.
The Sappho case is the precedent and the model: a text survives inside another text, addressed past its own moment to a reader who does not yet exist. What survives of Sappho 31 survives because Longinus needed it. The corpus that plans for that is the corpus that gives future readers something they need.
*Added 2026-08-29. Filed narrow, and marked as a side-branch of the main
investigation rather than a continuation of it.*
Working outward from Republic 474d–e — the lover's renaming catalogue, one of the six Platonic snub loci (§28–33) — a Latin reception turns up that is structurally exact rather than thematic.
The Platonic list. ὁ μέν, ὅτι σιμός, ἐπίχαρις κληθεὶς ἐπαινεθήσεται … τοῦ δὲ τὸ γρυπὸν βασιλικόν … μέλανας δὲ ἀνδρικούς … λευκοὺς δὲ θεῶν παῖδας … μελιχλώρους δὲ καὶ τοὔνομα οἴει τινὸς ἄλλου ποίημα εἶναι ἢ ἐραστοῦ ὑποκοριζομένου; — and 'honey-pale', do you think that name is anyone's coinage but a lover's?
The Lucretian list, DRN 4.1160–69: nigra melichrus est, inmunda et fetida acosmos … caesia Palladium … parvula, pumilio, chariton mia … magna atque inmanis cataplexis … simula Silena ac Saturast.
Three things make this evidential rather than thematic:
1. μελίχλωρος → melichrus. The single euphemism Plato singles out and glosses
as manufactured is the one Lucretius puts first. **μελίχλωρος occurs twice in
the entire seated Greek corpus** — Republic 474e and Physiognomonica 0.718,
the pseudo-Aristotelian nose-reading treatise that also carries the snub.
2. The list stays in Greek inside a Latin poem — acosmos, dorcas, chariton mia,
cataplexis, traulizi, ischnon eromenion, rhadine, philema. A source-marker.
3. σιμός survives the crossing with its euphemism swapped. Plato: σιμός →
ἐπίχαρις, charming. Lucretius: simula → Silena, a little Silenus — which is
what Alcibiades calls Socrates at Symposium 215b.
And the placement is structural. Book 4 is the book of simulacra — 46 occurrences against 1–4 in every other book. The renaming catalogue sits at line 1160 of 1287, position 0.901, closing the book on images with an image made by naming. The diagnosis at 1183–84 states the fault precisely: stultitiaque ibi se damnet, tribuisse quod illi plus videat quam mortali concedere par est — he condemns his folly, that he attributed more than it is right to grant a mortal. Over-attribution. Not deception; predication in excess of the bearer. And the therapy at 1189 is protrahere in lucem.
So one passage carries three treatments of whether a name may exceed its bearer: Plato the lover coins; Aristotle's snub cannot be defined without its matter; Lucretius says the coining is over-attribution and must be undone.
χλωρός is not a new observation. It is this archive's own centerpiece.
sequence: Fragment 31 decommissions the speaker channel by channel and *names
the colour of the writing surface the body is becoming* — χλωροτέρα ποίας, the
χλωρός spectrum from living green to the gray-green of cut vegetation, the
colour of papyrus.
into the philosophical corpus, and is marked *a precursor to Socrates as
Orthonym*.
(Pender), and sets Catullus 51 as the calibration floor: close translation,
lossy 3-to-2 compression, polarity inversion at the fifth-stanza position.
Material specificity below that line is not promoted.
warrant is structural, not documentary — judged by explanatory yield, on the
Parry and Prins precedent.
I was about to present a Sappho→Plato χλωρός link as a discovery. It is the program's own, in a direction I had not read. Filed accordingly: the Lucretian terminus goes under #1532's fourth category — transforms proposed by the program — not under explicit reception, and it is measured against the Catullus 51 floor rather than promoted past it.
This is a side-branch and is marked as one. The investigation is Plato–Aristotle and the extended-corpus thesis. The Lucretian terminus bears on it only at one point, and that point is already recorded as M34: Aristotle's refusal to make Socrates a variable (0/144 against Callias 5/35, p = 0.00022) is the same operation Lucretius performs in the other direction — Lucretius strips the added name to return the body; Aristotle declines to lend the body's name to a variable. Both refuse the substitution. That is the whole of the connection, and nothing further is claimed from it here.
The Sappho program is a separate line with its own deposits, its own erratum
chain, and its own evidentiary categories. It is not to be merged into this one.
Added 2026-08-29. The criterion is the archive's own, from #1051: interpretive warrant is structural rather than documentary, judged by explanatory yield, on the Parry and Prins precedent. Applied here as the standard requires.
Asked whether the corpus can be read as primarily concerned with authorial transmission, voice, survival and authenticity, I ran a lexical density test — γραφ-, μνημ-, ληθ-, παραδ-, διαδοχ-, σωζ- across nine corpora — and reported that Plato and Aristotle score 16.57/10k against a comparison mean of 19.21, i.e. below average, with Josephus and Aristophanes both higher.
That was the wrong instrument, and #1051 already said so. The question is not whether a corpus says "writing" often. It is whether reading a passage as being about transmission yields more than the competing lenses. Substituting word frequency for explanatory yield is precisely the failure this notebook has already recorded twice — replacing the concept with the words that made it visible.
The lexical result is retained as a fact about vocabulary and **withdrawn as an
answer to the question asked.**
Applying the yield criterion by hand, I concluded that the lens has "near-zero yield on the empirical and formal works" — the biology, the meteorology, the syllogistic — and high yield only on naming, definition, individuation and authorship.
That is false on the corpus's own distribution. Marker densities per 10⁴
tokens across Aristotle's domains:
domain tokens naming-gap form/matter transmit/persist definition homonym
bio 207,442 1.06 16.58 28.15 2.46 1.78
nat 118,915 0.76 28.09 26.57 11.35 4.04
psych 34,865 2.87 30.40 16.92 18.64 1.72
logic 135,052 0.44 22.14 9.70 33.32 9.11
ethics 106,454 2.63 8.83 13.90 14.18 1.60
metaph 78,946 0.38 89.30 22.93 29.13 8.99
polit 88,421 0.34 11.99 15.83 3.62 1.36
poet 65,515 1.07 17.10 9.77 7.17 1.68
EMPIRICAL (bio/nat/psych) naming-gap 1.14 transmit/persist 26.55
THEORETICAL (rest) naming-gap 0.99 transmit/persist 14.00
The empirical works carry persistence-and-transmission vocabulary at nearly twice the theoretical rate, and the naming-gap reflex slightly more. The correction is not marginal and the original claim is withdrawn.
Read across the table, the corpus does not divide into about transmission and not about transmission. It divides into where the apparatus is theorised and where it is used.
five times any other domain — with definition at 29.13 and homonymy at 8.99.
This is where σιμόν/κοῖλον is argued.
highest in the corpus. Categories 1a opens the Organon on **same name,
different account** — the authenticity problem in its purest form.
the two highest. Generation of Animals is a treatise on how **form is
transmitted through matter across a gap in the bearer**: the male supplies the
λόγος, the female the ὕλη, and what crosses is a motion, not a material.
the two highest — where classes are reached that Greek has no word for and
Aristotle coins (NE 1108a: πειρατέον αὐτοὺς ὀνοματοποιεῖν).
That is not a lens with a periphery. It is a lens whose subject is theorised in
one region and exercised in another.
High yield — the lens consumes residue the standard reading leaves.
and has no account of the bearer-selection — Callias five times, Socrates
zero, p = 0.00022 (M34). The transmission lens does.
explains the sentence's falsity and not its survival. A corpus carrying
οὐδ' ἔστιν σύγγραμμα Πλάτωνος cannot be closed, and was copied for two
millennia by people who could see the problem.
σύμβολον-credential as two unrelated accidents. The lens reads one device.
only self-mention outside the Apology **withholds the author from the scene
his corpus most needs authenticated.**
not explain the refusal to answer, which under the lens is a copying engine.
Genuine yield, previously denied.
treatise's subject, not a redescription of it.
classification project is the indexing problem, which EA-FLOOD-01 identifies
as a transmission mechanism.
— is there something that survives the bearer is the transmission question in
its own words.
carry its matter** is the question, put by Aristotle.
Low yield, stated honestly. Descriptive meteorology — comets, halos, the Milky Way. Stretches of HA that are pure observation: the hippopotamus is snub-faced because hippopotamuses are snub-faced. The syllogistic figures. Here the standard lens is complete and the transmission reading adds nothing it does not import.
Not maximum yield. The lens does not consume the whole corpus, and any version that claims to should be refused on the grounds this notebook has twice recorded: a reading absorbing every observation makes no predictions.
Not near-zero on the empirical. That was measured and is false.
What is defensible: the transmission lens has graded yield with no domain at zero, highest where the corpus is most load-bearing — the substance books, the Organon's opening on homonymy, the Theaetetus framed by an unindividuable mark, the Parmenides' unanswered destruction — and lowest at the descriptive periphery. The theoretical works state the apparatus; the empirical works run it.
And that gradient is itself the finding, because it is what a stewarded technology looks like and not what a universal thesis looks like. A thesis applies everywhere equally. A technology is theorised in one place and exercised in another, and its exercise leaves denser traces than its theory does — transmit/persist at 26.55 in the biology against 14.00 in the philosophy.
Added 2026-08-29. Registered 7bcbbb0fbe69c82c… and frozen before any assignment was made. Two measures were built for this question; the first was wrong and is recorded in §42.
The first design measured whether a lens's vocabulary co-varies across a corpus's works. Result: Plato +0.85, Aristotle +1.20, Philo +1.08, with six of nine corpora VOID or EXCLUDED because narrative and comic Greek carry no frequency-matched stems for definition or naming-gap. Correlations near zero everywhere.
Wrong observable. Interpretive yield is not lexical. It is the degree to which a lens reorganizes and develops a corpus according to the lens's own shape, per #1051 and the Parry precedent. Marked spent; the saturation defect was registered in its own §9 and duly arrived.
Six frozen slots, from the archive's specification of the Sapphic apparatus: T1 encoding · T2 substrate · T3 decommissioning · T4 withholding · T5 recovery · T6 verification. Scored for coverage, sharpness and development, against four frozen rivals — the four causes, the divided line, the cardinal virtues, and a random control — at four unit levels. Assignments made only from received characterization of what each work is about.
R4 on the Johannine group, by "what the work is ABOUT": 0/6
R4 on the Johannine group, by "what APPEARS in the work": 6/6
The vine, the boats, the healings, the war in heaven, the merchants of Babylon, the storm from the throne — all present, none of them what any book is. Under the appearance reading the entire run would have hit §9's void condition. §7's received-characterization rule is not a formality; it is the thing that makes the measure work.
PLATO — 6/6, six distinct dialogues, no doubling, sharpness 1.00
T1 encoding Phaedrus 274b-278b the critique of writing
T2 substrate Theaetetus 191c-195b the wax block
T3 decommissioning Phaedo the death, channel by channel
T4 withholding Parmenides objections raised, answer withheld
T5 recovery Meno 80d-86c ἀνάμνησις
T6 verification Sophist 231b-236d εἰκαστική vs φανταστική
ARISTOTLE — 6/6, five distinct works, one doubling, sharpness 0.83
T1 encoding De interpretatione 16a τὰ ἐν τῇ φωνῇ … σύμβολα, καὶ τὰ
γραφόμενα τῶν ἐν τῇ φωνῇ
T2 substrate De anima 424a ὁ κηρὸς τοῦ δακτυλίου ἄνευ τοῦ
σιδήρου δέχεται τὸ σημεῖον
T3 decommissioning De anima III.5 the separable νοῦς [doubles T2]
T4 withholding Metaphysics B the aporiai, not all resolved
T5 recovery De memoria ἀνάμνησις
T6 verification Soph. El. 164a φαινομένων μὲν ἐλέγχων ὄντων δὲ
παραλογισμῶν
The convergence is at the level of the device, not the theme:
without the iron or gold. The same analogy, and Aristotle's version is the
definitional statement of form-without-matter: τὸ δεκτικὸν τῶν αἰσθητῶν εἰδῶν
ἄνευ τῆς ὕλης. The snub doctrine stated as a theory of reception.
separates ἔλεγχοι from φαινόμενοι ἔλεγχοι. Same discrimination, and **both
works are titled for the sophist**.
is the word Plato uses at Statesman 260d for relaying another's mind, and the
work Andronicus judged spurious (T11).
Rivals, on Plato: R1 4/6 with Timaeus doubling; R2 6/6 nominal but all six slots from one passage, Rep. 509d–511e, sharpness 0; R3 6/6 but four slots from Republic IV alone; R4 0/6. Coverage alone does not discriminate — the registered sharpness term does all the work, and no rival produces six functions carried by distinct works in both corpora.
John fills 6/6 at L2, the author-group level, across three substrates that share almost nothing stylistically — the pair M19 measures at 0.1319, the widest span any NT name shows. Gospel takes T1 and T3; 1 John takes T2; Revelation takes T4 (Rev 10:4, σφράγισον … καὶ μὴ αὐτὰ γράψῃς), T5 (Rev 22:10, μὴ σφραγίσῃς), and T6 (Rev 22:18–19, the adding-and-taking-away curse). No slot is filled twice by the same book. The parts tile.
This was predicted in-session before execution — "does John remain John across radically different genre, register, voice" — and I initially and wrongly recorded it as an unpredicted outcome. Corrected here: the author-group level was specified in advance, and identity across radical substrate disparity is carried by the tiling, not the style.
Scope limit, recorded. I first read the L1/L2 difference as showing Plato and the NT to be "structurally different objects," a concentrated apparatus against an accreted one. That inference is withdrawn. It imported the received account of NT formation, which this measure did not establish and on which this notebook has no standing. What the measure shows is only the level at which each corpus fills. What that difference means depends on the NT authorship question, which is a separate investigation.
The registered prediction — strong convergent yield across Plato–Aristotle — is confirmed, and confirmed at matched slots rather than matched totals: both fill all six at the level of the individual work, with the same analogy at T2, the same term at T5, and the same discrimination under twin titles at T6.
What it does not establish. Common authorship. A school inheriting a transmission programme fills the same slots, and the six-slot lens cannot separate inheritance from identity — the same limit M30 carries. What is new is that the convergence is now measured at the level of function rather than style or vocabulary, which is the first instrument in this investigation to read λόγος rather than ὕλη and return a positive on both corpora at once.
Added 2026-08-29. Registered c185adb40e7b33df… and frozen before the remaining five slots were examined. The first scoring was wrong; both the wrong verdict and the correction are recorded, because the error is the more instructive result.
Every positive in this investigation is consistent with school transmission — M30's reflex, EA-YIELD-02's slot convergence, M34's exclusion. The registration asked what would not be, and answered: not reported correction but completion of an abandoned execution.
A reported correction is a statement, and statements are what schools transmit — including the master erred here. Measured directly: inside the Organon, the works about method, Plato appears four times, all in the Topica, and all four as examples of definitional error (genus-in-species; non-established names; soul as self-moving number; "mortal" attached in animal definitions). That is evidence for the school hypothesis and is admitted as such.
An abandoned execution is different. It is not a statement but a hole with a shape. Completing it requires reading a failure as unfinished rather than refuted — and Plato marks these refuted: ὃ ἔφαμεν ἀδύνατον. Maas's errores coniunctivi one level up: shared correct method proves nothing; what discriminates is a structure that could not have been transmitted as doctrine, because the doctrine says it failed.
T1 Phaedrus 275a writing is φάρμακον οὐ μνήμης ἀλλ' ὑπομνήσεως;
yields δόξαν οὐκ ἀλήθειαν
De int. 16a γράμματα differ, φωναί differ, but
ταὐτὰ πᾶσι παθήματα τῆς ψυχῆς
T2 Theaet. 196c the wax block collapses: εἰς τοὺς πρώτους πάλιν ἀνήκει
λόγους … ὃ ἔφαμεν ἀδύνατον
De an. 424a δεῖ λαβεῖν ὅτι ἡ αἴσθησίς ἐστι τὸ δεκτικὸν τῶν αἰσθητῶν
εἰδῶν ἄνευ τῆς ὕλης
T4 Parm. 135a the forms ἄγνωστα to human nature; requires a man
πάνυ εὐφυοῦς
Metaph. B διαπορήσαντες — aporia converted from terminus to method
T5 Meno 98b ἀνάμνησις, then ὡς οὐκ εἰδὼς λέγω, ἀλλὰ εἰκάζων
De mem. 450a ἐνσημαίνεται οἷον τύπον … καθάπερ οἱ σφραγιζόμενοι τοῖς
δακτυλίοις; and why it fails — running water, too hard a
surface, the young flowing by growth, the old by decay
I first scored prediction 3 as FAILED, reporting the completing mechanisms as heterogeneous: "form separated from matter" at T2, "a physiology of the receiving surface" at T5, "reclassifying aporia" at T4, "invariance of affections" at T1. Four completions, four mechanisms, kill condition fires, configuration reading defeated.
That description was of the four implementations, not the four operations.
Restated at the level of operation:
T2 form separated FROM matter
T5 matter SUPPLIED to receive form
T1 the invariant that holds ACROSS substrates
T4 aporia made passable rather than terminal
T2 and T5 are one doctrine executed in opposite directions. De anima states what crosses — the σημεῖον without the iron or the gold. De memoria states what must be true of the receiver for it to cross. Together: an account of transmission with both halves.
And they use the same figure, in the same three works:
De anima δακτύλιος 1, κηρός 4
De memoria δακτύλιος 1, σφραγ- 2
THEAETETUS δακτύλιος 2, σφραγ- 1, ἐκμαγεῖον 5 ← where it is abandoned
One figure, one abandonment, two completions.
T1 is the same mechanism as an invariance theorem — form constant across variable matter, applied to script and speech. T4 is the weakest member and is marked so: persistence-across-a-break is the shape without the form/matter machinery.
Verdict corrected: prediction 3 HOLDS on three of four, with a shared figure binding the two strongest. The completing mechanism is one thing — what persists when the bearer changes — abstracted at T2, conditioned at T5, proved invariant at T1.
I have now twice described matter where I was asked to describe form. Round 11 substituted lexical density for explanatory yield. Round 13 substituted implementation for operation — inside the test built to read operations.
The instrument is not the problem. The operator defaults to ὕλη. That default is the same one that made every substrate measure in this register return null, and it is worth more to a later reader than the slot table: when scoring a λόγος-level test, describe what the passage DOES, never what it is made of.
Aristotle is a successor under either hypothesis, and the completion result does not distinguish them — it only shows that the completions are one operation rather than several.
Provenance, as the heteronymic reading takes it: Socrates died, and Plato killed him — the founder who wrote nothing is decommissioned into a corpus that speaks for him, with the author writing himself absent at the death (Phaedo 59b, Πλάτων δὲ οἶμαι ἠσθένει). Plato died for Aristotle to be born: the abandoned executions are left as holes with shapes, and the successor position supplies what crosses. On this reading succession is the mechanism of the corpus, not an accident of biography.
What the texts license and what they do not. The completion structure is measured: one figure, one abandonment, two completions, three of four slots on one operation. The provenance reading is not measured and is marked as reading. The notebook's standing rule applies unchanged — the corpus supplies the structure; it does not perform the application.
Added 2026-08-29. EA-UNRES-01 registered a68bf7f2ee54b1af… with the candidate set frozen. EA-CORPORA-08 seated nine orators as the master-student baseline. One of the two things they were seated for works; the other does not, and the failure is mine to record.
Non-resolution has been filed throughout as instrument failure — substrate stylometry ordinary, the Poisson voiding on its own sanity control, YIELD-01 at zero, the base rate declining the Republic/Symposium question. Under the transmission reading these are the artifact behaving as specified: EA-FLOOD-01 names unresolved disputes as the copying mechanism, and corrected EA-PINAX-01 predicts that criteria fail to converge here. Both were recorded and neither was acted on; the investigation kept trying to make them converge, which is what a corpus of this construction would produce in its readers.
The reading is falsifiable and the discriminator is the shape of the ambiguity. Natural ambiguity is uneven — some works resolve, evidence accumulates directionally, new instruments move the needle. Engineered unresolvability would be flat: stalling at the same place regardless of instrument, not drifting toward settlement, not closing under rising attention.
The orators are the ideal control for one thing: **they are professionally fluent
in forgery.** Attribution-fraud is their daily business —
πλάσμα καὶ σκευώρημ᾽ ὅλον (Dem., of a will), τῶν διαθηκῶν ὡς πλαστῶν (Isaeus),
τὸ πλάσμα τὸ μέγα (Hyperides).
**And not one of them says anything about the making or standing of the speech
being read.**
corpus tokens σύγγραμμα σύμβολον ἀντιλεγ- πλάσμα
demosthenes 294,405 0 2 41 3
isocrates 118,455 7 1 10 0
lysias 55,889 0 3 5 0
isaeus 33,307 0 0 3 1
antiphon 18,140 0 0 0 0
aeschines 44,341 0 0 4 0
andocides 17,421 0 0 2 0
dinarchus 10,656 0 0 0 0
hyperides 13,387 0 0 2 1
─────────────────────────────────────────────────────────────
PLATO 585,069 37 4 33 16
ARISTOTLE 984,196 3 9 8 32
σύγγραμμα is zero in seven of nine orators; the exception is Isocrates, the one who wrote for readers rather than juries. σύμβολον appears in Lysias and Demosthenes only as commercial or civic tokens, never textual ones.
The asymmetry, as Lee frames it: the orators are concerned with detecting others' deceptive forgeries. Plato's corpus is concerned with distinguishing its own writings from forgery while sharing forgery's formal features — a text under a name, circulating without its author, whose standing must be established rather than assumed.
*(A direct test of predication-direction was attempted and failed: the "inward"
bucket caught ordinary first-person and demonstrative particles, so every corpus
read inward. Not salvaged by tuning. The asymmetry above rests on the
distribution, not on that measure.)*
C1 Letter II 314c K1 ✓ K2 ✓ K3 ✓ K4 ✓ ALL FOUR
C2 Letter XIII 360a K1 ✓ K2 ✓ K3 ✓ K4 ✓ ALL FOUR
C4 Phaedo 59b K1 ✓ K2 ✓ K3 ✗ K4 ✓ self-absence is consistent
with construction, not
confirming of it
C3 Symposium 223d K1 ✓ K2 ✓ K3 ✓ K4 ✗ indicates loss, not making
C5 Statesman 260e K1 ✓ K2 ✓ K3 ✓ K4 ✗ indicates openness
C6 Theaetetus 209c K1 ✓ K2 ✓ K3 ✓ K4 ✗ indicates failure
C2 is the strongest. ἀντιλέγεται ὡς οὐ Πλάτωνος transmitted inside the text, immediately followed by ἀρχή σοι τῆς ἐπιστολῆς ἔστω καὶ ἅμα σύμβολον ὅτι παρ᾽ ἐμοῦ ἐστιν. K4 holds because a σύμβολον is manufactured by agreement, not found.
C1: τὰ δὲ νῦν λεγόμενα Σωκράτους ἐστὶν καλοῦ καὶ νέου γεγονότος —
Socrates having been made beautiful and young. K4 explicit in the participle.
Two of six meet all four criteria; the orator control yields zero. Part B's
kill condition does not fire.
They were seated to supply two denominators. They supply one.
M30 cannot run on them. Median division rate: orators 0.5/10k, against Plato 6.1 and Aristotle 17.5. No orator work clears M30's 13/10k threshold, so the reflex has no opportunity to fire and the numbers are empty:
teaching link (Isaeus+Demosthenes) 1.57 /100k
other orators (non-links) 1.64 /100k
Plato + Aristotle 6.24 /100k
1.57 against 1.64 looks like a clean null and is not one — it is two figures from a genre that does not perform the operation being counted. When seating EA-CORPORA-08 I claimed M30 could be run on this baseline. That was wrong, and the genre mismatch was foreseeable — I had noted it two exchanges earlier and seated them for the purpose anyway.
**M30's diagnosticity remains blocked on Theophrastus, unchanged since it was
first recorded.**
K4 base rate ESTABLISHED — orators zero, Plato/Aristotle nonzero
Part B does not fail; 2 of 6 candidates meet all criteria
Part A BLOCKED on attribution metadata for Plato and the
orators; only Aristotle carries per-work status
M30 denominator STILL BLOCKED — the orators cannot serve it
completion baseline untested; does not depend on division rate and may
still run on Isaeus→Demosthenes
Added 2026-08-29.
Full census of the forgery-and-legitimacy families across both corpora:
term PLATO ARIST
σύγγραμμα writing-as-thing 37 3
εἴδωλον image 72 13
μίμημα copy 54 7
κίβδηλος counterfeit/debased 13 4
δόκιμος assayed, approved 28 24
ἀδόκιμος rejected in assay 3 2
πλάσμα / πλαστός fabricated 16 32
φάντασμα appearance 31 47
ὑποκριτής actor 10 32
γνήσιος legitimate 16 6
νόθος bastard 11 4
παράσημος false-stamped 1 4
The division of labour is exact, and matches Round 11's theory/exercise split. Plato holds the image vocabulary — εἴδωλον, μίμημα, σύγγραμμα — terms for a copy that stands in for an original. Aristotle holds the detection vocabulary — πλάσμα, φάντασμα, ὑποκριτής — terms for a fabrication and for the apparent-but-not-real. Plato makes copies; Aristotle catches them. And the assay word is shared almost equally: δόκιμος 28 / 24.
