The measurement register of the authorship investigation: §0A as the entryway and M1–M52 as the register, measurements only, each with the seated corpus and commit against which it was computed, the conditions that produced it, and the refutation recorded beside any claim it killed. Nothing is summarised out of existence because it came out negative. Source criticism, reception history and attestation audits are excluded by rule and belong in the notebook (EA-NOTEBOOK-01), a boundary the register records having got wrong twice. The register is the evidentiary basis for every figure in the derivative papers and for the technical report EA-MEASURES-REPORT-01.
Start at §0A. It is the entryway: a problem generated from three Aristotelian propositions, which arises before any hypothesis about authorship is introduced. The measurements in §1–§28 are evidence bearing on it. §0A was the last thing found and is placed first because it is where a reader should begin.
Findings are numbered and stated with their conditions. Where a finding was later refuted, the refutation is attached to it, not filed separately. Where a null is uninformative rather than negative, that is said.
Unless stated otherwise: distance = cosine distance on particle-transition profiles (function-word bigrams, top 300, diacritics stripped, final sigma normalised), computed per work, on corpora seated under EA-CORPORA.
The entry below is placed first because it is the entryway to the whole investigation. It is not the strongest statistic and it is not a correspondence found by searching. It is a problem Aristotle leaves, generated from three of his own propositions read without letting his nouns slide — and it arises before any hypothesis about authorship is introduced. Everything after §1 is evidence bearing on it. This was the last thing found and belongs at the front.
Rewritten 2026-08-29 — the third rewrite, and the first that survives its own sources. The two earlier versions overstated the problem. This one gives the conventional reading its strongest form first, and then asks the question that survives it.
(i) The Platonic corpus presents division-and-collection as a repeatable procedure. Phaedrus 266b — spoken by Socrates, not Plato in propria persona:
τούτων δὴ ἔγωγε αὐτός τε ἐραστής, ὦ Φαῖδρε, τῶν διαιρέσεων καὶ συναγωγῶν,
ἵνα οἷός τε ὦ λέγειν τε καὶ φρονεῖν
Of these I am a lover — the divisions and collections — in order to be able to speak and to think. And Meno 86e names investigation ἐξ ὑποθέσεως, ὥσπερ οἱ γεωμέτραι; the Parmenides runs the exhaustive derivation.
(ii) Aristotle characterises that procedure in syllogistic terms.
APr I.31, 46a31:
ἡ διὰ τῶν γενῶν διαίρεσις μικρόν τι μόριόν ἐστι τῆς εἰρημένης μεθόδου …
ἔστι γὰρ ἡ διαίρεσις οἷον ἀσθενὴς συλλογισμός
(οἷον licenses "a kind of," not species-membership. The weaker claim is what the argument needs.) And of its users: οὔτε ὅ τι ἐνδέχεται συλλογίσασθαι διαιρούμενοι ξυνίεσαν — they did not understand what can be syllogized by dividing.
(iii) Aristotle denies prior partial working-out. SE 183b–184b, after supplying the succession case by name — οἷον ἐκ διαδοχῆς, Tisias, Thrasymachus, Theodorus:
**Ταύτης δὲ τῆς πραγματείας οὐ τὸ μὲν ἦν τὸ δ᾽ οὐκ ἦν προεξειργασμένον, ἀλλ᾽
οὐδὲν παντελῶς ὑπῆρχεν.**
Two earlier versions of M40 called the narrow reading "an interpretive repair" and "an unmarked restriction of the scope of πραγματεία." Both are WRONG, and the Greek of the same paragraph shows why.
(a) ARISTOTLE DRAWS THE ART/PRODUCTS DISTINCTION HIMSELF, in the next sentence:
οὐ γὰρ τέχνην ἀλλὰ τὰ ἀπὸ τῆς τέχνης διδόντες παιδεύειν ὑπελάμβανον, ὥσπερ
ἂν εἴ τις … σκυτοτομικὴν μὲν μὴ διδάσκοι, δοίη δὲ πολλὰ γένη παντοδαπῶν
ὑποδημάτων
*They supposed they taught by giving not the art but the products of the art —
as if one did not teach shoemaking but handed out many kinds of shoes.*
**"Practice versus art" is not a modern rescue. It is Aristotle's own frame, in
this passage.**
(b) THE COMPARISON CLASS IS NAMED, AND IT IS NOT PLATO:
καὶ γὰρ τῶν περὶ τοὺς ἐριστικοὺς λόγους μισθαρνούντων ὁμοία τις ἦν ἡ
παίδευσις τῇ Γοργίου πραγματείᾳ
The "nothing existed" is glossed by reference to **paid eristic teachers and
Gorgias**, who gave arguments to be memorised.
(c) THE CONTRAST IS WITH A WRITTEN TRADITION:
περὶ μὲν τῶν ῥητορικῶν ὑπῆρχε πολλὰ καὶ παλαιὰ τὰ λεγόμενα, περὶ δὲ τοῦ
συλλογίζεσθαι παντελῶς οὐδὲν εἴχομεν πρότερον ἄλλο λέγειν
And the close uses πραγματεῖαι in the plural for disciplines *ἐκ παραδόσεως
ηὐξημένας*.
(d) AND APr I.31 SUPPLIES RETROSPECTIVE SUBSUMPTION ITSELF: the dividers ξυνίεσαν — did not understand — what could be syllogized. Plato may perform D without possessing D as syllogistic theory.
1 ART vs PRACTICE people reasoned; nobody had a τέχνη of deduction.
ARISTOTLE'S OWN FRAME (0A.2a).
2 LOCAL DIALECTICAL SCOPE ταύτης τῆς πραγματείας = the Topics-SE inquiry,
dialectical reasoning from ἔνδοξα. The epilogue
recalls that programme immediately before.
⭑ THE STRONGEST HOSTILE READING.
3 RETROSPECTIVE SUBSUMPTION Plato devised division; Aristotle later discovers
it is a weak syllogism. Supported by I.31's own
ξυνίεσαν.
4 DIACHRONIC ARISTOTLE Topics/SE predates the mature syllogistic
(SEP's probable order; Hitchcock). The two
propositions need not be held simultaneously.
5 PLATONIC GENESIS division materially generated the syllogistic
(Maier; de Strycker; Vlasits — against Shorey,
Cherniss, Ross). Vlasits reads I.31 as almost a
backhanded acknowledgement of dependence.
INTENSIFIES the historical puzzle; explains I.31
as genealogy, not leakage.
6 PRIORITY RHETORIC an inventor's overstatement. Cheap, stubborn,
and not defeated by elegance.
Readings 1–4 work TOGETHER, and together they are what M40 must beat:
In the early Topics–SE *Aristotle claims novelty for a systematic dialectical
art. Earlier thinkers reasoned and Plato had division, but none possessed a
theory of syllogizing as such. Later, having developed the Analytics, Aristotle
retrospectively classifies division as a weak syllogism while saying explicitly
that its practitioners did not understand its syllogistic character. No
contradiction requiring unusual provenance.*
The central problem is no longer "Aristotle contradicts himself." It is:
**What exactly counts, for Aristotle, as a prior PARTLY WORKED-OUT contribution
to an inquiry?**
Because Platonic division is not a pile of shoes. It is:
NAMED διαίρεσις καὶ συναγωγή
REPEATABLE held as a method, ἵνα οἷός τε ὦ λέγειν τε καὶ φρονεῖν
GENERAL applied across kinds, arts, capacities, the soul
and CLASSIFIED BY ARISTOTLE as μόριον τῆς εἰρημένης μεθόδου —
a PART of the method
So the exact question the conventional reader must answer:
**Can a predecessor possess a general, named, repeatable methodological art D,
which Aristotle calls a PART of S, while Aristotle truthfully says that no part
of S was previously worked out?**
That is answerable. But the answer must SHOW THE BOUNDARY — must say what distinguishes having a part of the method from having part of it worked out. The conventional reading has a real answer available and has not been made to state it.
**Reading 2 restricts πραγματεία to the dialectical programme. The next clause is
broader:**
περὶ δὲ τοῦ συλλογίζεσθαι παντελῶς οὐδὲν εἴχομεν πρότερον ἄλλο λέγειν
**Is περὶ τοῦ συλλογίζεσθαι still governed by the dialectical programme just
recalled, or does Aristotle widen the claim to syllogizing as such?**
That is a genuine philological question and it must not be settled by assertion. It is where the M40 station does its work: the conventional account has to draw its boundary somewhere, and the paper's task is to make it draw it and then examine whether it holds.
**Round 18's DISCOVERY reading supersedes the encoding reading and requires far
less:**
P_earlier performs the operation
P_later discovers what the operation is and what its range includes
same maker there was no predecessor from whom the discovery was inherited
No plan, no prescience, no concealed confession, no chronological adjustment. οὔτε ὅ τι ἐνδέχεται συλλογίσασθαι διαιρούμενοι ξυνίεσαν read inclusively means the earlier position did not know the range of its own operation — which makes the priority claim literal rather than equivocal.
And the classificatory move is not isolated. Aristotle calls recollection οἷον συλλογισμός τις (De mem. 453a) exactly as he calls division οἷον ἀσθενὴς συλλογισμός. Two Platonic operations absorbed into the syllogistic family. (This does not independently reproduce the priority denial and is not counted as a second pillar.)
M40 is a PROVENANCE QUESTION generated from inside the Aristotelian corpus, not a demonstrated contradiction. Its force is that it makes the conventional account specify a boundary it has not had to specify — and that it does so before any hypothesis about authorship is introduced.
Three earlier overstatements are withdrawn: equivocation on his own technical term; self-refuting; unmarked restriction of scope. The restriction is marked — by the shoemaker, by Gorgias and the paid eristics, and by the rhetoric/syllogizing contrast.
What is not withdrawn: Platonic division is a named repeatable general method; Aristotle calls it a part of the method; and the denial of prior partial working-out is stated in the strongest available terms — παντελῶς, twice.
These are the reference bands everything else is read against. They are the most
reusable thing in this document.
M1 — Same-author band, Greek prose. ⚠ RECOMPUTED — Josephus only. 0.0148 – 0.0323, n=6 pairs, mean 0.0248. Josephus across four works and four registers (BJ, AJ, Vita, CA) — war narrative, antiquities, autobiography, polemic — and he stays inside 0.0323.
→ The original band cited Xenophon Socratic-vs-history alongside Josephus.
**Xenophon is inside the candidate cluster (M33) and cannot calibrate a
same-author band without assuming the answer.** His within-author dispersion
is mean 0.0530, max 0.1598 — including him would have raised the ceiling from
0.0323 to 0.1598, a fivefold inflation of what counts as "one author."
→ Josephus is **the only clean same-author reference entirely outside the
hypothesis space.** Every band comparison in this register should be read
against 0.0148–0.0323.
→ Plato|Xenophon at 0.0235 remains inside the corrected band; the M8 observation
survives the recomputation, its inference does not (M33).
M2 — Within-author dispersion, per work, by author.
thucydides 0.0097 (n=8)
herodotus 0.0103 (n=9)
plotinus 0.0130 (n=6)
philo 0.0197 (n=29)
josephus 0.0264 (n=4)
aristophanes 0.0358 (n=11)
xenophon 0.0533 (n=9)
plato 0.0570 (n=24)
ARISTOTLE 0.1033 (n=30) ← highest in the sample by a factor of two
M3 — Genre versus author effect. ⚠ CIRCULAR — the author term is void. Genre effect +0.0023, from Xenophon's fourteen works across Socratic dialogue, history, encomium and technical treatise. That term stands: it is within-author.
The author term of +0.0078 does not. It was computed as Plato-dialogue vs Xenophon-Socratic minus Plato-self — i.e. from a Plato|Xenophon comparison, treating them as two authors. Xenophon is a candidate position in the configuration under test (seated held apart, EA-CORPORA-06/01), so using him as the definition of "a different author" assumes the answer. See M33 for the clean baseline and the standing rule.
→ The "authorship moves this 3.4× more than genre" claim is withdrawn. It
rested on the contaminated term.
M33 — THE CANDIDATE CLUSTER, AND WHAT MAY NOT BE ASSUMED ABOUT IT. (2026-08-29. Voids the author term of M3; corrects M8's framing; constrains all future control selection.)
The cluster is {plato, aristotle, aristophanes, xenophon}. No pair inside it
may be used as a control of any kind — known-different OR known-same — without
an explicit defence of the assumption. The rule is bidirectional: if the
hypothesis is live, a within-cluster pair's sameness is as much in question as
its difference. Xenophon's within-author dispersion calibrating a same-author band
(M1, corrected) was exactly this error in the other direction. The hypothesis under test would readily
expand to include all four, and — the point that makes this binding — **if one
configuration were established, more than one becomes more likely, not less**,
since an existence proof lowers the prior for the practice. Nothing here supposes
the expansion. It refuses the assumption in either direction.
**Clean baseline, author effect = between-author distance minus the mean of the
two within-author dispersions:**
OUTSIDE the cluster n=10 mean +0.0760 range +0.0316 to +0.1358
CROSS cluster/outside n=20 mean +0.0729
WITHIN the cluster n= 6 mean +0.0587 range +0.0299 to +0.0787
Plato|Xenophon = +0.0299 — the lowest value in the table, below the out-of-cluster minimum of +0.0316. The pair used as the operational definition of "two authors" is the closest pair in the sample.
The cluster is not uniform. Plato, Xenophon and Aristophanes are mutually close (+0.0299, +0.0356, +0.0583); Aristotle is far from all three (+0.0716 to +0.0787), inside the ordinary out-of-cluster range. Consistent with register — three dialogic writers against one treatise writer — but no longer an authorship finding, because the baseline it was read against was contaminated.
Consequently contaminated and marked:
withdrawn; the distance stands, the inference from it does not.
Plotinus and Josephus — **two cluster members inside a pool labelled
multi-author.** The M28 conclusion (subject explains the dispersion) is
unaffected, since it rests on within-Aristotle domain partitions, but the pool
comparison should be rebuilt from outside the cluster.
Xenophon|Aristophanes as known-different truths; both are within-cluster, so the
reported 1-of-3 agreement rate is computed against an invalid panel and must be
redone.
Unaffected and verified: the 61-pair empirical null pairs Xenophon only with Philo, Thucydides, Herodotus, Josephus, Longinus, Plotinus and the NT authors — all cross-cluster — and excludes Aristophanes as verse. Plato and Aristotle are not in it. The null is clean.
M4 — Raw separation. Plato|Aristotle per work: 0.1578. Rank 29 of 45
author-pairs — the far half.
M5 — Standardised separation. Between-author distance divided by the mean of the two authors' internal dispersions. Plato|Aristotle = 1.975, rank 7 of 36 — among the LEAST separated pairs in the corpus. Herodotus|Thucydides, exact contemporaries in one genre, = 8.867. Plotinus|Herodotus = 12.574.
→ This corrected an error made earlier in the same session. "Twelve times a
normal author gap" compared a raw distance against an effect size. Adjusted
for dispersion, the pair is unremarkable and on the close side.
M6 — Declared-author partition strength. Plato vs Aristotle by declared author, 65 works: within/between 0.548, and the true partition sits at the 100th percentile of random partitions, Z = +39.18. The names are overwhelmingly strong containers.
M7 — Reference point for M6. The New Testament by listed author, same test: within/between 0.869, Z = +2.83. NT names carry signal but do not contain their voices; "John" spans Gospel 0.0923 / Epistles / Revelation up to 0.1842, and Matthew|Luke (0.0167) is closer than Luke|Acts (0.0183).
M8 — Plato is closer to Xenophon than to Aristotle. Plato dialogues |
Xenophon Socratic works = 0.0235 — inside the M1 same-author band, tighter
than Josephus's Vita from his Antiquities.
→ ⚠ The original gloss, "two men nobody has ever proposed as one," is
withdrawn (M33). The measurement stands; the independence it assumed does
not.
M9 — Aristotle's isolation. His self-distance (0.1033) exceeds his distance to Plotinus (0.0810), a man born six centuries later. He is further from Xenophon and Thucydides than Plato is from Homer.
**M10 — Exceeds multi-author pools. ⚠ SUPERSEDED BY M28 — the anomaly is the
subject, not the author.** Aristotle's mean internal dispersion (0.1033) sits at
the 100th percentile of 200 synthetic five-author pools drawn from Plato, Philo,
Xenophon, Plotinus and Josephus and matched to n=30 (pool mean 0.0665, sd 0.0062,
max 0.0819).
