The literary reading accompanying the sixth erratum. Sappho 31 breaks its metre at two points and the manuscript is damaged at both. At line nine the poem says the tongue has broken and a hiatus between the two words makes the reader's own mouth perform the gag — an effect Nagy reads as intended and the tradition admires, while obelizing the words around it and generating three separate papers by Bonanno on two syllables. At line seventeen the poem says all is to be dared and the metre breaks again, and the identical phenomenon is filed as corruption. The essay argues the difference between those verdicts is not evidence but completeness: the tradition admires a break it can see the far side of and pathologizes one it cannot, because an apparatus that calls a disruption corrupt has work to do while an apparatus that calls it deliberate has only to stop. Both lines name an act and hand its performance to the reader — once in the mouth, once in the hand, the second by a verbal adjective in -tos that cannot govern a nominative agent. The dialect count is reported with its insufficiency stated.
Object-level reading against the Perseus TEI with counts stated and their limits carried in the accompanying erratum. Grammatical claims checked against the forms in the text; editorial history read in public-domain editions rather than through citations of them.
Falsified if the hiatus at line nine is shown not to produce the effect Nagy describes, or if the two ruptures can be shown to be produced by a common scribal mechanism. The reading of the verbal adjective is falsified if the form is shown to admit a nominative agent. The essay makes no attribution claim and cannot be falsified on that axis.
This essay reports what can be counted and says what the counting looks like. Its philological constraints, its corrections and its limits are held in the accompanying erratum — eeeeee, the sixth in the Sappho 31 chain — which states plainly that the dialect count is not significant on its own and that the attribution remains undecided. The kernel-and-decompression model is #1052's, not this essay's.
There is a moment in Sappho 31 where the poem says the tongue breaks, and the metre breaks with it.
γλῶσσα ἔαγε. My tongue has broken. Between the final short vowel of γλῶσσα and the initial short vowel of ἔαγε there is a hiatus — two vowels meeting without elision, a gap where Greek song does not permit one. Nagy saw it in 1974 and named it: the term hiatus, in the original Latin, conveys the idea of gagging. The sequencing has produced a non-sequencing. The line does not describe a tongue breaking. The line breaks, at the tongue, in the mouth of the reader.
And that is where the manuscript is damaged. Voigt sets a dagger at ἀλλὰ †κὰμ† in the same line. Maria Grazia Bonanno published three papers on those two words — 1993, a postilla in 1996, a seconda e ultima postilla in 1998 — and the question is not settled. The words for the tongue's breaking are the words the transmission could not carry. The wound is at the wound.
That is the first rupture, and it is made. There is a second, and the tradition has never allowed it the same dignity.
Four stanzas survive. Then a fifth begins and stops:
ἀλλὰ πᾶν τολματόν, ἐπεὶ καὶ πένητα
but all is to be dared, since even a poor —
Voigt daggers it: †καὶ πένητα†. Cruces desperationis — the crosses of despair, set where an editor declines to guess. D'Angour records that whether the words belong to the poem at all is debated, and that they are metrically flawed. Arieti translates and then gives up mid-clause: since as a wretch…
So the poem is daggered in exactly two places. At the tongue, and at the ending. At the moment the voice fails and at the moment the voice stops.
Two cruces, and both fall where the poem names an act — and both hand that act to the reader.
γλῶσσα ἔαγε is a finite verb with its subject expressed: the tongue has broken. The rupture is in the sound — a hiatus, two short vowels colliding where Greek song forbids it — and the reader's own mouth performs the gag. The line says a breaking and makes you break.
πᾶν τολματόν is not a finite verb at all. It is a verbal adjective in -τός: all [is] to-be-dared. Neuter, predicative, and incapable of taking a nominative doer. The form names an action as available and leaves the agent slot structurally open. The rupture here is in the agency — and what fills it is the reader.
The line says a daring and makes you dare.