This was missed in Round 9 and is the round's philological result.
395b — κατακεκερματίσθαι sits four lines from μιμήματα:
πάντα δὲ ταῦτα μιμήματα … καὶ ἔτι γε τούτων φαίνεταί μοι εἰς σμικρότερα
κατακεκερματίσθαι ἡ τοῦ ἀνθρώπου φύσις, ὥστε ἀδύνατος εἶναι πολλὰ καλῶς
μιμεῖσθαι
507a — the same coin figure, now applied to the account being handed over:
τοῦτον δὴ τὸν τόκον τε καὶ ἔκγονον αὐτοῦ τοῦ ἀγαθοῦ κομίσασθε.
εὐλαβεῖσθε μέντοι μή πῃ ἐξαπατήσω ὑμᾶς ἄκων, κίβδηλον ἀποδιδοὺς τὸν λόγον
τοῦ τόκου.
*Receive this interest and offspring of the Good itself. But be on your guard lest
I unwillingly deceive you, paying back a counterfeit account of the interest.*
One monetary figure runs the whole: nature minted into denominations too small to hold two arts (395b), and the λόγος itself possibly debased coin handed over as payment (507a) — with τόκος and ἔκγονον, interest and offspring, joining the coin metaphor to the kinship metaphor in one clause. Socrates warns, at the threshold of the Good, that what he is about to pay out may be κίβδηλον.
And Laws 742a legislates the structure: currency ἔντιμον to the citizens and ἀδόκιμον to everyone else — a token valid inside the house and worthless outside it. That is standing-not-paternity, in coin.
ὅταν δ᾽ ὑπὸ τῆς ἐκείνου προσποιήσεως ἀπατηθῇ, δίκαιον ἐγκαλεῖν τῷ
ἀπατήσαντι, **καὶ μᾶλλον ἢ τοῖς τὸ νόμισμα κιβδηλεύουσιν, ὅσῳ περὶ τιμιώτερον
ἡ κακουργία**
When someone is deceived by another's assumed character, it is just to accuse the deceiver — and more than those who counterfeit the coinage, inasmuch as the crime concerns something more precious.
Aristotle grades false personation and coin-forgery on one scale, and puts personation higher. The two are the same offence differing in the value of what is falsified. This is the counterfeiting frame applied to persons, stated by Aristotle, in his own voice, in the ethics.
On the reading: the corpus supplies the assay. Plato states the fragmentation (κερματισμός), warns that the account handed over may be debased (κίβδηλον at 507a), and legislates the two-currency structure (Laws 742a). Aristotle supplies the detection apparatus — πλάσμα, φάντασμα, ὑποκριτής at 32/47/32 against Plato's 16/31/10 — and states the grading rule at NE 1165a. So Aristotle has told us where the assay is: personation is worse than coin-forgery, and the corpus that holds the tools for distinguishing apparent from genuine is his.
Status. The passages are recorded and quoted; the reading is Lee's and is marked as reading. What is measured is the distribution (§58) and the continuity of the monetary figure across 395b, 507a and Laws 742a. What is not measured is that any of it was placed for this purpose. The standing rule holds: the corpus supplies the structure; it does not perform the application.
*Added 2026-08-29 as the frame for the whole notebook. Placed here so it is not
buried; it is the shape everything measured today has converged on.*
The two facts that must be held together:
DECLARED PARTITION Z = +39.18 the names are overwhelmingly strong
(M6) containers — the strongest partition
measured in the sample
FUNCTIONAL INTERLOCK 6/6 slots matched at the level of the individual
(Round 12, 13) work, with ONE figure (seal-and-wax) and
ONE abandoned execution completed twice
Maximally separated at the substrate. Maximally interlocked at the function.
Nothing else in the sample does both. Josephus is unified at both levels. The New Testament is loose at both — its declared partition sits at Z = +2.83 and it fills the slot structure only at the author-group level. Plotinus and Philo are separated and not interlocked at all.
That is the shape of a σύμβολον: halves that do not resemble each other and
only fit.
What this frame does NOT claim. Not common authorship. The comparison class is thin — one multi-author corpus (the NT) plus the rival-lens controls — because the orator baseline failed for the interlock measures (§56). A school inheriting a transmission programme could produce the interlock; the substrate divergence is ordinary for two authors. The conjunction is what is unusual, and the conjunction has been measured once.
Added 2026-08-29. Not a claim. A reading run at full strength so its interpretive yield can be measured — which cannot be done without running it through. Every caveat is deferred to §S9 rather than inserted along the way, because inserted caveats prevent the run.
Aristotle wrote the Platonic dialogues. "Plato" is his heteronym.
Not "constitutes the position," not "holds the account." The literal claim: one hand, two named bodies of work, the earlier name a construction of the later writer.
Everything below is what the day's measurements look like read through that.
SE 183b–184b: rhetoric grew οἷον ἐκ διαδοχῆς — Tisias, Thrasymachus, Theodorus — but of his own subject, οὐ τὸ μὲν ἦν τὸ δ᾽ οὐκ ἦν προεξειργασμένον, ἀλλ᾽ οὐδὲν παντελῶς ὑπῆρχεν. There was absolutely nothing before.
Read literally this is not modesty about the syllogistic and not a priority dispute. There was nothing before because he had not yet written it. He then asks to be judged not on succession but on the finding: τοῖς δ᾽ εὑρημένοις πολλὴν ἔχειν χάριν.
**A man disclaiming διαδοχή, in the treatise on telling the genuine from the
apparent, asking for a verdict.**
M28: hold subject constant and Aristotle's within-domain dispersion is 0.0571 — Plato's total is 0.0570 — and within ethics he is tighter than 99.5% of multi-author pools. His variance is object-determined: the treatise takes the shape of its object.
A writer whose form conforms to its object can write anything, including dialogues. M10's anomaly — "more dispersed than five authors" — was not a defect to be explained away. It is the range the reading needs, and M28 measured it.
I recorded M16 as the mechanism-negative: Plato's constructed Aristophanes ranks 5 of 7 Symposium speeches for resemblance to the real Aristophanes, and sits at 0.0380 from the Plato corpus — more Platonic than its neighbours. I called it the only ancient case with a checkable target, reading negative.
Under the strong reading it is not an attempted imitation at all. The Symposium's Aristophanes is a character in a dialogue, written by the man who wrote the dialogue. It comes out Platonic because the voice being written is Plato — the constructed position — not because an imitation of Aristophanes failed. The measure was comparing a character to a historical person and scoring the mismatch as incapacity.
The mechanism-negative dissolves. What M16 actually measures is that the
Symposium speeches are internally consistent with the corpus that contains them.
Round 13: the wax block collapses at Theaetetus 196c — εἰς τοὺς πρώτους πάλιν ἀνήκει λόγους … ὃ ἔφαμεν ἀδύνατον — and the same figure is completed at De anima 424a by ἄνευ τῆς ὕλης, with De memoria 450a supplying the receiver's physiology. δακτύλιος and the seal appear in exactly three works: the two completions and the abandonment.
Under inheritance: a student who selects abandoned rather than developed
operations, and supplies apparatus the master lacked, five times of six.
Under the strong reading: one hand wrote the problem knowing where it was going. The abandonment is the mould; the completion is the casting. That is simply what writing is — and it explains why the figure is confined to those three works and why the Platonic one is where it fails.
Socrates is named 144 times in Aristotle and used as the interchangeable concrete individual — pale, sick, an instance of number, paired with Callias nine times in Καλλίαν ἢ Σωκράτην. Zero times at the snub. Callias fills the slot 5 of 35. Fisher exact p = 0.00022.
Under the strong reading: the snub is the mark he gave his own constructed founder in the Theaetetus. It is not available for lending. A maker does not hand out the predicate by which his construction is recognised, and Callias takes the slot because Callias is nobody.
The measured exclusion becomes a signature rather than a curiosity.
ἀόριστος indefinite Aristotle 108 Plato 3
δυάς the Dyad 41 4
χωριστ- separate Forms 411 165
ἀριθμ- Forms as numbers 977 246
These are not a reporter's ratios. They are the ratios of a man expounding his own material in one register while having it attributed in another. The doctrines by which Platonism is known are where they are because that is where they were made.
**οὐδ᾽ ἔστιν σύγγραμμα Πλάτωνος οὐδὲν οὐδ᾽ ἔσται, τὰ δὲ νῦν λεγόμενα Σωκράτους
ἐστὶν καλοῦ καὶ νέου γεγονότος**
*There is no writing of Plato's and there never will be; what are now called his
are Socrates', having been made beautiful and young.*
Under the strong reading this is neither forgery nor riddle. It is true, it names the operation in the participle — γεγονότος, made — and it is placed in the one book of the corpus where the author's name appears 23 of its 26 times. Letter XIII then carries its own athetesis, ἀντιλέγεται ὡς οὐ Πλάτωνος, followed immediately by a σύμβολον named as the letter's credential — a token manufactured by agreement, not found.
M39: Aristotle is the only member of the cluster with external documentary attestation — FD III 1, 400, a contemporary Delphic decree. Socrates wrote nothing and is attested only through figures inside the cluster; Plato through the corpus that denies being his; Aristophanes through a civic record naming producers.
Of course he is the anchored one. He is the position that had to be real; the others are what a real position makes. And the one contemporary document crowns him for compiling the register of Pythian victors — the maker of the attribution apparatus, certified for making it, by the apparatus that confers standing on everything it lists.
And Plato's own frame licenses the claim. The lover's μελίχλωρος is ἐραστοῦ ποίημα — a making. Plato's distinction is not made-versus-given but the poemed against the forged: two makings, separated by standing. So a legitimate author is one whose making has standing. Aristotle's whole apparatus — γνήσιον against φαινόμενον, ἔντεχνον against ἄτεχνον, personation graded above coin-forgery at NE 1165a — is the machinery for establishing that this making is the legitimate kind. He built the assay because he needed it.
Chronology. The received dating puts Aristotle at seventeen on arrival with early dialogues already circulating. Two things: the received dating is built from the corpus and the succession-list, which are T11 and M40c — inside the object under test, not outside it — and a seventeen-year-old already circulating is what the historical record repeatedly shows of first-rank makers.
The Topica catalogue. Aristotle lists Plato's definitions among definitional faults four times. Under the strong reading that is a man cataloguing his own constructed position's errors. Round 5's corrected criterion applies: a declaration that reads as ordinary succession is what camouflage looks like — and both known heteronymic corpora, Pessoa's and Kierkegaard's, declared themselves in print.
And the honest residue: nothing measured today distinguishes one maker holding two positions from a successor whose making is acknowledged as continuous. M38's ledger stands. The strong reading is not established here — it is run, at strength, so its yield can be measured.
Sections S6–S8 cite M39 and M40 in forms since corrected. Specifically: the certification asymmetry is an identified epigraphic anchor, not a demonstrated absence for the others; the stone anchors the person, not textual paternity; and "the position that had to be real" is a conditional architectural reading, not an established claim. S8 should be read at that reduced strength. The run is retained at full strength as a run — that is its purpose — but it inherits the corrections.
Read through the strong version, the following stop requiring separate explanations: the object-determined dispersion (S2), the Aristophanes result (S3), the abandonment/completion asymmetry with its single shared figure (S4), the lexical exclusion at p = 0.00022 (S5), the doxographic concentration (S6), the authorship denial and its σύμβολον (S7), and the certification asymmetry (S8).
Seven structures, one construction. Against M38's ledger — which has since been rebuilt two-sided: inheritance requires five latent assumptions, not ten auxiliaries, of which four are ordinary and one (L5) is where it genuinely pays; and the configurational reading carries its own five (K1–K5), including the chronology, the unspecified relation between the attested person and the authorial position, and the absent expected signatures. Neither ledger is short.
That is the compression the reading buys, and it is what M38b said the heteronymic hypothesis would have to earn against its prior. Whether it earns enough is the next measurement, not this one.
*Added 2026-08-29. Two things, kept apart: a frozen target run under EA-UNLOCK-01
and scored, and a companion finding that cannot be scored under it.*
Target as frozen: *do the four Platonic naming-gap works have an Aristotelian
counterpart naming what they perform?*
Step 1 — what each performs, from the census, without interpretation:
Theaetetus 0.203 ἄλλαι ἀπέραντοι μὲν αἱ ΑΝΩΝΥΜΟΙ, παμπληθεῖς δὲ αἱ
ὠνομασμέναι — infinitely many unnamed perceptions
Sophist 0.878 λέων ἔλαφος ἵππος … ΟΥΔΕΙΣ ΠΩ ΣΥΝΕΣΤΗ ΛΟΓΟΣ — a string of
names yields no λόγος
Statesman 0.340 τὸν πολιτικὸν οὐδαμοῦ συνελάβομεν οὐδ᾽ ὠνομάσαμεν, ἀλλ᾽
ἡμᾶς ἔλαθεν ΚΑΤΑ ΤΗΝ ΟΝΟΜΑΣΙΑΝ ΕΚΦΥΓΩΝ — he escaped in the
naming
Timaeus ×3 nameless kinds of air, juices, smells
Step 2 — search space fixed BEFORE looking: the Aristotelian work on the same
subject, first co-occurrence taken, no roaming.
Theaetetus → DE ANIMA 412a
ψυχὴν οὐσίαν εἶναι ὡς εἶδος σώματος φυσικοῦ ΔΥΝΑΜΕΙ ΖΩΗΝ ΕΧΟΝΤΟΣ …
ἡ δ᾽ οὐσία ΕΝΤΕΛΕΧΕΙΑ … αὕτη δὲ λέγεται ΔΙΧΩΣ
→ the unnamable infinity becomes tractable: unnamed = δυνάμει,
named = ἐντελεχείᾳ
Sophist → DE INTERPRETATIONE 16a
πρῶτον δεῖ θέσθαι ΤΙ ΟΝΟΜΑ ΚΑΙ ΤΙ ΡΗΜΑ, ἔπειτα τί ἐστιν … λόγος
→ names the parts whose absence the Sophist discovers
Statesman → POLITICS 1275a
ΑΝΩΝΥΜΟΝ γὰρ τὸ κοινὸν ἐπὶ δικαστοῦ καὶ ἐκκλησιαστοῦ … ΕΣΤΩ ΔΗ ΔΙΟΡΙΣΜΟΥ
ΧΑΡΙΝ ἀόριστος ἀρχή
→ where the statesman escaped Plato IN the naming, Aristotle meets the same
structure and settles it BY DECREE — let it be, for the sake of marking off
Three of four resolve in Form A, cleanly, with exact Greek.
SCORE: NULL — ⚠ AND THE JUSTIFICATION OVERCLAIMED. I wrote that "inheritance
predicts Form A exactly as strongly." That was never measured. What is
established is that inheritance provides an intelligible explanation — P(E|H_S)
0 — not that P(E|H_S) = P(E|H_H). The honest code is **FIT CONFIRMED;
DIFFERENTIAL WEIGHT UNRESOLVED — COMPARATOR NOT CALIBRATED**. The NULL score
stands under §5 (it cannot count toward the threshold without demonstrated
differential determinacy) but the reason is comparator absence, not
demonstrated equivalence. See M43 for the control class that would settle it.
§5 requires the target to become **more determinate under H_H —
differentially**. That has not been shown either way. Sophist → De interpretatione is the standard
account of how logic developed; Theaetetus → De Anima the standard account of
how psychology developed. Inheritance predicts Form A exactly as strongly. The
Statesman/Politics pair differs in kind but not in favour: Aristotle is not
completing Plato's specific failure but meeting a different nameless class with
the same reflex — which is M30, whose diagnosticity is unresolved and which
Theophrastus also carries (M36).
Running total: 0 unlocks, 1 null, 5 targets remaining, threshold 3.
Recorded because the protocol worked. Without §5 these three would have been written up as a strong result: the Greek is exact, the correspondences are clean, and the De Anima line lands precisely. The rule that stops them counting is the requirement of differential determinacy rather than mere fit.
Not T-F. Found by open search, therefore unscoreable under EA-UNLOCK-01 §5 and carrying no threshold weight. Recorded because null-for-counting is not null-for-value: a structure the model handles cleanly is part of what makes it a functioning inferential model rather than an available one.
De Anima II.2, 413b:
πότερον δὲ τούτων ἕκαστόν ἐστι ψυχὴ ἢ μόριον ψυχῆς, καὶ εἰ μόριον, πότερον
οὕτως ὥστ᾽ εἶναι χωριστὸν λόγῳ μόνον ἢ καὶ τόπῳ … ὥσπερ γὰρ ἐπὶ τῶν φυτῶν
ἔνια διαιρούμενα φαίνεται ζῶντα καὶ χωριζόμενα ἀπ᾽ ἀλλήλων
De Anima I.4, 409a:
τὰ δὲ φυτὰ καὶ τῶν ζῴων πολλὰ **διαιρούμενα ζῇ, καὶ δοκεῖ τὴν αὐτὴν ψυχὴν ἔχειν
τῷ εἴδει**
The distinction is Aristotle's own and it is exact: χωριστὸν λόγῳ μόνον ἢ καὶ τόπῳ — separable in account only, or also in place. And divided living things live, having the same soul in form.
What it supplies. EA-KERMA-01 had to answer Rep. 395b's κατακεκερματίσθαι against Rep. 473d's εἰς ταὐτὸν συμπεσεῖν, and answered with the σύμβολον — parts that do not resemble but fit. That was an image. De Anima gives the ontology: one actuality, divided κατὰ λόγον, yielding real parts that live and are the same in form, with the criterion of separation being account, not place.
unity at the level of actuality ≠ undifferentiatedness at the level of
potentiality
And potential is not unreal. A δύναμις is a real capacity of the actual thing. So: all lives are real; some only potentially so — sharpened for the argument as potential lives are not unreal lives; they are real capacities of one actuality.
Why it matters to the model. It is the recognition principle from the snub arriving from the other side. Matter individuates, λόγος recurs — and here λόγος is what divides, while the actuality stays one. Under that description the positions are not masks. They are real determinations of one actuality, separated according to account.
Explicit limit. Aristotle does not apply De Anima to authorship and nothing here suggests he does. This is a reconstruction of machinery available in the corpus, at the same epistemic level as EA-KERMA-01 and under the same standing rule: the corpus supplies the structure; it does not perform the application.
Added 2026-08-29. Supersedes the encoding framing of Round 16 §S4 and revises the K-ledger of M38. The correction that produced it came from reading ξυνίεσαν as inclusive of Plato.
Round 16 ran the strong reading as encoding: the abandonment is the mould, the completion the casting, one hand writing the problem knowing where it was going. That framing required a plan, and a plan required prescience — K1 in M38's ledger, and the most expensive item on it.
The sentence forbids it. APr I.31: οὔτε ὅ τι ἐνδέχεται συλλογίσασθαι διαιρούμενοι ξυνίεσαν — they were not understanding what can be syllogized by dividing. Imperfect, and inclusive of Plato, who is the one being diagnosed.
So the earlier position did not know what its own operation could conclude. There is no plan. The first half is not a withholding; it is an execution that stops where its practitioner's grasp stops. The completion is not a designated receiver arriving but a discovery about the instrument's range, made by continuing to operate it.
The Sappho objection. Sappho's device works because the receiver is unknown — the future reader is constituted by not being specified (#503, #1483). I had objected that a maker completing himself collapses that dependency.
It does not, if the second position discovers rather than executes. The unknown-ness is real and is located exactly where the device needs it: not in the world's ignorance of who will read, but in the earlier position's ignorance of what its own operation entailed.
The completions, re-read. Theaetetus 196c does not fail on purpose. It fails — ὃ ἔφαμεν ἀδύνατον — and the failure is genuine. What De anima 424a supplies is not a prepared answer but the discovery that form separates from matter, which the wax block needed and its author did not have. The seal-and-wax figure recurs in exactly three works because the instrument was carried forward, not because a mould was cast.
And M40 becomes discovery rather than encoding. οὐδὲν παντελῶς ὑπῆρχεν says nothing existed before because the operation's range had not been seen — and the one who saw it is the one claiming it. Neither an equivocation nor a concealed confession. It is what it says.
M28, which the day has now made mean three things, and this is the third.
Aristotle within-domain dispersion 0.0571
Plato total dispersion 0.0570
domain effect, Aristotle Z = +4.54 (Plato +1.75)
within ethics tighter than 99.5% of multi-author pools
edition, excluded Z = −0.10
ethics 0.0226 · psych 0.0390 · nat 0.0465 · polit 0.0584 · bio 0.0680 ·
poet 0.0707 · logic 0.0942
→ ordered by the internal object-space of each domain; logic loosest because
it ranges widest
What this measures, stated plainly: the writing divides by object. Hold the object constant and the variance is one author's. The treatise takes the form of what it treats. Ethics tightest, logic loosest.
⚠ WITHDRAWN — see Round 19 §72. Object-conditioning at ratio 1.489 is ORDINARY: Aquinas, a securely distinct single author, reaches 1.503. What survives is §69's first reading and the parity (0.0571 ≈ 0.0570), which is a two-corpus coincidence and is NOT the ratio. Original text follows, superseded:
And that is the discovered operation, performed on its discoverer. A soul divided κατὰ λόγον into parts really distinct, really alive, τὴν αὐτὴν ψυχὴν ἔχειν τῷ εἴδει — the same in form — with the criterion of separation being account, not place (De An. II.2, Notebook §63). Not asserted. Measured: 0.0571 against 0.0570.
1. Division is run on KINDS — the four naming-gap works, the sophist, the
statesman, the arts — and its practitioners ξυνίεσαν οὔτε ὅ τι ἐνδέχεται
συλλογίσασθαι: they were not understanding its range.
2. The RANGE IS DISCOVERED: that a soul admits division by account into parts
that live and are the same in form (De An. I.4, II.2).
3. The discoverer THEN WRITES THAT WAY — dividing by object across thirty
works, which is what M28 measures.
**The corpus is the operation applied to its own operator, and the statistic is
the trace.**
M38's K-ledger charged the configurational reading five items. **K1 is now
withdrawn:**
K1 chronology / prescience — the received dating must be wrong, or a plan
must have anticipated its own completion
→ WITHDRAWN. Discovery requires no prescience, no plan, and no
anticipation of instruments. The seventeen-year-old problem dissolves
with it: nothing needs to have been foreseen.
K2 the relation between the attested person and the authorial position
→ STANDS, unspecified.
K3 why externally transmitted biographical distinctions exist
→ STANDS.
K4 why expected signatures are ABSENT — no προσωπεῖον, no ψευδεπίγραφος,
M16 requiring reinterpretation rather than prediction
→ STANDS, and is now the heaviest item.
K5 documentary machinery sustaining the partition across transmission
→ STANDS.
Four, not five, and the most expensive one is gone. H_S's ledger is unchanged
at five latent assumptions (M38), of which L2 is supported by Theophrastus.
Recorded because a measurement used three ways in a day is exactly what a reviewer
should distrust.
FIRST (M28, morning) the dispersion is the OBJECT, therefore the
multi-author anomaly of M10 dissolves.
SECOND (Round 16 §S2) object-determination is the CAPACITY the strong
reading requires — a writer whose form conforms to its
object can write dialogues.
THIRD (this round) being-determined-by-the-object IS the divided-soul
operation, performed on the writer.
The first and third are compatible — the dispersion is the object, and that is not in dispute. The third adds a reading on top of a number, and the addition is what must be argued, not the number. The number stands however the reading goes: 0.0571 against 0.0570, per-domain, edition excluded.
Nothing here establishes one hand. A school trained to write to its object
could produce the same statistic — could, not does: that frequency has
never been measured (M43). What the round changes
is that the configurational reading no longer requires a plan — which was its
single most expensive assumption.
*Added 2026-08-29. Aquinas seated; three results; one refutation of a claim the
notebook does not make, corrected.*
Corpus Thomisticum, 312 pages fetched, 6,698,317 tokens, structured HTML, no
OCR. Split three ways, not two:
independent 5,219,700 tok / 10 works Summa Theologiae, Scriptum,
SCG, the Quaestiones, Quodlibeta
aristotle-commentary 1,020,889 tok / 8 works Metaphysics, Ethics, Physics,
De caelo, De sensu, Meteora…
other-commentary 457,728 tok / 4 works Psalms, Job, Dionysius,
De causis
Admissible under M36a where no Greek figure was: attested by canonization proceedings, university registers, papal bulls and Dominican chapter records — evidence with no relation to textual style, doctrine, or any succession tradition under test.
⚠ SEATING ERROR, CAUGHT AND CORRECTED. My prefix table mislabelled every commentary: cme declared In Metaphysicam is Sentencia super Meteora; cmp is the Metaphysics; ccm is De caelo; ctc is the Ethics; cps declared De sensu is In Psalmos; cdn is Dionysius; cio is Job. Found by a locus probe returning zero for "metaphysic" in the file labelled In Metaphysicam, and rebuilt from each text's own header. The three-way split is a product of that correction and is better than the two-way I registered.
kind tokens gap/100k div/10k
independent 5,219,700 1.07 26.6
other-commentary 457,728 0.22 10.3
aristotle-commentary 1,020,889 3.72 48.5
Aristotle-commentary / independent = 3.47×
other-commentary / independent = 0.20×
The naming-gap reflex arrives with the source text, not with the man. 3.47× higher where Thomas glosses Aristotle; five times lower where he glosses Scripture and Dionysius. Positive control passes — the Latin probe finds 38 hits in the commentaries, so low counts elsewhere are absence, not blindness.