→ The figure is correct and the inference drawn from it was wrong. **Holding
subject constant removes it entirely** (M28). Do not quote 0.1033 as an
authorship anomaly.
M11 — What the dispersion tracks. Partition tests inside Aristotle, 30 works,
3,000 random partitions each:
DOMAIN (subject matter) ratio 1.373 100th pct Z = +4.25
received authenticity ratio 1.506 100th pct Z = +3.47
EDITION (Perseus / First1K) ratio 0.993 50th pct Z = −0.10
→ Edition excluded as a confound, at the work level, confirming the
corpus-level test at ratio 1.00.
→ Subject dominates. The leading explanation is one man across an
exceptionally wide register range.
→ The authenticity signal is real but confounded: Andronicus drew that
partition on internal-consistency grounds, so it may be recovering the
stylistic structure it was built from. The two hypotheses predict the same
measurement.
M12 — Conservative headline. ⚠ REVISED BY M28. The earlier form — the most internally heterogeneous single-author object in the sample, and that heterogeneity is primarily register — was directionally right and understated the cause. Corrected: the heterogeneity is the object. Aristotle-within-object is ordinarily consistent, and in ethics is one of the most consistent writers in the sample.
M28 — THE SUBJECT DETERMINES THE SHAPE. *(2026-08-29, supersedes M10, revises
M11 and M12.)*
Domain effect, measured per author with 2,000 random partitions:
author n within-domain cross-domain ratio Z
Aristotle 26 0.0667 0.0993 1.489 +4.54
Plato 20 0.0499 0.0592 1.186 +1.75
Subject determines Aristotle's shape about 2.6× more strongly than Plato's.
And holding subject constant, the M10 anomaly disappears. Aristotle's within-domain dispersion overall is 0.0571 — Plato's total figure is 0.0570. Domain by domain, against 400 synthetic five-work multi-author pools (mean 0.0669, sd 0.0208):
ethics 0.0226 → 0.5th pct of multi-author pools
psych 0.0390 → 6.5th
nat 0.0465 → 16.2th
polit 0.0584 → 39.0th
bio 0.0680 → 56.8th
poet 0.0707 → 61.2th
logic 0.0942 → 88.5th
Within ethics, Aristotle is tighter than 99.5% of multi-author pools — tighter than Josephus (0.0264), tighter than Philo (0.0197) relative to the pool baseline. What made him look like many authors is that he wrote about many things.
The reading this supports. This is mimesis in the Republic III sense moved from voice to form: the tragedian disappears into his character, the treatise disappears into its object. On that reading Aristotle has no style because having one would be a failure of the method — a consistent manner across ethics and biology would mean the manner was not taking its shape from the thing.
The residual ordering agrees: ethics tightest, logic loosest. Ethics has one object and one register; logic ranges over categories, interpretation, analytics and sophistical refutations — the widest internal object-space of the group. The remaining dispersion tracks the heterogeneity of the subject matter within each domain, not the author.
What this does not establish. That the corpus is by one hand. A school trained to a method that conforms to its object would produce the same measurement, and the object-conformity is a property of the method, not of a person. It removes an anomaly rather than supplying a positive finding — but the anomaly it removes was the strongest one the investigation had.
M13 — Plato's variance is temporal, Aristotle's is topical. Same partition
test, 24 Platonic works:
CHRONOLOGY (early/middle/late) ratio 1.442 100th pct Z = +6.92
DOMAIN (subject) ratio 1.130 91st pct Z = +1.42
FORM (dialogue / expository) ratio 1.225 85th pct Z = +1.08
→ The strongest partition effect measured in the entire session.
→ Orthogonal axes. Aristotle's dispersion is indexed by *what he is writing
about; Plato's by where along a sequence*. Magnitudes fit: Plato 0.0570 is
roughly half of Aristotle's 0.1033 — a subset, not an expansion.
→ Caveat, standing. The E/M/L stratification was built partly from
stylometry, so this risks the circularity of M11. Denyer's argument (that
Plato could write any style at any time) makes E/M/L a repertoire rather
than a record — in which case Z=+6.92 detects a staged development, which is
what "artificed" would claim. The measurement does not distinguish these.
M14 — Plato's internal speaker structure. From TEI <said who=…> markup: 56 speakers, 519,743 spoken words. Socrates 252,416 · Athenian Stranger 102,795 · Eleatic Stranger 29,497 · Timaeus 20,691.
Distances to Socrates: Athenian 0.0298, Eleatic 0.0338 — both at the edge of the M1 same-author band. Timaeus 0.0772 — nearly the full Plato–Aristotle distance, inside one author.
M15 — The equilateral. Socrates | Athenian | Eleatic form a near-perfect
equilateral: balance (min/max displacement magnitude) 0.953, angles 116.8° /
121.4° / 121.8°, spread under 5°. Nothing else measured comes close. **Not a size
artifact** — the three corpora are 252k / 103k / 29k words and their displacement
magnitudes are 11 / 11 / 11. Substituting Timaeus for the Eleatic collapses it to
0.719 / 41°.
M26 — Plato writes in the form he banishes. Republic 392d–394c defines the wholly mimetic mode as speech with no reporting poet, and names it with tragedy and comedy. Reporting-formula density per 10⁴ tokens: 20 of 31 works, 61% of the corpus by volume, carry no reporting voice (Crito 0.0, Statesman 1.1, Sophist 1.7, Laws 2.4, Gorgias 6.1). The Republic itself is narrated at 144.1 — Socrates reports the whole of it. The book performing the expulsion is written in the mode that escapes it. Textual, not interpretive.
M27 — Aristotle on the same axis. Reporting 0.1 (lower than Plato's dramatic dialogues), self-speaking 29.0 against their 25.3. Caveat: a treatise has no reporting formulae because there is no conversation to report, and speaking in his own person places Aristotle in the un-banished second mode. What survives the caveat is the near-equality of self-speaking rates across the two most formally opposed prose bodies in Greek — recorded as a measurement without an account.
M16 — Plato does not construct heteronymically. Symposium speeches by Stephanus range, against the real Aristophanes: Plato's Aristophanes speech is 0.0854, rank 5 of 7 — Pausanias (0.0428), Socrates/Diotima (0.0556), Phaedrus and Alcibiades are all closer to actual Aristophanic Greek. Meanwhile every speech sits near the Plato corpus, Aristophanes' at 0.0380. Given the one chance to construct against a checkable target, the speech comes out more Platonic than its neighbours, not less. The only ancient case with an independently attested target, and it reads negative.
M17 — ∮ = 1 on Pessoa. Pooling the three heteronyms recovers the orthonym at 0.0258 — inside the M1 same-author band — while the best single heteronym is Campos at 0.0498, then Caeiro 0.0903, Reis 0.2205. Pooled beats best individual by 1.93×. Individually the voices do not look like Pessoa; summed, they do.
Comparators: Josephus BJ+AJ+CA → Vita 1.08×; Kierkegaard A+B → signed 1.07×;
Plato parts → Aristotle 1.01×; **Aristotle parts → Plato 0.78× (pooling makes
it worse — his Ethics alone is closer to Plato than his whole corpus is).**
Limits: ratio statistics carry a size confound; cross-language absolute
comparison is shaky; Kierkegaard's seat holds four pseudonymous works of ~12.
M18 — Ownership statistic. Feature-ownership concentration, n=3 groups: Plato-self 0.1025 · Kierkegaard 0.1188 · Xenophon-self 0.1349 · Pessoa 0.1442 · Plato|Aristotle|Xenophon 0.1890 · three independent Greeks 0.2044. One mind gives low ownership; many minds give high. But the matched 2×2 design collapses it: Plato(2)+Aristotle(2) 0.1151 sits between two independent controls (0.1038, 0.1242) and inside the within-author range (Plato alone 0.0830, Aristotle alone 0.1179). Register, not authorship.
M19 — Dionysius replicated. Revelation | Gospel of John = 0.1319 / 0.1189,
far outside anything the NT shows internally. The third-century stylistic
authorship judgment holds on this instrument.
M20 — Revelation's nearest neighbour is Mark, at 0.0329 / 0.0460 — inside the M1 same-author band. Almost certainly shared Semitizing register rather than shared hand, and the sharpest demonstration in the session that this measure reads register as authorship.
M21 — The Pauline ordering inverts the consensus. Undisputed Paul | Pastorals
= 0.0494; undisputed Paul | disputed (Eph/Col/2Thess) = 0.0873. The
letters the field is most confident are pseudepigraphic score closer than the
ones it is least sure about. John's Gospel | Johannine epistles = 0.0665/0.0927 —
further apart than Paul from the Pastorals.
M22 — NT criteria applied to Plato/Aristotle: one of five holds. cardinality mismatch ✗ · cross-name convergence ✗ · redundancy-with-complementarity ✗ (they are content-complementary and function-redundant, the inverse) · structured cross-attestation ✗ (one direction only) · absent centre ✓. The one that holds is the one not measurable from texts, because its subject left none.
M23 — No whole-text operator. Monotone alignment (DTW on 40 blocks, 60-shuffle null): Timaeus→Metaphysics Z = −0.46; Symposium→Poetics −0.68; Republic→Politics +0.96; every candidate inside ±1.2. No order-preserving transform maps any Platonic dialogue onto any Aristotelian treatise.
M24 — Mediation through Aristophanes fails. 100 routes, 2,000-shuffle direct null: Symposium→Poetics direct = −0.82, rank 15 of 20 dialogues; nothing in the sweep crosses 2 (max +1.89, the expected max of 20 draws). The mediation sweep's apparent crossings sit inside a distribution whose mean is +0.76, i.e. the statistic is biased upward by the DTW path's own selection.
M25 — Corpus-layer leak deficit: TEST VOID. EA-LEAK-DEFICIT-02 (sha256 493514a2f142385f…). Sanity control fired: Josephus pooled = 9881.2, Plato|Aristotle pooled = 9912.5, both at the 100th percentile. One author's works pooled and two authors' works pooled are indistinguishable, because pooling homogenises blocks by construction while the null was built on single works. No number from that run may be quoted.
Each was constructed, produced an apparent crossing, and was then found to be an
artifact. Four were reported as findings before the artifact was located.
1. Deviation-from-group — forced to r = −1 by the sum-to-zero constraint.
2. Shared-region correlation — dominated by the shared frequency baseline;
high-frequency words are high for everyone.
3. DTW mediation null — biased upward by the path's own selection of
high-similarity cells; mean Z +0.76 across 100 routes, not 0.
4. Coherence-destroying null — inverted in sign; homogenised blocks are
more similar to each other, not less.
5. Within-row permutation — left the global multiset unchanged, so real and
null were identical by construction (all Z = 0.00 exactly).
Consequence, now standing policy: nulls for this work are empirical — real
matched author-pairs — not synthetic.
EA-LEAK-DEFICIT-01 fbbd13a53a5b29af… SPENT — tau at the 99th percentile
gave a degenerate lower tail (8 of 61
pairs leak exactly 0), so the
predicted deficit was unobservable.
Found before the test pair was
computed.
EA-LEAK-DEFICIT-02 493514a2f142385f… DISCHARGED VOID — sanity control
fired (M25).
The null set built for these — 61 matched cross-author Greek prose pairs across 19 author-pairings, capped at 6 per pairing, from Xenophon, Philo, Plotinus, Josephus, Longinus, Herodotus and Thucydides by book, and the NT authors — is reusable and is the most expensive artifact produced.
1. ~~Base rate for the Republic 395a / Symposium 223d contradiction~~ —
RUN 2026-08-29, VERDICT: NOT ADJUDICATED (EA-SNUB-01 §19-20). Loose
criterion 653/16,449 = 4.0%, target unremarkable; strict criterion
61/16,449 = 0.37%, target excluded on an adjacency technicality. Propositional
contradiction is not separable by proximity-negation. The observation stays
philological and cannot be promoted by statistics.
New M26 (see §4): by Republic III's own criterion, 61% of the Platonic
corpus is written in the banished mode, and the Republic is not.
2. Authenticity within domain, inside Aristotle. Hold subject constant and
test spurious-vs-genuine. Breaks the M11 confound. Counts are thin (6 spurious
of 30) so power may not exist.
3. External dating against stylistic position for M13's circularity — a
dating source not downstream of style.
4. A recognition measure reading λόγος rather than ὕλη (EA-SNUB-01 §6).
Untouched, hardest, and the one the whole line of work is missing.
Added 2026-08-29. Everything attempted, including what failed, what was superseded, what was left half-run, and what was never started. A later instance should be able to see the whole floor, not only the standing results.
*Everything attempted, including what failed, what was superseded, what was left
half-run, and what was never started.*
T1 — Ending profiles (aporia vs closure). Genuine Aristotle 7/7 CLOSED (aporia 1, closure 23); Plato majority OPEN (aporia 22, closure 15); the single Aristotelian work with an open ending is Oeconomica, which is spurious. → CIRCULAR. The result was sharpened by excluding the spuria, and the spuria are classified spurious partly because they read differently. A category built from the consistency being measured was used to measure it. Recorded because the error is instructive, not because the number stands.
T2 — Coverage complementarity on the virtues. Courage: Laches 0 definitional
formulas, Aristotle 7 — and he closes it explicitly (σχεδὸν εἴρηται) while the
Laches ends with an appointment to meet tomorrow. Temperance 1 vs 8. Friendship
0 vs 4. But: piety is dropped by both (ὁσιοτ- appears 0 times in Aristotle's
genuine ethical corpus), and the beautiful is duplicated (Hippias Major 17 vs
Aristotle 15).
→ Inconclusive, and it favours influence. Systematic design predicts complete
coverage without duplication; what was found is incomplete coverage with one
dropped question and one duplication — the profile of a successor finishing what
interested him.
T3 — The nine-author landscape on aggregate profiles. Produced Aristotle|Plotinus 0.0310 as the second-closest pair in the set. → SCALE ERROR. Aggregate profiles (mean across an author's works) and per-work profiles are not comparable, and averaging destroys exactly the variance that matters. Superseded by M4/M5, computed per work. The 0.0310 figure must not be quoted against per-work numbers.
T4 — Tiling / exclusivity. Vocabulary exclusivity: Kierkegaard A|B — the purest symbolon available — is the least exclusive pair measured (0.4900). → Correct but the wrong operationalisation. Two halves of a broken token share the break line; exclusivity measures distance in topic, not shape.
T5 — Shared-region correlation. Designed pairs occupy both extremes (Pessoa Caeiro|Reis 0.4479, Pessoa heteronyms|orthonym 0.8203) while independent pairs cluster at 0.54–0.62 and Plato|Aristotle sits at 0.6160, the base rate. → Does not separate. Designed configurations have high variance in pairwise similarity, not a characteristic level. Suggestive and unused.
T6 — Effective rank / span of deviations. Kierkegaard 0.895, Josephus 0.952, Plato-4-groups 0.839, Greek-4-independent 0.932, Plato+Aristotle 0.820. → Does not separate. All within 0.82–0.95.
T7 — Aristotle's attribution practice. Across 48 works, unnamed attribution
(τινές, φασί, λέγεται, οἱ μέν…οἱ δέ) runs 3.83 per named predecessor; in the
Metaphysics alone 4.48, with λέγεται at 309. The systematizing position
doing the doxography that grounds most of what we think we know about who held
what attributes anonymously four times for every naming.
→ Measured, never used. Relevant to any argument about how attribution behaves
inside the corpus.
T8 — Folding / uptake retention. Source-vocabulary retention when one text is folded into another: Sappho 31 → Longinus 100% (the seat states Longinus is the sole direct carrier); LXX Pentateuch → Philo 76%; LXX → Josephus AJ 44%; Plato → Aristotle Metaphysics 33%; Plato → Xenophon 0%. → Interesting and unfinished. Ezekiel/Daniel → Revelation failed on a filename glob and was never rerun.
T9 — Authorial identity was obscurable in this exact period and circle. Aristophanes produced his first three plays under other men's names — Banqueters (427), Babylonians (426), Acharnians (425), all διὰ Καλλιστράτου — describes the practice himself at Wasps 1017–22, and continued it after he had a name (Lysistrata 411 via Callistratus, Frogs 405 via Philonides). The civic didascalic record carries the producer, not the poet. Cleon's prosecution over the Babylonians cannot be resolved as against Aristophanes or Callistratus. Biographical facts about "Aristophanes" — baldness, the Aegina connection — are extracted from plays whose named author may be a different man.
