eeeeee — the sixth erratum: The Dialect Seam, and Decompression as the Mechanism the Chain Had Not Named

eeeeee — the sixth erratum: The Dialect Seam, and Decompression as the Mechanism the Chain Had Not Named

Methodology

Dialect markers counted programmatically against the Perseus TEI with the quoted body delimited from the poem's opening to the trailing line; markers tabled with Attic equivalents so the count is reproducible or refutable. Binomial computed on the body's own marker rate. Apparatus consulted directly in public-domain editions on archive.org: Roberts 1907, the TEI's base text, and Bergk Poetae Lyrici Graeci fourth edition volume III.

Falsification Conditions

The dialect count is falsified by a recount against the TEI returning different totals. The significance claim is falsified by nothing — it is a computation, and it already tells against the argument. The attribution reading is falsified if the metrical objection, read in Russell's own terms, rests on grounds independent of dialect. The framing is falsified if an edition can be shown to drop the disputed clause without loss.

eeeeee — THE SIXTH ERRATUM

The dialect seam, and decompression as the mechanism the chain had not named

0. Position in the chain

The Sappho 31 erratum sequence runs e → ee → eee → eeee → eeeee, from stanza numbering through the structure of the Catullan transform to the hanging line read as a transmission instrument. This is eeeeee. It takes its name, as #1052 did, from Tao Lin's Eeeee Eee Eeeee (2007), in acknowledgment of a recursion the sequence made explicit before it made it embarrassing.

What the chain already established, and this erratum does not disturb. #1052 §3 states the governing finding:

Whatever the crux at καὶ πένητα proves finally to be, the direction is legible… The completion's shape is not adjudicated by the crux. The crux modifies the comparandum's identity, not the operator the completion executes.

The hanging line inscribes three constraints — adversative turn · universalizing move · comparative illustration by a low case. Those constraints are the kernel. A reader who reads the line correctly decompresses the same shape in whatever material they possess. #1052 demonstrates this on Catullus 51.4, which preserves the recoverable functional shape while reversing every pole: low case to high, endurance to destruction, durability to vulnerability — functioning, in the erratum's phrase, like a solved-and-doubled checksum.

This erratum's first task is to record that its own author spent a working session ignoring that finding. The whole apparatus below was first assembled to answer whose words are ἐπεὶ καὶ πένητα — a question #1052 had already retired, on the ground that the completion's shape does not depend on the answer. The archive-first precept would have caught it at the opening search. The search was run at the end.

0.1 What decompression means for the attribution question

The operative correction is that "Sappho's words" and "Longinus's invention" are not the available options.

A third relation holds, and the chain already named it for Catullus: decompression. A sufficiently constrained kernel carries its completion's shape; supplying that shape is neither quotation nor invention but reading correctly. Catullus decompresses into Latin and inverts. Longinus, if ἐπεὶ καὶ πένητα is his, decompresses into koine and does not invert — πένης remains the low case, the even-this logic runs forward.

                  SAPPHO'S KERNEL   LONGINUS          CATULLUS 51.4
adversative       ἀλλά              carried           implicit — otium
universalizing    πᾶν τολματόν      carried           otium destroys
comparative       ἐπεὶ καί + low    πένητα — LOW      reges et urbes — HIGH
direction         endurance         endurance         destruction

Two decompressions of one kernel, one straight and one inverted, both preserving the three constraints.

And this is the same operation Peri Hypsous calls ἀποτύπωσις, stated in a second vocabulary: the stamp carries the form into new material and does not carry the surface. The three constraints are form. Aeolic is surface. Surface is not supposed to survive an impression, which is why the dialect finding below is evidence of decompression rather than evidence of appropriation.

I. The finding that reframes the question

Parisinus graecus 2036 writes Sappho 31 as prose.

The tenth-century codex is the only manuscript of On the Sublime and therefore the only manuscript of the poem. Arieti's commentary states it plainly: "The quotation in the Greek manuscript is written out as if it were prose, as are the other poetical quotations."

The consequence is total and is rarely stated as such:

There is no hanging line in the manuscript. There are no stanzas, no line-breaks, no lineation, no quotation marks, and no closing formula. Every visual boundary in every printed text of Sappho 31 is editorial.

The trailing line "hangs" because editors set it hanging. What the scribe wrote is an unbroken run of letters from φαίνεταί μοι κῆνος through ἐπεὶ καὶ πένητα into οὐ θαυμάζεις — with the modern section number [3] falling mid-phrase, between πένητα and οὐ θαυμάζεις, as Perseus still prints it.