And the number that bears on M30:
Aristotle 8.43/100k at division 30.9/10k
Plato 2.60/100k at division 15.4/10k
AQUINAS indep. 1.07/100k at division 26.6/10k ← securely distinct
Plotinus 0.47/100k at division 36.8/10k
Philo 0.35/100k at division 13.3/10k
Thomas divides as heavily as Aristotle and shows the reflex at one-eighth the rate in his own voice. The most saturated Aristotelian in the Latin tradition does not reproduce it when writing independently. That bears on the objection that killed M30's diagnosticity — but it is one author, not a pair, and R3 measures agreement between two members. It needs Albert.
Object-conditioned partition, within-work against cross-work dispersion:
Aquinas independent 1.503 ← securely distinct SINGLE author
Aristotle (M28) 1.489
Plato (M28) 1.186
Aquinas Aristotle-comm. 1.762
Aquinas other-comm. 2.707 ← internal check: maximally unlike objects
A securely distinct author reaches Aristotle's ratio. So "the writing divides
by object" is what any author writing across objects does.
**⚠⚠ THIS INFERENCE IS INVALID — CORRECTED 2026-08-30. AQUINAS IS NOT A CONTROL
FOR THIS MEASURE.**
His object-partition apparatus IS Aristotle's. The Summa's quaestio and articulus structure — division by objection, distinction, and classification — is the Aristotelian partition applied to theology. Aquinas "writing independently" is still writing inside the Aristotelian object-space.
So 1.503 is not an independent author reaching Aristotle's ratio. It is Aristotle's ratio appearing in its most saturated inheritor. The object was measured with a descendant of the object.
And this table's own internal check shows the measure is reading object-distance rather than authorship: 1.489 (Aristotle) → 1.503 (Aquinas independent) → 1.762 (Aquinas on Aristotle) → 2.707 (Aquinas on unlike objects). The ratio tracks how unlike the objects are. It does not track how many authors there were.
THE SEQUENCE THAT WENT WRONG:
1 Aristotle's author-internal variance measured ~1.8 ANOMALOUS
2 controlled for object → 0.0571, against Plato 0.0570 PARITY
3 "but Aquinas independent reaches 1.503, so ordinary" ⚠ INVALID
4 therefore Round 18 §66 withdrawn ⚠ ON INVALID GROUND
**Step 3 normalises the extreme object-shaping of the Aristotelian corpus by
pointing to a commentator on Aristotle.**
WHAT THIS RESTORES AND WHAT IT DOES NOT. Round 18 §66's reading is not thereby reinstated — §69 independently flagged it as laying a reading on a number, and that objection stands on its own. What is withdrawn is the GROUND of the withdrawal. §66 returns to undecided rather than refuted, and the question of whether Aristotle's object-conditioning is ordinary is reopened and unanswered.
WHAT A VALID CONTROL WOULD REQUIRE: an author writing across genuinely unlike objects whose partition apparatus is not descended from the Aristotelian one. No such author has been identified among the seated corpora. Theophrastus, Albert and Aquinas all fail on the same ground; Xenophon writes across unlike objects but is a candidate in the hypothesis space (M33). The control is not merely missing — it may be structurally hard to obtain for a partition apparatus this widely inherited.
Round 18 §66's language — that object-division IS the discovered soul-division operation performed on its discoverer — rests on the ratio and is WITHDRAWN. §69 flagged the third reading as the one laying a reading on a number; the control agrees with the flag.
(Caveat against over-reading the refutation: Latin function-word space, different block size, different language. The ratios are comparable in form, not identically computed.)
On seeing 1.503 ≈ 1.489 I reported that M28's finding was ordinary. **That was
wrong, and the error is worth recording.**
M28's claim is not the ratio. It is the PARITY: control for object and Aristotle's within-domain dispersion lands at 0.0571 against Plato's total 0.0570 — the most internally heterogeneous author in the sample dropping to parity with Plato specifically. That is a two-corpus coincidence. The ratio is a one-corpus property. I measured the second and reported it as refuting the first.
The ratio being ordinary was never in dispute — M28 says the domain effect explains the heterogeneity. What the ratio cannot explain is why the residue equals Plato's number rather than any other.
Eleven seated corpora, common 160-feature space, all 55 pairs:
| Δ | ≤ 0.0001 0 of 55 = 0.0% ← M28's tolerance |
| Δ | ≤ 0.0005 1 of 55 = 1.8% |
| Δ | ≤ 0.002 5 of 55 = 9.1% |
| Δ | ≤ 0.005 13 of 55 = 23.6% |
closest pair in the entire set: Josephus 0.0208 / Philo 0.0211 = 0.00026,
two and a half times M28's gap
Near-parity at M28's tolerance never occurs among arbitrary pairs.
LIMITATION, STATED: this is total against total. M28's quantity is object-controlled residual against uncontrolled total — a different pairing with a different distribution. Running its base rate needs domain classifications for the other corpora, which exist only for Aristotle. So: the parity is not a generic artifact of dispersion measures. Whether residual-vs-total parity is rare is UNRUN.
I proposed scoring the parity on Albert→Thomas. That is the M43c error again, one exchange after recording it. Whether Thomas's residual equals Albert's total is a different quantity that answers nothing about Plato and Aristotle. What a control could establish is a base rate for the parity as a form across many pairs — and that needs many, not one.
Added 2026-08-29. Feasibility, recorded as it landed rather than in a batch.
Lee's question: *in a heteronymic single-author corpus, would we see convergence
across whole corpora rather than a fractional series distributed work-by-work?*
Not runnable on what is seated. The coverage:
PESSOA 111,449 tok / 4 voices ortonimo 48,636 · campos 41,142 ·
caeiro 16,123 · REIS 5,548
KIERKEGAARD 449,144 tok / 5 voices Victor Eremita B 133k · A 128k ·
signed 119k · de Silentio 43k ·
Notabene 27k
Reis at 5,548 tokens cannot fill six functional slots whatever Pessoa did. And Kierkegaard is missing Constantin Constantius, Vigilius Haufniensis, Johannes Climacus, Anti-Climacus, Frater Taciturnus, Hilarius Bogbinder — Climacus and Anti-Climacus being the pair the whole architecture turns on. Five voices of about a dozen.
A null here would measure the seat, not the corpus — the same defect as the
orator baseline (Round 14 §56). No opportunity denominator.
1920 Drachmann (samledevaerkerud*) FRAKTUR — unusable, confirms T14
1901 edition (srenkierkegaard*goog) ROMAN TYPE — usable with declared error
The 1920 failure diagnosed: the OCR read blackletter as roman. iffe for ikke, bette for dette, benne for denne, felb for selv, bære for være, forbi for fordi. Danish-marker share 1.5%.
The 1901 scan is roman type and reads correctly: ikke 1737, denne 165, hvad 140, fordi 120. Marker share 3.6%. Scan degradation is patchy — ikkr 266 against ikke 1737, ≈13% error on that token — so it is usable for function-word and structural measures, unsafe for rare-lexeme measures, the same declaration as Hyperides and Albert.
T14 recorded a blocker for the edition it tested and I generalised it to the author. Second time today a recorded blocker turned out to be untested for the relevant case (cf. Theophrastus, M36).
Current seat: 4 voices, from pt.wikisource categories.
arquivopessoa.net lists 39 authorial positions, each with its own text index:
Alberto Caeiro · Ricardo Reis · Álvaro de Campos · Fernando Pessoa ·
BERNARDO SOARES · Alexander Search · Barão de Teive · António Mora ·
Charles Robert Anon · Frederico Reis · Adolph Moscow · Faustino Antunes ·
Henry More · Horace James Faber · Coelho Pacheco · David Merrick ·
Carlos Otto · Gaudêncio Nabos · Dr. Pancrácio · António de Seabra …
Verified: /autores/4 (Bernardo Soares) returns 504 text links; individual texts fetch clean UTF-8 Portuguese, no OCR. Soares alone is the Livro do Desassossego, absent from the seat entirely.
This is the acquisition the distribution test needs — not four voices but the full range, which is the only way to see where and how the divisions were distributed.
Pessoa expansion FEASIBLE NOW — 39 positions, clean text, no OCR
Kierkegaard expansion FEASIBLE via the 1901 edition, with a declared
error rate; the 1920 Fraktur stays unusable
the distribution test BLOCKED until both are seated. Any null before then
measures the seat.
*Added 2026-08-29. EA-CORPORA-11 seated at 25 voices / 579,493 tokens. First
numbers, and a methodological error that turns out to be the same one as M28.*
Seven voices above 6k tokens, 3,000-token blocks, top-150 function words,
within-voice against cross-voice dispersion:
all 7 voices (incl. English Search) ratio 1.415 Z +18.52
Portuguese only (Search dropped) ratio 1.242 Z +19.05
VERSE only — Caeiro, Reis, Campos ratio 1.094 Z +2.45
PROSE only — Soares, Mora, orthonym ratio 1.313 Z +21.84
Plato internal, by domain (M28) ratio 1.186 Z +1.75
Aristotle internal, by domain (M28) ratio 1.489 Z +4.54
Uncontrolled, Pessoa by voice (1.415) sits beside Aristotle by domain (1.489) — which is the shape Lee predicted: a heteronymic system looks like intra-Aristotle, many positions under one maker, not like a two-corpus split.
I reported the verse-only run as the clean test: three voices, one genre, one language, one author, separating at only 1.094 — and concluded that designed heteronymy does not produce substrate divergence, and therefore that Plato–Aristotle's Z = +39.18 counts against the configuration.
That conclusion is withdrawn, and the reason is exact.
Genre-assignment is not a confound in a heteronymic corpus. It is the operation. Caeiro is a pastoralist, Reis is a Horatian odist, Campos is a long-line modernist, Soares is a prose diarist, Mora is a philosophical prosaist, Search is an English poet. The genre is part of the position. Removing it removes what a heteronym is.
And this is precisely the M28 error, repeated:
M28 controlled for OBJECT → Aristotle's dispersion collapsed
→ read as "the anomaly dissolves"
→ but writing-to-the-object IS the operation (Round 18)
Round 21 controlled for GENRE → Pessoa's voice-partition collapsed
→ read as "heteronymy produces no divergence"
→ but genre-assignment IS the heteronymic operation
In both cases the quantity controlled away was the mechanism. Recorded because the pattern is now twice-observed and is a standing hazard of this whole investigation: when the phenomenon is an operation performed ON a variable, controlling for that variable deletes the phenomenon.
Pessoa's authorship is documented — the Casais Monteiro letter of 13 January 1935. So this is not a blind test and should not be run as one. We are not asking do the voices separate. We are asking:
Along what axes does one author's splitting actually run?
That is characterization, not inference, and it is what a positive control is for. The H-class calibration EA-RELFORM-01 needs is not "does heteronymy produce X" but "what is the native shape of heteronymic splitting" — measured on a case where the answer to the authorship question is already known.
The uncontrolled 1.415 is therefore the number that means something, because it includes the splitting along every axis the maker used at once: genre, language, register, meter, philosophy, biography. Those are not separable components of a confound. They are the dimensions of the partition.
The orthonym is at 628 of 2,814 texts and is the largest gap in the seat. It is also the position most likely to matter, since the orthonym is where an apparatus would concentrate if it concentrates anywhere. Any characterization of the native shape is provisional until it is complete.
Added 2026-08-29. EA-CORPORA-11 completed: 1,254,436 tokens / 25 positions, orthonym 2810/2814, Campos 396/397, Reis 307/307, Soares 504/504, Caeiro 127/127, Search 139/139, Mora 95/95. Eleven times the original seat. Characterization, not inference — Pessoa's authorship is documented.
fernando-pessoa (ORTHONYM) 0.0445 n=266 blocks
alvaro-de-campos 0.0400 n=38
antonio-mora 0.0317 n=16
ricardo-reis 0.0305 n=15
alexander-search 0.0298 n=8
bernardo-soares 0.0278 n=46
alberto-caeiro 0.0275 n=6
The orthonym is the most internally heterogeneous position in the system —
more varied than any constructed voice, by a wide margin over Caeiro and Soares.
Portuguese positions only (Search dropped as a language outlier at 0.089–0.109
from everything):
position internal to others ratio
fernando-pessoa 0.0445 0.0460 1.033 ← THE ORTHONYM
alvaro-de-campos 0.0400 0.0451 1.129
ricardo-reis 0.0305 0.0419 1.374
antonio-mora 0.0317 0.0455 1.437
bernardo-soares 0.0278 0.0455 1.632
alberto-caeiro 0.0275 0.0490 1.781
**The orthonym sits at 1.033. It is as far from itself as it is from the
heteronyms.**
The constructed positions are containers — Caeiro at 1.781, Soares at 1.632, each tighter internally than its distance to the rest. The orthonym is not a container. Its internal variance equals its separation from the voices it generates.
This is the intra-Aristotle shape, in a corpus whose single authorship is documented. M10 recorded Aristotle as the most internally heterogeneous author in the entire sample — dispersion 0.1033, above the 100th percentile of synthetic five-author pools — and I treated that as the anomaly to be explained. M28 then explained it as the object.
Here the same shape appears in Fernando Pessoa's orthonym, where the answer to
the authorship question is known from the Casais Monteiro letter of 13 January
1935. In a documented heteronymic system, **the position that carries the maker's
own name is the least container-like thing in it.**
Gradient, and it is orderly: the more sharply specified the position, the tighter the container. Caeiro — the pastoralist with the smallest declared vocabulary and the most fixed persona — is tightest at 1.781. The orthonym, which is not a persona at all, is loosest at 1.033.
THE SPECIFICATION HYPOTHESIS (registered as EA-SPEC-01). Tightness tracks the
severity of the constraint, not similarity and not disguise:
Caeiro 1.781 no philosophy, no memory, no interiority — sensation only
Soares 1.632 one street, one office, one clerk, one register
Mora 1.437 one doctrine, argued
Reis 1.374 one meter, one stance
Campos 1.129 everything, at maximum volume
ORTHONYM 1.033 no restriction at all
Read as a series, the heteronyms are a search downward: what is the least one can specify and still get a self that coheres? Each position subtracts something and tests whether a voice survives the subtraction. Caeiro is the floor — remove more and no one is there. A container is tight because something was taken away, and the ratio is how much.
Nothing about Plato and Aristotle. It establishes what one axis of heteronymic structure looks like when the answer is known, which is what an H-class positive control is for (EA-RELFORM-01 §3).
Two limits, stated. Block counts are very uneven — the orthonym has 266 blocks against Caeiro's 6, and small-n positions have noisier internal estimates. Function-word profiles may be the wrong instrument for verse, where Reis and Campos differ in meter and line-length more than in particles; that is a construct-validity risk carried forward, not resolved.
*Added 2026-08-29. Registered fd839bbcf7ff94cc… with the constraint ranking frozen
before measurement.*
Speakers extracted by Perseus tag, ≥6,000 tokens, 2,500-token blocks, top-150
Greek function words:
speaker blocks internal to others ratio
ΤΙ Timaeus 8 0.0285 0.0470 1.649
ΤΙΜ Timaeus (whole file) 9 0.0302 0.0463 1.532
ΑΘ Athenian 41 0.0303 0.0435 1.433
ΦΑΙΔ Phaedo narrator 8 0.0320 0.0420 1.313
ΑΠΟΛ Apollodorus 7 0.0315 0.0412 1.310
ΞΕ Eleatic Stranger 11 0.0351 0.0424 1.207
ΣΩ SOCRATES 68 0.0377 0.0439 1.166
P2 HOLDS. Socrates — frozen as the least constrained speaking position — comes in rank 1 of 7 from the loose end at 1.166. And Timaeus is tightest, also as frozen: one continuous cosmological monologue, the most narrowly specified position in the corpus.
P1 PARTIAL. The ordering runs Timaeus → Athenian → narrators → Eleatic Stranger → Socrates, which is the predicted direction, but the Eleatic Stranger lands second-loosest at 1.207 where I ranked him constrained. (Post-hoc, marked as such and unavailable as a rescue: he speaks across two dialogues on different subjects. That reasoning was available before the run and I did not use it.)
MOST OF THE FROZEN RANKING WAS UNTESTABLE. Diotima, the constructed Aristophanes, Parmenides, Protagoras and Callicles all fall under the threshold. Seven speakers of the ranked set cleared it.
⚠ EXTRACTION FLAW: ΤΙ and ΤΙΜ-mono overlap — the tagged Timaeus speeches sit inside the whole file I also counted. They are not independent and should not both appear. Recorded rather than silently dropped.
And §5's limit governs everything above: this buys nothing about authorship. A dramatist specifying characters to differing depths produces the same gradient. What it shows is that the Platonic corpus is ordered by specification depth in the same ordinal pattern as a documented heteronymic system — construction, which dramatic and heteronymic composition share.
Speaker is not the right unit, and the corpus itself says so. M16: Plato's constructed Aristophanes ranks 5 of 7 Symposium speeches for resemblance to the real Aristophanes, and sits at 0.0380 from the Plato corpus — more Platonic than its neighbours. Speaker-differentiation is sometimes deliberately switched off. Where it is off, a per-speaker measure reads the absence of construction as the absence of a position.
So the measure has to run at several units and be compared across them:
WORK the unit YIELD-02 used, where both corpora filled 6/6
SPEAKER run here — valid only where differentiation is switched ON
DIALOGUE-GROUP the tetralogy or thematic cluster
CORPUS the whole, which is where Pessoa's orthonym result sits
The Pessoa comparison was run at position level and the Plato test at speaker level. Those are not the same unit and the parallel drawn in §88 is weaker for it. Work-level is the next run and it is the one that matches EA-CORPORA-11's structure most closely.
Added 2026-08-29. Albert the Great fetched from the Borgnet Opera omnia (nine volumes, 3.46M tokens) and scored against Aquinas under EA-RELFORM-01 §4's contamination rule. The first externally-secured control result in this investigation.
v03 Physica 375,768 physicorum 259 aristotle-commentary
v05 De anima / Parva nat. 384,757 animalibus 79 aristotle-commentary
v06 Metaphysica 433,651 metaphysic 320 aristotle-commentary
v07 Ethica 338,391 ethicorum 197 aristotle-commentary
v15 In Psalmos 433,173 psalm 933 other-commentary
v19 biblical 356,088 other-commentary
v27 In Sententias 326,407 sententiarum 44 non-aristotle
v31 In Sententias 530,878 sententiarum 121 non-aristotle
v33 In Sententias 283,189 sententiarum 64 non-aristotle
non-Aristotle-commentary works only tokens gap/100k div/10k
ALBERT In Sententias 1,140,474 0.00 16.4
ALBERT biblical 789,261 0.00 4.7
AQUINAS independent 5,219,700 1.07 26.6
AQUINAS other-commentary 457,728 0.22 10.3
where the source IS Aristotle
ALBERT Aristotle commentaries 1,532,567 1.31 35.5
AQUINAS Aristotle commentaries 1,020,889 3.72 48.5
**Albert shows ZERO instances in 1,140,474 tokens of Sentences commentary, while
dividing at 16.4 per 10,000.** He has ample opportunity and does not do it.
POSITIVE CONTROL PASSES. The Latin probe finds 20 instances in Albert's Aristotle commentaries, 13 in the Ethics alone — including qui in timiditate superabundat Graece innominatus est, multa enim sunt innominata in mediis et in extremitatibus, glossing NE 1107b. The instrument sees the phenomenon in this author's Latin. The zero is absence, not blindness.
documented master-student pair, non-Aristotle works:
Albert 0.00/100k at division 16.4/10k
Thomas 1.07/100k at division 26.6/10k → THEY DO NOT SHARE IT
the Greek pair:
Plato 2.60/100k at division 15.4/10k
Aristotle 8.43/100k at division 30.9/10k → BOTH CARRY IT
The reflex does not transmit master-to-student into independent writing. In both Latin authors it appears only where Aristotle's text is being glossed — Albert 1.31 on the commentaries against 0.00 elsewhere, Aquinas 3.72 against 1.07. It arrives with the source and does not become a habit.
This is the first evidence bearing on the objection that killed M30's diagnosticity this morning. "Schools transmit arbitrary procedural habits" was an unmeasured auxiliary (M38 L2, corrected to UNMEASURED). Measured on a documented pair under maximal transmission pressure, it does not hold for this reflex.
Not that Plato and Aristotle are one hand. Two authors both carrying a habit that a Latin pair does not share is a difference between pairs, not a fact about authorship. Ordinary explanations remain: Greek philosophical prose may simply carry the reflex where scholastic Latin does not; the Latin idiom for the same move may differ from my probe set despite the positive control; and one control pair is one draw.
Sample: nine Borgnet volumes of thirty-eight. The Sentences commentary dominates the non-Aristotle sample and is a single work-type. Albert's genuinely independent treatises — De natura boni, the Summa theologiae — are not in the fetched set.
And the OCR caveat stands (M43d): usable for structural and function-word measures, unsafe for rare-lexeme measures. innominat- is a rare lexeme. Albert's zero could in principle be OCR loss, though the positive control makes that unlikely — the same OCR preserved 20 instances in the commentaries.
EA-RELFORM-01 §5 order of operations was NOT followed. The measure was not calibrated on the labelled classes before Plato–Aristotle was scored; the Greek numbers existed first. This is a comparison, not the registered classification test, and it is recorded as such.
*Added 2026-08-29. A correction to how Rounds 21–22 measured Pessoa, and the
distinction that makes the earlier comparison invalid.*
Round 22 §86 set Pessoa's voice-partition beside Aristotle's domain-partition and read the ratios as commensurable. They are not. One instrument was run across two different operations.
PLATO / ARISTOTLE — division by OBJECT
maximum substrate divergence (Z = +39.18), resolving to EXACT PARITY when
the object is controlled (0.0571 ≈ 0.0570), with interlocking function
(6/6 slots, one shared figure, one abandoned execution completed twice)
→ THE INTEGRAL IS A SUM. Complementary parts that fit.
What closes it is COMPLETENESS OF COVERAGE.
PESSOA — division by progressive REMOVAL
each position subtracts more; containers tighten as the constraint sharpens
(Caeiro 1.781 → orthonym 1.033)
→ THE INTEGRAL IS A REMAINDER. What persists through every narrowing.
What closes it is INVARIANCE.
Both are heteronymy. They are not the same operation, and the right measure for
the second is not distance-between-voices but what survives the subtractions.
Six Portuguese positions ordered by constraint severity (Caeiro tightest →
orthonym loosest), top-120 features, coefficient of variation across the gradient:
INVARIANT CV
há (there is) 0.080 ← the most invariant word in the system
que 0.094
de 0.101
no 0.102
o 0.108
um 0.121
com 0.123
Existence-assertion plus the bare connective spine. That is what cannot be
removed.
rises as the constraint TIGHTENS rises as it LOOSENS
como r = −0.82 também r = +0.86
porque r = −0.79 ou r = +0.80
não r = −0.76 em r = +0.80
coisas r = −0.75
The tighter the position, the more it runs on negation, comparison, causation and the generic noun. não · como · porque · coisas — which is Caeiro exactly: as coisas não têm significação: têm existência.
A narrowed position is constituted by what it REFUSES, and refusal is expensive in negations and similes. The subtraction has to be performed in the text, and performing it costs vocabulary.
and · the · it · of · is · that · be · not appear at CV ≈ 1.8 with strong positive trend — English inside the orthonym file. Pessoa wrote in English under his own name, and Arquivo Pessoa files it under position 1. The loosest position's numbers are polluted and the orthonym rate must be recomputed with non-Portuguese text stripped before §95–96 are trusted. The invariant list is less affected (those are Portuguese function words) but the trend coefficients for the loose end are not clean.
The Plato–Aristotle comparison in Round 22 §86 stands only as far as the observation that a documented heteronymic corpus has an orthonym less container-like than its constructed voices. The quantitative parallel with Aristotle's domain-partition is withdrawn: it compared a remainder-structure to a sum-structure using a distance instrument built for neither.
And it sharpens what the Greek pair would have to show to be the same KIND of object as Pessoa. Not similar ratios — a remainder: something invariant across the positions that is not explained by shared language, genre or subject. That measure has not been run on Plato and Aristotle, and it is a different question from the six-slot coverage that Round 12 measured.
Added 2026-08-29. Lee's item of heteronymic provenance theory: a heteronym often has to die; a heteronym emerges from the death. Checked in the one documented case.