T10 — Base rate around T9. Ps.-Xenophon's Athenaion Politeia sat in Xenophon's corpus from antiquity, questioned by Demetrius of Magnesia but quoted as Xenophon by Stobaeus, author never identified. The Hippocratic corpus: ~60 treatises under one name, multiple hands, this exact window.
T11 — The Aristotelian corpus boundary is an inference. Andronicus of Rhodes constituted it in the 1st c. BCE, ~250 years after, working from the cross-references the texts themselves contain plus school documents. His catalogue is not extant. He judged De interpretatione spurious; the field now receives it as genuine. Metaphysics A's authenticity was questioned in Alexander's time, with Asclepius reporting attribution by some to Pasicles of Rhodes — and A is the book the paper's worked case is built on.
T12 — The recursion. The Athenian didascalic records descend from Aristotle's lost Didascaliae. The one surviving contemporary inscription naming a principal (FD III 1, 400, ante 329) crowns Aristotle and Callisthenes for compiling the register of Pythian victors. The compiler of the record that assigns plays to authors is the figure whose own corpus's partition is in question, and that record by convention sometimes names the producer instead of the poet.
T13a — RESOLVED 2026-08-29. Herodotus and Thucydides, listed as unseated dependencies in the first version of this register, are now seated as EA-CORPORA-07/01 and 07/02 (deposit #1563). The empirical null's seventeen units from these authors now rest on archive-held, commit-pinned material rather than a census checkout.
T13 — Unavailable at any pin. Iamblichus and Damascius: absent from canonical-greekLit and First1KGreek under every author string. Prior attempts failed for this reason, not for want of searching.
T14 — Kierkegaard's Samlede Værker is unusable. All 25 volumes are on archive.org, unrestricted, with full text — and the OCR has failed completely because the editions are Fraktur. Danish common-word rate 0.13–0.14 uniformly across sampled segments, against 0.28 for genuine Danish. 0/12 clean segments in either volume tested. A naive quality gate at 0.10 passes them; it must not be. → The seven missing pseudonyms — Constantin Constantius, Vigilius Haufniensis, both Climacus positions, Hilarius Bogbinder, Frater Taciturnus, Inter et Inter — require a source outside reach (SKS at sks.dk, not allowlisted).
T15 — Pessoa's minor heteronyms are unavailable. pt.wikisource carries only Caeiro, Reis, Campos and the orthonym. Bernardo Soares, Alexander Search, António Mora, the Barão de Teive, Vicente Guedes: no categories, nothing usable on archive.org.
T16 — Two silent retrieval failures, both recorded in their seats.
(a) The Wikisource extracts API returns empty bodies for short lyrics whose
text sits inside <poem> templates, while longer prose-poems extract normally —
so an extracts-based fetch yields Campos nearly complete and Caeiro and Reis
nearly empty, and reports no error. Raw wikitext via prop=revisions is
required.
(b) Danish Wikisource pages for Enten–Eller are djvu transclusions:
prop=revisions returns only a <pages index=…> stub, so a raw-wikitext fetch
yields nothing, also with no error. action=parse resolves them.
T17 — A structural limit on the calibration set. The two strongest NT criteria — name-to-voice cardinality mismatch, and cross-name convergence exceeding within-name — require multiple positions under one declared name. Pessoa and Kierkegaard are one-name-one-voice by construction (the sole exception being Victor Eremita carrying A and B). So those criteria cannot be calibrated on the known-designed corpora. The corpora that have the structure are the New Testament and, on the hypothesis, the Platonic corpus.
genre-controlled instrument was never built on him.
the orthonym. The design calls for withholding each of the four in turn, so
that recovery can be checked on voices nobody considers special. Three folds
unrun.
never rerun.
share a semantic character — technical vocabulary Aristotle needed — or is the
list arbitrary? συμβεβηκός (2× → 449×) suggests the former.
important open item; see §10.1.
(2026-08-29). See M29.** Three unprompted positives; diagnosticity unresolved.
position to technical Platonic use? Under the received chronology it cannot; on
any reading that questions the chronology, the absence is a datum. The 31-term
class supplies the denominator.
the M11 confound. Power may not exist (6 spurious of 30).
only. A systematically reordered mapping would not register.
move, figure. The alignment scaffold accepts any unit and any similarity
function; swapping the block-content vector is a small change.
Nothing measured marks the Plato–Aristotle pair as anomalous for two contemporaneous authors. The declared partition is overwhelming (M6). The standardised separation is on the close side (M5). One of five NT criteria holds (M22), and it is the one not measurable from texts. No whole-text operator exists (M23). The one ancient case with a checkable target reads negative (M16).
What was anomalous was Aristotle's internal heterogeneity (M10) — and as of M28 it is not. Holding subject constant removes it: within-domain 0.0571 against Plato's total 0.0570, and within ethics tighter than 99.5% of multi-author pools. The heterogeneity is the object, which is a claim about method rather than about hands. The authenticity signal remains confounded by the criteria that produced the authenticity labels, and the corpus boundary remains an inference by an editor whose judgments have been overturned (T11).
With M28 the investigation had no standing statistical anomaly. Every
substrate-reading lead was closed or explained.
M30 reopened it, on a different axis. The naming-gap reflex groups Plato and Aristotle at 9.4× the rate of other high-division authors, P ≈ 0, and survives the controls that killed everything else because it is a count of an arbitrary procedural habit rather than a measure of similarity. Plotinus divides harder than Aristotle and never does it; the multi-hand Hippocratic corpus does not do it at all.
The standing position is therefore:
substrate stylometry no anomaly; the pair is ordinary or close
Aristotle's heterogeneity explained — it is the object (M28)
whole-text transform negative (M23)
MECHANISM NEGATIVE — Plato's constructed Aristophanes
ranks 5 of 7 against the real one (M16)
procedural reflex POSITIVE — 10.5x at work level, p = 0.0007;
diagnosticity UNDECIDABLE-PENDING, not merely
blocked — no admissible control exists, since
any corpus performing the operation is
Academy-descended (M30, M36)
forgery/legitimacy census ψευδεπίγραφος and προσωπεῖον BOTH ZERO in
every seated corpus; the frame is kinship and
standing, not inscription (M35)
lexical exclusion POSITIVE — Socrates 0/144 against Callias
5/35 in snub contexts, p = 0.00022 (M34)
philological M26, M31, M32, and the Republic/Symposium pair
And the reflex's diagnosticity is unresolved for one reason: a school transmits procedures, and the successor control that would separate personal from Academic transmission — Theophrastus dividing — is not extant in the available corpora.
M16 belongs in this table and was previously buried. 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. It is the only ancient case with an independently attested target, and it is the direct test of the mechanism the unity hypothesis requires: sustained voice separation across a large corpus. It reads negative. Continuity (M30, M34) and capacity-for-separation (M16) are different claims, and the investigation currently has positives on the first and a negative on the second.
The strongest unrefuted observation in the entire line of work is a philological one: the same formula, asserted and denied, on the exact question at issue, with the affirming instance placed where its demonstration is destroyed in transmission (EA-SNUB-01 §16–18). It has not yet met a base rate.
Added 2026-08-29. Supersedes §15's "never attempted" entry for the λόγος measure.
EA-SNUB-01 §6 derived a recognition criterion from the snub passage itself: matter individuates, form recurs, so what persists across substrates is the account and not the instance. §7 and §15 both listed a measure reading λόγος rather than ὕλη as untouched and as the gap in the whole line of work.
It is no longer untouched. Three searches built in Rounds 4–7 of EA-SNUB-01 read operations rather than substrate, and all three returned unprompted positives in a session where every substrate-reading instrument returned null or artifact.
The three hits.
1. Terminal aporia → technical uptake. σιμόν at Theaetetus 209c, the failed
individuating mark, becomes Aristotle's canonical case of enmattered form.
Found by lexical search; the positional claim was refuted by base rate
(EA-SNUB-01 §9), the semantic reading survives.
2. Self-application regress. Soph. El. 182a — if snub is concavity-of-nose
and there is a snub nose, there is a concave-nose nose — returns Parmenides
132a, the Third Man, by argument shape with zero lexical overlap. Found
without being shown where to look.
3. Divide, flag, coin. Run a division far enough and you reach a class the
language has no word for. Both corpora stop, say so, and coin a name.
Platonic instances occur in exactly the four works that classify — Sophist
(×3), Statesman, Theaetetus, Timaeus (×3) — and nowhere else in Plato.
Aristotle states the policy outright at NE 1108a: πειρατέον αὐτοὺς
ὀνοματοποιεῖν σαφηνείας ἕνεκα.
Why it works where the others failed. Method is what maintains its shape across shifting matter. Every substrate instrument reads ὕλη — particles, frequencies, blocks, register — and ὕλη is what individuates. The procedural searches read the operation, which is what recurs.
The problem, stated as strongly as it can be put against the instrument. Method is precisely what a school transmits. A student acquires his master's procedures more reliably than his vocabulary, his register, or his opinions. So procedural continuity is simultaneously the most detectable thing across the boundary and the least diagnostic for authorship: its presence is predicted equally well by ordinary teaching.
The sharpening that would make it diagnostic. Not whether both use division — everyone in the Academy divided — but whether they share arbitrary features of the method: reflexes not determined by the method's purpose. Nothing requires a divider to stop and coin; leaving the class unnamed and proceeding is available and costs nothing. Both stop. Both flag. Both coin.
The test, and its denominator. RUN 2026-08-29 for reflex 3 — see M30. Plato and Aristotle share the naming-gap reflex at 9.4× the rate of other high-division authors, P ≈ 0, with Plotinus dividing harder than Aristotle and never doing it, and the multi-hand Hippocratic corpus at zero. That is the first positive result the investigation has produced. Reflexes 1 and 2 remain untested against a denominator, and the decisive successor control (Theophrastus dividing) is not extant in the available corpora.
Added 2026-08-29. The first measurement in this investigation that groups Plato and Aristotle rather than separating them. Partly answers M29's diagnosticity question.
The reflex. Run a division far enough and you reach a class the language has no word for. Nothing about διαίρεσις requires you to stop and say so — leaving the class unnamed costs nothing and proceeds. Plato and Aristotle both stop, both flag it, and both coin. Aristotle writes the policy down at NE 1108a: πειρατέον αὐτοὺς ὀνοματοποιεῖν σαφηνείας ἕνεκα — we must coin names ourselves for the sake of clarity.
Measured as instances per 100,000 tokens, counting ἀνώνυμ-, οὐκ ἔχει ὄνομα,
οὐ κεῖται ὄνομα, οὐκ ὠνόμασται and cognates:
corpus gap/100k raw / tokens division/10k
ARISTOTLE 8.43 81 / 960,910 30.9
Diogenes Laertius 2.74 3 / 109,679 16.1
PLATO 2.60 15 / 576,894 15.4
Herodotus 2.16 4 / 184,935 3.7
Aristophanes 1.96 2 / 102,058 4.0
Josephus 0.64 3 / 467,569 5.4
Plotinus 0.47 1 / 214,028 36.8
Philo 0.35 2 / 575,592 13.3
Xenophon 0.32 1 / 310,278 1.7
Thucydides 0.00 0 / 150,157 1.2
Hippocratic (multi-hand) 0.00 0 / 98,175 2.4
Greek NT 0.00 0 / 116,244 1.3
The reflex is not a function of dividing. Plotinus has the highest division density in the sample — 36.8 against Aristotle's 30.9 — and one instance in 214,000 words. Philo divides at 13.3, comparable to Plato, and has two in 575,000.
The test. ⚠ INFERENCE CORRECTED — the token-level p-value was indefensible.
Token level, descriptive only. Restricting to authors dividing above 13 per 10⁴:
Plato + Aristotle 96 in 1,537,804 = 6.24 /100k
Plotinus + Philo + Diogenes 6 in 899,299 = 0.67 /100k
RATIO 9.4x
The Poisson figure originally recorded here — P(X≥96 | λ=10.3) ≈ 0 — **is
withdrawn.** It treats 96 tokens as independent arrivals under one homogeneous
rate. They are not independent: they cluster by work (NE 11, HA 12, Sophist
3, Timaeus 3) and by construction. The effective experimental unit is the
work, not the token, and an infinitesimal p on clustered events gives a false
impression of replication.
Work level, the honest unit. A work counts once if it carries the reflex at all,
restricted as before to works dividing above 13 per 10⁴:
Plato + Aristotle 16 of 32 works 50.0%
Plotinus + Philo 1 of 21 works 4.8%
RATIO 10.5x
Fisher exact, two-sided p = 0.0007
The effect survives the correct unit and the ratio is marginally larger. The defensible statement is p = 0.0007 across 53 works, not P ≈ 0. Per-corpus: Plato 4/9, Aristotle 12/23, Plotinus 1/6, Philo 0/15.
Why this survives what killed everything else.
same philosophical prose in the same tradition and do not do it.
ends of the measure.
multiple hands, classifying diseases — has zero in 98,175 tokens.
What it does not show. A school transmits reflexes, and an arbitrary procedural habit is the most transmissible thing there is. This is consistent with ordinary teaching and does not establish common authorship.
The decisive control is missing. Theophrastus — Aristotle's actual successor — survives in the available corpora only as Characters, which does not divide. Xenocrates and Speusippus are not extant. So whether the reflex is personal or Academic cannot presently be determined, and that is the question that would make it diagnostic.
Status. The first result in this investigation that distinguishes Plato and Aristotle jointly from every available control, by a mechanism that is neither register nor distance. Every other measurement either separated them or dissolved. This one groups them, at p = 0.0007 on 53 works.
Never to be quoted without the missing control. ⚠ AND THE CONTROL IS NOT THEOPHRASTUS — see M36. The claim recorded here, that Theophrastus survives only as Characters, was false and never searched; the philosophical works were in First1KGreek all along and are now seated. But when run, the result is void as a control result: Theophrastus cannot be established outside the configuration (M36), being defined by his relation to Aristotle, writing continuations of the abandoned Aristotelian programme, and sharing a single transmission channel with it. M30's diagnosticity is not resolved by him and may be undecidable in principle (M36b). School transmission remains live; it has not been confirmed.
Denominator, stated. Of five Aristotelian snub-arguments given Platonic-vocabulary signatures and searched (Notebook Part I, Round 4), one mapped — the self-application regress. Of the procedural reflexes since pursued, three were tested and three hit. The full inventory of candidate operations — elenchus, doxography, myth-making, definition by genus and difference — has not been built, and until it is, three-of-three is a numerator without its denominator.
Added 2026-08-29. See EA-SNUB-01 §24–27 for the passage and its cautions.
Statesman 260d–e, position 0.064 — one of the four Platonic works carrying the M30 naming-gap reflex. The Eleatic Stranger enumerates the directive arts that relay ἀλλότρια νοήματα, another's thoughts — ἑρμηνευτική, κελευστική, μαντική, κηρυκική — finds that σχεδὸν ἀνώνυμον ὂν τυγχάνει τὸ τῶν αὐτεπιτακτῶν γένος, and coins αὐτεπιτακτική for the position that issues from itself. The remainder is dismissed with ὄνομα ἕτερον αὐτοῖς παραχωρήσαντες θέσθαι τινά — let them set themselves some other name.
Recorded as a fact about the corpus, not an inference about its authorship. The distinction between speaking from oneself and relaying another's mind was available, named, and coined by Plato in the act of finding Greek short of a word for it.
⚠ CORRECTED same day. The first version of this entry said Plato assigns
the coined art to kings, and that the class is political rule. That inverted the
containment. τὸ μὲν τῶν βασιλέων γένος εἰς τὴν αὐτεπιτακτικὴν θέντες — τίθημι
+ εἰς: kings are placed into the self-commanding, which is the container. And
at 267b: **ζῳοτροφικὴ δὲ πάλιν αὐτεπιτακτικῆς οὐ τὸ σμικρότατον τῶν γενῶν
ἀπεσχίζετο** — animal-rearing split off from the self-commanding, *not the
smallest of the genera*. Other genera exist inside the class and Plato says so,
declining them with τοῦ γὰρ ἄρχοντος ἕνεκα ἡμῖν ἡ μέθοδος ἦν — our method was for
the sake of the ruler.
So 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.
Standing caution, corrected. Plato does not apply this to authorship, and the unentered genera are unentered — supplying them is an inference the text does not license. The ἑρμηνευ- link to Περὶ ἑρμηνείας remains a homonym; Aristotle's sense is expression. Nothing is built on it.
Added 2026-08-29. See EA-SNUB-01 §28–33.