So the parallel is not merely positional. It is one operation in two materials: a line that names an act and then requires the reader to perform it. Once in the mouth, once in the hand. And at line 17 the performance is legible in the text itself — the daring named is the daring done, six syllables later, by someone taking up a dead woman's open form in his own dialect.
An apparatus built to record damage has no way to record either.
And here is what almost no reader of a printed Sappho knows.
In the manuscript there is no ending. Parisinus graecus 2036, the sole witness, tenth century, writes the poem as prose — as it writes every verse quotation. No stanzas. No line-breaks. No lineation. The hanging line does not hang. It is set hanging, by editors, nine hundred years later. The scribe wrote a continuous run of letters, and somewhere inside that run a woman stops speaking and a critic begins.
There is no mark. You cannot see the seam. You can only measure it.
Longinus keeps her Aeolic with care. Not approximately — scrupulously. κῆνος for ἐκεῖνος. γελαίσας. κὰμ, κὰδ, apocopated. ὀππάτεσσι with its doubled consonant. ἐμμί for εἰμί. τεθνάκην, ὀλίγω, ʼπιδεύην — infinitive, genitive, all Lesbian. φωνᾶς with the long alpha where Attic wants eta.
Seventeen dialect markers in seventy-four words. One every four and a half words, sustained without a lapse to the last syllable of φαίνομαι.
Then it stops. Six words follow — ἀλλὰ πᾶν τολματόν, ἐπεὶ καὶ πένητα — and not one of them is Aeolic. At the rate the poem has been running, one and a half markers were due. Zero arrive.
And the word at the seam has no settled form at all. P reads παντόλματον — one word. The familiar πᾶν τολματόν is vulgo, an editorial split. Voigt prints πὰν τόλματον, split and re-accented. Robortello omitted the clause after it; Ahrens, Bergk, Hermann and Heller each proposed something different.
Seven editors have split, joined, accented or emended these five syllables. There is no neutral form of them to point at. What can be said is narrower and still worth saying: whatever the base form, the sixteen unambiguous markers stop at φαίνομαι, and to print the trailing line as Sappho's an editor must decide her dialect for her.
Now the objection everyone raises against line 17: they are metrically flawed.
Read that against line 9.
At line 9 the metre also breaks. Nagy read it as intended — the breaking of regularity by way of interrupting the flow of the music is an intended effect. The poem says the tongue gives out and the verse gives out in the same breath. Nobody calls that a defect. It is quoted as one of the finest things in Greek.
At line 17 the metre breaks again. The poem is again about breaking — all is to be dared, said by someone with nothing left. And here the identical phenomenon is filed as corruption, daggered, and offered as grounds for doubting the words are Sappho's at all.
Two ruptures. Both at the poem's own subject. One called art, the other called error.
The difference between those verdicts is not evidence. It is completeness. We possess the whole of line 9, so its rupture reads as control. Line 17 stops mid-clause, so its rupture reads as accident. The tradition admires a break it can see the far side of and pathologizes one it cannot.
This is the oldest reflex in the reception chain and it is a failure of nerve about what verse is. An editor who calls a disruption corrupt has work to do — collate, conjecture, emend, publish. An editor who calls it deliberate has only to stop. The apparatus is built to find damage, so it finds damage, and a poet who breaks form on purpose is unreadable to an instrument that treats every broken form as a wound.
Sappho 31 has already broken its metre once, at the failure of the voice, on purpose. It is the last poem in Greek whose metrical break at its ending should be assumed to be a scribe's mistake.
So the dialect count is not evidence that the metre failed. They are two different ruptures landing on the same six words. The metrical break is composition — hers. The dialect shift belongs to the mouth that carried her. To fold them together is to hand the poem's deliberate act back to the copyists.
But they are not unrelated, and this is the whole of it: one taught the other.
Sappho could break metre. She could not break dialect. There was no second dialect inside her poem's world to break into — Aeolic was the entire material. Metrical rupture was available to her craft and she used it, at the tongue, at the precise point where the voice gives out.