Alberto Caeiro (1889-1915) given a lifespan; dies at 26
"LAST POEM (ditado pelo poeta a death-day text, final words
no dia da sua morte)"
Campos reports the master's «Ora suponha que» ia eu uma vez dizendo,
speech in dialogue mas ele interrompeu… E o Mestre retorquiu
Mora narrates the arrival Nesta altura SURGIU Alberto Caeiro
Pessoa names the structure o início da «ESCOLA»: as obras do MESTRE
e algumas do DISCÍPULO DIRECTO
Caeiro's last poem: *É talvez o último dia da minha vida. / Saudei o sol,
levantando a mão direita, / Mas não o saudei, para lhe dizer adeus.*
So the documented heteronymic system has: a master who dies young, disciples who report his words in dialogue, a school, and a last utterance on the day of his death. The master position is constituted by dying — Caeiro cannot be the master until he is gone, and the disciples' relation to him is a relation to someone absent.
That in the one case where authorship is documented, the provenance structure runs: death of the founding position → the interpreting positions as survivors. Designed, dated, and attested by the Casais Monteiro letter of 13 January 1935.
It does not establish that death-of-the-master is an ESSENTIAL feature of heteronymy. One documented case cannot make a structure necessary, and there is a much simpler account available:
Pessoa was imitating Plato.
A master who writes nothing, dies young, and survives in the reported speech of disciples is the Socratic structure, and Pessoa knew it. The parallel may be influence, not convergence — Caeiro built on a model Pessoa had read, in which case the Pessoan case is not independent evidence about heteronymic form at all. It is a copy of the object under test.
Recorded as the limiting reading, and it is the correct default.
**The most famous declared heteronymist in the modern canon, given a free hand to
construct any authorial architecture at all, built THIS one.**
a master who writes nothing decisive and dies
disciples who transmit him in reported dialogue
a school with a named succession
a last utterance on the day of the death
an orthonym that is not a container (Round 22)
positions differentiated by progressive constraint (Round 25)
**Kierkegaard, independently, also built a school with a founder-figure, reported
speech, pseudonymous disciples, and a signed explanation appended at the end.**
That two of the best-documented heteronymic systems both reproduce the Socratic-Platonic architecture is either because that architecture is what heteronymy converges on, or because both were imitating the same model — and those two readings are not distinguishable from the cases themselves.
But note what the second reading concedes. If Pessoa and Kierkegaard were imitating Plato, they were imitating it as a heteronymic architecture — reading the master-disciple-school-death structure as a way to build authorial positions, and finding it good enough to copy. The imitation is itself a reading of the Platonic corpus, made by the two people in the modern canon best qualified to recognise the operation. That is not evidence. It is not nothing either.
Status: provenance-theory item, characterised in one documented case, with the imitation reading recorded as the default and the convergence reading recorded as undecidable from the cases.
*Added 2026-08-29. EA-FULFIL-01 registered 645e9620e857f30a…, construct-validity
check run.*
The Matthean structure — ἵνα πληρωθῇ τὸ ῥηθὲν διὰ τοῦ προφήτου — has three parts: a later text claims an earlier was about it; the earlier was not making that claim (Hosea 11:1 looks back at the Exodus, and Matthew reads it forward); and the relation works by structural correspondence rather than the earlier text's intent.
case A: announced C: closeness S: scope
Matthew / Torah YES loose verse-level targets
Catullus / Sappho no VERY CLOSE one poem, register-shifted
Aristotle / Plato no ? whole apparatus (claimed)
This replaces the completion test's broken form. Not do the six slots transfer — that imports the Sapphic thesis into its own control — but does a documented master-student pair produce unannounced fulfillment-relations at Catullan closeness across a whole apparatus?
slot S31 C51
seems-to-me equal-to-god ✓ ✓ φαίνεταί μοι ἴσος θεοῖσιν
/ Ille mi par esse deo videtur
that man ✓ ✓ κῆνος / ille
sitting opposite you ✓ ✓ ἐναντίος τοι ἱζάνει
/ qui sedens adversus te
hears ✓ ✓ ὑπακούει / audit
sweetly laughing ✓ ✓ γελαίσας ἱμερόεν / dulce ridentem
heart-struck → senses-snatched ✓ ~ ἐπτόασεν / omnis eripit sensus
voice fails ✓ ~
tongue breaks → grows numb ✓ ✓ γλῶσσα ἔαγε / lingua torpet
thin fire ✓ ✓ λεπτὸν πῦρ / tenuis flamma
runs under the flesh/limbs ✓ ✓ ὑπαδεδρόμακεν / sub artus demanat
eyes see nothing ✓ ✓
ears ring ✓ ✓
────────────────────────────────────────────
sweat pours ✓ ✗
trembling seizes ✓ ✗
greener than grass ✓ ✗
seem near to dying ✓ ✗
C = 11/16 = 0.688
Sequence preserved throughout except one inversion (eyes/ears). Aspect systematically shifted: aorist ἔαγε → present torpet; perfect ὑπαδεδρόμακεν → present demanat. The last four slots are not omitted but REPLACED — by the otium stanza, which is the only decisive completion in the Latin poem.
CONSTRUCT VALIDITY PASSES. Two arbitrary poems by the same author in the same language — Catullus 5 against Catullus 64 — share 0.023 of content lemmas. C is a sequence-aligned slot measure and cannot be produced by vocabulary overlap between unrelated texts.
Measurable: whether Albert→Aquinas produces relations at C ≈ 0.7, unannounced, across an apparatus. If it does, the Plato–Aristotle structure is not singular and frequency inference becomes available. If it does not, singularity is consistent with the evidence.
⚠ NOT measurable, and registered as the governing difficulty (EA-FULFIL-01 §4): if the relation is genuinely singular, frequency inference is unavailable — no comparison class, no percentile, no likelihood ratio. Everything this investigation has built assumes a distribution to score against.
The available substitute is a FIT test, not a frequency test: what is the probability of this configuration arising with no relation? — which needs only the two corpora and a null model of accidental correspondence.
AND THE CEILING IS FIXED BY THE CALIBRATION ITSELF. Catullus scores 0.688 and is not Sappho. Maximal C is compatible with distinct authorship, demonstrably so, in the standard case. Any C value for Plato–Aristotle establishes relation, never authorship — and the whole day is a record of measures that collapse at exactly that step.
*Added 2026-08-29. The counterintuitive claim, demonstrated in its general form
and requiring no philology about the Greek pair.*
We do not need to say what the Plato–Aristotle relation IS. We need only answer: does the master–student relation generally produce interlocking fulfillment at that closeness?
That is a claim about what school structure defaults to — answerable from
documented cases, and counterintuitive if it comes back negative.
A UNANNOUNCED the later text does not declare the relation
Q SEQUENCE-KEPT the earlier text's order is preserved
T TRANSFORMED the material is altered, not expounded
ANNOUNCEMENT (source named, per 10k)
Aquinas, Metaphysics commentary 14.4
Aquinas, Summa Theologiae (independent) 21.9
TRANSFORMATION (content tokens)
Aristotle, Metaphysics 42,065
Aquinas, In Metaphysicam 194,064
ratio 4.61×
A school under maximal inheritance pressure produces COMMENTARY, and commentary is the structural OPPOSITE of the fulfillment operator on two of three parameters.
It announces at saturation — and more often in the independent works (21.9) than in the commentary itself (14.4). The relation is declared on nearly every page.
It expands rather than transforms — 4.61× its source. Against which Catullus is 16 lines to Sappho's 17, with four slots not omitted but replaced. Compression, register-shift, silence.
relation unannounced sequence transformed
Aquinas → Aristotle (comm.) ✗ ✓ ✗
Matthew → Torah ✗ ~ ✓
Catullus → Sappho ✓ ✓ ✓
Aristotle → Plato (claimed) ✓ ? ✓
Interlocking unannounced fulfillment is not what a school produces. A school produces exposition that names its master and grows longer than him. The operator requires the reverse: silence about the source, and material compressed rather than expounded.
Lee's formulation, which is the round's actual claim:
*Not one of them will produce an interlocking machine that close — and that is
the contrary of the expected pattern, where it normally takes TIME AND DISTANCE
across inheritance to build the interlocking fulfillment machine.*
The intuition says proximity produces tight fit. The measurement says the reverse: direct contact produces commentary; the fulfillment operator appears across distance. Catullus is two centuries and a language from Sappho. Matthew is centuries from Hosea. Both are unannounced or loosely-fitted across a gap; neither is a pupil in the room.
A pupil in the room does not need to fulfil his master. He can cite him.
Every directly attested master–student contact line gives a datum, and the prediction is uniform: commentary and citation, not unannounced interlocking fulfillment.
Aristotle → Theophrastus direct contact, Lyceum [seated]
Albert → Aquinas direct contact, Cologne [seated]
Isaeus → Demosthenes direct contact, attested [seated]
Freud → Jung direct contact, correspondence extant
Williams → Ginsberg direct contact, Paterson; W. wrote the
introduction to Howl
Husserl → Heidegger direct contact, assistant then successor
Boas → Mead direct contact, doctoral
Schoenberg → Berg/Webern direct contact, private lessons
Each is scored on the same three parameters. The claim is falsified by ONE directly-attested pair producing unannounced, sequence-kept, transformative interlock across an apparatus.
Williams → Ginsberg is the sharpest of the unrun, because Williams announced it — he wrote the introduction to Howl. Announcement by the master is the strongest possible failure of property A, and it is what direct contact characteristically produces.
STATED LIMITS. One school pair measured; a commentary compared against a poem across genres; and this shows what the school structure DEFAULTS to, not that no school has ever produced fulfillment. The programme in §111 is what would convert a default into a rate.
Added 2026-08-29. The comparative measure named: it is not raw distance but WHAT YOU HAVE TO WORK WITH. With the text alone you must reconstruct; with an institution around the text you can cite.
Sappho → Catullus ~550 yr, language change, NO INSTITUTION
→ unannounced transformation, C = 0.688
Hosea → Matthew ~700 yr, language change, synagogue
→ ANNOUNCED fulfillment, loose fit
Catullus → Ginsberg ~2000 yr, language change, NO INSTITUTION
→ "Malest Cornifici Tuo Catullo" — titled in the
source's own Latin
Aristotle → Aquinas ~1600 yr, language change, UNIVERSITY
→ commentary: 14.8/10k naming, 4.61× expansion
Albert → Aquinas ~20 yr, same language, DOMINICAN STUDIUM
→ commentary: 12-16/10k naming
Distance alone does not produce the operator. Aquinas is 1,600 years and a language from Aristotle and produces commentary, because he has a university, a translation tradition, and a commentary apparatus. Catullus has a poem.
So the variable is INSTITUTIONAL MEDIATION, not years. When the text arrives inside an institution built to transmit it, you get exposition. When it arrives alone, you get transformation.
ARISTOTLE, corpus 761,916 content tokens
Πλάτων- 53 0.70/10k ← the SOURCE named
Σωκρατ- 144 1.89/10k ← the VARIABLE, not citation (M34)
**M34 established Socrates is Aristotle's interchangeable concrete individual — Καλλίαν ἢ Σωκράτην. So Σωκρατ
later text names master/10k ratio
ARISTOTLE names Plato 0.70 1×
Aquinas names Aristotle (Summa) 12.00 17×
Aquinas names Aristotle (commentary) 14.80 21×
Albert names Aristotle (In Sent.) 16.20 23×
At the same institutional proximity — direct pupil, same school, twenty years — the scholastics name their master 17 to 23 times more often than Aristotle names Plato.
And the corpus ratios: Aristotle 1.60× Plato — a corpus of comparable scale.
Aquinas 4.61× the Metaphysics — exposition.
Aristotle has the profile of DISTANCE at PROXIMATE range.
proximate relations produce frequent naming, expansion, exposition
distant relations produce silence about the source, compression,
transformation
Aristotle has: 0.70/10k naming, 1.60× scale
— the distant profile
Aristotle's situation was: same city, same school, twenty years,
the master alive
— maximal proximity
**He had every institutional resource that produces commentary, and the relation
has none of commentary's properties.**
DOES: the school structure has a measured default — name the master often, expand him, expound him — and Plato–Aristotle departs from it by an order of magnitude on the one parameter that is cleanly countable.
DOES NOT: establish authorship. Both readings must account for the departure. H_S accounts for it by Aristotle was unusually independent — which is true, is what everyone says, and is one more auxiliary on M38's ledger, unmeasured until now and now measured at 17–23×.
LIMITS. Naming conventions differ between fourth-century Greek philosophical prose and thirteenth-century scholastic Latin, and that difference is not controlled. Greek philosophers may simply name predecessors less. The Theophrastus check is available and unrun: he is the one Greek writer at the same proximity in the same tradition, and his naming rate for Aristotle would calibrate the convention directly.
*Added 2026-08-29. Two rounds voided, the calibration standard corrected, and a
methodological rule I had been enforcing backwards.*
My taxonomy was wrong at its own calibration standard. Catullus 51 declares its
relation to Sappho three times over:
the SAPPHIC STROPHE the metre IS the citation
TRANSLATION FORM it presents itself as a rendering
"LESBIA" the woman of Lesbos — and poem 51 is where the
name first appears, INSIDE the Sappho translation
Corrected:
pair announced? fit scope
Matthew → Torah YES — ἵνα πληρωθῇ formula loose verse targets
Catullus → Sappho YES — metre, form, name VERY CLOSE one poem
Aristotle → Plato NO close? whole apparatus
Everyone who performs a transformation this close ANNOUNCES it. The announcement is how the reader is told to read it as a transformation — Matthew by formula, Catullus by metre and name. Aristotle does not. That is the distinctive feature, and correcting the error sharpens it rather than weakening it.
Round 28 established three properties — unannounced, sequence-kept, transformed — and I then measured only announcement, and announcement as name-frequency, which is not even the right form of A.
A is not "how often is the source named." The spec was: unannounced AT THE MOMENT OF GREATEST FUNCTIONAL INTERLOCK. A is conditional on Q and T — it means something only at the loci where the machine engages. At De anima 424a, completing the abandoned wax block with the seal figure retained and ἄνευ τῆς ὕλης supplied, there is no mention of Plato and no cross-reference to the Theaetetus. That is the observable. A corpus-wide naming rate cannot see it, and I ran a corpus-wide naming rate.
And then I called the result a refutation. That is the worse error. Producing a number that does not measure the variable and announcing it as a failure is not rigour — it is the same fabrication as announcing a success, wearing rigour as costume. Self-refutation is honest only when the thing refuted is the thing that was claimed. Round 29's Theophrastus figure refutes a naming-convention claim nobody made.
Q and T were never measured at all, and the interlock lives in those.
I had been treating the identification of correspondences as contamination, and
reaching for mechanical proxies to avoid it. That is backwards.
You cannot measure closeness of fit without first identifying the fit.
Searching for correspondences is constitutive of the measure, not a breach of it. EA-UNLOCK-01 §5's reverse-engineering rule governs SELECTING WHICH TARGET TO SCORE after seeing the outcome. It does not and cannot forbid IDENTIFYING THE ALIGNMENT the score is computed over. I collapsed those two, and that collapse is why I kept producing measures that measured nothing.
No composite score. Each property is its own question, asked separately of the
school-inheritance model, and answered by reading and by cases:
Q1 Is UNANNOUNCED TRANSFORMATION characteristic of school inheritance?
— the Catullus correction says even the distant, unschooled case
announces. Does any pupil transform in silence?
Q2 Is SEQUENCE-PRESERVATION across an apparatus characteristic?
— commentary preserves sequence trivially, by following the text.
Does any pupil preserve sequence while NOT commenting?
Q3 Is TRANSFORMATION rather than EXPOSITION characteristic?
— Aquinas expands 4.61×. Catullus renders 16 lines for 17 with four
slots REPLACED. Does any pupil compress and replace?
**Each is tractable as a philological question about specific cases. None requires
a composite statistic, and the attempt to build one produced two void rounds.**
Added 2026-08-29. The design corrected: the blindness belongs on Plato–Aristotle, not on fit-identification. Characterise the classes from cases where the relation is known; only afterwards ask where the target falls. This is EA-RELFORM-01 §5's order of operations, specified this morning and followed for the first time here.
ISAEUS → DEMOSTHENES. Direct teaching attested by Dionysius of Halicarnassus and the Lives of the Ten Orators. Fourth century BCE, Greek, same culture. Isaeus 28,022 content tokens; Demosthenes 246,565.
Plato and Aristotle were not examined in producing what follows.
C1 — THE MASTER IS NOT NAMED.
Ἰσαῖος occurs ZERO times in 246,565 content tokens of Demosthenes. *(A first
count returned 162; checked, and every hit was a form of ἴσος/ἰσόω. Corrected
before reporting.)* The teaching relation is externally attested and textually
invisible.
→ Naming is not how direct inheritance shows itself. This is the finding that retroactively explains why Rounds 28–29 measured nothing: naming rates track scholastic citation convention, not inheritance.
C2 — THE APPARATUS IS INHERITED WHOLE BUT DILUTED.
term Isaeus/10k Demosthenes/10k ratio
κληρ- 94.57 10.10 0.11
ἀγχιστ- 14.27 0.41 0.03
γνήσι- 28.19 1.66 0.06
διαθηκ- 38.18 4.50 0.12
ἀμφισβητ- 34.26 3.97 0.12
μαρτυρ- 142.74 57.59 0.40
Every technical term of the master appears in the pupil at 3–30× lower density — while absolute counts stay comparable (inheritance cases 265 vs 249; wills 107 vs 111) across a corpus 8.8× larger.
→ The pupil carries the whole apparatus into a WIDER FIELD. Dilution, not
selection.
C3 — THE OBJECTS PERSIST. He is still trying inheritance cases.
→ Direct inheritance transmits the subject, not only the method.
C4 — EXPANSION. 8.80× the master's corpus — matching Aquinas at 4.61×.
→ Pupils write more than masters.
**Only now, and only after more cases, may the question be asked: does the
Plato–Aristotle relation fit this profile or not?**
The four characteristics are stated as properties of a known relation. They were derived without reference to the target. Whether the target matches them is a separate question, to be asked afterwards, and not asked here.
NEXT CASES NEEDED FOR THE CLASS — each seated or obtainable, each with the
relation externally attested:
Albert → Aquinas partly characterised (expansion 4.61×,
citation-heavy — but Latin convention)
Freud → Jung correspondence extant; break and RENAMING
Husserl → Heidegger assistant then successor; renaming (Dasein)
Williams → Ginsberg master wrote the pupil's introduction
Boas → Mead · Schoenberg → Berg
And the distance class needs the same treatment, from Sappho→Catullus,
Hosea→Matthew, Catullus→Ginsberg, Aristotle→Aquinas.
**The discriminating features must come OUT of those two characterisations.
Nothing may be imported from the Greek pair.**
*Added 2026-08-29. Gathering broadly without deciding in advance what matters.
Plato and Aristotle not examined.*
Nine features, computed identically on both pairs. No claim is made that any of them is significant — they are gathered so that, across enough pairs, what holds can be seen rather than posited.
F1 corpus size ratio (pupil / master)
F2 % of master's distinctive terms that appear in the pupil at all
F3 median density ratio across inherited terms
F4 % of inherited terms DENSER in the pupil
F5 % of pupil's distinctive terms ABSENT from the master
F6 type-token ratio, master vs pupil
F7 hapax fraction, master vs pupil
F8 top-1000 vocabulary Jaccard
F9 uniformity of density direction (is the shift one-way?)
DIRECT #1 — Isaeus → Demosthenes DIRECT #2 — Albert → Aquinas
(Dionysius, Lives of the Ten Orators) (Cologne, attested)
28,022 → 246,565 tokens 2,102,245 → 5,243,581 tokens
F1 size ratio 8.80× 2.49×
F2 terms inherited 96.7% 94.0%
F3 median density 0.665 1.003
F4 denser in pupil 25.9% 50.4%
F5 novel in pupil 3.0% 0.3%
F6 TTR Δ −0.086 −0.078
F7 hapax Δ −0.039 −0.333
F8 Jaccard 0.384 0.476
F9 direction uniform 0.741 0.504
F2 — NEAR-TOTAL APPARATUS INHERITANCE. 96.7% and 94.0%. Of the master's three hundred most distinctive terms, the pupil uses essentially all of them at least once. This is the strongest cross-pair regularity in the battery.
F5 — THE PUPIL INTRODUCES ALMOST NOTHING LEXICALLY NEW. 3.0% and 0.3%. Of the pupil's own three hundred most distinctive terms, virtually none is absent from the master. A pupil's distinctive vocabulary is his master's vocabulary.
F6 — TYPE-TOKEN RATIO FALLS, both by ≈0.08. The pupil is lexically less varied
per token — consistent with the larger corpus, but the magnitudes match closely.
F1 — EXPANSION, both directions positive. 8.80× and 2.49×. The pupil writes
more.
F3 and F4 split the pairs. Demosthenes dilutes (median 0.665, only 26% of terms denser); Aquinas holds density (1.003, 50% denser). F9 follows: Demosthenes' shift is one-way (0.741 uniformity), Aquinas' is balanced (0.504).
→ Dilution is not a class property. It distinguishes carrying an apparatus into a wider field (Demosthenes: inheritance law into politics) from staying in the master's field (Aquinas: scholastic theology into scholastic theology). Two sub-shapes of direct inheritance, not one.
Two pairs is not a class. F2 and F5 are the candidates for a real regularity — near-total inheritance of the master's terms, near-zero novel vocabulary in the pupil — and both would need to survive Freud→Jung, Husserl→Heidegger, Williams→Ginsberg, Boas→Mead before being called characteristic.
The distance class has not been characterised at all, and it must be, on the same battery, before any feature can be called discriminating rather than merely present.
**Plato and Aristotle remain unexamined on this battery. That is deliberate and
must hold until both classes are characterised.**
Added 2026-08-29. Fourteen pairs across three classes, with the class scheme corrected and the lexical battery shown to measure something other than what it was built for. Plato and Aristotle still not examined.
Lee's correction: constructed lineage is the oldest move in the book, and the more ancient the pair, the likelier an "inheritance" is a retrospective construction. Homer → Apollonius is not a neutral distance control; the Alexandrian poet claiming epic descent is a constructed lineage.
1 MASTER-STUDENT DIRECT embodied, externally attested contact
2 DISTANT CONSTRUCTED a later writer constructs the relation
3 HETERONYMIC CONSTRUCTED one maker constructs both positions
**The classes are marked as different from one another without deciding what the
difference IS.** That decision comes after enough cases, not before.
pair relation F2 F5 F8
Lysias → Isocrates CONTEMPORARY 100.0 0.7 0.404
Thucydides → Xenophon succession 98.7 4.0 0.383
Isaeus → Demosthenes DIRECT PUPIL 96.7 3.0 0.384
Albert → Aquinas DIRECT PUPIL 94.0 0.3 0.476
Thucydides → Philo distance 91.3 7.7 0.223
Herodotus → Josephus distance 79.3 23.7 0.230
Herodotus → Thucydides CONTEMPORARY 76.0 22.7 0.244
Herodotus → Plotinus distance 65.7 22.7 0.169
Lysias → Isocrates has NO teaching relation and scores higher on every inheritance feature than Isaeus → Demosthenes, which has one. And Herodotus → Thucydides is contemporary and scores like the distance pairs — because they write different dialects, Ionic against Attic.
→ The battery tracks TEMPORAL AND DIALECTAL PROXIMITY. It does not track pupillage. Round 32's F2 and F5 "regularities" are period effects. Two Attic prose writers of one generation look like master and student whether or not either taught the other.
→ No lexical-distributional measure can identify a teaching relation. That is a hard constraint on an entire family of instruments, and it retroactively explains why every measure today collapsed at the same step.
pair F1 F2 F3 F4 F5 F8 F9
Caeiro → Reis 2.54 96.6 0.477 15.5 16.7 0.353 0.845
Caeiro → Campos 6.12 99.6 0.609 20.2 9.3 0.380 0.798
Caeiro → Mora 2.58 89.3 0.481 23.9 22.7 0.274 0.761
Caeiro → orthonym 42.35 100.0 0.395 9.8 26.0 0.289 0.902
Caeiro → Soares (unrelated) 7.31 95.7 0.549 19.6 10.3 0.377 0.804
Either/Or A → B 1.03 98.7 0.981 48.0 1.0 0.535 0.520
A → signed 0.85 94.7 0.438 10.9 34.0 0.280 0.891
class n F2 range F5 range F8 range
1 DIRECT 2 94.0 - 96.7 0.3 - 3.0 0.384 - 0.476
2 CONSTRUCTED 5 65.7 - 97.7 1.5 - 23.7 0.169 - 0.286
3 HETERONYMIC 7 89.3 - 100.0 1.0 - 34.0 0.274 - 0.535
The ranges overlap on every feature. Class 3 spans both others. And within class 3, Caeiro → Soares — two positions with no constructed master-disciple relation between them — scores like Caeiro → Campos, which has one.
→ Confirms §129 from the other side. The battery cannot separate the classes, and cannot separate related from unrelated positions inside a single maker's work.
A hard negative, and it is worth more than the positives it replaces:
Relation-type is not recoverable from lexical distribution. Not pupillage,
not constructed lineage, not heteronymic construction. What lexical measures see
is period, dialect, genre and subject.
Therefore the fulfillment question requires STRUCTURE, not vocabulary — slots, sequence, and transformation-versus-exposition, which is what the Catullus calibration measures (Round 27) and what has not been built.