κερματίζω, to mint a unity into small change, is Plato's standing metaphor for the one broken into many (Meno ×2, Parmenides, Republic 0.675, Timaeus ×3, Cratylus). It is the stated ground of the denial at Republic 395b: εἰς σμικρότερα κατακεκερματίσθαι ἡ τοῦ ἀνθρώπου φύσις — human nature is minted into denominations too small to hold two arts. The incapacity is a fact about fragmentation, not about poets.
Symp 191d 0.366 one cut into two; each a σύμβολον; Eros makes ἓν ἐκ δυοῖν
and ἰάσασθαι τὴν φύσιν τὴν ἀνθρωπίνην
Rep 395b 0.237 κατακεκερματίσθαι ἡ τοῦ ἀνθρώπου φύσις → οἱ αὐτοὶ οὐ δύνανται
Rep 473d 0.509 εἰς ταὐτὸν συμπέσῃ, δύναμίς τε πολιτικὴ καὶ φιλοσοφία;
those travelling χωρὶς ἐφ᾽ ἑκάτερον excluded by necessity
Symp 223d 0.996 τοῦ αὐτοῦ ἀνδρός — the same man must hold both arts
The lexical link between the first two is exact: Republic 395b says ἡ τοῦ ἀνθρώπου φύσις is minted into pieces; Symposium 191d says Eros attempts ἰάσασθαι τὴν φύσιν τὴν ἀνθρωπίνην by making one from two. Same subject, opposite verbs.
Republic 473d demands the inverse of κερματισμός and is introduced as what will drown its speaker in γέλως — in the book that has just ruled no one can hold comedy and tragedy together.
Not established: that any of this concerns authorship. Both recombinations are stated of their own subjects — political power and philosophy in a ruler, two poetic arts in a maker. What the corpus supplies is the structure, not its application.
Added 2026-08-29. sha256 51be41d0b6ee194c7a65654f96387395…
A separate document assembles the M32 loci into an argument and states where
heteronymy would sit in it. Recorded here as a pointer, not a finding.
The move: the σύμβολον is not a broken object but an object broken by agreement — the break line is the credential, not the damage. Rubble cannot be recognised; a token is recognised precisely at its fracture. So: you cannot un-cut human nature, you can cut it on purpose. Heteronymy would convert κερματισμός from a wound into a method — the fragmentation ratified rather than reversed, and thereby made legible.
Why it resolves 395b against 473d: it accepts the denial as premise (no single undivided nature holds two arts) and satisfies the requirement in a coordinated multiplicity whose parts sum rather than resemble. ---
Added 2026-08-29. sha256 cb14e25a66fb30d36f6964b1e2c7721c… Pointer, not a finding.
The standing objection was that antiquity built an authenticity apparatus and would therefore have caught a construction. The hypothesis inverts it: the technology is the output. Criteria must be fitted to a case whose answer is known, and a deliberately structured corpus is a calibration object — the Pinakes with incipit and stichometric checksum, the obelus, the νόθος/ἀντιλεγόμενα vocabulary, and the practice of transmitting corpora with their disputes are what working on one produces.
It reframes two measured results. M11's authenticity signal at Z = +3.47 was flagged circular because Andronicus sorted partly by style; under this hypothesis the circularity is the datum. And M30's naming-gap reflex reads as a cataloguer's discipline rather than a school habit — an unnamed class is an unshelvable class, and the Poetics' unnamed-genre problem is a cataloguer's problem.
Its danger is stated in the document, at length. It explains everything, including every null; both hypotheses predict the Pinakes. §4 exists so the property cannot operate silently.
The one asymmetric prediction: an instrument built on a calibration object overfits it. The registered test measures the history of judgements, not the texts — verdict density, decisiveness (stated grounds), and reversal rate, across the Platonic-Aristotelian, Hippocratic, Homeric, Pythagorean and Orphic traditions. It counts verdicts rather than measuring similarity, so none of the five broken nulls (§8) applies. Kill condition binding on decisiveness; void condition if survival cannot be normalised against citation volume.
Added 2026-08-29. The cleanest single measurement in this register: one name against its own placeholder-twin, inside one corpus, needing no register control, no genre matching and no constructed null.
The bearers of the snub in Aristotle, counted across all 46 σιμ- tokens by
what stands within ±30 tokens:
ῥίς nose 102 ΚΑΛΛΙΑΣ 5
σάρξ flesh 14 ἄνθρωπος 4
ὀφθαλμός eye 5 ζῷον, σκέλος 3, 3
χαλκός, κύκλος 1, 1 ΣΩΚΡΑΤΗΣ 0
Σωκράτης appears 144 times in Aristotle — 45 of them in the Metaphysics, the very work where the snub is defined five times — and nine times paired with Callias as the stock placeholder individual: οὐ γὰρ ἄνθρωπον ὑγιάζει ὁ ἰατρεύων ἀλλὰ Καλλίαν ἢ Σωκράτην (Metaph. 0.004); ὁδὶ μὲν ὁ ἀριθμὸς ἄνθρωπος, ὁδὶ δὲ Σωκράτης, ὁδὶ δὲ Καλλίας (0.087).
Zero within forty tokens of any σιμ-, anywhere in the corpus.
The base-rate test says nothing, and must be stated first. Expected
Socrates-near-snub if the two were independent: 144 × 46 × 80 / 960,910 = 0.55.
Observed 0. P(0 | Poisson 0.55) = 0.58. On raw base rate the absence is
unremarkable, and a reviewer will reach for this first. Correctly.
The right comparison is against his co-placeholder, because Callias and Socrates are the same kind of item in Aristotle's apparatus — the interchangeable concrete individual, named together in the pattern Καλλίαν ἢ Σωκράτην:
Callias in snub contexts 5 / 35 = 14.3%
Socrates in snub contexts 0 / 144 = 0.0%
Fisher exact, two-sided p = 0.00022
At Callias's rate, expected 20.6. Observed 0.
Both readings, and the innocent one is not weak. Socrates already carries the snub as a biographical fact, so using him would import unwanted associations into a technical example; Aristotle picks the neutral twin because Socrates is not neutral. That fully explains the distribution.
But it concedes the premise: Socrates is the one individual whose snubness is unavailable for use as a variable. Everywhere else in Aristotle he is exactly a variable — the man who is pale, or sick, or an instance of number — freely interchangeable with Callias. At the snub, and only there, he stops being substitutable. And the mark he cannot be substituted at is his own face.
The Platonic side, for the record. Plato's three snub instances are about naming, not defining. Theaetetus 209c requires that this snubness, differing from all others seen, διάφορόν τι μνημεῖον παρ᾽ ἐμοὶ ἐνσημηναμένη καταθῆται — deposit a memorial in another mind — and it fails to. Republic 474d is a renaming: the lover calls the snub one ἐπίχαρις, charming, and the hooked one βασιλικόν, kingly. Plato names; Aristotle defines; and the bearer changes hands between them.
Status. A measured, non-circular exclusion at p = 0.00022, with a sufficient innocent explanation attached. Not evidence of design. It establishes only that the corpus treats one individual's snubness as unusable where every comparable individual's is usable.
Added 2026-08-29. Notebook Round 15 carries the passages; this is the lookup form.
Full census of the authenticity 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
ψευδεπίγραφος 0 0
προσωπεῖον mask 0 0
Two nulls. ψευδεπίγραφος is zero in every seated corpus — Plato, Aristotle, the nine orators, Philo, Josephus, the NT. There was no word for a falsely-titled book. προσωπεῖον, the theatrical mask, is likewise zero in both corpora — the mask is not their tool for authorial voice, which constrains any persona reading.
The division of labour — ⚠ SUGGESTIVE, NOT MEASURED. Plato leans to the image vocabulary (εἴδωλον, μίμημα, σύγγραμμα); Aristotle to the detection vocabulary (πλάσμα, φάντασμα, ὑποκριτής); δόκιμος is shared at 28/24. These are raw counts. Establishing a division of labour as a quantitative result requires semantic disambiguation of each family, normalisation, a frozen vocabulary set fixed before counting, and comparison corpora. None of that has been done, and the phrase "division of labour" is used descriptively here, not as a finding.
The frame is kinship, not inscription. In Isaeus — γνήσιος at 23.72/10k, the densest in the sample — legitimacy is litigated by μαρτυρ- (279), ἐγγύη (91), εἰσάγω (45) and φράτηρ (35), with zero blood-vocabulary. A νόθος can be the father's certain child. Legitimacy is standing, not paternity, and it is re-litigable on testimony. Applied to texts, the ancient mechanisms were asking does this work have standing in the corpus, not whose hand wrote it.
M35a — the coin runs continuous. Rep. 395b: κατακεκερματίσθαι sits four lines from μιμήματα — nature minted into denominations too small to imitate many things well. Rep. 507a: εὐλαβεῖσθε … κίβδηλον ἀποδιδοὺς τὸν λόγον τοῦ τόκου — the account handed over may be debased coin, with τόκος and ἔκγονον joining coin to kinship in one clause. Laws 742a: currency ἔντιμον to citizens, ἀδόκιμον to everyone else — standing-not-paternity, in coin.
M35b — NE 1165a. δίκαιον ἐγκαλεῖν τῷ ἀπατήσαντι, καὶ μᾶλλον ἢ τοῖς τὸ νόμισμα κιβδηλεύουσιν, ὅσῳ περὶ τιμιώτερον ἡ κακουργία — Aristotle grades false personation above coin-forgery on one scale.
M35c — orator control. The orators are professionally fluent in forgery (πλάσμα, πλαστόν, σκευώρημα, of opponents' documents) and say nothing about the standing of the speech being read: σύγγραμμα is zero in seven of nine, the exception being Isocrates, who wrote for readers.
*Added 2026-08-29. Supersedes the same-day claim that school transmission was
confirmed.*
The blocker was false. M30's diagnosticity was recorded three times as blocked because Theophrastus "survives only as Characters." That was an assumption, never a search. First1KGreek was checked out locally all day and contains Historia Plantarum, De causis plantarum, De sensibus, De igne and the Metaphysica — now seated as EA-CORPORA-09/01, 15 works, 208,324 tokens.
The numbers. Theophrastus 5.43/100k; high-division works carrying the reflex 2/5 = 0.400 against Plato+Aristotle 16/32 = 0.500; Fisher exact p = 1.0000. Historia Plantarum alone carries nine instances at division 29.9/10k.
⚠ AND THE RESULT IS VOID AS A CONTROL RESULT. On seeing it I reported that school transmission was confirmed. That was inherited inference, not measurement, and it violates M33 — I used a figure as a control without first asking whether he can be inside the configuration.
Theophrastus is the Chaerephon shape. The Chaerephon problem, as recorded in the Aristophanes seat: the Suda's entry on Chaerephon is sourced from the scholia to Clouds 144 — the biographical record is downstream of the text it is meant to attest. Theophrastus is defined by his relation to Aristotle (successor as scholarch, heir to the library and manuscripts by Aristotle's will), and his major works are continuations — HP extending HA, De causis extending the causal analysis, his Metaphysica a fragment picking up where the aporiai stop. That is the profile of another completed abandonment, which is what EA-COMPLETE-01 exists to detect. And the Aristotelian corpus reaches us through Theophrastus's library — Neleus, Scepsis, Apellicon, Sulla, Andronicus — so the two texts share one transmission channel and their agreement in editorial habit is not independent.
Seat stands as a corpus. ⚠ CORRECTED PHRASING: the measurement stands; its use as an authorship discriminator is void. Conditional on the received succession account, Theophrastus's hit is evidence that this procedural reflex can propagate down an intellectual lineage — which matters to H_S even though it does not separate H_S from H_H.
M36a — the control criterion, stated. A control must be **impossible to be
inside the configuration, which requires either (i) no causal contact** — a
corpus that could not have received the method — or (ii) **independent attestation
of the author's existence** by sources outside the configuration.
→ ⚠ See M39. I stated criterion (ii) and did not notice that exactly one
member of the candidate cluster satisfies it: Aristotle, by FD III 1, 400.
Socrates, Plato and Aristophanes do not. Theophrastus
fails both. Demosthenes passes (ii) via inscriptions and decrees. **The Hippocratic
corpus passes (i)** — largely pre-Platonic, Coan and Cnidian, medical rather than
Academic — and stands at 0 of 15, the one control nobody can absorb.
**M36b — the reflex is not identifiable from presently available within-tradition
controls.** *(Originally written "undecidable in principle" — too strong, corrected.
"In principle" would require excluding external, cross-cultural, documentary and
later known-author calibration routes as well, which has not been done.)* The operation requires
division, and in Greek prose division is Academy-descended. **Any corpus that
performs the operation is downstream; any corpus that is not downstream does not
perform it.** If that holds, M30's diagnosticity is not merely blocked pending a
corpus — it is undecidable, which is what EA-UNRES-01 predicts an engineered
object produces. Recorded as a conjecture requiring its own test, not as a result.
*Added 2026-08-29. Records a proposed discriminator, its failure at the level
first tried, and what the existing data already says at the level that works.*
The proposal. Inheritance should produce redundancy — the student possesses what the teacher taught, so overlap is high and the union barely exceeds either part. Deliberate partition should produce the opposite: small intersection, union far more complete than either part. Formally, coverage of the union over redundancy between the parts.
It fails on content vocabulary. Top-1500 content words, redundancy =
| A∩B | / | A∪B | , union gain = | A∪B | /max( | A | , | B | ): |
pair redund union gain
PLATO | ARISTOTLE 0.340 1.493
Isaeus | Demosthenes (succession) 0.310 1.527
Herodotus | Thucydides (independent) 0.198 1.669
Aristotle | Theophrastus 0.323 1.511
Plato | Xenophon 0.302 1.537
Philo | Plotinus (independent) 0.296 1.543
Everything lies in 0.20–0.34. Documented succession and documented independence are indistinguishable, and Plato–Aristotle is the most redundant pair in the set — the opposite of the partition prediction. Content vocabulary tracks shared subject matter, so the measure reads topic overlap and nothing else. Do not rebuild it at this level.
And the object is level-dependent, which is the finding.
At the slot level the pair is REDUNDANT. Each fills 6 of 6 (Round 12). Union gain 1.0. That is inheritance behaviour. The Johannine group is the one that partitions — six slots across three books, no slot doubled, union gain 3.0 over its parts.
At the execution level within a shared function they are COMPLEMENTARY (M35):
εἴδωλον 72/13 · μίμημα 54/7 · σύγγραμμα 37/3 ← Plato: representation
πλάσμα 16/32 · φάντασμα 31/47 · ὑποκριτής 10/32 ← Aristotle: detection
δόκιμος 28/24 ← shared: the assay
Statement of the object, at the size the evidence carries: redundant in which functions they perform, complementary in how they execute them. Both cover the whole apparatus; each holds a different end of the operation. That is not tiling and is a different signature from the Johannine partition. It needs its own specification before it gets a measurement, and no measurement of it is registered here.
M37a — Theophrastus was never logically decisive. M30's "decisive successor control" language overstated what any such control could do, independently of M36's exteriority problem. A hit is explained by school transmission and by an extended configuration — if one configuration were demonstrated, the prior on further differentiated positions rises. A miss is explained by four ordinary considerations: students need not inherit every reflex; doctrine can transmit without craft; genre can suppress; the surviving Theophrastean material may sample the wrong register. Neither outcome discriminates. The language is corrected.
M37b — the design that replaces it. Fix labels outside the phenomenon under measurement. An externally secured different-person control requires evidence not derived from style, doctrine, literary tradition or attribution history — autographs, independently dated correspondence, documentary provenance. An externally secured one-person/many-voice control requires the heteronymic relation to be documented rather than inferred, which is why Pessoa is valuable. Then the ancient objects — Plato, Aristotle, Xenophon, Aristophanes, Theophrastus — enter unlabelled, and the question becomes a classification with the labels fixed elsewhere. This moves the methodological problem out of the corpus instead of shifting the arbitrary boundary one name outward.
M37c — the inferential standard. No single observation need distinguish school from configuration; that standard is too strong for historical inquiry. The contest is comparative likelihood over a bundle: substrate divergence, functional tiling, arbitrary procedural reflex, negative exclusion (M34), completion structure, the counterfeit/personation apparatus (M35), authorial self-erasure. School transmission may accommodate each item singly. The question is whether it predicts their joint configuration, and how many auxiliary assumptions it accumulates in doing so — against a heteronymic hypothesis carrying an enormous prior penalty that must be earned by compression: does one latent configuration explain many otherwise independent peculiarities with fewer auxiliary assumptions?