So the question where does Sappho end and Longinus begin has no external answer. There is no diplomatic evidence to appeal to. Only internal criteria remain: metre, dialect, syntax, and doctrine.

II. The editorial history, granular

The text as transmitted (P, s. X):

… τεθνάκην δ' ὀλίγω 'πιδεύην φαίνομαι ἀλλὰ πᾶν τολματόν ἐπεὶ καὶ πένητα οὐ θαυμάζεις ὡς …

Voigt (1971), Sappho et Alcaeus: fragmenta — the standard critical edition — prints:

ἀλλὰ πὰν τόλματον ἐπεὶ †καὶ πένητα†

Three editorial acts in one line, each worth naming:

1. πὰν τόλματον is accepted as Sappho — printed as fr. 31.17, undaggered.

2. †καὶ πένητα† is obelized — marked corrupt, sense unrecoverable.

3. The accentuation is re-Aeolicized. P gives πᾶν τολματόν; Voigt prints πὰν τόλματον — barytone, recessive, Aeolic. The editor must repair the dialect to print the line as Sappho's.

That third act is not usually remarked and it is evidence. Woolerton's translation registers the obeli directly: "But everything must be endured, †since a poor man, too†."

D'Angour, Classical Quarterly, states the crux:

"Whether the words ἀλλὰ τόλματον ἐπεὶ καὶ πένητα, which follow in [Longinus] belong to the poem or not, is debated. They are metrically flawed, although a relatively slight emendation … 'but since everything is endurable for the poorly-off', is at least conceivable."

Two things follow. The attribution is openly contested in the scholarship — it is not a settled reading this document is disturbing. And the standing objection is metrical: as transmitted, the words do not scan as a Sapphic hendecasyllable.

D'Angour also argues πὰν τόλματον means "all must be dared", not "endured" — turning the fifth stanza from resignation toward resolve.

Lardinois (Center for Hellenic Studies) takes the words as Sappho's and reconstructs from them: "The remainder of this line suggests that Sappho illustrated this gnomic thought with the example of a poor man (πένητα), but here the citation of the poem in the text of Longinus unfortunately breaks off."

Arieti concurs on the corruption — "the sense can only be approximate, since the text is corrupt here" — and translates with an ellipsis: "All must be endured, since as a wretch…"

Supporting the existence of a fifth stanza: Catullus 51 carries a fourth stanza (the otium strophe) beyond its Sapphic model, which most take as evidence that Sappho's poem continued. Apollonius Dyscolus preserves a variant opening, φαίνεται κῆνος. Paulus Silentarius echoes ἀλλὰ πὰν τόλματον.

The state of the question, then: editors accept πὰν τόλματον as Sappho, dagger καὶ πένητα, repair the accents to do so, and record that the metre does not work.

II-bis. The apparatus, consulted — and what it corrects

Russell 1964 remains in copyright and unread. It was not needed. Roberts 1907 — the base text of the TEI these counts run against — is public domain at archive.org/details/onsublimegreekte0000long, and it carries the note that answers §VII's outstanding question. Bergk's own apparatus is at poetaelyricigra11berggoog, Poetae Lyrici Graeci⁴ III (1882), p. 90.

The question is answered: yes, and by five scholars

§VII asked whether any editor has argued in print that ἐπεὶ καὶ πένητα is Longinus's own. It has been argued since the sixteenth century.

V. 17. πὰν τόλματον, P παντόλματον, vulgo πᾶν τολματόν. Quae sequuntur ἐπεὶ καὶ (hoc codd., ut videtur omnes, praet. cod. Spengelii, om. Robortell) πένητα uncis inclusi, nam videntur haec ad Longini orationem pertinere, fort. ἀλλὰ πὰν τόλματον, ἐπεῖπεν· εἶτα (vel κάτα) οὐ θαυμάζεις κτλ.

Antea conieceram versu 16 integrum carmen absolvi, ut deinde sit scribendum πᾶν τὸ ἀσμάτιον ἐπεῖπον· εἶτα οὐ θαυμ., similiter Ahrens πᾶν τὸ ἀσμάτιον ὡς ἔπος εἰπεῖν, ἦ γὰρ οὔ; θαυμάζεις κτλ. Hermann vero coni. φαίνομαι οὐδέν, ἀλλὰ παντόλματον ἀπεῖπεν ἦτορ, Heller πᾶν δὲ τολματέον ἐπὶ τἀγαπάτᾳ.