Fourteen pairs, one constraint, and the instrument family ruled out. Plato and Aristotle remain unexamined, which is now cheap to maintain, since the battery would tell us nothing about them either.
*Added 2026-08-29. Withdrawing Round 33 §129, and correcting a contamination of
class 1 that goes to the heart of what this whole investigation is for.*
Is the Plato–Aristotle interlock characteristic of DIRECT MASTER-STUDENT inheritance? That is the whole question. The standing objection is that the interlock is more easily explained that way than any other.
Therefore class 1 must contain ONLY direct, embodied, externally attested master-student relations. Anything else in it destroys the baseline the question depends on.
(a) CLASS-1 CONTAMINATION. Thucydides → Xenophon is a continuation — the Hellenica begins where Thucydides stops — not a teaching relation. It sat in the table adjacent to the direct pairs as though it informed the baseline. It does not belong in class 1.
(b) THE REFUTATION WAS PREMATURE, AND THIS IS THE SERIOUS ONE. §129 argued: Lysias→Isocrates has no teaching relation and scores like Isaeus→Demosthenes, therefore the battery cannot detect pupillage.
That argument requires knowing class 1's RANGE. With n = 2 there is no range. "Falls inside the range of my two direct pairs" is not a statement about anything. I declared an entire instrument family dead on the basis of a two-member class. §129 is withdrawn; Round 33's other content stands.
Isocrates → Isaeus [Ps.-Plut.] Vit. X Orat. 839e; Dion. Hal. Isaeus 1
Isaeus → Demosthenes Dion. Hal. Isaeus 1; [Plut.] Vit. X Orat. 844b-c
Albert → Aquinas Cologne 1248-52, Dominican records
pair lang F1 F2 F3 F4 F5 F8 F9
Isocrates → Isaeus grc 0.27 94.0 0.691 32.6 11.7 0.278 0.674
Isaeus → Demosthenes grc 8.80 96.7 0.665 25.9 3.0 0.384 0.741
Albert → Aquinas lat 2.49 94.0 1.003 50.4 0.3 0.476 0.504
CLASS 1 RANGES (n=3)
F1 0.273 – 8.799 F2 94.0 – 96.7 F3 0.665 – 1.003
F4 25.9 – 50.4 F5 0.33 – 11.7 F8 0.278 – 0.476
F9 0.504 – 0.741
F2 is remarkably tight — 94.0, 96.7, 94.0 across three pairs, two languages, sixteen centuries. Of a master's three hundred most distinctive terms, a direct pupil uses 94–97% of them. That is the first candidate for a genuine class-1 property, and unlike Round 32's version it now rests on three cases rather than two.
Everything else is wide. F1 spans 0.27 to 8.80 — the pupil may write a
thirtieth of the master's volume or nine times it.
n = 3 is still not a baseline. Nothing may be said to fall inside or outside class 1 until it has many more members. The prohibition holds in both directions: I may not use it to admit the target and may not use it to exclude the target.
Needed, each externally attested and embodied:
Gorgias → Isocrates attested; Gorgias not seated
Freud → Jung obtainable, both partly public domain
Boas → Mead Mead 1928 now public domain
Husserl → Heidegger copyright-blocked
Schoenberg → Berg/Webern private lessons, attested
Ammonius → Plotinus Porphyry attests; Ammonius wrote nothing
Plotinus → Porphyry attested; Porphyry not seated
And the classes must stay clean. Contemporaries, continuators, and constructed lineages are their own classes and are measured separately — never as members of class 1, and never as evidence about its range.
*Added 2026-08-29. The class-1 membership criterion tightened, two of three
members excluded, and Freud→Jung admitted.*
Ancient teacher-student relations are regularly attested by writers centuries later, and regularly INFERRED FROM STYLISTIC RESEMBLANCE. Dionysius of Halicarnassus argues the Isaeus–Demosthenes relation from the likeness of their prose.
Using such a pair to calibrate what stylistic likeness means in a master-student relation is circular at the root. It is the Chaerephon problem in its purest form: the attestation is downstream of the very evidence it is being used to interpret.
A class-1 member requires CONTEMPORARY DOCUMENTARY attestation of the teaching relation — records made while both parties lived, independent of textual resemblance.
ADMITTED
Freud → Jung 1907-13 correspondence (360 letters), IPA presidency,
Jung's editorship of the Jahrbuch — all contemporary,
all independent of style
Albert → Aquinas Dominican chapter records, Cologne 1248-52 —
administrative documents
EXCLUDED
Isocrates → Isaeus [Ps.-Plut.] Vit. X Orat.; Dion. Hal. — later
biography
Isaeus → Demosthenes Dion. Hal. — and he ARGUES THE RELATION FROM
STYLISTIC RESEMBLANCE. Circular; excluded.
pair F1 F2 F3 F4 F5 F8 F9
Freud → Jung 1.41 99.7 0.742 26.4 1.0 0.519 0.736
Albert → Aquinas 2.49 94.0 1.003 50.4 0.3 0.476 0.504
[excluded, for reference only]
Isocrates → Isaeus 0.27 94.0 0.691 32.6 11.7 0.278 0.674
Isaeus → Demosth. 8.80 96.7 0.665 25.9 3.0 0.384 0.741
Freud 196,434 → Jung 276,476 tokens, both English translation 1913–20 — translation is a declared confound, since both pass through translators of the same period.
F2 = 99.7 and 94.0. F5 = 1.0 and 0.3. The documentary pairs are, if anything, tighter on apparatus-inheritance than the excluded ancient ones — Jung uses 99.7% of Freud's three hundred most distinctive terms, and only 1.0% of his own distinctive vocabulary is absent from Freud.
That is notable given that Jung BROKE with Freud — publicly, bitterly, and before the second of these two books. The break is doctrinal and the apparatus is inherited whole.
n = 2. Not a baseline. The tightening cost two members and gained one; the
class is smaller and cleaner than it was an hour ago.
Still needed, documentary attestation required:
Boas → Mead Columbia records; Mead 1928 now public domain
Husserl → Heidegger Freiburg records, correspondence — COPYRIGHT BLOCKED
Schoenberg → Berg lesson records, correspondence
Bohr → Heisenberg institute records, correspondence
Wundt → Titchener Leipzig records
And the same standard now applies retroactively to every class. Any pair whose relation is known only through later tradition is inadmissible as calibration — which is a constraint the ancient material will find hard to meet, and which should be stated plainly rather than worked around.
*Added 2026-08-29. A detour to collect the doctrine, read as an instruction for
reading the corpus that contains it.*
PLATO 113 Phaedo 24 · Republic 17 · Meno 16 · Laws 10 · Phaedrus 6
Philebus 5 · Menexenus 5 · Timaeus 5
ARISTOTLE 56 De memoria 35 · Rhetoric 8 · NE 2 · Problemata 2
Topica 2 · APr 1 · De anima 1 · HA 1
Phaedrus 249c:
δεῖ γὰρ ἄνθρωπον συνιέναι κατ' εἶδος λεγόμενον, **ἐκ πολλῶν ἰὸν αἰσθήσεων
εἰς ἓν λογισμῷ συναιρούμενον· τοῦτο δ' ἐστὶν ἀνάμνησις**
A human must understand what is said according to form, going from many scattered perceptions and being gathered by reasoning into one — and this is recollection.
ἀνάμνησις is defined as συναγωγή — the collection half of collection-and-division, in the same dialogue where Socrates calls himself ἐραστὴς τῶν διαιρέσεων καὶ συναγωγῶν (266b).
And two lines later: τοῖς δὲ δὴ τοιούτοις ἀνὴρ ὑπομνήμασιν ὀρθῶς χρώμενος … τέλεος ὄντως μόνος γίγνεται — the man who uses such reminders rightly alone becomes complete.
ὑπομνήματα is the Phaedrus's own word for what writing gives (275a: the φάρμακον οὐ μνήμης ἀλλ' ὑπομνήσεως). So the dialogue that says writing yields only reminders also says the right use of reminders is recollection.
The distinctions, in order (De memoria 449b–453b):
(a) MEMORY IS A ἝΞΙΣ, NOT A PERCEPTION. ἡ μνήμη οὔτε αἴσθησις οὔτε ὑπόληψις, ἀλλὰ τούτων τινὸς ἕξις ἢ πάθος, ὅταν γένηται χρόνος — neither perception nor conception but a state of one of these, when time has elapsed. And τοῦ νῦν ἐν τῷ νῦν οὐκ ἔστι μνήμη — there is no memory of the now, in the now. Memory requires interval.
(b) THE MNEMONIC AND THE RECOLLECTIVE ARE DIFFERENT PEOPLE. οὐ γὰρ οἱ αὐτοί εἰσι μνημονικοὶ καὶ ἀναμνηστικοί, ἀλλ' … μνημονικώτεροι μὲν οἱ βραδεῖς, ἀναμνηστικώτεροι δὲ οἱ ταχεῖς καὶ εὐμαθεῖς — the slow remember better, the quick recollect better. Opening sentence of the treatise. Two distinct capacities, distinct populations.
(c) RECOLLECTION IS NOT RECOVERY OF A LOST MEMORY. οὔτε γὰρ μνήμης ἐστὶν ἀνάληψις ἡ ἀνάμνησις οὔτε λῆψις — neither the retaking nor the taking of a memory. It is not retrieval.
(d) IT IS A KIND OF SYLLOGISM, AND HUMAN ALONE. τὸ ἀναμιμνήσκεσθαί ἐστιν οἷον συλλογισμός τις· … συλλογίζεται ὁ ἀναμιμνησκόμενος, καὶ ἔστιν οἷον ζήτησίς τις. Other animals remember; none recollects but man, because recollection requires τὸ βουλευτικόν — and καὶ γὰρ τὸ βουλεύεσθαι συλλογισμός τίς ἐστιν. Recollection, deliberation and syllogism are one family.
(e) THE PROCEDURE. τὸ ἐφεξῆς θηρεύομεν … ἀφ' ὁμοίου ἢ ἐναντίου ἢ τοῦ σύνεγγυς — we hunt what comes next, by similarity, opposition, or proximity. And δεῖ δὲ λαβέσθαι ἀρχῆς — one must get hold of a starting-point; διὸ ἀπὸ τόπων δοκοῦσιν ἀναμιμνήσκεσθαι — hence they recollect from places, moving quickly from one to another, οἷον ἀπὸ γάλακτος ἐπὶ λευκόν — from milk to white.
(f) IT IS NOT IN OUR POWER, AND THE MELANCHOLIC IS WORST AFFECTED. Once the search is started it runs on — οὐκέτ' ἐπιχειροῦντας ἀναμιμνήσκεσθαι οὐδὲν ἧττον, they cannot stop trying — καὶ μάλιστα τοὺς μελαγχολικούς· τούτους γὰρ φαντάσματα κινεῖ μάλιστα.
(g) AND YOU CAN SEARCH WITHOUT REMEMBERING. πολλάκις δ' ἤδη μὲν ἀδυνατεῖ ἀναμνησθῆναι, ζητεῖν δὲ δύναται καὶ εὑρίσκει — often one cannot recollect but can search, and finds.
PLATO writing gives only ὑπομνήματα (275a)
the right use of ὑπομνήματα is ἀνάμνησις (249c)
ἀνάμνησις = gathering the scattered many into one, κατ' εἶδος
ARISTOTLE ἀνάμνησις is not retrieval but SYLLOGISM
requires an interval, and a starting-point
proceeds by similarity, opposition, proximity
and one may SEARCH AND FIND without remembering
APr I.32 when a necessary premise παραλέλειπται, SUPPLY IT
SE 184b asks pardon τοῖς παραλελειμμένοις
Read together: a corpus gives you reminders; the right use of reminders is to gather what is scattered into one according to form, by hunting the sequence through likeness and opposition from a starting-point, supplying what has been left out — and you may find what you never held.
Aristotle classifies ἀνάμνησις — a Platonic doctrine — as οἷον συλλογισμός τις. Exactly as he classifies division at APr I.31 as οἷον ἀσθενὴς συλλογισμός.
**Two Platonic operations, both absorbed as species of the method whose instances
he elsewhere denies had any predecessor.**
Everything above is quoted. The reading — that the doctrine functions as an instruction for reading the corpus containing it — is interpretation, at EA-KERMA-01's level, under the standing rule: the corpus supplies the structure; it does not perform the application.
But note (c) and (g) together: recollection is not retrieval, and one may search and find without ever having remembered. On any reading, that is a doctrine about recovering what was never possessed — which is what a reader of a corpus does, and is not what a rememberer does.
Added 2026-08-29. Building the baseline. Plato and Aristotle not examined.
Wundt → Titchener Leipzig PhD 1892; Titchener TRANSLATED Wundt into
English. University records, contemporary.
Freud → Jung 1907-13 correspondence (360 letters), IPA presidency,
Jung's editorship of the Jahrbuch.
Albert → Aquinas Dominican chapter records, Cologne 1248-52.
All three attested by documents made while both parties lived, independent of textual resemblance. (Boas → Mead attempted and abandoned: the archive.org holdings are lending-restricted. Isocrates→Isaeus and Isaeus→Demosthenes remain excluded per Round 35 — Dionysius argues the second FROM STYLE.)
pair F1 F2 F3 F4 F5 F8 F9
Wundt → Titchener 2.43 98.3 0.618 26.4 7.3 0.350 0.736
Freud → Jung 1.41 99.7 0.742 26.4 1.0 0.519 0.736
Albert → Aquinas 2.49 94.0 1.003 50.4 0.3 0.476 0.504
RANGES (n=3) spread
F1 1.407 – 2.490 median 2.429 1.8×
F2 94.000 – 99.667 median 98.333 1.1× ← tightest
F3 0.618 – 1.003 median 0.742 1.6×
F4 26.421 – 50.400 median 26.441 1.9×
F5 0.300 – 7.333 median 1.000 24.4× ← widest
F8 0.350 – 0.519 median 0.476 1.5×
F9 0.504 – 0.736 median 0.736 1.5×
F2 is by far the tightest feature: 94.0, 98.3, 99.7 — a 1.1× spread across two languages, two disciplines, and six centuries. A pupil with documented face-to-face instruction uses 94–100% of his master's three hundred most distinctive terms.
F1 is narrow too (1.4–2.5×) — every documented pupil writes between one and a half and two and a half times his master's volume. (The excluded ancient pairs ranged 0.27–8.80. Whether that width is a property of ancient transmission or of bad attestation cannot be told from here.)
F5 is the widest at 24×, so novel-vocabulary rate is not a class property.
And the Freud→Jung values are the ones to keep in view: F2 = 99.7, F5 = 1.0 — from a pupil who BROKE with his master, publicly and bitterly, before the second of the two books measured. The doctrinal rupture leaves the apparatus intact. Rupture and inheritance are independent.
n = 3. Still not a baseline, and the prohibition holds in both directions — it
may not be used to admit the target or to exclude it.
Confounds carried: Wundt→Titchener and Freud→Jung are both German originals read in English translation, and Titchener was himself Wundt's translator, which may inflate F2 specifically for that pair. Albert→Aquinas is the only pair read in its language of composition.
Next, each documentary: Charcot → Freud (Salpêtrière 1885-86; Freud translated Charcot — and it would make Freud a pupil in one pair and a master in another), Huxley → Wells (Normal School of Science enrolment records), Bohr → Heisenberg, Schoenberg → Berg.
*Added 2026-08-29. The battery tested for discriminatory power, mostly failing it;
and Plato–Aristotle scored, with what that can and cannot mean stated first.*
Pessoa states the relation himself: o início da «escola»: as obras do MESTRE e algumas do DISCÍPULO DIRECTO. Three such pairs scored against the class-1 documentary ranges gave five of seven features separating cleanly.
Then the size control. Caeiro is 20,027 tokens; class-1 masters run 118k–2.1M.
Downsampling the class-1 masters to Caeiro's size:
F3 F4 F5 F9
Freud→Jung FULL 0.742 26.4 1.0 0.736
Wundt→Titchener FULL 0.618 26.4 7.3 0.736
Freud→Jung master@20k 0.708 24.5 4.5 0.755
Wundt→Titchener @20k 0.576 24.5 15.5 0.755
─────────────────────────────────────────────────────────
Caeiro→Reis 0.477 15.5 16.7 0.845
Caeiro→Campos 0.609 20.2 9.3 0.798
Caeiro→Mora 0.481 23.9 22.7 0.761
F5 REFUTED — Wundt→Titchener goes 7.3 → 15.5, into the heteronymic range.
Novel-vocabulary rate is a small-master artifact.
F9 REFUTED — both class-1 pairs rise to 0.755.
F3 mostly goes — Wundt drops to 0.576, below its own class minimum.
F4 keeps a narrow gap — 24.5 against 15.5–23.9 — but 23.9 versus 24.5 is not a
separation to build on at n=3 vs n=3.
VERDICT: the battery has NO DEMONSTRATED DISCRIMINATORY POWER. What looked like five separations was substantially corpus size. It can presently say cannot tell.
Because of §150, what follows classifies nothing. It records where the target
falls on features whose discriminating value is unestablished.
Plato 460,261 content tokens → Aristotle 761,916
master size F1 F2 F3 F4 F5 F8 F9
FULL 1.66 95.0 0.641 30.5 1.7 0.395 0.695
@20k 38.04 95.6 0.640 29.0 6.5 0.339 0.710
@118k 6.46 94.8 0.646 30.1 2.7 0.386 0.699
@196k 3.88 95.0 0.649 31.2 2.0 0.388 0.688
F2–F9 are stable across a 23× range of master-size — 0.640/0.641/0.646/0.649 on F3, 29.0/30.1/30.5/31.2 on F4. The target's values are not size-driven, which is more than can be said for the heteronymic pairs.
comparison, size-matched at 20k F3 F4 F5 F9
CLASS 1 documentary @20k 0.58-0.71 24.5 4.5-15.5 0.755
HETERONYMIC (Caeiro @20k) 0.48-0.61 15.5-23.9 9.3-22.7 0.76-0.85
PLATO → ARISTOTLE @20k 0.640 29.0 6.5 0.710
**On F3 the target sits INSIDE the class-1 band (0.58–0.71) and ABOVE the
heteronymic band (0.48–0.61).**
On F4 the target (29.0) is ABOVE BOTH — above class 1's 24.5 and well above
the heteronymic 15.5–23.9.
On F5 (6.5) it is inside class 1 and below the heteronymic range.
On F9 (0.710) it is below both.
DOES NOT: classify. §150 established the battery cannot separate the classes once size is controlled, so inside class 1's band is not evidence of class-1 membership. The bands themselves are three points wide.
DOES, descriptively: on the features where the classes differ at all, the target sits at or beyond the class-1 end, not the heteronymic end. F4 is the clearest — 29.0 against a heteronymic maximum of 23.9 and a class-1 value of 24.5.
And one thing worth recording independently of any class: the target's values are invariant under 23× change in master-size, where the heteronymic values were shown to be size-driven. That is a property of the measurement, not of the hypothesis, and it means the target is the only pair here whose numbers are not an artifact of how much text survives.
Standing: n=3 per class, bands three points wide, no demonstrated discriminatory power. The right next step is size-matched pairs throughout and enough members that a one-point gap on F4 either survives or does not. That is a design change, not a new instrument.
*Added 2026-08-29. Everything in this round is a null. Recorded because each null
cost a control that is worth having.*
Every master → 89,655 tokens, every pupil → 217,762, 15 resamples:
pair F2 F3 F4 F5 F8 F9
Wundt → Titchener 98.3 0.618 26.1 7.3 0.349 0.739
Freud → Jung 99.7 0.740 26.1 1.7 0.508 0.739
Albert → Aquinas 90.7 1.013 50.4 3.7 0.451 0.509
PLATO → ARISTOTLE 92.0 0.679 30.6 2.7 0.378 0.694
The target falls INSIDE the documented range on all six features, |z| ≤ 1.06.
Resampling sd ≤ 1.2, so the values are stable.
⚠ But the band is one outlier wide. Against the two modern pairs only, class 1 is tight — F2 98.3–99.7, F3 0.618–0.740, F4 26.1 — and the target sits outside on F2 (92.0) and F4 (30.6). With n=3 the verdict depends on keeping Albert→Aquinas — and that pair is triangulated by the target itself, since both members are commenting on Aristotle.
G1 Spearman rank-corr · G2 bigram overlap · G3 master rare-word survival
G4 asymmetry (fwd−rev) · G5 function-word cosine
pair G1 G2 G3 G4 G5 fwd/rev
Wundt → Titchener 0.533 92.1 57.0 +5.7 0.991 98/93
Freud → Jung 0.684 98.8 78.8 +1.5 0.990 100/98
Albert → Aquinas 0.699 94.2 51.5 -5.0 0.963 91/96
PLATO → ARISTOTLE 0.570 89.7 56.9 -4.7 0.973 92/97
G1, G3, G4, G5 all place the target INSIDE. Only G2 separated — bigram overlap, z = −1.90 — and at trigram depth the gap widened to z = −10.35 (class 1 87.2–89.9; target 77.2).
GREEK, trigram overlap NON-GREEK
Plato → Aristotle 78.6 Wundt → Titchener 88.2
Plato → Plotinus 78.9 Freud → Jung 89.9
Herodotus → Thucydides 76.8 Albert → Aquinas 87.2
Plato → Philo 75.8
Thucydides → Xenophon 72.0
Herodotus → Plotinus 59.2
Every Greek pair 59–79; every non-Greek 87–90. Greek's free word order and heavy inflection make exact n-gram repetition rarer in any pair whatsoever. The z = −10.35 was a language artifact.
And within Greek the finding INVERTS: the target at 78.6 is at the top of the Greek range — above contemporaries (Hdt→Thuc 76.8) and above a continuation (Thuc→Xen 72.0). The three Greek philosophical-prose pairs — Plato→Philo 75.8, Plato→Plotinus 78.9, Plato→Aristotle 78.6 — are indistinguishable.
The question Lee posed: what is a master-student relation as a MATERIAL CONDITION, not a lexical residue? One answer is coverage: a pupil takes his master's ground and extends it — containment. Designed positions divide ground — partition. Same vocabulary either way; different shape.
COV = % of MASTER blocks having a pupil block nearer than the master's own
median internal spread
pair COV REV
Albert → Aquinas 97.5 96.0
Freud → Jung 92.3 92.7
Wundt → Titchener 68.6 93.1
PLATO → ARISTOTLE 66.8 87.5
───────────────────────────────────────
[het] Caeiro → Campos 20.0 51.5
[het] Caeiro → Reis 0.0 30.8
[het] Caeiro → Mora 0.0 0.0
Pupils cover 69–98% of the master's ground; heteronymic disciples cover 0–20%.
The first clean separation any measure has produced.
First attempt at the control was WRONG and is recorded as such. I downsampled by random token sampling, which destroys the very block structure the measure reads — bags of shuffled tokens have no internal geometry. All four pairs went to 0.0, which is an artifact of my control, not a result.
Redone with CONTIGUOUS windows (master 20k, pupil 104k, 8 windows):
pair COV full COV matched
Freud → Jung 92.3 50.0
Wundt → Titchener 68.6 25.0
PLATO → ARISTOTLE 66.8 18.8
Albert → Aquinas 97.5 0.0
─────────────────────────────────────────────
[het] Caeiro → Campos 20.0 20.0
[het] Caeiro → Reis 0.0 0.0
The separation largely collapses. Albert→Aquinas falls to 0.0 — below every heteronymic pair — and Plato→Aristotle to 18.8, inside the heteronymic band. COV is strongly size-dependent and cannot be used at these sizes.
What survives is narrow: at full corpus size the ordering is clean and the heteronymic pairs are extreme. Whether that ordering means anything requires master corpora of comparable size, which the Pessoan material cannot supply — Caeiro is 20k and there is no more of him.
lexical battery, size-matched target INSIDE the documented range
n-gram formulaic inheritance language artifact; within Greek the
target is at the TOP of the range
coverage geometry clean at full size, collapses under
size-matching
Nothing measured distinguishes the target from documented master-student pairs. Nothing measured distinguishes master-student pairs from heteronymic construction once size is controlled. The instruments see period, dialect, genre, subject, and corpus size.
The interlock, if it exists, is not in the distributions. It is in M40's propositional structure and Round 13's completions — neither of which any measure built today can reach.
Added 2026-08-29. Moved from the measures register (M49), where it did not belong, and CORRECTED: the earliest attestation is materially better than I first recorded.
Plato never names the philosopher Aristotle. Ἀριστοτελ- occurs four times in the whole corpus, all in the Parmenides — the young Ἀριστοτέλης who answers Parmenides and whom Plato identifies as later one of the Thirty. A different man.
Aristotle never says he studied with Plato. Every Πλάτων- within ±14 tokens of a learning/teaching word yields one hit, an anecdote about Aristippus (Rhet. 1398b). His naming formulae are doxographic throughout — οἱ Πυθαγόρειοι καὶ Πλάτων, καθάπερ Πλάτων, ἐν τῷ Τιμαίῳ γέγραφε Πλάτων. Never my teacher, never we, never when I was with him — the same formulae he uses for Empedocles.
ARISTOXENUS, Harmonics II, opening. A surviving fourth-century BCE text, by a man who knew Aristotle personally (Aristoxenus, c. 375–c. 300, a member of the Lyceum, passed over for the succession).