Added 2026-08-29. Methodological standing, extending M37c. The contribution is not a measurement; it is an accounting rule that the investigation has been applying asymmetrically, by me, all day.
The asymmetry, stated. Inheritance has been permitted to be a very high-capacity explanation; unity has not. Every structure this investigation found was absorbed by the received model and the absorption was never charged:
same arbitrary procedural reflex → school habit
same six-slot functional architecture → inherited programme
Plato abandons, Aristotle completes → student solves master's problem
Aristotle catalogues Plato's method → student catalogues master's errors
as definitional fault (Topica ×4)
complementary vocabulary and roles → specialisation
substrate separation at Z = +39.18 → genre, register, object
Each is a good explanation. Several are almost certainly true. But the model being invoked is no longer P → A. It is P → A plus a selective-transmission account in which the student retains some arbitrary reflexes and not others, radically changes substrate, picks up specifically abandoned rather than developed problems, completes them with apparatus the master lacked, preserves the master's figure while completing what it failed at, distributes rather than repeats functions, and negates inherited formulations while preserving the operations beneath them.
⚠ THE ORIGINAL LEDGER WAS WRONG IN TWO WAYS AND IS REBUILT BELOW. It listed ten items as ten independent auxiliaries, when several flow from one latent assumption; and it charged only H_S, which is the same asymmetry this entry exists to correct, merely reversed.
REBUILT — grouped by the smallest latent assumption that generates each set.
H_S — inheritance. Latent assumptions, not observation-counts:
L1 Aristotle is an unusually close successor systematically working through
UNRESOLVED Platonic problems.
→ generates: the abandoned-execution selection, completion by apparatus
the master lacked, retention of the master's figure at its failure
point, and the six-slot co-filling. [was A3, A4, A5, A8]
L2 Arbitrary procedural habits transmit within a school alongside method.
→ generates: the naming-gap reflex (M30). [was A1]
→ SUPPORTED: Theophrastus carries it (M36), which is direct evidence for
L2 conditional on the received succession.
L3 Register and object diverge sharply between dialogue and treatise.
→ generates: the Z = +39.18 partition, and much of the vocabulary split.
[was A2, part of A6]
L4 A successor specialises where the master did not.
→ generates: the representation/detection lean (M35). [was A6]
L5 Local causes, individually unexplained by L1–L4.
→ the Socrates/Callias exclusion at p = 0.00022 (M34); Phaedo 59b; the
authorship denial at Letter II 314c. [was A7, A9, A10]
Five latent assumptions, not ten auxiliaries. ⚠ AND THE STATUS LINE WAS WRONG.
The original read "L1–L4 are ordinary and L2 now has direct support." Corrected:
L1 successor completes abandoned executions UNMEASURED
L2 arbitrary procedural reflexes transmit
across SECURELY DISTINCT persons UNMEASURED
(Theophrastus shows the reflex extends one position further into the
Peripatetic lineage; he is not externally secured, so this is not
evidence of transmission between distinct makers — M36)
L3 register/object divergence produces
substrate separation PARTLY SUPPORTED
L4 successors divide functional labour
complementarily UNMEASURED
L5 local causes for M34, Phaedo 59b, Letter II UNSPECIFIED
A plausible auxiliary is not an empirically supported auxiliary. Calling L1–L4
"ordinary" imported common sense as a denominator. H_S retains an enormous
prior advantage; that does not license inflating P(E | H_S) without
measuring it.
H_H — configuration. Its own ledger, which must be equally ugly:
K1 Chronology / prescience. ⚠ WITHDRAWN 2026-08-29 — see Notebook Round 18.
Reading ξυνίεσαν at APr I.31 as inclusive of Plato makes the second
position a DISCOVERY of the operation's range rather than the execution of
a plan. No prescience, no anticipation, no chronological adjustment
required. This was the ledger's most expensive item.
K2 A historical-person architecture. What the attested Aristotle IS relative
to the authorial position — patron, writer, appropriated name, occupied
role — must be specified, and is not.
K3 Why externally transmitted biographical distinctions exist at all, and in
the detail they do.
K4 Why expected signatures are ABSENT: no προσωπεῖον vocabulary (M35), no
ψευδεπίγραφος, and M16's Aristophanes result requiring reinterpretation
rather than prediction.
K5 The documentary machinery required to sustain the partition across
centuries of transmission, cataloguing and athetesis.
Neither ledger is short. That is the point. The contest is which set of latent assumptions is smaller and less tailored — not which model can accommodate the observations, since both can.
M38a — the null's kill condition is missing, and this is the real defect. This investigation has pre-registered kill conditions for the heteronymic reading repeatedly, and they have fired: EA-LEAK-DEFICIT-01 spent, -02 void, EA-YIELD-01 spent, M30's Poisson withdrawn, the snub position refuted, Poetics namelessness refuted, my own completion verdict reversed twice. The ledger is built so failed predictions stay attached.
No equivalent has ever been demanded of the received model. So it must be
asked directly:
What observation would make "Plato influenced Aristotle" insufficient?
And the worrying property is that the received model assimilates all four
directions of evidence without adjustment:
different style → different authors
same method → teacher transmitted it
contradiction → the student rejected the master
continuity beneath → the student developed the master
the contradiction
Difference, similarity, contradiction and continuity are all absorbed. That does not make the model false — a mature historical model should explain many kinds of evidence — but a hypothesis that no configuration of evidence can strain has stopped discriminating and become the grammar in which results are narrated.
I do not currently have a kill condition for H_S, and I could not construct one when I tried: the best candidate — that Aristotle faults Plato for coining (Topica) while stating coining as his own policy (NE 1108a) — is absorbed by both models. That absence is recorded as a defect of the investigation, not resolved.
M38b — the correct contest. Not can inheritance tell some story about E? —
it can, and elastic hypotheses always can. The contest is:
P(E | H_S) versus P(E | H_H)
with the likelihood under H_S not assessable by inventing the best inheritance
story after seeing each datum, since that is P(E | H_S, A₁…Aₙ) with every Aᵢ
tailored to its observation. The auxiliaries must be charged.
And H_H is charged in the opposite direction: an enormous prior penalty, because the record presents Socrates, Plato and Aristotle as distinct persons in succession. It can only earn that penalty by compression — one latent configuration making many otherwise independent peculiarities consequences of a single construction.
H_S : historical prior advantage, auxiliary-expensive
H_H : prior-expensive, compression advantage
The question the paper has to answer is which model makes the observed total architecture less surprising after both are charged for everything they assume — and specifically why two extraordinarily well-separated textual containers behave as unusually complementary machinery. Teacher and student is an answer to that. It is not automatically a complete one.
Status. An accounting rule, not a result. Recorded because the investigation applied the asymmetry silently toward the received model for most of the day — and then, on correcting it, began applying the mirror-image asymmetry toward the configurational reading (see M39, M40b, M40c, all corrected).
The governing danger, stated as the rule this entry now enforces: do not cure an unfalsifiably elastic H_S by making H_H equally elastic. The required architecture is: nothing simply assumed, nothing simply discarded, every datum retains provenance, both models pay for auxiliaries, and both models must be capable of losing.
Added 2026-08-29. A correction to how M36a's own criterion was applied. I stated the criterion and did not notice that exactly one figure in the candidate cluster meets it.
**M36a requires of a control: independent attestation of the author's existence by
sources outside the configuration.** Applied to the cluster itself:
SOCRATES wrote nothing. Attested through Plato, Xenophon and
Aristophanes — all three INSIDE the cluster (M33). The
Chaerephon problem in its purest form: the biographical record
is downstream of the texts it is meant to attest.
PLATO attested through his own corpus and the tradition that corpus
generated. Letter II 314c denies the corpus is his; Letter XIII
transmits its own athetesis. No external documentary anchor.
ARISTOPHANES produced under other men's names for a decade (T9); the civic
didascalic record carries the PRODUCER. Biographical facts are
extracted from plays whose named author may be another man.
XENOPHON held apart as a candidate position (M33).
─────────────────────────────────────────────────────────────────────────
ARISTOTLE FD III 1, 400 (ante 329) — a contemporary Delphic decree
crowning Aristotle and Callisthenes. Epigraphic, non-literary,
outside the configuration. Passes M36a.
⚠ CORRECTED. Established: Aristotle is presently the only candidate position with an identified external epigraphic anchor. The stronger form originally written here — that he is the only member with such certification — is a negative documentary claim (¬E_Socrates ∧ ¬E_Plato ∧ ¬E_Aristophanes ∧ ¬E_Xenophon) requiring an exhaustive survey that has not been performed. ∃E_A is established; the negations are not.
And a second correction: the stone anchors the PERSON, not textual paternity. Aristotle existed ⇏ Aristotle authored the Aristotelica. Under any configurational reading a real Aristotle could be patron, recipient, collaborator, appropriated identity or named position. The inference from certification to authorship does not run.
This inverts the frame the investigation has been using. Aristotle has been treated throughout as the anomaly requiring explanation — the heterogeneous corpus (M10, dissolved by M28), the excluded twin (M34), the completions (Round 13). He is the anchored position. On any configurational reading he is the terminus: the one member certified to have been a person.
M39a — what the completions then say. M35 established that ancient legitimacy is standing, not paternity — conferred by acknowledgment and witness, and re-litigable on testimony. Under that frame, completing what a predecessor left undone is how an heir takes the estate. Aristotle finishing the abandoned wax block with ἄνευ τῆς ὕλης, converting aporia into method, supplying the receiver's physiology at De mem. 450a — each is a claim to standing over the corpus completed. Not I succeeded Plato but what Plato left undone is mine and I can discharge it.
Methodologically, the completions assert that Aristotle is the authentic Plato.
M39b — and this does NOT discriminate. A student who completes his master's unfinished work is equally the legitimate heir, and both readings license the same claim. This is M38's problem arriving in a new location, and it is recorded as such rather than resolved.
What is not symmetric between the readings is the anchoring. Under inheritance the certification is incidental — of course the latest of the three has the best documentation; he was a court figure and a compiler of civic records. Under the configurational reading it is structural: the certified position is the one that must be real, and it is the position that ends holding everything.
M39c — T12's recursion, sharpened. The single external anchor in the entire configuration is FD III 1, 400 — and it crowns Aristotle for compiling the register of Pythian victors. The one contemporary document certifying the existence of any figure in the cluster certifies the man who built the attribution apparatus, for building it. The Athenian didascalic records descend from his lost Didascaliae; that record by convention sometimes names the producer rather than the poet (T9, T12).
Status. A documentary asymmetry, verified against the record, with its non-discriminating character stated. It licenses no authorship conclusion. What it changes is which position the investigation should treat as given: not Plato, whose corpus denies its own authorship in his name, but Aristotle, who has a stone.
Added 2026-08-29. A cross-reference entry: today's coin census (M35) is the classical substrate of an archive framework built downstream of it. Recorded so the two are not developed separately again.
What the archive already held. Deposit #642 established operative numismatics: a token compresses image, superscription, issuing authority, value, and the power to include or exclude from circulation — with two rival regimes, Rev 13's mark (externally authorized, publicly calculable, coercive) against Rev 2:17's ψῆφον λευκήν bearing a new name ὃ οὐδεὶς οἶδεν εἰ μὴ ὁ λαμβάνων. Deposit #610 gives compressed portraiture and the mint's design surface; #126 gives the inverse principle; #165 the Pergamon counter-archive.
What today supplies that those lacked. The framework was built on Revelation and on modern currency. Plato and Aristotle already run the assay vocabulary on λόγοι and on persons (M35): κίβδηλος 13/4, δόκιμος 28/24, παράσημος 1/4, with Rep. 507a handing over a possibly κίβδηλον λόγον, Laws 742a legislating a currency ἔντιμον within and ἀδόκιμον without, and NE 1165a grading personation above coin-forgery. Operative numismatics has a classical layer and was constructed without it.
Three joints, recorded:
(a) Rev 2:17 is the σύμβολον in token form. A stone whose name is known only to the receiver is authenticated by reception, not by public calculability — the same structure as M40's claim authenticated by performance rather than declaration, and as Letter XIII's σύμβολον, manufactured by agreement rather than found.
(b) #126's inverse principle is M35's standing frame, independently derived.
The surface stamp is institutional; the underlying inscription is operative — γνήσιος-as-standing
against γνήσιος-as-paternity, reached from modern academic publishing rather than
from Isaeus.
(c) The Scepsis books do double duty, and neither deposit knew of the other. #165 reads the concealment of the Aristotelian books at Scepsis as preservation under institutional pressure. M36 establishes the same chain from the opposite end — Theophrastus's library, Neleus, Scepsis, Apellicon, Sulla, Andronicus — and uses it to void Theophrastus as a control, because the Aristotelian and Theophrastan texts share one transmission channel. Same event: in one place a model of preservation, in the other the reason no Aristotelian control can be independent.
Also noted: deposit #701, "Snub-Poemed" — a portrait of Socrates made of the questioner's own text, titled on the fusion of snub-nosed with being poemed — predates today's finding that Plato's snubs are namings and Aristotle's are definings (M35, M34).
*Added 2026-08-29. An interpretation, marked as one throughout, of a sentence
whose documentary content is already established at §0A.1b.*
The sentence. APr I.31: the users of division
οὔτε ὅ τι ἐνδέχεται συλλογίσασθαι διαιρούμενοι ξυνίεσαν
did not understand what can be syllogized by dividing.
What is established (§0A): the complaint is not that they had no apparatus. It concedes an apparatus and faults its scope — they possessed the operation and did not grasp its range of possible conclusions.
The reading. Under the configurational hypothesis the sentence is about the range of objects division can be applied to, and the object they missed is the one running the division.
Plato's divisions in the four naming-gap works are run on capacities and positions, not only on kinds: the Sophist on the appearer-who-is-not, the Statesman on the self-commanding as against those relaying ἀλλότρια νοήματα (M31), the Phaedrus holding division ἵνα οἷός τε ὦ λέγειν τε καὶ φρονεῖν — in order to be able to speak and to think.
And the companion finding (Notebook §63) supplies what division of that object yields: De An. II.2 asks whether the parts of soul are χωριστὸν λόγῳ μόνον ἢ καὶ τόπῳ — separable in account only, or also in place — and I.4 records that divided living things live, τὴν αὐτὴν ψυχὴν ἔχειν τῷ εἴδει, having the same soul in form.
So on this reading the missed conclusion is: a soul can be syllogized by dividing — separated κατὰ λόγον into positions that are really distinct, really alive, the same in form, and not several substances. One in actuality, many potentially. Heteronymy would then be the actualization of a division the practitioners performed on everything except themselves.
Why the fit is exact where it is exact. It answers the Republic 395b / Symposium 223d problem with a mechanism rather than an image (EA-KERMA-01): not parts that fail to resemble but fit, but one actuality divided by account. And it makes the complaint specific rather than generic — not they divided badly but they did not see what this operation could conclude about the one performing it.
⚠ STATUS — INTERPRETATION, AND THE WEAKEST LINK IS NAMED. Aristotle's stated complaint at I.31 is technical: division begs what it must prove (αἰτεῖται) and takes the universal as middle term. He gives his own reason, and it is not this one. The configurational reading supplies a different content for "what they failed to see" than the content Aristotle supplies four lines later.
That is a real cost and it is charged here: the reading requires either that the technical complaint and the configurational one are compatible, or that the stated reason is the exoteric form. Neither is demonstrated. Recorded as reading, at the same epistemic level as EA-KERMA-01, under the standing rule — the corpus supplies the structure; it does not perform the application.
**⚠ CORRECTION 2026-08-30. The title of this measure claimed the control problem
was SOLVED. It is not, and the error is M36's Theophrastus problem at one remove.**
M36a demanded a control impossible to be INSIDE THE CONFIGURATION. That is two
requirements, and this entry satisfied one while claiming both:
PERSON-DISTINCTNESS SATISFIED. Canonization proceedings, chapter
records and university registers attest Albert
and Thomas as distinct men, on evidence with no
relation to textual style.
TRANSMISSION-INDEPENDENCE NOT SATISFIED. Albert and Aquinas are
COMMENTATORS ON ARISTOTLE. Their procedural
apparatus IS the apparatus under test. They sit
downstream of the object.
Consequence for the Round 24 result. That the naming-gap reflex fires only where Aristotle's text is being glossed (Albert 1.31, Aquinas 3.72 on the commentaries; 0.00 and 1.07 in independent composition) remains a valid finding about whether a reflex transmits between documented distinct persons. It is not an independent observation about how procedural shaping propagates in general, because the corpus measured is shaped by the very object whose shaping is in question. The object was measured with a copy of itself.