Three of these conjectures restore a citation formula — ἐπεῖπεν, ἐπεῖπον, ὡς ἔπος εἰπεῖν. §IV of this erratum observes that the treatise carries 74 citation verbs and none here. The nineteenth century observed the same absence and concluded that one had been lost. That is the same datum reached independently, and it should be credited as such rather than presented as new.

The convergence that is clean

Bergk's earlier conjecture places the poem's end at line 16 — which is φαίνομαι, which is where the Aeolic stops.

He reached it from syntax and sense, without a dialect count. The count arrived at the same word from marker density, without knowledge of the conjecture — which was not consulted until after the count was published in this chain. Unlike the convergence with the archive's own reconstruction, whose claim was in context before the analysis began, this one is not contaminated. It is the first clean corroboration the dialect argument has.

The correction to §III's premise

P reads παντόλματον — one word. The split πᾶν τολματόν is vulgo: the vulgate, editorial, not the manuscript.

§III as first written contrasted "the manuscript's πᾶν τολματόν" with "Voigt's re-Aeolicized πὰν τόλματον" and treated the first as the transmitted form. It is not. The transmitted form is a compound whose accent is recessive — which is Aeolic-looking rather than unmarked.

Consequences, stated exactly.

• The παῖσαν/πᾶν minimal pair is withdrawn in the form given. It compared a manuscript reading against an editorial one.
• The re-Aeolicization argument is weakened, not destroyed. Voigt's πὰν τόλματον still differs from P's παντόλματον — she splits the compound and accents the result. But the contrast is now editor-against-editor, not editor-against-manuscript, and it is correspondingly weaker.
• The count over the remaining sixteen markers stands, as does P(0 in 6) ≈ 0.209 and everything §III says about its insufficiency.
• A new fact enters: this word has been split, joined, re-accented and emended by Robortello, Ahrens, Bergk, Hermann, Heller, Crusius and Voigt. There is no neutral form of it to count. Any dialect argument about πᾶν τολματόν is an argument about a word no manuscript contains.

II-ter. The framing: same premise, opposite editorial consequence

The nineteenth-century position and this erratum's share a premise and diverge completely on what follows.

Bergk, Ahrens, Robortello: the words are Longinus's — therefore they are not part of Sappho 31, and are bracketed, omitted, or emended away. Attribution determines inclusion.

This erratum: the words are Longinus's — and they are the ending of Sappho 31, and any edition that drops them drops the ending.

The divergence is the decompression model doing work. Under §0.1, ἐπεὶ καὶ πένητα is Longinus's as material — his koine, his mouth, his dialect — and Sappho's as form: her kernel, her three constraints, her open agent slot. Authorship is not exclusive here, and an apparatus that assumes it is will make exactly the error the tradition made: reasoning from whose words to whose poem, as though the second followed from the first.

It does not follow. A decompression belongs to the compressor's work and is executed in the decompressor's material. Bracketing it out on the ground that the material is his removes the completion the kernel required — which is not a conservative act but a destructive one, performed under the appearance of caution.

On this reading the editions rank differently than the tradition ranks them.

Voigt is closest to right, for the wrong reason. She retains what the others discard, but records it as corruption — damage to be someday repaired. Under the decompression model it is not damage. It is the impression: form carried into new material, surface not carried, exactly as ἀποτύπωσις at 13.4 specifies.

The editorial proposal that follows, and it is modest. Where a passage is neither sound text nor crux but a completion executed in a transmitter's material, the apparatus has no diacritic for it — and in the absence of one, it defaults to the dagger, because the dagger is what an apparatus does with anything it cannot assign. The cruces at line 17 are that default firing. They mark not a text that is broken but a category the apparatus lacks.

III. What this study adds: dialect density

If the only criteria are internal, one of them can be counted.

Longinus preserves Sappho's Aeolic scrupulously through the quoted body — κῆνος, γελαίσας, κὰμ, ἔαγε, χρῶ, ὀππάτεσσι, ἄκουαι, κὰδ, κακχέεται, παῖσαν, ἀγρεῖ, ἐμμί, τεθνάκην, ὀλίγω, ʼπιδεύην, βροχέως, φωνᾶς.

poem body              74 words · 17 Aeolic features · 1 per 4.4 words
ἀλλὰ πᾶν τολματόν       3 words ·  0 features
ἐπεὶ καὶ πένητα         3 words ·  0 features
                        ─────────────────────────
expected at the body's rate for 6 words:  1.4
observed:                                 0

The markers, tabled rather than asserted

A count is not checkable unless the counted items are listed.