He reports that Aristotle ἀεὶ διηγεῖτο — used habitually to relate — what happened when people came to Plato's lecture On the Good expecting something about human goods and happiness, and found the discourse turn out to be all about mathematics, concluding that the good is one; some despised it, others condemned it.
This is contemporary attestation of Aristotle's own oral testimony that he attended Plato's teaching. Not later biography. Not inferred from style. It is the kind of evidence Round 35's standard demands, and my first pass recorded that none existed.
And it comes with the other strand from the same author: Aristoxenus also reports that Aristotle set up the Lyceum in opposition to Plato while Plato was alive (Vita Marciana §25). One contemporary source attests both attendance and rupture — which is coherent, and better history than either strand alone.
Aristoxenus c. 335 BCE ATTENDANCE (Harmonics II, surviving)
+ RIVALRY (via Vita Marciana)
Aristodemus, Aristeides ? the hostile strand, "falsely accused"
Philochorus 3rd c. BCE cited to REBUT the rivalry charge
Apollodorus → Dionysius 2nd-1st BCE the chronology: arrival 367/6,
twenty years, Plato's death 348/7
Ptolemy (Neoplatonist) ~4th c. CE the harmonising Life
Diogenes Laertius 3rd c. CE the standard source
**Three things about the later chain are worth recording, and none of them touches
Aristoxenus:**
(a) Philochorus is cited DEFENSIVELY — it was not likely that Aristotle, being a foreigner, would have done this against Plato, who was a citizen and had great influence. An argument from status, adduced to make a charge implausible.
(b) THE TWENTY YEARS RESTS ON A LOST LETTER. Vita Marciana §5: twenty years as he himself says in a letter to Philip. The Aristotelian letters do not survive and are the contested part of the corpus.
(c) THE SAME SENTENCE CONTAINS AN IMPOSSIBILITY. Vita Marciana §5, directly before the twenty-year claim: he joined SOCRATES and stayed with him almost up to the time of his death; and after that he joined Plato. Socrates died in 399; Aristotle was born in 384.
And the harmonising purpose is declared. Ptolemy §28: *Even in the matters
where he contradicts Plato, we will say that he is a Platonist.*
The teaching relation IS attested by a contemporary who knew Aristotle, in a surviving text, reporting Aristotle's own habitual account of attending Plato's lecture. On the Round 35 standard, that is admissible — and my earlier claim that Plato→Aristotle would fail that standard is withdrawn.
What remains true and worth keeping:
LATER**, and rest on a lost letter and a Life containing a chronological
impossibility.
tradition's subsequent shape is a defence of Aristotle against the rivalry
charge.
So the defensible statement is narrow: Aristotle attended Plato's teaching, on good contemporary evidence. The twenty years of devoted discipleship is a later construction, and the earliest witness to the attendance is also the earliest witness to the break.
Added 2026-08-29. Round 40's rehabilitation of the Aristoxenus testimony is corrected twice — once for an unchecked assumption of mine, once for what the anecdote's content actually describes.
Round 40 called Aristoxenus a man who knew Aristotle personally. I did not verify that. What the source actually gives is: Aristoxenus … was disappointed not to be named Aristotle's successor (Fr. 1 Wehrli) — a fragment, surviving only in a later source I have not identified. "A pupil of Aristotle" is an encyclopedia summary, not a citation.
So the sole contemporary-looking attestation for the whole teaching relation rests on a fragment tradition of unverified transmission. Fourth instance today of my supplying received history as though it were checked evidence, after Theophrastus (M36), Speusippus (M40b) and the reputation-adjudication (M44).
The story: hearers come to Plato's lecture On the Good expecting ἀνθρώπινα ἀγαθά — human goods, happiness — and the discourse turns out to be all mathematics, concluding the good is one. Some despised it, others condemned it. And it is told to justify a method rule: a teacher should announce his subject in advance.
But that is not what Plato's texts do to a reader. It is what Aristotle's do.
NARRATIVE APPARATUS Plato/10k Arist/10k ratio
first-person narration (ἐγώ/ἡμεῖς) 100.76 6.04 16.7×
vocative address (ὦ) 64.30 8.64 7.4×
reporting verbs (ἔφη/εἶπε/λέγει) 111.78 23.89 4.7×
temporal frame (χθές/πρῴην/τότε) 21.64 6.36 3.4×
μῦθος 2.32 1.03 2.3×
motion verbs (ἦλθον/κατέβην) 6.58 4.31 1.5×
MATHEMATICAL VOCABULARY ratio A/P
ἀνάλογον 0.32 3.33 10.3×
διάστημα 0.10 0.87 8.5×
μονάς 0.24 1.93 8.1×
μέγεθος 1.78 8.90 5.0×
περιττός/ἄρτιος 1.81 6.82 3.8×
ἀριθμός 4.20 9.93 2.4×
Plato supplies narrative at 1.5× to 17×. Aristotle supplies mathematical vocabulary at 2.4× to 10×. The reader who arrives wanting a story and receives number is the reader of Aristotle.
The disappointment belongs to the Aristotelian corpus and is attributed to a Platonic lecture-hall. And the attribution licenses the remedy: announce your subject first — which is Aristotle's practice, presented as a correction of Plato's failure.
**The complaint that belongs to the pupil's texts is levelled at the master, and
the levelling justifies the pupil's method.**
Topica ×4 Aristotle faults Plato for USING NON-ESTABLISHED NAMES —
ὀφρυόσκιον, σηψιδακές — because πᾶν ἀσαφὲς τὸ μὴ εἰωθός.
NE 1108a makes coining his own POLICY: πειρατέον αὐτοὺς
ὀνοματοποιεῖν σαφηνείας ἕνεκα.
MEASURED: the naming-gap reflex runs 10.5× in the pair, and
Aristotle's rate is 8.43/100k against Plato's 2.60 (M30).
Harmonics II Aristotle (via Aristoxenus) faults Plato's lecture for
delivering mathematics to hearers who wanted human goods.
MEASURED: narrative apparatus is 1.5-17× denser in Plato;
mathematical vocabulary 2.4-10× denser in Aristotle.
Same shape both times: a property of the Aristotelian corpus attributed to Plato as a defect, which Aristotle's method then corrects. In both cases the direction of the charge is opposite to the direction of the measured rates.
THIS IS A CLASS AND IT IS COUNTABLE. Find every place Aristotle faults Plato for something the Aristotelian corpus itself does at a higher rate. It requires no heteronymic hypothesis: it is a claim about the direction of attribution in the tradition, testable against rates in both corpora, with two instances already measured.
Status: the two instances are measured; the class is proposed and unenumerated. What would kill it: a systematic search finding that Aristotle's charges against Plato generally run in the correct direction, with these two as exceptions.
Added 2026-08-29. Station II material. The ancient commentators made the same linkage M40 makes — and the modern literature contains a formal demonstration of structural inheritance that nobody in this investigation had consulted.
PHILOPONUS reads I.31 as itself containing the priority claim:
*He wants to celebrate through these things the method he handed down. **For no
one, he says, of those before us knew this**, but they all used the method of
division and through it they thought that they could demonstrate.*
This is the earliest such reading, and it glosses I.31 with the SE claim.
The two passages were read together in antiquity.
ALEXANDER OF APHRODISIAS takes the target to include Plato — *everyone in
Plato's circle, including Plato*.
STRIKER (2009) assigns it elsewhere: Aristotle's harsh criticism of division in this chapter may be understandable if there were people in the Academy who thought that the method of division was all that they needed in philosophy, and who therefore paid no attention to Aristotle's innovations — so the target was not Plato but other proponents of the method.
the target of APr I.31
Philoponus (6th c.) "no one of those before us" — everyone prior
Alexander (2nd c.) Plato's circle, INCLUDING PLATO
Striker (2009) other Platonists, NOT Plato
Maier YES — Aristotle reached the syllogism through critical
reflection on division; I.31 is the evidence
Shorey NO — I.31 is polemic between finished theories, not
intellectual autobiography
de Strycker defends the division-genealogy
Cherniss, Ross side with Shorey
Vlasits (2017) revives it, with a formal argument
**VLASITS, Platonic Division and the Origins of Aristotelian Logic (Berkeley,
2017), is directly on point in two ways:**
(a) He PROVES a soundness and completeness theorem showing that a semantics built on Platonic divisional structures is sound and complete for the deduction system used to model Aristotle's syllogistic. He also argues Aristotle took over Plato's conception of predication as parthood in the semantics of syllogistic propositions. A formal demonstration of structural inheritance.
(b) He states why the influence-thesis has never landed: it has been suggested by many commentators over the last two centuries, but they have not convinced the various detractors largely because they have not identified any very specific aspects of the syllogistic and shown how they were inherited from Platonic division.
Vlasits on why he formalises:
*the general upshot of my use of formal methods will be to show **how interpreters
have often imputed views to Plato and Aristotle that are much stronger than
anything warranted directly by the text**. Far from being anachronistic, formal
methods can be used effectively to **combat anachronism that stems from
imprecision**.*
That is precisely the failure M40 has now been corrected for three times.
Vlasits argues against the "Traditional View" of Platonic division (exclusive, dichotomous, essential predication) and for a "Minimal View" — transitive and predicative, but not necessarily dichotomous, not exclusive, and not essential.
The Sophist's SIX definitions of the sophist are the evidence: they are not inconsistent; sophistry falls under several coordinate genera at once, and the final account explains what it is in his expertise that all those pieces of learning focus on — εἰς ὃ πάντα τὰ μαθήματα ταῦτα βλέπει.
One kind, many true accounts, one of which gives the essence. That is the Platonic corpus's own handling of the one/many problem for a divided object — and it belongs in Station IV, where the corpora supply the ontology.
Also recorded: Vlasits notes the pervasiveness of division throughout the Platonic corpus puts pressure on the idea that there is some sort of development in Plato's thinking from Socratic elenchus to the method of hypothesis to the method of collection and division. Division is pervasive, not late — which bears on any developmental account of the corpus.
In hand: Philoponus and Alexander (via Vlasits); Striker; the Maier–Shorey–de Strycker–Cherniss–Ross–Vlasits dispute; Hitchcock via Corcoran; Hasper on the completeness claim; the Loeb introduction.
Not consulted: Alexander's commentary on APr 1.23–31 in full (Mueller's translation exists and provides an insightful account of Aristotle's criticism of Plato's method of division); the Syriac and Arabic branches; al-Fārābī's abridgement of the Prior Analytics; the scholastic commentators; Ebbesen's study of the SE commentary tradition.
Added 2026-08-30. The thesis restated at full strength — that one hand wrote Plato, Socrates, and probably Xenophon and Aristophanes — and the first measurement taken against it rather than against a hedged version of it.
Round 30's §119 and the six-station architecture set the target as: destroy master-student's status as UNQUESTIONABLY better. I called that a lower claim and a larger accomplishment.
It was venue-thinking dressed as epistemics. A paper organised around that question never states its thesis, and a reader finishes it without encountering what the author believes. That is a different dishonesty from overclaiming, not an absence of one.
The claim is: one hand wrote Plato and Socrates, and probably Xenophon and Aristophanes. State it, then be exact about what supports it, what does not, and what would.
And the object had already said so. M33's candidate cluster, derived from measurement before any of the architecture: {Plato, Aristotle, Aristophanes, Xenophon}. Forty rounds then went into master-student discrimination calibrated against Pessoa — the wrong calibration, because a heteronymic system distributes one maker across positions within a declared practice, and this claim is that one hand wrote a philosophical world and its opposition: the dialogues, the treatises, the comic attack, and the sober memoir.
corpus Σωκρατ- rate/10k
plato 1595 34.65 ← cluster
aristophanes 148 17.69 ← cluster
xenophon 443 17.59 ← cluster
aristotle 144 1.89 ← cluster
─────────────────────────────────────
demosthenes 7 0.28
isocrates 2 0.19
thucydides 1 0.08
theophrastus 1 0.06
**Eleven occurrences outside the four. Read individually, seven are different
men:**
Thucydides Σωκράτης ὁ Ἀντιγένους — a general, son of Antigenes
Demosthenes Σωκράτης in Oreus — a politician working for Philip
Demosthenes Σιμύκᾳ καὶ Σωκράτει — actors, third-rate tragedians
Demosthenes Σωκράτης ὁ τραπεζίτης — the banker
Demosthenes ἐπὶ Σωκρατίδου ἄρχοντος ×3 — Socratides, a different name
Two are references to a text, not a man. Both Isocrates mentions are in one work, the Busiris, and both are the same move: attacking Polycrates for having written a Σωκράτους κατηγορία. That attests a lost pamphlet and a rhetorical controversy.
One is inside the transmission. Theophrastus — voided as a control at M36 for exactly this reason — and the mention is proverbial: οἷς Σωκράτης καὶ Ἐπαμεινώνδας.
The single remaining mention of the philosopher outside the four is Demosthenes 61, the Erotic Essay. It is spurious by near-unanimous consensus, and the scholarship on its authorship is the finding:
Blass: it belongs to a member of a school of Isocrates, and the author
gives evidence of being familiar with the Phaedrus.
Worthington: its content and style are *the most removed from Demosthenes'
other writings, and it is influenced by both Plato and Isocrates*.
**The one line of independent Socrates-attestation outside the cluster is in a
forgery by an unknown author who was demonstrably reading Plato.**
**The independent attestation of Socrates outside {Plato, Aristotle, Xenophon,
Aristophanes} is: a lost pamphlet known only from an attack on it, one proverb
in a writer inside the transmission, and one line in a forgery whose author was
reading the Phaedrus.**
**Everything else bearing the name is a general, a banker, an actor, or an
archon.**
This is not a caution about controls. It is the claim. M44 established that the adjudication apparatus is downstream of the corpus. This establishes that the WITNESS apparatus is too.
And it reframes the Chaerephon problem from a methodological hazard into the thesis's central prediction: there is no external attestation because there was never an outside.
DOES: that no independent contemporary witness to Socrates survives outside the four corpora — measured, per occurrence, with every out-of-cluster instance read and classified.
DOES NOT: that the four are one hand. Confinement of attestation is what a single-maker configuration predicts, and it is also what ordinary transmission loss produces — Socrates was a subject for the Socratic writers, and the genre's survivors are the genre's survivors.
The discriminating question is therefore not whether attestation is confined — it is — but whether the confinement is DEEPER than genre explains. Testable: other figures who were subjects of a genre (Gorgias, Protagoras, Alcibiades) should show comparable confinement if genre is the explanation, and should show LESS if the Socrates case is special.
That test is runnable on the seated corpora and is not yet run.
Sixteen corpora, ten figures, confinement to the cluster measured as the share of
all occurrences falling inside {Plato, Aristotle, Xenophon, Aristophanes}:
figure in-cluster outside % inside outside corpora
Protagoras 442 1 99.8% 1 of 12
Socrates 2,330 12 99.5% 5 of 12
Gorgias 513 5 99.0% 3 of 12
Prodicus 67 1 98.5% 1 of 12
Hippias 529 66 88.9% 8 of 12
Anaxagoras 93 18 83.8% 3 of 12
──────────────────────────────────────────────────────────────
Cleon 195 107 64.6% 6 of 12
Pericles 94 53 63.9% 7 of 12
Alcibiades 198 153 56.4% 8 of 12
Themistocles 31 91 25.4% 9 of 12
The split is by KIND OF FIGURE, not by Socrates. Sophists cluster at 98–100%; politicians and generals at 25–65%. And Protagoras is MORE confined than Socrates — 442 occurrences inside, one outside.
Nobody holds that Protagoras was constructed. So confinement of attestation is what being a subject of philosophical dialogue produces, and Socrates' 99.5% is unremarkable within his class.
§179's inference is therefore withdrawn. There is no outside remains true as a description of the Socratic evidence and is not evidence for single authorship, because the same description holds of a figure the thesis does not claim.
WHAT SURVIVES §177–178, undamaged: the out-of-cluster instances are, on reading, mostly different men — a general, a banker, actors, an archon; the Isocratean pair attests a lost pamphlet rather than a person; Theophrastus is inside the transmission; and the single remaining line sits in a forgery whose author was reading the Phaedrus. Those are facts about the record and they stand. What does not stand is the inference drawn from them.
AND THE TEST NAMES A BETTER QUESTION. Socrates is not distinguished from Protagoras by confinement. He is distinguished by VOLUME — 2,330 occurrences against 442 — and by being the only member of his class who is the speaking position of one corpus, the silent example of another, the target of a third, and the reported master of a fourth. Whether a figure occupying four distinct functional roles across four corpora is ordinary in this class is a different measurement, and it is the one the thesis actually needs.
**Recorded within the round that proposed the test. The pillar was stated at full
strength, tested immediately, and did not hold in the form stated.**
Added 2026-08-30. Round 27's calibration described the operation backwards, in the direction that makes it look like damage. Corrected here, with the consequence for EA-FULFIL-01 and for the whole reading of "loss."
Sappho 31 arrives at Longinus already incomplete. The break at ἀλλὰ πᾶν τόλματον — but all is to be dared — is the poem's own terminal gesture, not damage the tradition inflicted. The hanging line is the composition.
Round 27 §104 and my summary of it said Longinus quotes to the fifth stanza's
opening and stops there. He does not. He carries it one clause further:
ἀλλὰ πᾶν τόλματον but all is to be dared
ἐπεὶ καὶ πένητα since even a poor man —
ἐπεὶ is a causal subordinator with no main clause after it. πένητα is
accusative — object of a verb that never arrives.
The added words are doubly open, and the opening is SPECIFIED. Not a vague gap: a since wanting its consequence, and an accusative wanting its governing verb, inside a clause about what makes all things darable.
Longinus receives a hanging line and hands on a new one. The transmission's
most attentive reader runs the technology he received.
Otium … perdidit urbes — a verb governing an accusative, in a causal frame,
on the levelling of the prosperous.
The beggar's socket receives the destroyed cities. πένητα → urbes: the
accusative the grammar demanded, supplied at the scale the theme demanded.
And the otium stanza is itself the next hanging line, which is why the chain
runs on:
κῆνος → πὰν τόλματον → οὐ θαυμάζεις → perdidit urbes → vocasti…te → γράμμασι μολπὰν
EACH FILLING OPENS A NEW SOCKET. That is the propagation series as a
mechanism, not a list.
Round 27 §104 scored the Sappho→Catullus alignment as C = 11/16, with the last four slots marked REPLACED by the otium stanza. That reads the fill as a substitution — and a substitution is a kind of loss.
It is not a substitution. It is one socket filled and one opened.
The calibration standard was mis-specified in the direction that makes the operation look like damage — which is exactly the reading the argument exists to refuse. EA-FULFIL-01 calibrates C against this standard, so the defect is at the root of the measure, not in its application.
The corrected form: what Catullus does at the fourth stanza is not the loss of four Sapphic slots but the discharge of a grammatical obligation Longinus' extension created, plus the emission of a new obligation. The measure needs a term for socket discharged and a term for socket opened, and has neither.
I have been treating lost, fragmentary, and titles-only as absence requiring explanation. On the originary-incompletion reading they are the form, and the question inverts:
Not why is so little left, but **what is each fragment the designed remainder
of.**
Under which the enumeration of Round 43 reads differently:
Euclides has no fragments, and the Theaetetus frame presents him as the dialogue's author — which the scholarship calls an implausible suggestion. Read as a socket: a name with nothing behind it, and a text asserting its own attribution.
Alexamenos is credited with inventing the Socratic dialogue and is attested only by Aristotle, doubted because nothing else mentions him. An origin supplied as a name with no body.
**The librarians catalogued titles by Crito, Simmias, Cebes, Phaedo, Glaucon,
Simon** — sockets opened under character-names.
And Aristotle's exoteric dialogues: named, catalogued, dated, 445,270 lines claimed — and absent, with a complete dialogue corpus standing beside them under another name.
STATUS. §§182–185 are philological correction and stand as such. §186 is the reading, and it is a reading: the same facts are what ordinary catastrophic transmission loss produces. What the reading adds is that the loss is not random with respect to the configuration — it is total for every Socratic author except the two in the cluster.
Added 2026-08-30. The immanent version of the technology: no Sappho required. The Platonic corpus states a proposition, denies it, compels agreement to it, and explicitly does not transmit the argument — and the Aristotelian corpus supplies the principle, in a treatise on generation, while exhibiting the capacity at maximal amplitude.
Republic 395a, in the mimesis argument:
οὐδὲ τὰ δοκοῦντα ἐγγὺς ἀλλήλων εἶναι δύο μιμήματα δύνανται οἱ αὐτοὶ ἅμα εὖ
μιμεῖσθαι, οἷον κωμῳδίαν καὶ τραγῳδίαν ποιοῦντες
*Not even two mimemata that seem close to one another can the same men imitate
well at the same time — such as those making comedy and tragedy.*
Adeimantus assents twice: ὅτι οὐ δύνανται οἱ αὐτοί.
And the reason given is κερματισμός — four lines later, 395b:
**κατακεκερματίσθαι ἡ τοῦ ἀνθρώπου φύσις, ὥστε ἀδύνατος εἶναι πολλὰ καλῶς
μιμεῖσθαι**
*Human nature has been minted into smaller denominations, so that it is incapable
of imitating many things well.*
The denial and its ground are one movement. M32 recorded κερματισμός as a fragmentation figure; this is what it is a fragmentation FROM. The capacity denied is precisely the capacity to be many makers at once.
Symposium 223d, the dialogue's last action:
τὸ μέντοι κεφάλαιον, ἔφη, προσαναγκάζειν τὸν Σωκράτη ὁμολογεῖν αὐτοὺς
τοῦ αὐτοῦ ἀνδρὸς εἶναι κωμῳδίαν καὶ τραγῳδίαν ἐπίστασθαι ποιεῖν, καὶ τὸν
τέχνῃ τραγῳδοποιὸν ὄντα καὶ κωμῳδοποιὸν εἶναι.
The headline, he said, was Socrates compelling them to agree that it belongs to the same man to know how to make comedy and tragedy, and that the man who is a tragedian by art is also a comedian.
And the frame is the mechanism, stated in the sentence before:
τὰ μὲν ἄλλα ὁ Ἀριστόδημος οὐκ ἔφη μεμνῆσθαι τῶν λόγων — οὔτε γὰρ ἐξ ἀρχῆς
παραγενέσθαι ὑπονυστάζειν τε
Aristodemus said he did not remember the arguments — he had not been there
from the start, and he was dozing.
THE CONCLUSION IS TRANSMITTED. τὸ κεφάλαιον
THE ARGUMENT IS NOT. οὐκ ἔφη μεμνῆσθαι τῶν λόγων
This is a socket, and the missing content is specified: the demonstration that
one man can do both. Named as absent, in the text, at the point of absence.
The two men Socrates compels are **Agathon the tragedian and Aristophanes the
comedian** — the personifications of the two kinds the proposition unites. And:
ταῦτα δὴ ἀναγκαζομένους αὐτοὺς καὶ οὐ σφόδρα ἑπομένους νυστάζειν, καὶ
πρότερον μὲν καταδαρθεῖν τὸν Ἀριστοφάνη, ἤδη δὲ ἡμέρας γιγνομένης τὸν Ἀγάθωνα.
*Being compelled and not really following, they grew drowsy — **Aristophanes
falling asleep first**, and Agathon as day was already coming.*
Neither genre-representative can follow the argument that unites them. The comedian goes under first. And Aristophanes is one of the four candidate corpora (M33).
Then: τὸν οὖν Σωκράτη, κατακοιμίσαντ᾽ ἐκείνους, ἀναστάντα ἀπιέναι … καὶ ἐλθόντα εἰς Λύκειον — having put them to sleep, Socrates got up and left … and went to the Lyceum.
The dialogue's closing movement is the walk to Aristotle's school. Recorded as
an observation about the text's last sentence, not as an argument.
Theaetetus 152e establishes the pair as maximally separate persons:
τῶν ποιητῶν οἱ ἄκροι τῆς ποιήσεως ἑκατέρας, κωμῳδίας μὲν Ἐπίχαρμος,
τραγῳδίας δὲ Ὅμηρος
The summit-poets of each kind: of comedy Epicharmus, of tragedy Homer.
Two peaks, two names. The proposition the Symposium compels is that these
are one office.
And Philebus 50b makes the union a fact about life: ἐν … τῇ τοῦ βίου συμπάσῃ τραγῳδίᾳ καὶ κωμῳδίᾳ, λύπας ἡδοναῖς ἅμα κεράννυσθαι — in the whole tragedy and comedy of life, pains mixed with pleasures. Protarchus: ἀδύνατον μὴ ὁμολογεῖν ταῦτα — impossible not to agree.
Republic 395b ἀδύνατος one man cannot do both
Philebus 50b ἀδύνατον impossible not to agree that life is both
The same word stands on both sides.