AND THE ANALOGY FAILS WHERE IT IS LOAD-BEARING. Aquinas's object-shapedness is DECLARED COMMENTARY — announced, titled, structurally explicit. Aristotle's shaping by the Platonic material is presented as independent inquiry carrying an explicit priority denial (§0A). Using the announced case to normalise the unannounced one assumes the relation under test.
**One may not establish that object-shapedness of discourse is ordinary — and
thereby discount the extreme object-shapedness of the Aristotelian corpus — by
pointing to a commentator on Aristotle.**
WHAT SURVIVES: the scholastic pairs remain the best available control for person-level questions — does a reflex cross between two documented men in a teaching relation. They do not license any claim about what degree of object-shaping is normal, and every such claim resting on them is withdrawn.
WHAT IS STILL NEEDED: a control performing heavy division-and-classification whose object is not the Aristotelian corpus and which stands outside its transmission. No candidate has been identified. The negative class M36a demanded remains empty.
Original entry follows, with its title's claim withdrawn.
*Added 2026-08-29. The externally-secured negative class M36a demanded and no
Greek figure could supply.*
Why no Greek control worked. M36a requires a control impossible to be inside the configuration — either no causal contact, or independent attestation of the author's existence from outside. Theophrastus fails both (M36). Xenophon and Aristophanes are held apart as candidates (M33). The orators fail on opportunity: median division 0.5/10k against Plato 6.1, no work clearing M30's threshold (Round 14 §56). Every Greek candidate is either inside the hypothesis space or does not perform the operation.
The solution. Scholastic and religious master–student pairs supply all three
requirements at once:
SECURELY DISTINCT PERSONS attested by canonization proceedings, university
registers, papal bulls, chapter records, monastic
administration — evidence with NO relation to
textual style, doctrinal similarity, or any
succession tradition under test
DOCUMENTED TEACHING master–student relations recorded institutionally
HEAVY PROCEDURAL APPARATUS division, classification, distinction, commentary
— enough opportunity for the reflex to fire
And the null becomes hard rather than easy. The correct negative class is not random distinct authors — Plotinus and Philo barely knew each other, so their nulls are uninformative. It is securely distinct authors under MAXIMAL transmission pressure: a disciple who spent years in the master's household.
Candidate pairs:
Albert the Great → Thomas Aquinas Latin, vast, Aristotelian apparatus
Ibn ʿArabī → Ṣadr al-Dīn al-Qūnawī systematizes rather than repeats
Tsongkhapa → Gyaltsab / Khedrup elaborate inherited procedures
Maimonides → Abraham Maimonides familial + intellectual, maximal case
Luther → Melanchthon separate class: COLLABORATION
Relationship classes to label externally, not by text:
H documented one-person / multiple authorial positions (Pessoa, Kierkegaard)
S documented master–student succession
C documented distinct-author collaboration
I independent authors, same tradition
O documented single-author corpus under one name
**The decisive comparison is H vs S; C, I and O identify what drives the
classification.**
M43a — FEASIBILITY, CHECKED (2026-08-29).
Thomas Aquinas corpusthomisticum.org HTTP 200, structured HTML,
ISO-8859-1, one file per question-block, NO OCR
→ FEASIBLE NOW
Albert the Great archive.org, 222 items incl. the Borgnet Opera omnia
→ PAGE SCANS. Latin OCR risk, the same failure mode that
made Hyperides token-unsafe and Kierkegaard's Samlede
Værker unusable (T14). NOT yet feasible; requires an OCR
quality gate before seating.
Doc. Cath. Omnia HTTP 403 — blocked
So the pair is half-feasible. Thomas is seatable; Albert needs a quality gate. That is a corpus-building task with a known shape, and it is the highest-value acquisition remaining, because it converts "schools transmit procedures" from a story into a number.
M43b — the instrument must be built WITHOUT Plato–Aristotle in view. Derive the measure on the labelled classes; validate out-of-sample that it separates H from S; only then apply it to Plato–Aristotle. If it cannot distinguish Pessoa's relationships from documented teacher–successor relationships, it has failed and says nothing about Plato.
**M43c — ⚠ THE TRANSFER SET CONTAINS NO GREEK CONTENT. A correction to my own
framing.**
I first wrote that the controls should show whether they "reproduce the Plato–Aristotle signature." There is no such thing to reproduce, and the phrase was a category error. The six observables are content-bound to this corpus: six transmission-engineering slots derived from these works, an exclusion about one Greek predicate, a seal-and-wax figure in three named books. Testing whether Albert and Thomas reproduce that would measure whether Albert and Thomas are themselves inside a Platonic configuration — which is the hypothesis extended, not a control.
Fifth substitution of a particular for the operation it instantiates, in one session, and the same error as counting συλλογ- tokens instead of locating the apparatus.
What transfers is the RELATIONAL FORM, stated with no Greek in it:
R1 COMPLEMENTARY COVERAGE — is each side internally tight while the union is
more complete than either? (inheritance predicts redundancy: the student
possesses what the master taught)
R2 ABANDONED-EXECUTION COMPLETION — does the successor complete operations
the predecessor explicitly ABANDONED, rather than extending ones the
predecessor finished? At what rate, over the denominator of abandonments
available to him?
R3 ARBITRARY PROCEDURAL AGREEMENT — shared habits the method does not
require, defined without reference to any particular habit.
Each must be defined on the labelled classes BEFORE Plato–Aristotle is scored, or the definition merely re-encodes the target. The Greek observables then become one instance scored by the general measure, not the template it is cut from.
And the outcomes: if securely-distinct pairs under maximal transmission pressure routinely show R1–R3, the thesis takes a serious hit. If they do not, the Plato–Aristotle values become diagnostic. Both outcomes can hurt something.
Retrieval, verified 2026-08-29. Borgnet Opera omnia via archive.org DjVuTXT — 220 items carry extracted text. Volume 7 (Ethics commentary): 338,391 tokens; volume 19: 356,088.
OCR quality gate, run rather than assumed. Scattered errors — divinitatc, intcr, ct for et, iuebrianiin for inebriamini — but the frequent vocabulary is clean: scilicet 2596, quod 2542, quia 1489, enim 1403, sicut 1000. Verdict: usable for function-word and structural measures; UNSAFE for rare-lexeme measures. Any measure resting on a low-frequency form must be declared unsafe on this seat, as with Hyperides (EA-CORPORA-08).
⚠ AND THE CONTROL HAS A CONTAMINATION THE GREEK CASE ALSO HAD.
Volume 7 carries the naming-gap reflex — innominat- ×13:
qui quidem in timiditate superabundat Graece innominatus est, (multa enim
sunt innominata in mediis et in extremitatibus), Latine autem formidolosus
But that is Albert glossing NE 1107b–1108a — the exact passage where Aristotle writes πειρατέον αὐτοὺς ὀνοματοποιεῖν. He carries the reflex because he is commenting on the text that performs it.
This is the Theophrastus problem in a new form: contamination through a shared source. Albert and Thomas both write Aristotle commentaries. Any Aristotelian reflex appears in both by construction. Scoring R3 on their commentaries would measure the presence of Aristotle in commentaries on Aristotle.
**M43e — THE DESIGN THIS FORCES, and it comes from the feasibility check rather
than from theory:**
COMMENTARY WORKS unusable for R3 (shared-source contamination);
usable for R1 and R2, which are properties of the
RELATION rather than of inherited content
INDEPENDENT WORKS the clean test. Thomas's Summa Theologiae and the
Quaestiones Disputatae are not glosses on Aristotle.
IF THE REFLEX TRANSFERS INTO INDEPENDENT WRITING, that
is transmission. If it appears only where he is
commenting, it is the source showing through.
That split is the instrument. It is available for Albert→Thomas, and it generalises: every scholastic pair must be scored commentary and independent work separately, with the contamination declared on the commentary half.
Added 2026-08-29. The background assumption the configurational reading rests on had never been checked. Checking it produced one solid finding and one correction to my own framing, which was the third instance today of supplying the tradition's own product as external evidence.
Pseudepigraphic composition was extensive, deliberate, and stylistically engineered. The pseudo-Pythagorean authors did not merely attach borrowed names: they gave the writings a Pythagorean veneer to convey authenticity, most notably through an artificial Doric dialect, with locally appropriate examples ('in Tarentum' as the specimen of the where category in Ps.-Archytas, On the Whole System). Constructed stylistic markers, chosen to authenticate.
And the practice reaches into the Academy in the fourth century. Scholars argue that Hicetas and Ecphantus of Syracuse were not historical figures at all, but characters in dialogues by Heraclides of Pontus, a fourth-century member of the Academy. Invented speaking positions, received by the later tradition as real philosophers with doctrines.
It also happened to a Platonic speaker. Timaeus of Locri is one of the two most influential Pythagorean pseudepigrapha; all other ancient reports about him are likely based on Plato's Timaeus or on the spurious works in his name, and some scholars hold that Plato uses him as a mask for Archytas. A speaking position generated a philosopher, who then received a forged corpus.
I first recorded the finding as: the economy is not anonymity by default but adjudication by reputation rather than by law — citing that antiquity lacked copyright yet defended authorship through reputation, rivalry and public judgment, with exposure as the enforcement mechanism resting on expert authority and audience discernment.
That is circular and I offered it as a corrective.
"Reputation" in a manuscript culture is not a fact about a man. It is a fact about texts. The reputation adjudicating attribution is constituted by the corpus, the catalogues, the biographies drawn from the works, the anecdotes generated by the dialogues. Diogenes reports what the tradition preserved; the tradition preserved what the cataloguers admitted; the cataloguers judged by fit with a corpus whose boundary they were drawing. "Expert discernment" means adjudication by readers whose sense of how an author sounds came from the works under dispute.
And the two cases above demonstrate the circularity rather than escaping it. The reputational mechanism did not detect Hicetas, Ecphantus, or the Timaeus of Locri who may never have existed. It manufactured them.
**Third instance today of supplying the tradition's own product as external
evidence** — after Theophrastus (M36) and Speusippus (M40b).
SURVIVES: the practice existed, was extensive, involved deliberate stylistic construction, and was performed by Academy members in the fourth century. These are facts about texts and their production and do not depend on the reputational apparatus being reliable.
DOES NOT SURVIVE: any claim that the culture could or did reliably detect
false attribution.
H_S's prior advantage rests partly on the same apparatus. Not entirely — Aristotle has an epigraphic anchor (M39) — but the succession-lists, the corpus boundaries (T11: Andronicus's inference) and the biographical distinctions all issue from it. That does not collapse the prior, but it means the prior and the evidence are not independent, which is what the comparison in M38b assumes. Recorded as a defect in the comparison, not as a correction to the prior's magnitude.
Thesleff, H. (1965) The Pythagorean Texts of the Hellenistic Period
— the standard collection; Prorus pseudepigrapha at 154.13
Centrone, B., "Authority and Doctrine in the Pseudo-Pythagorean Writings"
— the artificial-Doric strategy, name-selection as intentional
Zhmud, L., "What is Pythagorean in the Pseudo-Pythagorean Literature?"
— Heraclides Ponticus, Speusippus, Xenocrates, Aristotle, Aristoxenus
in relation to the pseudepigrapha; argues secure traces of early
Hellenistic pseudepigrapha attributed to Pythagoras' students are
NOT to be found
Huffman, C., "Pythagoreanism", Stanford Encyclopedia of Philosophy
— Hicetas/Ecphantus as possible Heraclides characters; Timaeus of Locri;
the Archytas-mask proposal
Diogenes Laertius 8.7
— Heraclides Lembus, Epitome of Sotion's Successions: the Mystikos logos
assigned to Hippasus rather than Pythagoras, "written to discredit
Pythagoras"; books by Aston of Croton attributed to Pythagoras
Speyer, W. (1971) Die literarische Fälschung im heidnischen und
christlichen Altertum — the standard treatment of ancient forgery
Baumgarten/Bremmer et al. on ancient pseudepigraphy — to be located
Dillon, J. (2003) The Heirs of Plato — the Academy under Speusippus,
Xenocrates, Polemo; ch. 5 on Heraclides of Pontus
Ehrman, B. (2013) Forgery and Counterforgery — Christian material, but the
taxonomy of ancient forgery-detection is transferable
Geue, T. (2019) Author Unknown: The Power of Anonymity in Ancient Rome
— via BMCR 2020.03.53; Latin side
Love, H. (2002) Attributing Authorship — "Attributionists see the anonymous
or pseudonymous work as a vacuum in nature which it is their moral duty
to fill with an author"
Gow, A.S.F. (1958) The Greek Anthology: Sources and Ascriptions, 20-29
— terminology of unknown authorship
Reynolds & Wilson, Scribes and Scholars (4th ed. 2013)
— transmission mechanics
Pfeiffer, R., History of Classical Scholarship (1968-76)
— Alexandrian cataloguing practice
TO CHECK: whether any ancient source describes a method for determining authorship that does not reduce to stylistic fit with an already-constituted corpus. If none exists, the circularity in 33.2 is not merely a defect in my framing but a structural feature of the period's attribution apparatus.
*Added 2026-08-29. The paper's endpoint, stated as a demand rather than a claim,
with a status column that keeps it honest.*
The question is not can the single-author hypothesis survive the objections but:
Given the observed configuration, what does H_S explain BETTER than H_H?
Every form of the standing objection — a student could inherit this, a successor could complete unfinished problems, a school could transmit the apparatus — establishes only
P(E | H_S) > 0
and never
P(E | H_S) ≥ P(E | H_H).
Could is not a likelihood estimate. Mere compatibility must stop counting as
explanatory superiority.
Polansky's objection — the same results follow without the hypothesis — is answered thus: perhaps; but can follow is not the comparison. The question is whether the observations are expected with comparable probability and comparable auxiliary cost under the multi-author model.
P(H | E) ∝ P(E | H) · P(H)
**The received model gets its prior. It does not get its likelihood automatically
from tradition.** And:
historical default ≠ explanatory model
observation H_H H_S STATUS
─────────────────────────────────────────────────────────────────────────
M40 priority structure literal requires scope ASSERTED
repair both sides
naming-gap reflex same operator requires ⭑ MEASURED
habit transmission of Round 24
arbitrary
procedure
reception capture interpretive later framing ⚠ OCCURS IN
(M46 / M47) vocabulary displaces BOTH CLASSES;
supplied by earlier terms NEITHER MODEL
the other SHOWN TO EXPLAIN
position THE PLATO CASE
BETTER (§36.9)
abandoned-execution continuation of successor ASSERTED;
completion own unfinished recovers a NO DENOMINATOR
operation shaped failure
functional interlock distributed unusually ASSERTED
construction coordinated
inheritance
substrate divergence position ordinary distinct CONTESTED and
differentiation authorship ARCHITECTURE-
DEPENDENT (34.3)
chronology, biographical COSTLY ⭑ ADVANTAGE CONCEDED
tradition, prima facie to H_S
declared partition
The status column is what stops this being rhetoric. As it stands the argument rests on one measured row and one conceded row, with four asserted. The next work is named by the table: convert the four.
And H_S must be allowed its wins, plainly. The received account keeps the chronology, the external biographical tradition, and the prima facie declared partition. A table in which H_H wins every row would be rigged and should be disbelieved.
Round 25 established that Plato–Aristotle closes by sum (complementary
positions, coverage completeness) while Pessoa closes by remainder (progressive
subtraction, invariance under narrowing). **Heteronymy is therefore not a single
generative model, and P(E | H_H) is undefined until the architecture is
specified:**
H_H1 SUM-CLOSURE — differentiated positions whose union is complete;
divergence high, function interlocking; integral = coverage
H_H2 REMAINDER-CLOSURE — nested subtraction, containers tightening with
constraint severity; integral = what cannot be removed
This cuts both ways and both must be recorded. The Pessoa calibration cannot be transferred to the Greek pair as a single expected morphology — that was Round 22's error. But H_S can no longer be scored against a straw H_H either: each architecture has its own predicted signature and must be scored separately.
Round 24 is narrower and sharper than "Aquinas doesn't reproduce the reflex".
Albert (non-Aristotle works) 0.00/100k in 1,140,474 tokens, division 16.4
Thomas (independent works) 1.07/100k in 5,219,700 tokens, division 26.6
→ THE PAIR DOES NOT SHARE IT
Plato 2.60/100k, division 15.4
Aristotle 8.43/100k, division 30.9
→ BOTH CARRY IT
And in both Latin authors it appears only where Aristotle is being glossed
(Albert 1.31, Aquinas 3.72 on the commentaries).