• for βραχ
• |

Counting note, recorded because it affects the denominator. ἐπιρρόμβεισι (Aeolic -εισι for -ουσι) is present in the quotation but is hyphenated across a line-break in the TEI as ἐπιρρόμ- βεισι, so a stem search does not match it. The table above therefore lists 17 forms while the body arguably carries 18. Counting it raises the marker rate and slightly raises P(0 in 6) — it does not help the argument, and is recorded for that reason.

Counts were computed programmatically against urn:cts:greekLit:tlg0560.tlg001.perseus-grc2, with the quoted body delimited from φαίνεταί μοι κῆνος to ἀλλὰ πᾶν τολματόν and word-split on whitespace and Greek punctuation. Any reader with the TEI can reproduce or refute both numbers.

The significance test, which the count fails on its own

This must be stated before the argument goes any further. With p = 17/74 = 0.230, the probability that six words drawn from the same distribution contain zero markers is

P(0 in 6) = (1 − 0.230)⁶ = 0.209

Roughly one in five. The dialect absence is not significant in isolation, and no weight should be placed on it as though it were. At this marker rate a zero-run of twelve words would be needed to reach p < 0.05. There are six.

An earlier draft of this erratum called the count "the strongest new evidence in the chain." That was overclaiming and is withdrawn. The count is a constraint, not a result.

What the count contributes is its conjunction with three independent facts, none of which is a p-value and which therefore cannot be multiplied into one:

• Voigt's re-Aeolicization — an editorial acknowledgment, arrived at independently, that the transmitted form is dialectally unmarked
• The prose manuscript — which removes the visual boundary altogether and forces the question onto internal criteria
• ~~παῖσαν against πᾶν~~ — WITHDRAWN, §II-bis. P reads παντόλματον, one word; the πᾶν this pair compared against is vulgo, not manuscript. The conjunction is correspondingly weaker.

The conjunction is an argument, not a test, and it should be presented and cited as one.

The Aeolic stops at φαίνομαι and does not resume.

Three qualifications, stated because the argument is weakened without them. Six words is a small sample and absence in a small sample is weak on its own — the base rate is what gives it force. And the base form is itself contested — see §II-bis. P reads παντόλματον as one word; πᾶν τολματόν is the vulgate split; Voigt prints πὰν τόλματον. Robortello, Ahrens, Bergk, Hermann, Heller, Crusius and Voigt have each split, joined, accented or emended this word. There is no neutral form of it to count, and a dialect argument resting on it is an argument about a word no manuscript contains. The remaining sixteen markers do not depend on it.

The contribution is to make the repair visible and to attach a number to it.

On the metrical objection — a correction to how this study first framed it.

An earlier draft argued that the metrical objection and the dialect observation were one objection, the metre failing because the dialect had failed first. That is wrong, and it concedes a premise it should contest.

Sappho 31 breaks its metre at line 9 as well, at γλῶσσα ἔαγε, in the hiatus between the two words. Nagy reads that break as intended — the breaking of regularity by way of interrupting the flow of the music is an intended effect — and it is universally admired. The poem says the tongue gives out; the verse gives out.

At line 17 the metre breaks again, and the poem is again about breaking. The identical phenomenon is filed as corruption and offered as grounds for doubting Sappho's authorship.

The difference between the two verdicts is not evidential. It is that line 9 survives entire, so its rupture reads as control, while line 17 stops mid-clause, so its rupture reads as accident. The apparatus is constructed to detect damage and therefore detects damage; a poet who breaks form deliberately is invisible to it. Emendation is work an editor can do; deliberate rupture leaves an editor nothing to do but stop.

The iconicity argument made for line 9 has never, so far as this study has found, been made for line 17. That is a live gap in the scholarship and it is where a contribution would go.

So the two observations must be kept apart as events. The metrical break is composition — Sappho's. The dialect shift belongs to the transmitting mouth. Collapsing them hands the poem's deliberate act back to the copyists.

They are not, however, independent, and the asymmetry between them is the argument. A dialect shift was not available to Sappho as a compositional device: there was no second dialect within the poem's world to break into, and Aeolic was the whole material. Metrical rupture was available, and she used it at line 9, at the failure of the voice. Longinus possessed an instrument she did not — and deployed it at the point her poem had established as the site of rupture.

This is 13.4's ἀποτύπωσις executed rather than described: the stamp carries the form into new material and does not carry the surface. The hiatus is surface and is not reproducible in prose koine. The principle — break the form where the form fails — is form, and is.