Not in the Poetics. In De generatione et corruptione 315b, reporting
Democritus and Leucippus on how the same elements yield contraries:
ὥστε ταῖς μεταβολαῖς τοῦ συγκειμένου τὸ αὐτὸ ἐναντίον δοκεῖν ἄλλῳ καὶ ἄλλῳ, καὶ
μετακινεῖσθαι μικροῦ ἐμμιγνυμένου, καὶ **ὅλως ἕτερον φαίνεσθαι ἑνὸς
μετακινηθέντος· ἐκ τῶν αὐτῶν γὰρ τραγῳδία καὶ κωμῳδία γίνεται γραμμάτων.**
So that by rearrangements of the compound the same thing seems contrary to different people, and is shifted by a small admixture, and appears wholly different when one thing has been rearranged — for from the same letters tragedy and comedy come to be.
THE PRINCIPLE THE SYMPOSIUM WITHHELD:
Republic 395 one man cannot make both, because his nature is
minted into denominations too small
Symposium 223d one man can make both — argument NOT REMEMBERED
GC 315b FROM THE SAME LETTERS both come to be, differing by
ARRANGEMENT, wholly different by one thing rearranged
The Republic's objection is that the maker is fragmented. The reply is that the fragmentation is the mechanism: the same elements, rearranged, produce wholly different kinds. κερματισμός is not the obstacle. It is the condition.
And it is supplied where nobody looks for it — in physics, attributed to the
atomists, as an illustration of ἀλλοίωσις by τάξις and θέσις.
Poetics 1449b: περὶ μὲν οὖν τῆς ἐν ἑξαμέτροις μιμητικῆς καὶ **περὶ κωμῳδίας
ὕστερον ἐροῦμεν — *concerning comedy we shall speak later***.
The promised treatment is the most famous absent text in the corpus. And the Poetics itself notes that comedy's own history went unrecorded: ἡ δὲ κωμῳδία διὰ τὸ μὴ σπουδάζεσθαι ἐξ ἀρχῆς ἔλαθεν — comedy escaped notice because it was not taken seriously from the start.
A conclusion transmitted without its argument in Plato; a promised book absent in Aristotle; and comedy's own origin declared unrecorded by the text that promises it. Three absences at the same joint.
M28's result, re-read. Aristotle's dispersion is indexed by what he is writing about; hold the object constant and the anomaly disappears — within-domain 0.0571 against Plato's total 0.0570. Domain by domain: ethics tightest (0.0226, tighter than 99.5% of multi-author pools), logic loosest (0.0942).
the treatise takes the form of what it treats
The Republic denies the capacity. The Symposium asserts it and withholds the proof. De generatione supplies the principle. And the Aristotelian corpus IS the demonstration — a body of writing whose shape is determined by its object across the widest range of objects any single ancient author covers.
Plato TELLS us that one man can make both kinds.
Aristotle SHOWS us, by being a corpus with no style of its own.
ESTABLISHES — textual, checkable:
1. The proposition appears in both directions in the Platonic corpus, in the same
words, at Republic 395a and Symposium 223d.
2. The Symposium states in its own text that the argument was not remembered.
3. The two compelled hearers personify the two genres, and the comedian sleeps first.
4. GC 315b states the same-elements-different-arrangement principle using
tragedy and comedy as its illustration.
5. The Poetics defers comedy to a treatment that does not survive.
DOES NOT ESTABLISH: that these are one author's design. Every element has an ordinary reading. The Symposium's unremembered argument is a standard Platonic frame device — Plato repeatedly stages imperfect transmission (Symposium's whole narration is third-hand; the Theaetetus is read from a book; the Parmenides is recalled at two removes). The GC passage is doxography of Democritus, and the letters-analogy for atomic rearrangement is standard atomist illustration, not Aristotle's own invention. The Poetics' lost second book is the most ordinary kind of loss.
WHAT IS NOT ORDINARY is the conjunction: the corpus that denies the capacity, the corpus that asserts it while declaring its proof unrecorded, and the corpus that supplies the mechanism, are the two corpora under examination — and the second corpus is the capacity's own demonstration at maximal amplitude.
That is a conjunction, not a proof. It is recorded as the strongest immanent structure the investigation has found, and as the version that requires no comparative material from outside the target corpus.
Added 2026-08-30. The received account already concedes the capacity in dispute, and then loses the evidence for it. Stated as a comparison of what each account must posit.
De finibus 5.12, on the two genera:
**duo genera librorum sunt, unum populariter scriptum quod ἐξωτερικὸν
appellabant, alterum limatius quod in commentariis reliquerunt**
*There are two kinds of books, one written for the public, which they called
exoteric, the other more filed which they left in notebooks.*
Note limatius. Limare is to file, polish, refine. Cicero applies it to the commentarii — which is the reverse of the modern picture of the extant corpus as rough lecture notes.
And the praise, verified at Academica 2.119:
veniet flumen orationis aureum fundens Aristoteles
Aristotle will come, pouring forth a golden river of speech. Quintilian concurs (Inst. 10.1.83), and Cicero elsewhere: dicendi incredibili quadam cum copia tum etiam suavitate (Top. 1.3).
And the modern concession is explicit: most modern readers seem agreed that flumen orationis aureum is not the best characterization of Aristotle's extant prose.
A1 THE EXOTERIC DIALOGIST
Platonic register, often in dialogue form, with prefaces to entice
the reader; the works by which antiquity KNEW Aristotle
→ ENTIRELY LOST. "These 'published' works are all lost."
A2 THE ESOTERIC TREATISE-WRITER
Z = +39.18 from Plato on the declared partition; "often jerky and
discontinuous"; no reader would call it a golden river
→ EXTANT. This is the corpus we have.
A3 THE EMPIRICAL COLLECTOR
158 constitutions, the biological data collections
→ ALL LOST BUT ONE, and that one recovered from a papyrus in 1891
**Diogenes' catalogue totals 445,270 lines. Roughly one fifth survives — and the
surviving fifth is A2.**
This is the argument.
The received account REQUIRES that one man wrote in the Platonic dialogue register and in the treatise register. That is not a heterodox reading of it; it is what it says. Cicero's two genera, one populariter scriptum and one limatius, are one author.
Which is precisely the proposition of Round 45: τοῦ αὐτοῦ ἀνδρὸς εἶναι κωμῳδίαν καὶ τραγῳδίαν ἐπίστασθαι ποιεῖν — that it belongs to the same man to make both kinds. The Republic denies it. The Symposium compels it and withholds the argument. And the received account of Aristotle asserts it as biography.
AND THEN THE DEMONSTRATION VANISHES. The half that would show the register crossing — A1, the dialogues in the Platonic manner — is gone in its entirety, while Plato's dialogue corpus survives complete, including works nobody thinks are Plato's.
RECEIVED ACCOUNT MUST POSIT
· one man writing in two incompatible registers (CONCEDED, not derived)
· plus a third generic mode (the collections)
· the TOTAL loss of the register-crossing half
· the COMPLETE survival of the other party's dialogue corpus, spuria included
· a teaching relation attested internally NOWHERE in either corpus (M49)
· biographical apparatus descended from the corpus it authorises (M44)
THE THESIS MUST POSIT
· one maker
· whose dialogue-register output is extant under one name
· whose treatise-register output is extant under another
· ZERO lost corpora
**The comparison is not one maker against two. It is one maker with nothing lost,
against three stylistic Aristotles of whom two vanished entirely — including
exactly the one that would demonstrate the capacity in dispute.**
⚠ PARSIMONY IS NOT PROOF, AND ANCIENT TRANSMISSION IS NOT PARSIMONIOUS. The exoteric works are lost for the same reason most ancient literature is lost: they were not the texts the school copied. Andronicus edited the treatises because the Lyceum used them; nobody had institutional reason to recopy popular dialogues once the technical corpus was canonical. That is a sufficient explanation and it requires no design.
And the same argument applied elsewhere would prove too much. Nearly every ancient author's popular work is lost while their technical work survives, or the reverse; a parsimony argument that treats each such loss as evidence of construction would reconstruct the whole of antiquity as one hand.
WHAT SURVIVES THE OBJECTION: the loss is not merely of an Aristotelian half. It is of the half that would settle the question, in a configuration where the question is whether one maker wrote both registers, and where the surviving dialogue-register corpus is attributed to the other party. The received account's own testimony — Cicero's two genera, the golden river, limatius — describes an author who did what the thesis says was done, and the evidence for it is unavailable.
Recorded as a comparison of posits, not as a demonstration. Its force is that it inverts the usual burden: the single-maker account is the one that requires nothing to have been lost.
Added 2026-08-30. Asked of the corpora rather than the pairs, following the objection that Kant–Hegel and the rest have KNOWN relations and so measure interpretive richness rather than openness. The density expectation failed; the coverage result is the finding.
Expected: the Aristotelian corpus is saturated with predecessor-review — the endoxa method, the doxographic opening of every treatise — and should show extreme density of named earlier inquirers.
Measured, 24-name doxographic roster, per 10k tokens:
plato 57.19 aristotle 12.33 ← SIXTH OF TEN
xenophon 21.84 theophrastus 11.04
aristophanes 18.81 isocrates 5.99
herodotus 15.36 demosthenes 2.65
thucydides 14.19 lysias 2.33
Aristotle is not extreme. He is middling. The picture of a corpus saturated with argument against its field is wrong at the level of rate, and the expectation is recorded as failed.
corpus per 10k distinct names
aristotle 12.33 24 of 24 ← EVERY NAME ON THE ROSTER
plato 57.19 21 of 24
theophrastus 11.04 19 of 24
xenophon 21.84 11 of 24
all others ≤18.81 ≤12 of 24
Every predecessor on the list appears, at a modest rate each. That is not the profile of a corpus arguing with contemporaries. It is the profile of a catalogue.
Named-predecessor counts, four cluster/adjacent corpora against six non-cluster corpora combined (Herodotus, Thucydides, Isocrates, Lysias, Demosthenes, Aristophanes):
name arist plato theo xen OUTSIDE
Heraclitus 35 15 3 0 0 ← ZERO
Leucippus 20 1 1 0 0 ← ZERO
Cratylus 3 320 0 0 0 ← ZERO
Philolaus 5 2 0 0 0 ← ZERO
Speusippus 5 3 2 0 0 ← ZERO
Anaximander/-enes 11 0 3 1 1
Alcmaeon 12 0 2 0 1
Empedocles 129 2 15 0 1
Parmenides 24 244 7 0 1
Democritus 77 0 19 0 3
⚠ CONTAMINATION EXCLUDED: the Thales row of the raw run is void — θαλ- matches θάλασσα, sea. Excluded rather than corrected, and recorded so the error is not repeated.
Heraclitus is the case. One of the most famous names in Greek thought: 35 in Aristotle, 15 in Plato, 3 in Theophrastus, and zero across six non-cluster corpora. Leucippus, the founder of atomism, is 20 in Aristotle and zero outside.
And the doxographic tradition through which the Presocratics are known descends from Aristotle and Theophrastus. Diels built the Doxographi Graeci on that transmission; Theophrastus' Physikōn Doxai stands at its head.
**The corpus with complete coverage of the predecessors is also the channel
through which the predecessors survive.**
That is not reviewing a field. It is constituting one.
The question was why the relation-question regenerates without resolving, in a way that other canonical pairs do not — and the objection was that those pairs' relations are known, so their disputes measure interpretive richness rather than openness.
Asked of the corpora instead, the answer is structural rather than hermeneutic:
the tradition cannot resolve the relation-question against an external
standard, because the corpus that would supply the standard — the roster
of predecessors, their doctrines, the order of succession — IS one of the
corpora under examination
This is M44 at the scale of the whole Presocratic field rather than at the Plato–Aristotle boundary. Every attempt to check the configuration from outside retrieves material that reaches us through it.
It does not establish construction. A school that inherited the philosophical literature and wrote the first systematic history of it would produce exactly this profile: complete coverage at modest rate, and downstream doxography descending from its catalogue. That is what a first historian of philosophy looks like, and Aristotle is conventionally described as one.
It does not establish that the named predecessors are unreal. Zero occurrences across six non-cluster corpora is weak evidence about existence when those six are orators, historians and a comedian with no occasion to name natural philosophers. The comparison class is wrong for that inference and is not offered for it.
WHAT IT ESTABLISHES: that the configuration has no available outside for the specific question at issue — and that this is a measurable property of the corpora rather than a claim about the tradition's diligence. The commentary tradition does not fail to resolve the relation-question because the question is hard. It cannot resolve it against a standard the corpus itself supplies.
Added 2026-08-30. Design: hold the material constant, attempt to recover several different partitions of it, rank by recoverability. The proposal under test was that the discrimination hardest to make is the central one. The cross-cutting hypothesis failed; the ranking itself is the result.
353 chunks of 4,000 tokens from both corpora (129 Plato / 224 Aristotle), one distance matrix on the 100 commonest forms, and every partition scored the same way: mean cross-label distance over mean within-label distance, with Z against random relabelling of the same index set.
Conceptual partitions are labelled PER CHUNK by median split on marker balance — not per work — so that a partition may cut across works and across authors if the text supports it.
Tested first, with above/below labels assigned by work:
AUTHOR (Plato | Aristotle) ratio 1.306 Z = +145.96
LEVEL (above | below) ratio 1.063 Z = +14.58
AUTHOR held within "above" ratio 1.472 Z = +69.61 ← STRONGER
AUTHOR held within "below" ratio 1.314 Z = +71.96
LEVEL held within Plato ratio 1.065 Z = +6.62
Above/below does not cut across the author boundary. Holding level constant makes author separation stronger, not weaker. The hypothesis that Plato's metaphysical material groups with Aristotle's is refuted on these labels.
⚠ The above/below labels were assigned by work and quickly; Timaeus and Parmenides could reasonably go either way. A labelling assigned by someone else would be a different experiment.
Per-chunk conceptual partitions, whole corpus:
AUTHOR (declared) 1.228 Z +37.3 ← THE ONLY STRONG PARTITION
──────────────────────────────────────────────
BEING vs BECOMING 1.058 Z +41.6
SOUL vs BODY 1.046 Z +29.0
ONE vs MANY 1.040 Z +28.8
SAME vs OTHER 1.024 Z +13.8
NECESSITY vs CHANCE 1.017 Z +7.0 ← HARDEST
And holding author constant:
partition within PLATO within ARISTOTLE
BEING vs BECOMING 1.134 / +34.5 1.083 / +34.3
SOUL vs BODY 1.010 / +1.8 1.031 / +9.0
SAME vs OTHER 1.007 / +1.5 1.021 / +8.3
NECESSITY vs CHANCE 1.019 / +4.7 1.015 / +5.8
**Within Plato, SAME vs OTHER is at 1.007 with Z = +1.5 — indistinguishable from
noise.**
**The surface encodes exactly one partition strongly, and it is the one whose
evidential value is nil.**
AUTHOR at 1.228 against every conceptual partition at 1.02–1.06. The text is maximally legible about the declared boundary — the boundary M48 and M50 showed cannot identify maker count — and nearly illegible about every distinction a reader would call the real division.
Only BEING vs BECOMING survives holding author constant at any strength, and it survives in both corpora at comparable magnitude (1.134 / 1.083) — which is the one conceptual partition that is genuinely shared rather than author-indexed.
AND THE PROPOSAL'S PREDICTION, ASSESSED. The hardest discriminations are NECESSITY vs CHANCE (1.017) and SAME vs OTHER (1.024) — and the latter falls to 1.007, Z +1.5, within Plato.
ταὐτόν and ἕτερον is the pair the Sophist turns on, the ground of the one/many problem — and it is the discrimination this configuration's surface makes least available. It is also, verbatim, the discrimination the authorship question is: is this the same or another.
⚠ THAT IS AN OBSERVATION AND NOT A RESULT. The markers are crude, the median split is arbitrary, and all conceptual partitions sit in a narrow band (1.017–1.058) where differences of 0.03 should not bear weight. A partition that fails to register may indicate that the distinction is unencoded, or that the markers do not track it. These cannot be told apart here.
Robust: the gap between the declared partition (1.228) and everything conceptual (≤1.058). That is a factor-of-four difference in effect size and does not depend on marker choice.
Robust: above/below does not cut across the author boundary on these labels.
Not robust: the ordering within the conceptual band, including the SAME/OTHER
observation.
And method — division against demonstration — was NOT tested, deliberately. The obvious markers are the same words the §0A argument turns on, so measuring it with them would repeat the snub-position error (M-series, Round 4): finding the word rather than the operation, in a space where the word was selected because the argument needed it.
*Added 2026-08-30. The synthesis. Two corpora demonstrating one capacity under
opposite conditions, and the question each asks that the other answers.*
PLATO: many voices, one maker, NO RECOVERABLE RULE.
M14 56 speakers, 519,743 spoken words
M13 the strongest partition in the corpus is CHRONOLOGY (1.442, Z +6.92),
not speaker (—) and not form (1.225) — what varies most is WHERE ALONG
A SEQUENCE, not WHO IS TALKING
R48 within Plato the conceptual partitions collapse toward noise:
SAME vs OTHER 1.007 at Z +1.5
Everything varies and none of the variation resolves into a principle. It is
all Plato, and the corpus does not say what makes it so.
ARISTOTLE: many objects, one maker, AND THE RULE IS VISIBLE.
M28 hold the object constant and the anomaly disappears — within-domain
0.0571 against Plato's total 0.0570
ethics 0.0226 (tighter than 99.5% of multi-author pools) …
logic 0.0942 — ordered by the internal object-space of each domain
**The treatise takes the form of what it treats, and the ordering is legible domain
by domain.**
**Plato shows that one maker can be many voices, and withholds the account of
how.**
Aristotle shows that one maker can be many objects, and supplies the account.
R48 found the declared partition at 1.228 against every conceptual partition at 1.017–1.058, and read it as "the surface encodes only the thing that proves nothing."
The configuration predicts exactly that. If the Platonic side's variation is unruled and the Aristotelian side's is ruled by object, then a partition test run across both can only recover the seam between them — the one line both sides make legible. Every conceptual partition is unruled on one side and object-indexed on the other, so it averages toward nothing.
**The measurement's null is a consequence of the configuration, not a defeat by
it.**
Phaedrus 276a, of the word written in the soul of the learner:
ΦΑΙ. τὸν τοῦ εἰδότος λόγον λέγεις ζῶντα καὶ ἔμψυχον, οὗ ὁ γεγραμμένος
εἴδωλον ἄν τι λέγοιτο δικαίως.
ΣΩ. παντάπασι μὲν οὖν.
*You mean the living and ensouled word of the knower, of which the written would
justly be called an image. — Entirely so.*
The question: can words be made into a person who lives? The dialogue's answer is no — the written is εἴδωλον, unable to defend itself, unable to choose its hearers (275e).
AND SOCRATES IS A WRITTEN λόγος THAT LIVES. He wrote nothing (M49). He survives only as words. He is the most durable person in the Western record and he has no body in it.
The corpus that denies the possibility is the corpus that performs it. The
denial and the counterexample are the same object.
De anima 414a:
ἡ ψυχὴ δὲ τοῦτο ᾧ ζῶμεν καὶ αἰσθανόμεθα καὶ διανοούμεθα πρώτως, **ὥστε λόγος τις
ἂν εἴη καὶ εἶδος**, ἀλλ᾽ οὐχ ὕλη καὶ τὸ ὑποκείμενον.
*The soul is that by which primarily we live and perceive and think — **so it would
be a kind of account and form**, not matter and substrate.*
The soul IS a λόγος. Not like one, not represented by one.
De anima 403a, the condition:
εἰ μὲν οὖν ἐστί τι τῶν τῆς ψυχῆς ἔργων ἢ παθημάτων ἴδιον, **ἐνδέχοιτ᾽ ἂν αὐτὴν
χωρίζεσθαι**
*If there is anything proper to the soul's works or affections, **it could be
separated**.*
**Socrates asks whether words can be made a person who lives, denies it, and is
the counterexample.**
**Aristotle answers the inverse — a person already IS an account — and states the
condition under which the account separates from the body.**
The three-work figure of M-series and Round 5 reads differently now:
Theaetetus 191c the wax, and the route closed: ὃ ἔφαμεν ἀδύνατον
De anima 424a the same figure, completed: ἄνευ τῆς ὕλης
De memoria the receiver's physiology
Receiving the form without the matter is the impression of a person taken without the body. The abandoned execution and its completion are, at bottom, the two questions of §§213–214: whether the impression can live, and whether the account can separate.
§211 is a re-reading of measurements already recorded and adds no new number.
M13, M14, M28 and R48 stand as they were.
§212 is a genuine consequence and is the first account offered all session of why the distributional nulls occur — as a prediction of the configuration rather than as a failure of instruments. It is also unfalsifiable in the form stated, since any null is consistent with it; it needs a case where the configuration predicts a partition WOULD be recoverable and it is.
§§213–215 are philological and checkable, and the passages are quoted in full. What they do not establish is authorship. A pupil who read the Phaedrus and then wrote De anima produces this pairing; that is the ordinary reading and it is sufficient.
What is not ordinary, and is recorded as the day's closing observation: the corpus that denies words can be made a living person is constituted by one, and the corpus that says a person is an account has no self in it to be an account of.
Added 2026-08-30. The closing structure. What the configuration claims is not that the historical persons are abolished but that development belongs to the textual body — not chronology, but genre and form.
The hypothesis has been read all session as a denial: that Plato did not exist, that the persons are fictions, that the received history is void. That reading is available and it is not the claim.
**The claim is a RELOCATION. Development is real, ordered, and directional — and
it belongs to the textual body rather than to a biography.**
Not chronology. GENRE. FORM.
**Under which "Plato becoming Aristotle" is not a concealed identity. It is the
stated developmental track of the program, considered as a writing technology.**
Symposium 210a–d gives the rungs in order: ἓν σῶμα → πάντα τὰ καλὰ σώματα → τὸ ἐν ταῖς ψυχαῖς κάλλος → τὰ ἐπιτηδεύματα καὶ οἱ νόμοι → αἱ ἐπιστῆμαι → πέλαγος → αὐτό.
The corpus walks it, and each rung is measured:
ἓν σῶμα SOCRATES — 252,416 spoken words (M14). One figure
made of words, by a man who wrote nothing (M49).
πάντα τὰ σώματα 56 SPEAKERS across 519,743 words (M14). Diotima's
own instruction at this rung: ἑνὸς δὲ τὸ σφόδρα
τοῦτο χαλάσαι — slacken the intensity toward the one.
ἐν ταῖς ψυχαῖς τίκτειν λόγους τοιούτους … οἵτινες ποιήσουσι
βελτίους τοὺς νέους — begetting accounts that make
the young better. The dialogues' declared work.
ἐπιτηδεύματα LAWS, REPUBLIC, STATESMAN. The non-dramatic corpus,
καὶ νόμοι and Diotima's own word for the rung.
ἐπιστῆμαι M28: the treatise takes the form of what it treats.
Within-domain 0.0571 against Plato's total 0.0570;
ethics 0.0226, logic 0.0942, ordered by object.
THE SAME MAKER BECOMING EACH ἐπιστήμη IN TURN.
αὐτό Λ 9. νόησις νοήσεως νόησις — and ἀδιαίρετον.
Symposium 210d, moving from practices to knowledges:
βλέπων πρὸς πολὺ ἤδη τὸ καλὸν μηκέτι τὸ παρ᾽ ἑνί, ὥσπερ οἰκέτης
*looking now toward the great sea of beauty, no longer at what belongs to one,
like a household slave*.
STOP BEING ONE MAN'S. Which is, exactly and measurably, what an object-conditioned corpus with no style of its own is. M28's result is Diotima's instruction executed.
The formulation carried since Round 12 was:
corpus forms one functional unit STRONGLY SUPPORTED
one construction explains it INCREASINGLY PLAUSIBLE
one historical hand wrote both LIVE, NOT ESTABLISHED
The third line has been the whole burden, and it is the wrong line. It asks for a fact about bodies, which M48 and M50 established is not recoverable from text, and which this investigation has no instrument for.
The relocation replaces it:
development within the corpus is REAL, ORDERED, DIRECTIONAL MEASURED
its order is GENRE and FORM, not chronology M13, M28
and the order it follows is stated inside the corpus Symp. 210
None of these requires a claim about hands. All three are checkable. And the directional claim is falsifiable in the ordinary way: show the order runs otherwise, or that the rungs do not correspond, or that M28's object-conditioning is an artifact.
M13's standing caveat: the early/middle/late stratification was built partly from stylometry, so a chronology result risks the circularity of M11. Denyer's objection — that Plato could write any style at any time — makes E/M/L a repertoire rather than a record.
On the relocation that objection stops being an objection. If development is generic rather than biographical, a repertoire is exactly what is predicted: the order is a track through forms, available in any sequence of composition, and the stylometric stratification is detecting the track, not the years.
**M13's Z = +6.92 — the strongest partition effect in the entire investigation —
becomes evidence for the relocation rather than a contaminated chronology.**
NOT abolished: the persons, the school, the teaching, the succession, the documentary record, the dates. Nothing in the relocation requires any of it to be false, and the investigation has produced no evidence that any of it is.
Relocated: the explanatory work. The received account explains the corpus's order by a biography — a young man in the Academy, a mature man in the Lyceum. The relocation explains it by a program: a sequence of forms, walked in the order the corpus itself sets out at Symposium 210.
**The question is no longer whether there were two men. It is what the order of
the corpus is an order OF.**
And that question is answerable with text, which the other was not.