Measured claim: for this reflex, transmission into independent composition did not occur in the one documented master-student pair examined. Caveats standing: one pair, one draw; Latin idiom may differ from the probe set despite the positive control; nine Borgnet volumes of thirty-eight, with the Sentences commentary dominating the non-Aristotle sample.
The received model can no longer discharge its burden by saying *succession
explains it too*. It has to say:
**Here is the specific feature on which documented succession outperforms
documented heteronymy, and here is why the Plato–Aristotle value belongs with
that class.**
If that cannot be supplied, "multi-author" has stopped functioning as an explanation and is functioning as the default label — which earns a prior, and which this investigation concedes without argument.
Added 2026-08-29. A datum recorded earlier without being seen for what it was — an interlocking structure of a different kind from functional coverage, now with an H-class calibration and an S-class control.
One position supplies the interpretive vocabulary for another, and the vocabulary becomes TRANSPARENT — later readers take it for the interpreted position's own.
In the Greek pair (measured earlier, unexplained until now):
ἀόριστος the Indefinite Aristotle 108 Plato 3
δυάς the Dyad 41 4
χωριστ- separation 411 165
ἀριθμ- Forms as numbers 977 246
The doctrines by which Platonism is known are quantitatively Aristotle's.
The heteronyms interpret each other: Mora's Regresso dos Deuses — Prefácio a Caeiro, Campos's critical notes, Reis's preface. Arquivo Pessoa's bibliographic footers were stripped before counting (Caeiro mentions in Campos 186 → 172).
Terms the interpreters use OF Caeiro, against Caeiro's own text:
term Caeiro/10k interpreters/10k ratio
pagan- 0.00 15.88 ∞
paganismo 0.00 15.49 ∞
pagã 0.00 14.04 ∞
sensibilidade 0.00 6.61 ∞
mestre (the master) 0.00 5.89 ∞
objectiv- 0.00 5.79 ∞
sensacion- 0.00 1.21 ∞
────────────────────────────────────────────────────────
natureza 23.48 7.14 0.3
realidade 15.13 11.68 0.8
Every term by which Caeiro is known is ABSENT from Caeiro. Paganism, sensationism, objectivism, the master, sensibility — all supplied by Mora, Campos and Reis. What Caeiro writes about is nature and reality, at three times his interpreters' rate.
**Reception capture is a measured feature of documented heteronymic
construction.**
Does a documented successor supply vocabulary ABSENT from the master?
pair Aquinas/10k Aristotle/10k
esse / εἶναι 54.36 81.68
essentia / οὐσία 13.09 90.51
substantia / οὐσία 14.48 90.51
forma / εἶδος 24.11 27.83
materia / ὕλη 18.93 56.57
potentia / δύναμις 22.37 15.62
accidens / συμβεβηκός 5.55 8.17
participatio / μέθεξις 3.83 3.59
analogia / ἀναλογία 0.29 3.33
quidditas / τὸ τί ἦν εἶναι 0.29 1.23
actus purus / ἐνέργεια 0.14 6.67
NO. Every Thomist interpretive term has an Aristotelian counterpart present in Aristotle, most of them at higher rates in the master than the successor. Aquinas translates. He does not supply vocabulary the master lacks.
Added same day. Running headers stripped before counting (B's title "Ligevægten mellem det Æsthetiske og Ethiske" repeats as a page header; æsthetisk 251 → 243).
B addresses and diagnoses A directly — B.'s Papirer, Breve til A — so the
interpretive relation is internal to the work and documented.
term A/10k B/10k B/A
ethisk 1.33 18.73 14.0
fortvivlelse 1.49 12.64 8.5 (despair)
vælge 2.04 14.59 7.1 (to choose)
valget 0.55 3.84 7.0 (the choice)
æsthetisk 3.93 18.28 4.7
─────────────────────────────────────
umiddelbar 5.26 5.12 1.0
sandselig 6.99 2.93 0.4 (the sensuous)
ironi 2.51 0.90 0.4
forfører 5.73 1.05 0.2 (the seducer)
The aesthetic/ethical framework by which Either/Or is universally read is B's vocabulary. A writes about seduction, the sensuous, irony, immediacy. B supplies the aesthetic, the ethical, despair, the choice — and A is known by B's terms.
Partial capture, not total — A does use æsthetisk at 3.93. So the H class
itself has a range: Caeiro at ∞, Either/Or at 4.7–14.
CAEIRO / interpreters H, documented 0.00 in the interpreted TOTAL
PLATO / ARISTOTLE under test 25-35× thinner PARTIAL
EITHER-OR A / B H, documented 4.7-14× thinner PARTIAL
ARISTOTLE / AQUINAS S, documented counterparts present, NONE
often HIGHER in master
Both documented heteronymic cases show capture. The documented succession case shows none. The Greek pair falls inside the H range — and above the Either/Or values, below Caeiro's.
⚠ THE COMPARISON IS WEAKER THAN THE TABLE LOOKS. The term sets were chosen by me, separately for each case, after seeing each corpus. That is the reverse-engineering hazard EA-UNLOCK-01 §5 exists to score against, and it applies here. A defensible version requires the term-selection rule fixed in advance and applied identically — e.g. the N terms most characteristic of the interpreter's account, chosen before the interpreted position is counted. Until that is done, the ORDERING is suggestive and the MAGNITUDES are not comparable.
MEASURED, and this is a second row for M45's table, joining the naming-gap reflex. It is the first structure with both an H-class calibration and an S-class control, run on the same measure.
LIMITS, stated.
(a) Caeiro is small — 19,328 tokens — and zero counts on a small corpus are
weaker than zero counts on a large one.
(b) The interpreters write prose and Caeiro writes verse. Under Round 21's rule
the genre split is part of the heteronymic operation and not a confound to
remove — but it is not nothing either.
(c) The Aquinas control is cross-linguistic: a Latin term and a Greek term are
matched by translation, and the pairing is mine.
(d) One H case, one S case. Lee's proposed extensions — Borges and his
invented critics, Hegel through Marx's vocabulary, Kant through Hegel's — are
unrun and are the obvious next controls.
NOT ESTABLISHED: that reception capture indicates single authorship. A dramatist's characters could show it; so could any case where a later writer's framing displaces an earlier writer's own terms, which is ordinary in reception history. What is established is that the three cases separate on this measure, in the predicted order.
*Added 2026-08-29. Two controls run, one measure voided mid-run, and a control-class
error committed twice in opposite directions.*
Not do successors supply vocabulary but:
**Do a heteronymic system's ANALYTICAL positions control the terms by which its
NON-ANALYTICAL positions are discussed, at a higher rate than in multi-author
critical pairs?**
The pairing must therefore be CREATIVE ← CRITICAL. This is answerable by reading and the distinction is eyeball-able; it does not require the elaborate apparatus of EA-CAPTURE-01.
The registration specified term selection normalised against a same-language reference corpus. No such reference exists in the archive for Portuguese, Danish or English, so four of five pairs fell back to raw rate while Plato–Aristotle alone got reference-normalisation. T1's 12.20 against the others' 1.15–2.08 is a method artifact: the one pair scored by the better method scored higher. §5 forbids altering the rule mid-run; I altered it for four of five. Run void. A further bug: S1 passed Latin word-extraction over Greek, yielding "interpreted 109 tokens."
Aquinas ← Aristotle ANALYTICAL → ANALYTICAL wrong shape
Marx ← Hegel ANALYTICAL → ANALYTICAL wrong shape
Coleridge → Wordsworth CREATIVE → CREATIVE wrong shape
(Coleridge is a major poet criticising a
collaborator and rival, from inside the practice)
On the Coleridge result I withdrew M46 as evidence. That withdrawal is RETRACTED — it rested on a badly-shaped pair, exactly as the Aquinas control did. The control class is defined by the critic's relation to the PRACTICE, not by the fact of criticism, and I got it wrong in both directions inside one sequence.
First attempt failed on the same defect as EA-CAPTURE-01: drawing "Marx's most distinctive terms" returned commodities, prices, linen, silver, gold — his subject, not his interpretive vocabulary. Median 9.27 on that set is meaningless.
Run on the terms Marx actually uses OF Hegel:
term Marx/10k Hegel/10k M/H
negation of the negation 0.80 0.03 24.2
estrange- 8.33 0.54 15.5
alienat- 6.73 0.92 7.3
material 16.77 3.18 5.3
inver- 1.60 0.43 3.7
mystif
rational kernel ·
mystical shell
─────────────────────────────────────────────────────
dialectic 3.31 3.41 1.0
speculative 1.08 3.46 0.3
consciousness 12.03 41.73 0.3
self-consciousness 4.05 12.00 0.3
absolute 4.68 20.77 0.2
spirit 2.62 25.99 0.1
notion 1.71 16.47 0.1
Marx supplies CRITIQUE-METAPHORS — inversion, mystification, the rational kernel, the mystical shell — and nothing else. The substantive vocabulary of Hegel-reading is Hegel's own: dialectic even at 1.0, and spirit, notion, absolute, consciousness running 3–10× heavier in Hegel than in his most famous critic.
Hegel is read in Hegel's terms.
esemplastic · primary/secondary imagination · ∞ (Coleridge coinages)
poetic faith · willing suspension
objective 1.16 / 0.01 100.1
subjective 0.87 / 0.03 25.0
philosophic 4.48 / 0.22 20.4
diction 5.13 / 0.51 10.1
imagination 4.62 / 1.52 3.0
──────────────────────────────────────────────────────
fancy 3.40 / 3.68 0.9
soul 4.12 / 9.32 0.4
heart 6.86 / 17.67 0.4
Coleridge annexes the analytical register and Wordsworth keeps the poetry — which is two poets dividing labour, nearer the H shape than the S shape. Retained as data, not admissible as an S control.
Caeiro ← Mora/Campos/Reis H, documented key terms 0.00 in Caeiro
Either/Or A ← B H, documented ratios 4.7-14
PLATO ← ARISTOTLE under test ἀόριστος 108/3, δυάς 41/4,
χωριστ- 411/165
HEGEL ← MARX S, distinct, substantive vocabulary stays
analytical HEGEL'S; only metaphors captured
Wordsworth ← Coleridge mis-shaped creative→creative
The one properly-run control supports the observation: a securely distinct critic, however famous his critique, did not take over the terms by which his subject is read. Plato's are Aristotle's.
The clean case: Freud wrote no tragedies, had no rivalry with Sophocles, and the vocabulary by which Oedipus is now discussed is largely his. Freud 253,671 tokens (Dream Psychology, General Introduction, Three Contributions); Sophocles 55,715 (Jebb and Murray, English throughout).
term Freud/10k Sophocles/10k F/S
sexual 62.80 0.00 ∞
neuro- 26.02 0.00 ∞
psychoanaly- 14.98 0.00 ∞
libido 13.68 0.00 ∞
symbol 11.04 0.00 ∞
manifest / latent 10.33 / 4.97 0.00 ∞
infantile 9.89 0.00 ∞
censor 5.12 0.00 ∞
complex · Oedipus complex 3.35 / 0.12 0.00 ∞
dream 110.18 1.62 68.2
unconscious 15.10 0.36 42.1
repress- 5.16 0.18 28.8
────────────────────────────────────────────────────────
incest 1.70 1.44 1.2
mother 6.74 13.82 0.5
father 5.99 22.26 0.3
guilt 0.51 4.13 0.1
fate 0.83 8.62 0.1
blind 0.32 9.33 0.0
oracle 0.16 7.54 0.0
curse 0.12 9.87 0.0
parricide · prophecy 0.00 1.08/1.26 0.0
TOTAL CAPTURE, and it is the strongest in the whole set. Fourteen terms at ∞ — every one of them now standard in discussing the play. And the reverse column is what Sophocles actually wrote about: fate, oracle, curse, blindness, plague, prophecy, guilt — all of which Freud barely touches, several at zero.
**Sophocles wrote a play about an oracle and a curse. It is discussed as a play
about repression and the unconscious.**
**A securely distinct non-practitioner critic produces more total capture than
either documented heteronymic case.** Caeiro had 7 terms at zero; Freud has 14.
So capture is NOT DISTINCTIVE of heteronymy. It occurs in both classes.
On this result I recorded M46 as withdrawn as evidence. **That converted "not
discriminating" into "explained by succession", and those are different claims.**
What the Freud run establishes: the phenomenon is not unique to H. **What it does
NOT address: whether the PLATO–ARISTOTLE instance is better explained by
succession than by configuration.** Nothing in it speaks to that. By M45's own
rule — could is not a likelihood estimate — showing that S pairs also produce
capture establishes P(E | H_S) > 0 and nothing more. **I applied to my own measure
exactly the reasoning M45 forbids applying to H_S.**
And the two cases differ in a way the withdrawal ignored. Freud imports fourteen terms absent from Sophocles and everyone knows they are Freud's — libido is not mistaken for a Sophoclean concept. Aristotle's terms are taken as Plato's own doctrine. That difference is the original observation, and the measure never touched it: it counts distributions, not attribution.
CORRECTED STATUS: *reception capture occurs in both classes; neither model has
been shown to explain the Plato–Aristotle instance better.*
§36.9 proposed that the two cases differ in attribution: Freud's terms are known to be Freud's, while Aristotle's are taken as Plato's own doctrine. Checked against the reception literature, that is false — and false on my strongest example.
The Indefinite Dyad is attributed to Aristotle everywhere it appears:
revealed in Aristotle and other testimonia — Cambridge Elements,
Plato's Unwritten Doctrines
*the unwritten doctrines refer to **various things said by Aristotle and
others***, with (Aristotle, Metaphysics 987b20-22) cited inline
*the expression 'unwritten doctrines' … **was first used by his student
Aristotle*** — Wikipedia
Aristotle attributes to Plato the view that such ideal numbers constitute
substantial Forms
These are among the most explicitly sourced terms in the field. The entire Tübingen–Milan controversy exists because the attribution is visible. The tradition transmits them with their provenance attached.
So M46 does not survive — and it dies for a better reason than the first one. Not S pairs also do it (that was the M45 error, §36.9). But: the specific asymmetry proposed to distinguish the cases does not exist. Both are attributed capture; the reception layer tracks provenance in both.
WHAT SURVIVES IS THE DISTRIBUTIONAL FACT: ἀόριστος 108/3, δυάς 41/4, χωριστ
**Every measure in this sequence failed to discriminate. That is not a null for the
configurational reading alone.**
reception capture occurs in H pairs AND S pairs
attribution tracked in H pairs AND S pairs
presence (mechanical) 90-100% in EVERY pair, no separation
A multi-author corpus does not have the discriminative value the received model assumes it has. The received account treats distinct authors as an explanatory fact from which the observed structures follow. But on the measures run here, the distinct-author cases and the one-maker cases are not separable — Freud on Sophocles behaves like Mora on Caeiro; Coleridge on Wordsworth behaves like B on A; Marx on Hegel behaves like nothing in particular.
If the classes cannot be told apart by the phenomena, then "multi-author" is not doing explanatory work on those phenomena — for either side. It remains the default label (M45), holding a prior it has not converted into likelihood on any of these measures.
The honest summary of the whole sequence: reception capture is real, occurs across classes, and is evidence for neither model — which is a finding about the INSTRUMENT and about the received model's assumed discriminative power, not only about the configurational reading's failure to prove itself.
Lee's proposal: cross-genre, novelist criticised by poet, with the hypothesis that Kerouac supplied "Beat" and Ginsberg supplied most of the rest. Genuinely testable and the right shape — a theorist-poet reading a novelist who theorised as a novelist.
BLOCKED BY COPYRIGHT. Howl (1956) and On the Road (1957) are both in copyright; the corpora cannot be assembled here. Recorded as a blocker, not a refusal — the test is well-formed and runs the moment the texts are obtainable.
Also unrun: Bradley on Shakespeare, Heidegger on Hölderlin.
Added 2026-08-29. The complete numeric record of the class-characterisation programme, preserved so it need not be rebuilt. No discriminatory power is itself a finding, and the tables below are what a later attempt should start from rather than repeat.
1 MASTER-STUDENT DIRECT embodied contact, attested by CONTEMPORARY
DOCUMENTS independent of textual resemblance
2 DISTANT CONSTRUCTED a later writer constructs the relation
3 HETERONYMIC CONSTRUCTED one maker constructs both positions
Admission standard (Round 35). Ancient teacher-student attributions are regularly made centuries later and regularly inferred from stylistic resemblance — Dionysius argues the Isaeus–Demosthenes relation from the likeness of their prose. Using such a pair to calibrate what stylistic likeness means in a teaching relation is circular. Excluded on that ground: Isocrates→Isaeus, Isaeus→Demosthenes.