And it accounts for the shape of the data. Seventeen markers hold at 1 per 4.4 words and then cease at a single word. Erosion, scribal drift and fatigue are all gradual; a clean stop is more consistent with a boundary than with diffuse erosion. Gradual decay predicts a tapering distribution — 1 per 4 words, then 1 per 6, then 1 per 8, then nothing — and the observed distribution is 1 per 4.4, then nothing. That is a step function rather than a taper. It licenses boundary; it does not by itself license act, and the inference to composition is carried by the Longinian poetics of §V-bis rather than by the distribution.

Consequence for the metrical objection. Editors have tested whether the words scan as Sappho. If the break at 17 is the transmitter's, executed on the poet's principle, scansion is the wrong test: failure to scan would be the signature of the impression rather than evidence of damage. This does not establish the attribution. It establishes that the standard argument against it does not bear the weight placed on it.

IV. Longinus's own attestations

He gives none of the ordinary kind, in a treatise saturated with them.

The treatise uses 74 citation verbs — 52 φησί/φησὶ/φησίν/φησὶν, 14 λέγει, 5 λέγων, 3 γράφει — and interposes them inside quotations (γεννᾶται γάρ, φησί, τὰ μεγαλοφυῆ).

At Sappho 31: zero. She is named once at 10.1 — the only occurrence of her name in the treatise — and the poem begins directly off συνδῆσαι. with no verb of saying. No formula opens it and none closes it.

The one first-person marker falls after the analysis, and it claims continuity:

ἡ λῆψις δ' ὡς ἔφην τῶν ἄκρων καὶ ἡ εἰς ταὐτὸ συναίρεσις ἀπειργάσατο τὴν ἐξοχήν

ὡς ἔφην — as I said — points back across the quotation to 10.1, forming a ring:

10.1   ὅτε τὰ ἄκρα … δεινὴ καὶ ἐκλέξαι   καὶ εἰς ἄλληλα συνδῆσαι
       ⟨ the poem ⟩
10.3   ἡ λῆψις  ὡς ἔφην  τῶν ἄκρων        καὶ ἡ εἰς ταὐτὸ συναίρεσις

Longinus never attests that she ends. He attests that he has not stopped.

V. The doctrine performed on the voice

10.1 states the method for words: τὸ … ἐκλέγειν ἀεὶ τὰ καιριώτατα καὶ ταῦτα τῇ πρὸς ἄλληλα ἐπισυνθέσει καθάπερ ἕν τι σῶμα ποιεῖν — selecting the most vital constituents and composing them into as it were a single body.

Sappho performs it on symptoms. Homer performs it on storms — ἐκλαμβάνει τῶν παρακολουθούντων τὰ χαλεπώτατα. And Longinus performs it on her — ἡ λῆψις τῶν ἄκρων καὶ ἡ εἰς ταὐτὸ συναίρεσις.

One operation, three materials: words · symptoms · voice.

The figure is literalized in the quotation. σῶμα occurs three times in the treatise. At 10.1 it is the image — make the parts into one body. At 10.3 it is an item on the inventory: τὴν ψυχὴν τὸ σῶμα τὰς ἀκοὰς τὴν γλῶσσαν τὰς ὄψεις τὴν χρόαν, six holdings, articulated, asyndetic, every one ἀλλότρια and διοιχόμενα — another's, and gone.

And the stamping vocabulary is already present in this chapter. At 10.6, on Homer's forced compound ὑπὲκ θανάτοιο, Longinus says he almost stamped his own idiom for danger into a form — the ἐκτυπόω/ἀποτύπωσις family that 13.4 uses for transmission itself. Arieti's note on the root: "to blow, strike; then to imprint; then impression; then mould or matrix." The stamp is not introduced at 13. It is already working at 10, on Homer's words — and then on Sappho's voice.

ἀλλότριος runs the treatise as a line:

 4.1   ἀλλοτρίων ἁμαρτημάτων     Timaeus, censor of ALIEN faults, blind to his own
10.3   πάνθ' ὡς ἀλλότρια         her own senses become ALIEN
13.2   ἀλλοτρίῳ πνεύματι         possessed by ANOTHER'S spirit — transmission
44     ἀλλοτρίων θῆραι θανάτων   hunting OTHER MEN'S deaths, beside φιλοχρηματία

V-bis. The Homeric join, and the form of the seam

At 10.6 Longinus states an operation in his own words, of Homer:

τὰς προθέσεις ἀσυνθέτους οὔσας συναναγκάσας παρὰ φύσιν καὶ εἰς ἀλλήλας συμβιασάμενος … τῇ δὲ τοῦ ἔπους συνθλίψει τὸ πάθος ἄκρως ἀπεπλάσατο καὶ μόνον οὐκ ἐνετύπωσε τῇ λέξει τοῦ κινδύνου τὸ ἰδίωμα

Forcing together, against nature, elements that cannot be compounded; pressing them into one another; and the constriction performs the stamping. ἐνετύπωσε is the τύπος family, present at 10.6 — three chapters before the same family is used at 13.4 for the transmission of greatness between writers.