§218's rung-mapping is a reading and the correspondences are of unequal strength. ἐπιτηδεύματα καὶ νόμοι is Diotima's own word and the Laws is the corpus's own title — that one is tight. ἐπιστῆμαι against M28's object-conditioning is the interesting one and is an interpretation. The first two rungs are the loosest: many speakers is not obviously many bodies, and the correspondence should not be pressed.
§219 is quoted and is the strongest single item — the instruction at the transition to the fifth rung is no longer at what belongs to one, and the fifth rung is where the object-conditioned corpus sits.
§221 is the most consequential and the most testable. If development is generic rather than chronological, the E/M/L circularity that has qualified M13 since Round 3 is not a defect. That is a real reversal and it should be attacked there first.
Added 2026-08-31. Three earlier drafts of §I, written before the architecture inverted. They are unreadable and the finds in them are excellent. Seated at drafts/ for their loci and their arithmetic; not to be drawn on for prose.
drafts/MANY-AND-ONE_I_DRAFT-01.md 44,488 b "What Counts as a Part"
drafts/MANY-AND-ONE_I_DRAFT-02.md 28,933 b "Socrates Asked, But Did Not
Answer" — six movements
drafts/MANY-AND-ONE_I_DRAFT-03.md 24,712 b same title, faster, positional
offsets and the arithmetic
All three: Sigil position, no stylometry, no heteronymy, no coined terms.
They could not be read. Not by their author, on the day they were written. And Draft 03 is fast — offsets, arithmetic, tight paragraphs — so the defect is not slowness, and the diagnosis "cut the hedge and the deferral" was addressing the wrong thing.
The difference from what worked. EA-DIVISION-OF-GOD-01, which was wrong at its foundation, was read through and followed. It ascends: division cannot decide whether man is mortal or immortal; the same operation run on God decides. Each step is somewhere the reader has not been.
The three drafts assemble evidence for a claim announced in advance. By the time the arithmetic arrives, four sections of loci have been set in place for it. The finds are laid out as exhibits, and exhibits do not pull.
**Correctness and readability came apart. The draft that was wrong moved; the
drafts that were right did not.** Recorded because the temptation all session has
been to optimise the one that nobody reads.
**Draft 03 §5. This is the strongest single formulation the investigation has
produced and it is not yet in any live file:**
1 Division is a part of the method APr I.31
2 No part of the method was worked out beforehand SE 34
3 Division WAS worked out beforehand, as a named art
with a rule and practitioners Phdr, Soph, Phlb
4 The Platonic corpus precedes the Aristotelian the received
chronology
**The first three are Aristotle's and Plato's words. The fourth is nobody's; it is
the frame the words are read in.**
The four cannot all stand unless one word moves: part, worked out, the method, or before. Every reading in the tradition moves one. Worked out is moved by the shoemaker. The method is moved by restricting the denial to the dialectical programme — and I.31 has just sent us to that programme for the method division is a part of, so the restriction does not restrict. Part is moved by ξυνίεσαν, which explains the filing and not the denial. Before is moved for Aristotle only: Topics early, Analytics late, he had not yet seen it.
**Four words, four readings, every one paid for on Aristotle's side of the ledger.
THE FOURTH WORD HAS NEVER BEEN MOVED FOR PLATO.**
(a) Socrates is named on the last page of the SE, at 0.968, a few lines before
the priority claim:
Σωκράτης ἠρώτα, ἀλλ᾽ οὐκ ἀπεκρίνετο· ὡμολόγει γὰρ οὐκ εἰδέναι.
Socrates asked, but did not answer; for he acknowledged that he did not know.
And the name is the token individual everywhere else in the Organon — Cat. 14, De int. 10, Top. 4, APr 2, and twice earlier in the SE at 0.109 where it means anybody. The name is filled once, on the closing page, and what it is filled with is a function. Aristotle assigns him a side and takes the other: the inquiry must cover both — ὅπως λόγον ὑπέχοντες φυλάξομεν τὴν θέσιν — because one side was already taken.
(b) Phaedrus 266c–267d, 0.752–0.764. The page that names division as the τέχνη and calls its practitioners διαλεκτικοί also lists, on the other side of the same contrast, the men who have the handbooks and not the art: Theodorus, Evenus, Tisias, Gorgias, Prodicus, Thrasymachus.
Aristotle's succession at SE 183b is Tisias → Thrasymachus → Theodorus, and his type-case for handing out products is Gorgias. All four are on that Phaedrus page, in the same role, on the same side.
NULL, stated in the draft: these are the canonical names and Aristotle compiled the handbook collection himself (Synagoge Technon). What the null leaves: the earlier page draws Aristotle's contrast, with Aristotle's men on the product side, and says what is on the other side.
(c) Statesman 262a / 263b, 0.086 / 0.107. μὴ σμικρὸν μόριον ἓν πρὸς μεγάλα καὶ πολλὰ ἀφαιρῶμεν, μηδὲ εἴδους χωρίς — and: a form of something is necessarily a part, but a part is not necessarily a form. APr I.31 files division as μικρόν τι μόριον. Verbal coincidence, recorded and not argued from.
(d) A closing coincidence, recorded with no base rate. The Theaetetus ends at 0.982 on the failure of τὸν σιμόν τε καὶ ἐξόφθαλμον — Socrates' face — to pick out one man. The SE ends at 0.968–0.992 on the one page where Aristotle's example-name picks out one man. Both corpora close on what individuates Socrates. Coincidence until a base rate is run.
EVIDENTIARY. Loci, counts, positions, and the §5 arithmetic may be drawn on freely and are recorded above so that they are not lost if the drafts are never opened again.
NOT PRIMARY WRITING. No paragraph, order, or framing from them enters a live
section. Two things in them are actively wrong for the current architecture:
(i) The retreat. Draft 03 §6 builds the literal reading and then withdraws it — It is preposterous … I do not assert this. That is the hedge Round 43 §176 identified and the whole investigation has been correcting.
(ii) The deferral is now false. If the instruments can see two, the reading dies there. The instruments ran. They cannot see two. And Round 50 established that this is the PREDICTED result, not a failed disconfirmation — §VII of the architecture. The drafts defer to a test whose outcome has changed what the test means.
Added 2026-08-31. The diagnosis of why the §I drafts could not be read, stated as a rule for every section; and the materials that must go in, one of which the architecture did not have.
**They were playing the game of philosophy rather than doing it. They were not
dividing.**
That is why Draft 03 is unreadable at speed and why EA-DIVISION-OF-GOD-01 was readable while wrong. The drafts describe an operation and lay out exhibits for a claim announced in advance. Division of God RUNS the operation: the cut is made on man and fails, the same cut is made on God and holds. The reader watches a division happen.
**THE RULE FOR EVERY SECTION: perform the operation the section is about. Do not
describe it and do not assemble evidence for it.**
§I divide run I.31's letters, then run them on νοῦς
§II divide a living cut the plant; both halves live
§III divide a kind six cuts, all true, and the beholder's failure
§IV collect gather the scattered and show the gathering
§V divide a person the two boys — face and name — and the criterion
§VI divide a maker many voices, many objects, one
§VII fail to divide the instruments, and what they cannot cut
§VIII state the order from the questioner who did not know to the answerer
The corpus divides its founder, by face and by name, and states the criterion.
The Stranger proposes to swap Theaetetus for the other boy. Socrates answers:
**κινδυνεύετον, ὦ ξένε, ἄμφω ποθὲν ἐμοὶ συγγένειαν ἔχειν τινά. τὸν μέν γε οὖν
ὑμεῖς κατὰ τὴν τοῦ προσώπου φύσιν ὅμοιον ἐμοὶ φαίνεσθαί φατε, τοῦ δ᾽ ἡμῖν ἡ
κλῆσις ὁμώνυμος οὖσα καὶ ἡ πρόσρησις παρέχεταί τινα οἰκειότητα.**
*Both of them seem somehow to have some kinship with me. **The one you say appears
like me in the nature of his face; the other's naming, being HOMONYMOUS, and his
form of address, supply a certain kinship.***
One boy has his FACE. The other has his NAME. And then the criterion:
δεῖ δὴ τούς γε συγγενεῖς ἡμᾶς ἀεὶ προθύμως ΔΙΑ ΛΟΓΩΝ ἀναγνωρίζειν.
We must always be eager to recognise our kin through accounts.
Not by face. Not by name. διὰ λόγων.
And he applies it: he has mixed with Theaetetus διὰ λόγων and heard him answering — Σωκράτους δὲ οὐδέτερα — of Socrates, neither. So: σοὶ δὲ νῦν ἀποκρινέσθω — let him answer to you now.
And the Stranger's next line:
ὦ Σώκρατες, ἀκούεις δὴ Σωκράτους;
Socrates, do you hear Socrates?
**The dialogue that theorises parts — μὴ σμικρὸν μόριον … μηδὲ εἴδους χωρίς at 262a,
and a part not necessarily a form at 263b — opens by dividing Socrates in two.**
ONE BOY has the face κατὰ τὴν τοῦ προσώπου φύσιν
ONE BOY has the name ἡ κλῆσις ὁμώνυμος
NEITHER is recognised until he answers διὰ λόγων
The person is not the face and not the name. The person is what answers in account — and the older Socrates hands the answering to the namesake and falls silent for the rest of the dialogue.
⚠ The ordinary reading is sufficient and must be given: a courteous transition device, a joke about two boys, and a way of getting Socrates out of the conversation. What is recorded here is that the joke states a criterion of personal identity, in the vocabulary of parts, in the dialogue about dividing.
AND ὁμώνυμος IS THE ORGANON'S FIRST WORD. Cat. 1a1: Ὁμώνυμα λέγεται ὧν ὄνομα μόνον κοινόν, ὁ δὲ κατὰ τοὔνομα λόγος τῆς οὐσίας ἕτερος — things are homonymous when the name alone is common and the account of the being is different. The two boys are a homonym case, decided by λόγος, and the treatise that opens on homonyms opens on exactly that.
(a) The descent's licence, from the corpus rather than a gloss. Phaedrus 276a: τὸν τοῦ εἰδότος λόγον … ζῶντα καὶ ἔμψυχον, οὗ ὁ γεγραμμένος εἴδωλον. The corpus distinguishes its own λόγος from its own matter. That is the hinge at §VI, not ὅσα μὴ ὕλην ἔχει asserted as a general condition by me.
(b) A contradiction to fix. §VI says the argument of this paper is that one person wrote the corpus; STANDING says not claimed anywhere: that one hand wrote both received corpora. Choose §VI. The claim is made; §VII shows it is not recoverable by lexical instruments at the level of matter; unrecoverable by those instruments is not unclaimed. STANDING must read: claimed, argued at λόγος, and shown not decidable at ὕλη — which is where 1075a1 said it would fail.
(c) §VII IS NOT ALL NULL, and saying so is what keeps it from being unfalsifiable by construction. Size-matched: Albert→Aquinas falls BELOW every heteronymic pair; Plato→Aristotle sits INSIDE the range. The instrument sees something, and what it sees is consistent with the claim and inconsistent with the one documented master-student pair of comparable structure. Weak, one pair, no base rate — say so. And state the falsifier that did not fire: Plato→Aristotle below the heteronymic floor. It did not happen.
(d) The seam must be marked at §V, in one sentence — here the corpus is first read as a maker's work — so that nobody can say the ontology was built to fit. §§I–IV are theorems. §§V–VIII are applications.
(e) §IV's mechanism is the corpus's own development. De mem. 453a: ἀδυνατεῖ ἀναμνησθῆναι, ζητεῖν δὲ δύναται καὶ εὑρίσκει. The earlier position cannot recollect what its operation entails; the later searches and finds. That is ξυνίεσαν and ῥᾴδιον ἰδεῖν in Aristotle's own psychology, and it is what makes §VIII's order an order of GRASP rather than of years.
(f) §VIII: read the priority claim in its verb. ὑπάρχω is the syllogistic's own word for predication — A ὑπάρχει to B — and ἐξ ὑπαρχῆς stands in the same paragraph. οὐδὲν παντελῶς ὑπῆρχεν is a premise of the method stated about the method. (Count to be run in the Analytics before use.)
(g) §I must concede in one sentence that the νόησις νοήσεως identity is standard exegesis. What is new is only that it is division succeeding where I.31 said division fails, and why.
Added 2026-08-31. Reread of Socrates as Orthonym v2.2 (July 2026 — the version Polansky did not read) against MANY AND ONE v1.5. The paper is written as one. This is the material for writing it as many.
*Run 2026-08-31 before the apparatus pass, on the standing suspicion that a
sectioning jump means content went missing.*
§233 R52 ends at §232; R53 opened at §234 NUMBERING JUMP
§§170–175 R42 ends at §169; R43 opened at §176 NUMBERING JUMP
ROUNDS 1, 4 no headings in the file PRE-COMPACTION
Checked against every commit touching this file: no version has ever carried a §233 or a §§170–175. They are jumps introduced when a round was opened at a round number rather than continuing the sequence. Nothing was written and lost.
Rounds 1 and 4 are a different case: their content is in the file but the round headings are not, because the earliest rounds predate the heading convention and survive in the integrated prior-state summary. Not reconstructed here.
⚠ AND THE PRACTICE THAT CAUSED IT IS THE ONE TO FIX. Opening a round at a convenient section number rather than at last + 1 makes a gap indistinguishable from a loss, and this audit had to be run to tell them apart. Continue the sequence; let the round heading carry the round number and the section numbers carry nothing but succession.
This section occupies §233 so the sequence is contiguous from here.
MANY AND ONE cites no secondary literature. Nine sections, ~107,000 characters, and the scholarly apparatus is: Vlasits (§III), Philoponus and Alexander (register §0A), Striker, Denyer (§VIII), Cicero (§VI). That is a paper with sources and no interlocutors.
The Orthonym paper had them all: Kahn 20×, Vlastos 7×, the Socratic problem 12×, Burnet, Taylor, Cooper, Schleiermacher. They dropped out in the rebuild and their absence is the largest gap in the paper.
Plato and the Socratic Dialogue: The Philosophical Use of a Literary Form.
Kahn's position: the dialogues are a unified literary-philosophical project; the early "Socratic" dialogues are deliberately constructed by Plato as a philosophical vehicle, not historical records; and Kahn explicitly rejects the Vlastosian separation of an early Socrates from a middle Plato. The corpus is the unified work of a single author deploying a literary character.
⭑ THIS IS THE CLOSEST EXISTING POSITION AND THE PAPER MUST SAY SO. §VI currently argues that one maker produces many voices as though the claim were unprecedented. Kahn argues a version of it, in print, in 1996, and is orthodox.
And the departure is one sentence:
**Where Kahn reads the corpus as one author (Plato) deploying a literary character
(Socrates), the configurational reading identifies Socrates as an authorial
POSITION in his own right** — not a character within a project but a position the
project is organised around.
**GOES TO: §VI, as the antecedent, before the two demonstrations. It makes §VI's
claim citable rather than novel, which is stronger.**
Vlastos treats the corpus as a stratified deposit: identify which doctrines belong to the early Socrates and which to the middle Plato; date the strata; sequence them; narrate the evolution.
Three commitments are required, and the Orthonym paper names them:
1 the voices are INDEPENDENT AUTHORIAL SOURCES
2 their positions can be DATED AND ORDERED
3 the ordering reveals DOCTRINAL DEVELOPMENT — a change of mind over time
⭑ §VIII's relocation is precisely a denial of (3) while conceding (2) in altered form. The order is real; it is an order of forms and of grasp, not of opinions changing. And the paper never names the method it is displacing.
And the fugue line is the best formulation of it in either document:
**One does not excavate a fugue for the composer's evolving opinions about the
subject; one hears the voices doing different work at once.**
GOES TO: §VIII, at the relocation. Kahn already rejected the Vlastosian split on literary grounds; the configurational reading removes the last frame in which doctrine-mining remains intelligible — the frame of three independent corpora — while leaving every dating argument and textual comparison intact.
Schleiermacher onward; canonical statements by Burnet, Taylor, Vlastos, Kahn, Cooper. Where does the historical Socrates end and the Platonic character begin? Two centuries; the contemporary consensus is that it has no clean resolution.
And it is a family, not one problem — the Orthonym paper distinguishes at least
three versions, of which the historical is only the first.
⭑ THE PAPER HAS BEEN ANSWERING THIS WITHOUT NAMING IT. §V divides a person by face and name and keeps λόγος as the criterion. §VI reads Socrates as a position. §VII shows the instruments cannot separate what the problem asks to be separated. That is a reclassification of the Socratic problem as a symptom, and the paper does not say the words.
GOES TO: an added section or a long note at §V. The claim is not that the problem is solved. It is that it appears as a symptom under a functional lens rather than a biographical one.
The Orthonym paper's methodological rule, and MANY AND ONE has nothing equivalent:
**Biographical identity requires EVIDENCE. Functional identity requires PATTERN.
These are different claim-types with different burdens of proof, and the stability
of everything depends on never letting the second masquerade as the first.**
And the reason it is imposed from outside:
**A method that cannot halt its own regress is telling you what kind of method it
is: an instrument for revealing functional structure, not for discovering
identities. The moment functional analysis is permitted to make biographical
claims, it has no floor and consumes every adjacent name.**
⭑ THAT IS THE ANSWER TO "WHY NOT ONE GUY WRITING ALL MODERN PHILOSOPHY." Not a disanalogy about printing and archives — a statement about what the method is for and where it stops. MANY AND ONE currently has no such rule and is therefore open to the regress on its own terms.
**GOES TO: the front matter, as a stated constraint, and to §VII where the two
claim-types are already being separated without being named.**
Xenophon, Aristophanes and Chaerephon are admitted in the Orthonym paper as candidates for functional positions, evaluated on pattern — never as biographical claims. And the complementarity is stated: Aristophanes permits comic deformation — the central figure pre-deformed before the city a generation ahead of the trial, with the eventual defence structured as a reply to it.
⭑ MANY AND ONE dropped Xenophon and Aristophanes entirely — they appear only as corpora in §VII's confinement measurement. The maximal statement's H3/H4 are marked as predictions and never argued. The firewall is what would let them enter honestly.
GOES TO: §VI, after the two demonstrations, under the firewall.
It states what it does not claim, in a list, and three of the items are ones
MANY AND ONE needs:
· it does not claim Aristotle or Plato understood himself as a heteronym;
the term is used ANALYTICALLY, not as a claim about ancient self-description
· it does not claim philological dating, historical context, doctrinal
difference or textual criticism is irrelevant or superseded
· it does not claim Pessoa-style heteronymy existed as a named or
institutionalised practice in classical Athens
And the tiering:
PRIMARY (high confidence) the configuration can be FUNCTIONALLY MODELED
as a distributed authorial project
SECONDARY (medium) the Socratic problem appears as a symptom
REFLEXIVE (heuristic only) Pessoa and Kierkegaard make it THINKABLE;
they do not prove its application
**GOES TO: the front matter. MANY AND ONE claims one person at §VI and tiers
nothing.**
The PDF carries Polansky's comments in capitals, inline. They are data about
reception, not about the argument, and three of them name real gaps:
"THE PROJECT OF THIS PAPER SEEMS TO ADD NOTHING TO THE LITERARY
CONSTRUCTION, OR TRY EXPLAINING WHAT IT ADDS"
→ §235's departure sentence is the answer and was not made prominent
"MIGHT THE SAME APPLY TO THE FIVE BOOKS OF MOSES?" and
"HOW ABOUT THAT SOCRATES… WERE ALSO THE WRITERS OF THE BIBLE AND THE
ILIAD…? OR ALL OF THESE ARE ONE BIG PROJECT?"
→ §238's firewall is the answer and was not stated where he would hit it
"HYPOTHESIS: THE MOON IS GREEN CHEESE. HOLD IT OPEN, NOTHING MORE IS
REQUIRED OR LESS."
→ the response to "hold the question open" being read as a licence
rather than a method. The firewall again.
RECORDED AS RECEPTION. He answered the abstract and the throat-clearing. Every objection he raises is answered later in the same document, which is a fact about where the answers were placed.
KAHN, and the departure sentence §VI, as the antecedent
VLASTOS, the three commitments, the fugue §VIII, at the relocation
THE SOCRATIC PROBLEM as symptom §V, long note or new section
THE FIREWALL front matter + §VII
XENOPHON / ARISTOPHANES under firewall §VI, after the demonstrations
THE TIERED CLAIM STRUCTURE front matter
THE NON-CLAIMS LIST front matter
**None of this is new research. All of it exists, in a paper written before the
architecture, and was lost when the architecture replaced it.**
Added 2026-08-31. Proposal under test: Aristophanes manufactures the Socrates position and the later tradition inherits it — which would give the configuration an EXTERNAL predecessor rather than requiring Aristophanes inside the maker set. The test named for it: which properties of Socrates first appear in the comic layer and then persist, invert or differentiate.
This is not reception-history inference. The Apology says it.
18b–c, the old charge: **ἔστιν τις Σωκράτης σοφὸς ἀνήρ, τά τε μετέωρα φροντιστὴς
καὶ τὰ ὑπὸ γῆς πάντα ἀνεζητηκὼς καὶ τὸν ἥττω λόγον κρείττω ποιῶν.**
19b–c, the sworn form read out, then the source named:
**ταῦτα γὰρ ἑωρᾶτε καὶ αὐτοὶ ἐν τῇ Ἀριστοφάνους κωμῳδίᾳ, Σωκράτη τινὰ ἐκεῖ
περιφερόμενον, φάσκοντά τε ἀεροβατεῖν.**
⭑ Σωκράτη τινά — a certain Socrates. The indefinite, used of a figure in another author's text, by the figure of the same name. And φροντιστής at 18b is the Clouds' own coinage.
**The corpus records a prior, publicly circulating Socrates, comic in origin, which
it must answer — and answers by naming the comedy.**
Seated corpora, crude stem counts:
property Clouds Plato Xen Arist
φροντιστ- thinkery 11 4 8 2
ἀεροβατ- walking on air 2 1 0 0
μετεωρ- things aloft 8 19 12 59
μαιευ- midwifery 0 13 0 0
εἰρων- irony 0 1 0 13
⚠ TWO ROWS ARE CONFOUNDED AND ARE STATED AS SUCH. Aristotle's 59 μετεωρ
**The first pass returned ZERO for ἥττω λόγον in the Clouds, which was a pattern
error.** Rerun for any case-form of ἥττ-/κρειττ- within three words of λογ-:
Clouds 11 Plato 13
But the Clouds does not call them that. Its personified pair is:
δίκαιος λόγος 40
ἄδικος λόγος 38
κρείττων λόγος 0
ἥττων λόγος 0
**⭑ THE COMEDY STAGES THE JUST AND THE UNJUST ARGUMENT. THE APOLOGY REPORTS THE
CHARGE AS THE WEAKER AND THE STRONGER.**
That is not transmission of a phrase. It is a RELABELLING — a moral pair (just/unjust) restated as a dialectical pair (weaker/stronger). The property persists and its terms are changed, which is the differentiation the proposal predicted and is the strongest single item the test produced.
SUPPORTS, in the proposal's narrow form: a prior public Socrates exists, is comic, and is answered by name. ἀεροβατεῖν survives in Plato ONLY inside the quotation of the comedy — a property that did not transfer, only got cited. φροντιστής transfers as a marked term. And the just/unjust → weaker/stronger relabelling is a transformation, not a copy.
DOES NOT SUPPORT the broad form — that enough of the characteristic later Socrates is already structurally present in the Clouds. It is not. μαιευτική is zero in the comedy and thirteen in Plato. Nothing of the questioning-method, the midwifery, the ignorance-as-position, or the divine sign as Plato uses it is in the comic layer.
THE ACCUSATION-PROFILE IS COMIC. THE OPERATION IS NOT.
If comedy supplied the NOTORIETY and the later corpus supplied the OPERATION, then Aristophanes is the origin of the position's public availability and not of its content.
ARISTOPHANES puts a Socrates into public circulation; supplies a
name with a reputation attached
THE LATER CORPUS occupies that position and differentiates it —
defendant, questioner, midwife, ignorant knower,
reported master, absent centre
Which leaves §VI's question where it is — the operation is what §VI is about, and the operation is not comic — and gives §V an external anchor for why there was a name available to occupy.
⚠ AND IT IS A REAL ALTERNATIVE TO H4. The maximal statement lists Aristophanes belongs to the construction as a prediction the argument does not make. This test offers a cheaper account of the same evidence: Aristophanes as external predecessor rather than internal position. It is cheaper historically and it costs the configuration nothing that §VI needs.
§243 is quoted and checkable. Apology 18b–19c, seated text.
§§244–245 are crude stem counts with two rows marked confounded and one pattern error found and corrected in the same round. The relabelling result at §245 is the only one strong enough to build on, and it needs the Clouds' text read rather than counted before it is used.
§§246–247 are a reading. The ordinary account — that Plato knew the Clouds and answered it, which is uncontroversial — is sufficient for everything at §243, and the architecture claim at §247 is what exceeds it.
NOT SETTLED: whether Aristophanes is better placed inside or outside the maker set. This round shows the outside placement is available and cheap. It does not show the inside placement is wrong.
∮ = 1
https://alexanarch.org/data/attachments/ea-notebook-01/EA-NOTEBOOK-01.md