ADMITTED TO CLASS 1
Wundt → Titchener Leipzig PhD 1892; Titchener translated Wundt
Freud → Jung 1907-13 correspondence (360 letters), IPA presidency
Albert → Aquinas Dominican chapter records, Cologne 1248-52
F1 corpus size ratio (pupil/master)
F2 % of master's 300 most distinctive terms occurring in the pupil
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 delta F7 hapax fraction delta
F8 top-1000 vocabulary Jaccard
F9 uniformity of density direction
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
resampling sd ≤ 1.2 on every cell.
TARGET vs CLASS 1 F2 INSIDE z=−1.06 F3 INSIDE z=−0.67
F4 INSIDE z=−0.32 F5 INSIDE z=−0.66
F8 INSIDE z=−0.87 F9 INSIDE z=+0.29
⚠ 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 and F4. Albert→Aquinas widens it — and that pair is triangulated by the target itself, since both members are commenting on Aristotle.
pair class F1 F2 F3 F4 F5 F8 F9
Isocrates → Isaeus [excluded] 0.27 94.0 0.691 32.6 11.7 0.278 0.674
Isaeus → Demosthenes [excluded] 8.80 96.7 0.665 25.9 3.0 0.384 0.741
Lysias → Isocrates contemp. 2.18 100.0 0.712 32.3 0.7 0.404 0.677
Herodotus → Thucydides contemp. 0.82 76.0 0.705 34.2 22.7 0.244 0.658
Thucydides → Xenophon continuation 2.08 98.7 0.658 29.1 4.0 0.383 0.709
Herodotus → Josephus distant 2.71 79.3 0.493 21.8 23.7 0.230 0.782
Thucydides → Philo distant 3.91 91.3 0.346 20.1 7.7 0.223 0.799
Herodotus → Plotinus distant 1.11 65.7 0.545 33.5 22.7 0.169 0.665
Homer → Apollonius constructed 0.10 85.0 0.567 19.6 5.0 0.286 0.804
Hesiod → Apollonius constructed 2.47 97.7 0.521 27.3 17.0 0.214 0.727
Homer → Hesiod contemp. 0.04 83.7 0.587 23.9 1.5 0.277 0.761
Caeiro → Reis heteronymic 2.54 96.6 0.477 15.5 16.7 0.353 0.845
Caeiro → Campos heteronymic 6.12 99.6 0.609 20.2 9.3 0.380 0.798
Caeiro → Mora heteronymic 2.58 89.3 0.481 23.9 22.7 0.274 0.761
Caeiro → Soares (unrelated) heteronymic 7.31 95.7 0.549 19.6 10.3 0.377 0.804
Either/Or A → B heteronymic 1.03 98.7 0.981 48.0 1.0 0.535 0.520
A → signed heteronymic 0.85 94.7 0.438 10.9 34.0 0.280 0.891
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
n-GRAM DEPTH 2-gram 3-gram
Wundt → Titchener 95.0 88.2
Freud → Jung 100.0 89.9
Albert → Aquinas 93.8 87.2
PLATO → ARISTOTLE 90.9 77.2
COV = % of MASTER blocks with a pupil block nearer than the master's own
median internal spread (2,500-token blocks, top-150 features)
pair COV REV COV @size-matched (contiguous)
Albert → Aquinas 97.5 96.0 0.0
Freud → Jung 92.3 92.7 50.0
Wundt → Titchener 68.6 93.1 25.0
PLATO → ARISTOTLE 66.8 87.5 18.8
[het] Caeiro → Campos 20.0 51.5 20.0
[het] Caeiro → Reis 0.0 30.8 0.0
[het] Caeiro → Mora 0.0 0.0 0.0
(a) SIZE. Downsampling class-1 masters to Caeiro's 20k moved F3, F4, F5 and F9 into the heteronymic range. Wundt→Titchener F5 goes 7.3 → 15.5. Novel- vocabulary rate and direction-uniformity are small-master artifacts.
(b) LANGUAGE. Every Greek pair scores 59–79 on trigram overlap; every non-Greek 87–90. Greek's free word order makes exact n-gram repetition rarer in any pair. The apparent z = −10.35 was language. Within Greek the target is at the TOP of the range (78.6, above Hdt→Thuc 76.8 and Thuc→Xen 72.0).
(c) PERIOD AND DIALECT, NOT RELATION. Lysias→Isocrates — no teaching relation — scores F2 100.0, F5 0.7, F8 0.404, higher on every inheritance feature than Isaeus→Demosthenes, which has one. Herodotus→Thucydides is contemporary and scores like the distance pairs, because Ionic against Attic. And inside the Pessoan system, Caeiro→Soares — two positions with NO constructed relation — scores like Caeiro→Campos, which has one.
(d) A BROKEN CONTROL, RECORDED. The first size-control on COV used random token sampling, which destroys the block structure the measure reads; all four pairs went to 0.0. That was an artifact of the control. Contiguous windows are required for any block-based measure.
Relation-type is not recoverable from lexical distribution. Not pupillage,
not constructed lineage, not heteronymic construction. What these instruments
see is period, dialect, genre, subject and corpus size.
Consequently the target is not distinguishable from documented master-student pairs on any of fifteen features — and documented master-student pairs are not distinguishable from heteronymic construction once size is controlled. The null runs in both directions and is not evidence for either hypothesis.
What it IS evidence for: the whole distributional family is the wrong instrument for this question. M45's table gains no row from any of it, and should not.
Do not rebuild the battery. The numbers are above; the four killers are in
§37.7.
Constraints any new design must meet:
The binding scarcity: Caeiro is 20,027 tokens and there is no more of him. The only documented heteronymic system with an explicit master position cannot supply a size-matched master. Either a second such system is found, or the heteronymic class cannot be characterised at scale.
And the alternative direction, from Round 39 §158: the interlock, if it exists, is propositional — M40's suppressed premise and Round 13's completions — and no distributional measure reaches it.
Πλάτων-/Ἀριστοτελ- counts, which are measurements:
Ἀριστοτελ- in the whole Platonic corpus 4 (all Parmenides, the
young man who joined
the Thirty — a
different person)
Πλάτων- in Aristotle within ±14 tokens of a 1 (an anecdote about
learning/teaching word ARISTIPPUS, not
autobiographical)
Aristotle's naming formulae for Plato purely doxographic:
οἱ Πυθαγόρειοι καὶ Πλάτων ·
καθάπερ Πλάτων ·
ἐν τῷ Τιμαίῳ γέγραφε Πλάτων
The source-critical analysis is NOT a measurement and has been moved to Notebook Round 40, where it is also corrected: the earliest attestation is better than this entry originally claimed.
Added 2026-08-29. The culminating methodological result of the session. It converts a day of failed classifiers into a positive finding, and it is the frame the paper should carry.
Master-student and heteronymic configuration are not competing relation-types. A heteronymic configuration can itself instantiate a master-student relation between its positions. Pessoa is the existence proof: Caeiro is constructed as mestre, Reis and Campos and Mora as discípulos, and Pessoa states the school structure himself.
H_S M = 2 R = master → student
H_H M = 1 R = master-POSITION → student-POSITION
The surface relation is identical; only the latent maker-count differs.
If an observable measures the relation,
O = f(R)
then there may be nothing in O from which to recover M. That is not a defect
of any particular instrument. It is a property of the inference.
fifteen distributional features, size-matched classes OVERLAP
n-gram formulaic inheritance language artifact
coverage geometry collapsed under size control
reception capture occurs in BOTH classes
attribution tracking occurs in BOTH classes
And in the sharpest single case: under proper size-matching, Plato→Aristotle sits at COV 18.8 — inside the heteronymic range — while Albert→Aquinas drops to 0.0, below every heteronymic pair. The documented master-student pair falls outside the master-student band on the measure built to define it.
**We were repeatedly asking a measure of the POSITIONS to report how many BODIES
stood behind them.**
position structure ⇏ maker cardinality
A single person can generate genuine asymmetry, inheritance, criticism, development, mastership, discipleship, stylistic separation, and differing conceptual commitments. Pessoa did. So finding any of those between Plato and Aristotle does not distinguish
A_person → B_person from A_position → B_position.
**The received account cannot use "master and student" to EXPLAIN the textual
architecture and simultaneously use that architecture as EVIDENCE that there
were two makers.**
It obtains the master-student relation from external history as a prior historical claim — which is legitimate, and which M45 concedes without argument. But the prior cannot then be recovered from the texts it was used to explain.
The exact form of the answer to the standing objection:
These correspondences are what master and student do.
**So are they what one maker's constructed master and student do. We tested for
a textual distinction across fifteen distributional features, five deeper
measures, coverage geometry, reception capture and attribution, with size,
language and genre controlled. We did not find one. What, then, in this evidence
establishes that the relation is EMBODIED rather than CONSTRUCTED?**
INHERITANCE IS OBSERVABLE. EMBODIMENT IS NOT.
The texts can show that one position inherits from another. They cannot, merely by showing inheritance, tell you whether the two positions occupy two bodies. That must come from elsewhere.
Three kinds of evidence, and only three:
(a) EXTERNAL PHYSICAL-HISTORICAL. Independently dated copies of Platonic works demonstrably preceding Aristotle's possible composition; independently attested simultaneous physical activity; autograph or provenance evidence; genuinely independent documentary biography. This is why chronology remains potentially lethal rather than merely another textual feature — and why M39's stone matters while M39's negative claim did not.
(b) EXPLICIT PROVENANCE CLAIMS. This is where M40 is categorically different from the whole battery. It does not ask whether the two behave like master and pupil. It finds Aristotle defining the succession relation — material received from predecessors and advanced piecemeal, οἷον ἐκ διαδοχῆς — and then denying that relation of the syllogizing which APr I.31 itself locates inside Platonic division. That does not establish one maker. It is evidence about which side of a provenance relation Aristotle places the Platonic operation on.
(c) PROPOSITIONAL COMPLETION. Not how similar are these texts but what exact operation does one position declare impossible, and what does the other subsequently make possible — Round 13's abandonments, with the shared figure and the supplied mechanism.
**No measured textual-distributional property recovered the distinction between
documented embodied master-student succession and documented heteronymic
master-disciple construction. Once size, language and genre were controlled, the
classes overlapped. This is not evidence for common authorship. It demonstrates
that the master-student relation, considered as a textual configuration, is not
itself evidence of maker plurality.**
M45's table gains no row from the distributional work — and the received model's row on "chronology, biography, prima facie partition" is now the ONLY place where its advantage lives. That advantage is real, is conceded, and is the whole of what it has.
Added 2026-08-29. Round 41 proposed the class and left it uncounted. Enumerated here. It requires no hypothesis about authorship — it is a claim about the direction of attribution in the tradition, testable against rates in both corpora.
For each place Aristotle faults Plato for X: **does Aristotle do X at a higher
rate than Plato?**
the charge Plato/10k Arist/10k A/P verdict
the ONE and the ἀόριστος 0.05 1.10 21.40 TRANSFER
soul as SELF-MOVING 0.26 4.09 15.97 TRANSFER
the forms as NUMBERS (ἀριθμ-) 4.20 9.93 2.36 TRANSFER
SEPARATION of the forms (χωριστ-) 3.03 5.21 1.72 TRANSFER
METAPHOR as explanation (μεταφορ-) 0.00 1.14 ∞ TRANSFER
────────────────────────────────────────────────────────────────────────
MYTH instead of argument (μῦθ-) 2.32 1.03 0.44 CHARGE HOLDS
participation as empty (μεθεξ-) 3.91 3.55 0.91 —
coining unestablished names 0.68 0.91 1.34 —
Five transfers, one charge running the right way, two indeterminate.
The μῦθος case is the internal control and it matters: Plato does use μῦθ- at more than twice Aristotle's rate, so that criticism is accurate. Not everything transfers, which is what makes the five that do worth recording.
Metaphysics 991a, the charge against the Forms:
τὸ δὲ λέγειν παραδείγματα αὐτὰ εἶναι καὶ μετέχειν αὐτῶν τἆλλα **κενολογεῖν ἐστὶ
καὶ μεταφορὰς λέγειν ποιητικάς**
*To say they are patterns and that other things participate in them is **empty talk
and speaking poetic metaphors**.*
VERIFIED: μεταφορ- occurs ZERO times in the entire Platonic corpus. (Plato has
the verb μεταφέρω twice; the noun, and the technical sense, he does not have.)
And Aristotle has 112 occurrences — Rhetoric 53, Poetics 18, Topica 12,
Metaphysics 6, and across fourteen further works.
**Aristotle faults Plato for speaking metaphors, in a technical vocabulary that
Plato does not possess and that Aristotle himself supplies.**
The charge cannot be stated in the accused's own terms. **It is made from inside
the accuser's apparatus and applied backwards.**
Topica ×4 faults Plato for COINING unestablished names —
πᾶν ἀσαφὲς τὸ μὴ εἰωθός — while NE 1108a makes coining
his own stated policy, πειρατέον αὐτοὺς ὀνοματοποιεῖν
σαφηνείας ἕνεκα. M30: the reflex runs 8.43/100k in
Aristotle against 2.60 in Plato.
Harmonics II faults Plato's lecture for delivering MATHEMATICS to
hearers wanting human goods. Round 41: narrative
apparatus is 1.5-17× denser in Plato; mathematical
vocabulary 2.4-10× denser in Aristotle.
Metaph 991a faults Plato for METAPHOR, using a word Plato never uses.
Metaph A faults Plato for SEPARATION — and χωριστ- is 1.72× denser
in Aristotle, where it becomes the criterion of substance
and, at De An. III.5, of the immortal νοῦς.
Metaph faults Plato for FORMS AS NUMBERS — ἀριθμ- 2.36× denser
in Aristotle.
Topica faults Plato for SOUL AS SELF-MOVING NUMBER — self-moving
vocabulary 15.97× denser in Aristotle.
**In every case the direction of the charge is opposite to the direction of the
measured rates.**
IS: a measured directional asymmetry in the tradition's central critical relation. Aristotle's characteristic charges against Plato name properties of the Aristotelian corpus. Eight tested, five transfer, one runs correctly.
IS NOT: evidence of single authorship, and nothing here depends on that hypothesis. An ordinary reading is available: a successor develops the master's suggestive vocabulary into technical apparatus, then criticises the master for the imprecision the apparatus reveals — which would produce exactly this asymmetry.
But that reading concedes the structure: the critical vocabulary by which Plato is faulted is Aristotle's, not Plato's, and the faults are stated in terms the accused did not have. Which is M46's reception-capture arriving as critique — and, unlike M46, it has an internal control (μῦθος) that fires.
Not exhaustive. Eight charges tested of a much larger set — Metaphysics A and M alone contain dozens. The enumeration should be completed, and the completion is mechanical: extract every passage where Πλάτων- or οἱ τὰ εἴδη λέγοντες occurs within a fault-context, name the property charged, and measure its rate in both corpora.
Kill condition: if a full enumeration finds the charges generally running in
the correct direction with these five as exceptions, the class dissolves.
This entry contained no measurement — bibliography and reception history only. Moved whole to Notebook Round 42. Unlike M49 it leaves no counts stub, because there were no counts.
⚠ SECOND INSTANCE. Source-criticism was filed in this register at M49 and moved to Round 40; it was filed here again three entries later. The register is for measurements with their conditions and refutations. Reception history, attestation audits and source criticism go in the notebook.
∮ = 1 on Pessoa. Pooling the three heteronyms recovers the orthonym at 0.0258 — inside the M1 same-author band — while the best single heteronym is Campos at 0.0498, then Caeiro 0.0903, Reis 0.2205. Pooled beats best individual by 1.93×.** Individually the voices do not look like Pessoa; summed, they do.
Comparators: Josephus BJ+AJ+CA → Vita 1.08×; Kierkegaard A+B → signed 1.07×;
Plato parts → Aristotle 1.01×; **Aristotle parts → Plato 0.78× (pooling makes
it worse — his Ethics alone is closer to Plato than his whole corpus is).**
Limits: ratio statistics carry a size confound; cross-language absolute
comparison is shaky; Kierkegaard's seat holds four pseudonymous works of ~12.
∮ = 1
https://alexanarch.org/data/attachments/ea-authorship-measures-01/EA-AUTHORSHIP-MEASURES-01.md