The prepositional phrase is the chapter's refrain across three materials: εἰς ἄλληλα συνδῆσαι (10.1, Sappho's symptoms) · εἰς ταὐτὸ συναίρεσις (10.3, Longinus's gathering) · εἰς ἀλλήλας συμβιασάμενος (10.6, Homer's prepositions).

Applied to the seam, this predicts the form the seam should take, and the form is what the transmitted text exhibits:

1. τολματόν hangs — a fifth-stanza opening requiring completion.

2. The dialect breaks — the transmitter's available material, deployed at the site the poem established.

3. ἐπεὶ καὶ πένητα is positioned so that it can be construed as either a further hanging element of the quotation or the opening of the commentator's sentence.

4. The section division falls inside the join — Perseus prints ἐπεὶ καὶ πένητα [3] οὐ θαυμάζεις, the boundary landing mid-phrase because no cleaner boundary is available.

Consequence for the metrical objection, stated at its strongest. A compound made παρὰ φύσιν is not expected to scan; ὑπέκ scans neither as ἐκ nor as ὑπό, and that is why Longinus admires it. If the seam is a join of the Homeric type performed on voice, non-scansion is the predicted property of the artifact, not evidence against its authenticity.

And the two compounds are made at the same kind of line. Homer's τυτθὸν ὑπὲκ θανάτοιο — barely out from under death. Sappho's τεθνάκην δ' ὀλίγω 'πιδεύην — little short of having died. Both joins occur at the line where a subject is barely not dead.

Status of this section. The verbal parallels, the shared prepositional phrase, the presence of the τύπος family at 10.6, and the position of the section division are all verifiable in the text and are so verified. The claim that the seam was composed to be construable both ways is interpretive and unfalsifiable as stated; it is recorded as a reading the evidence permits, not as a finding. What it does establish is narrower and holds: the standard metrical argument against the attribution assumes a criterion the chapter's own poetics reject.

VI. The πένης, and what it costs the argument

πένης is a hapax in the treatise. One occurrence, in the disputed clause. The treatise's only φιλοχρηματία stands at 44.6, beside δουλαγωγοῦσι, in the passage on what destroys great natures — six sections before the terminal break.

If the clause is Longinus's, the treatise's only two economic terms sit at its two breaks, and 7.1 supplies the doctrine they would frame: nothing is great, contempt for which is great — wealth, honours, reputations, tyrannies.

Read continuously without a lacuna, ἐπεὶ καὶ πένητα οὐ θαυμάζεις, ὡς … ἐπιζητεῖ yields an a fortiori built on the sentence's own hinge word: you do not marvel at a poor man — a poor man seeks what is another's, that is what poverty is — how then she seeks her own soul, body, hearing, tongue, sight, complexion as another's. Poverty externalized against poverty internalized. πένητα is masculine and anarthrous, the only bare noun in a sentence where every other is articulated; it is generic, a foil, and she is never called poor. She is measured against poverty and exceeds it.

And this reading costs #1458 two claims. That deposit called πᾶν τολματόν an unexploited remainder — transmitted without being made to perform analytic labour. If ἐπεὶ καὶ πένητα is Longinus's, the exploitation is what is missing, not the exploitation that never happened. And "agentless, not subjectless" treated πᾶν τολματόν as the quotation's last words, which no edition supports: Voigt continues the fragment past it.

Both claims in #1458 are hereby marked corrected.

VI-bis. The errata proper

On #1458, EA-LONGINUS-01 — three claims

1 · "Unexploited remainder." #1458 reads πᾶν τολματόν as a line Longinus transmits without making it perform analytic labour. Under §0.1 this is doubly wrong: if the following clause is his decompression, the line is not unexploited — it is executed, and what is missing is the execution's continuation rather than an execution that never occurred.

2 · "Agentless, not subjectless." #1458's diagram treats ἀλλὰ πᾶν τολματόν as the quotation's last words. No edition supports this. Voigt continues the fragment past it; the agentlessness observation is sound but is attached to the wrong terminus.

3 · §VI's comparison table — the substantive one. #1458 places both ἀλλὰ πᾶν τολματόν and ἐπεὶ καὶ πένητα in the column headed SAPPHO 31.17 — last words, glosses the syntax as "INCOMPLETE — accusative reaching for a verb," and names the shape an APERTURE.

That reading requires πένητα to be Sappho's accusative, reaching for a verb she never wrote. If πένητα is governed by Longinus's θαυμάζεις, the accusative is not reaching. It already has its verb, four words later, across a section division that falls inside the join.

And #1458 is internally split on the point. Its Period 5 gloss reads "πᾶν τολματόν as the moment the poem stops speaking of its own and hands on" — placing the hand-off at τολματόν — while its table extends her last words two clauses past it. The prose and the table disagree, and neither was checked against the other. The aperture reading survives; its extent does not.

On #1052, eeeee — one editorial decision, on evidentiary grounds

This is not a correction to the erratum's argument, which stands entire. It concerns a surface decision recorded in its own front matter.

At v1.0, integrating Assembly-review consensus, #1052 standardized its Greek to Perseus's ordinary form — πᾶν τολματόν, not variant accentings. The stated reason is consistency of citation across the fleet, and as a citation policy it is correct.

But the "variant accenting" it set aside is Voigt's πὰν τόλματον — and that variant is not a variant. It is the editorial re-Aeolicization.

Voigt does not print the manuscript's accents. She restores barytone, recessive, Lesbian accentuation in order to publish the line as Sappho's. The difference between πᾶν τολματόν and πὰν τόλματον is precisely the evidence that the transmitted form is dialectally unmarked — the one signal that distinguishes a decompression from a quotation.

A standardization made for surface consistency erased the philological signal the chain most needed. The erasure is invisible in the deposit, because a standardized text looks like a clean text.

Correction of practice, not of argument. Where a deposit's argument turns on the dialectal or metrical character of a Greek line, the deposit must carry the critical edition's accentuation and note where it departs from the base text, or must state that its Greek is normalized for citation and is not evidentiary. #1052's Greek is normalized for citation. The dialect count in §III of this erratum is therefore run against the TEI and must never be run against the deposit.

On this erratum

Written by an author who reproduced the classicist reflex it criticizes — treating a metrical break at line 17 as a flaw while quoting, in the same document, Nagy's reading of the break at line 9 as intended. The reflex survives even in a reader holding the refutation. That is worth recording as a property of the reflex rather than of the reader.

VI-quater. On the two ruptures — a review adopted, then withdrawn

The Assembly review of 2026-08-15 held that the two ruptures are structurally parallel (break at extremity) but not thematically parallel, on the ground that line 9 names a physical failure while line 17 names endurance or resolve. That correction was adopted on 2026-08-15 and is withdrawn here on the same day. It rests on the sense of τολματόν and not on its form.

τολματόν is a verbal adjective in -τός. It is not a finite verb and does not name a state. Neuter and predicative — all [is] to-be-dared — the form cannot govern a nominative agent. It names an action as available and leaves the agent slot structurally open. This is the property #1458 recorded as "agentless, not subjectless", and it is doing the work here.

Set against line 9:

Both lines name an act and require the reader to perform it. Once in the mouth, once in the hand. The parallel is therefore not positional-only: it is one operation executed in two materials, which is the same claim §V-bis makes about εἰς ἄλληλα across three materials.

And at line 17 the performance is textually present. The daring named is the daring done — six syllables later, by a reader completing an open form in his own dialect. Under the decompression model of §0.1 that is not a coincidence between theme and event; it is the kernel operating as designed.

Status. The grammatical facts — the -τός formation, the impossibility of a nominative agent, the hiatus at line 9 — are secure and checkable. The claim that the poem was composed to be completed this way is not, and is not made. What is claimed is narrower: the two ruptures share a mechanism, not merely a position, and a review that reduces them to position has read sense where it should have read form.

Recorded as a failure mode of this document's author, not of the reviewer: the correction was accepted because it was well-argued and because reviewers catching over-readings is the ordinary case. The form was not checked before conceding. That is the same failure the chain diagnoses in the editorial tradition — deferring to a fluent account of a text instead of examining the text — committed here against a review rather than against an apparatus.

VI-ter. The consequence for the reconstruction

The reviews converge on a reframing this erratum should state rather than leave implicit.

If the dialect seam is evidence about a mechanism rather than about an author, the question the reconstruction chain has been asking changes shape. It has been:

Did Longinus later theorize transmission?

It becomes:

Has chapter 10 already performed the operation that chapters 13–14 theorize?

The materials for that question are all in §V and §V-bis: ἐκλογή and ἐπισύνθεσις at 10.1, λῆψις and συναίρεσις at 10.3, εἰς ἀλλήλας συμβιασάμενος and ἐνετύπωσε τῇ λέξει at 10.6 — the stamping verb three chapters before ἀποτύπωσις at 13.4.

The consequence for the lost πάθος section is substantial and should be recorded as a constraint rather than a result. If the mechanism is already distributed through the surviving treatise and demonstrated on Homer's words before it is named as doctrine, then a reconstruction of the lost ending does not need to invent the transfer doctrine from nothing. It needs to state what chapter 10 has already done.

This is a reduction in what a reconstruction must supply, not an increase in what it may assume. It does not license inventing more; it licenses inventing less.

VII. What is claimed, and what is not

Claimed. That the manuscript writes the poem as prose, so every boundary is editorial and only internal criteria remain. That the Aeolic density is 1 marker per 4.4 words in the body and 0 in the six disputed words, against 1.4 expected. That editors re-Aeolicize the accents to print the line as Sappho's, which is an acknowledgment of the same fact. That the metrical rupture and the dialect shift are two distinct phenomena converging at one seam — not one objection — and that the standard metrical test becomes less probative once the seam is modelled as decompression. That Longinus supplies no citation formula here in a treatise carrying 74 of them, and that his one first-person marker asserts continuity rather than closure. That the chapter's own vocabulary — ἐκλογή, ἐπισύνθεσις, λῆψις, συναίρεσις, σῶμα, and the stamping verb at 10.6 — describes one operation applied to words, to symptoms, and then to a voice.

Not claimed. That ἐπεὶ καὶ πένητα is Longinus's. The attribution is undecided and this document does not decide it — it supplies one countable criterion to a debate the scholarship already records as open. That the dialect argument is novel; the editorial practice shows otherwise. That six words are a sufficient sample on their own. That the hapax pattern at 10.3 and 44.6 is design rather than the ordinary coincidence of a damaged text — a hapax at a break looks like design from either side of a manuscript accident.

Outstanding, revised 2026-08-15. The blocking question — whether any editor has argued in print that the words are Longinus's — is answered in §II-bis: Robortello, Bergk, Ahrens, Hermann and Heller, from 1554 onward. Roberts 1907 and Bergk⁴ III were reached in minutes on archive.org; they were never unreachable, only unsearched.

Still outstanding: Russell 1964 (in copyright) for the modern metrical objection in its own terms; Mazzucchi; Halliwell 2022; and Robinson Ellis, Hermathena XXII.385, cited by Roberts as "another view" and not yet read. None of these is now blocking on the attribution question, which the nineteenth century already contested in print. The remaining blocker is narrower: whether any editor has argued that the words are Longinus's AND belong to the poem — the position of §II-ter, which the sources consulted do not record anyone holding.

SOURCES

Text. urn:cts:greekLit:tlg0560.tlg001.perseus-grc2 (Perseus/OGL TEI); counts computed against it 2026-08-15. Parisinus graecus 2036, s. X, sole witness.

Editions and scholarship consulted. Voigt, Sappho et Alcaeus: fragmenta (1971) — obeli and accentuation. D'Angour, "Fearless, bloodless … like the gods: Sappho 31 and the rhetoric of 'godlike'", Classical Quarterly — the crux stated, the metrical objection, the "dared" reading. Lardinois, "The New Sappho Poem (P.Köln 21351 and 21376): Key to the Old Fragments", CHS. Arieti, Longinus, On the Sublime: Translation and Commentary, ch. 10 — the prose manuscript, the corruption note, the root of the stamping verb, σύνοδος and συνεδρεύει. Woolerton trans. (Voigt text). Halliwell (2022) and Russell (1964) noted as standard and not consulted directly.

Corrects. #1458 (AXN:05E4) — three claims, §VI-bis. #1052 (AXN:0428) — one editorial decision on evidentiary grounds, §VI-bis.

Continues. #201 (AXN:0346) · #1048 (AXN:0424) · #1049 (AXN:0425) · #1051 (AXN:0427) · #1052 (AXN:0428). This is the sixth in the sequence.

Inherits. The kernel-and-decompression model from #1052 §3 and §6, and the holographic-kernel framework from #76 (AXN:023E).