Longinus had a tool she did not have — and on the decompression model, a reader who read the line correctly would have used it at the place her poem prepared. Whether this reader did is not settled here and is not settled anywhere; the erratum this essay accompanies declines to settle it. What follows describes the operation the model predicts, not a demonstrated act.
That is not analogy. It is his own doctrine performed. The stamp preserves the form in new material, not the surface. He does not copy the hiatus — he could not, and copying it would have been the wrong act. He takes the principle, break the form at the point where the form is failing, and executes it in the only material he has: his own Greek.
Aeolic cannot break in Aeolic. It can only break into something else.
And this explains the one thing an accident cannot. Why do all seventeen markers hold and then stop at a single word? Erosion is gradual. Copyist's drift is gradual. Fatigue is gradual. A clean stop is an act. He held her dialect as long as her poem held its own form, and released it exactly where she had already demonstrated that release is what the moment requires.
Which means the objection at line 17 has been asking the wrong question for four centuries. Editors test whether the words scan as Sappho. If the break is his, executed on her principle, it is not supposed to scan. Its failure to scan is the mark of the impression. The daggers are set on the evidence.
Longinus does not say. He does not say anywhere.
This is a treatise that cites constantly — seventy-four citation verbs, φησί and λέγει and γράφει, dropped even into the middle of quotations. At Sappho: none. He names her once, at 10.1, the only time in the whole work, and then the poem simply starts, off the end of his own sentence, without so much as a she says.
The one place he asserts himself comes after the analysis is finished:
ἡ λῆψις δ' ὡς ἔφην τῶν ἄκρων καὶ ἡ εἰς ταὐτὸ συναίρεσις ἀπειργάσατο τὴν ἐξοχήν
the taking, as I said, of the extremes, and the gathering-into-one, produced the eminence
As I said. Pointing back across her poem to his own words before it — she is skilled at selecting the extremes and binding them to one another — and closing a ring with the poem inside it.
He never tells us where she ends. He tells us he never stopped.
He opened the chapter by saying what sublimity requires: select the most vital constituents, and by composing them together make — καθάπερ ἕν τι σῶμα — as it were, a single body.
Then he quotes the one poem in Greek where that figure is not a figure.
τὴν ψυχὴν τὸ σῶμα τὰς ἀκοὰς τὴν γλῶσσαν τὰς ὄψεις τὴν χρόαν
Soul, body, hearing, tongue, sight, complexion. Six holdings, each with its article, no conjunctions — the form of an inventory, an estate itemized. And every one of them ἀλλότρια διοιχόμενα: another's, and gone.
The body that was his metaphor for composition is an item on her list of losses.
And in that same sentence there are eight third-person verbs and no subject. ἐπιζητεῖ, ψύχεται, κᾴεται, ἀλογιστεῖ, φρονεῖ, φοβεῖται, τέθνηκεν, φαίνηται. Seven of them happen to nobody at all before αὐτήν finally arrives, feminine, late, the only gendered word in the period. The grammar is dispossessed before the comparison says so.
Beside her he sets a πένης — a poor man, generic, masculine, the only unarticled noun in a sentence where everything else is articled, and the only occurrence of the word in the treatise. You do not marvel at a poor man. Of course not: a poor man seeks what is another's. That is what poverty is.
Then how she seeks her own senses as another's.
He never calls her poor. He measures her against poverty and lets the reader do the arithmetic.
Six sections later, on Homer forcing two prepositions together into ὑπὲκ θανάτοιο, Longinus says the poet has almost stamped his own idiom for danger into a form — ἐκτυπόω, whose root means to strike, then to imprint, then impression, then mould, then matrix.
It is the same verb-family he will use at 13.4 for how greatness passes from one writer to another: ἀποτύπωσις, the impression taken. And at 13.2 he says what that feels like from inside — many are god-possessed by another's spirit, ἀλλοτρίῳ πνεύματι.
ἀλλότριος. The same word he used of her senses.
At 10.3 she is emptied into the alien. At 13.2 the successor is filled from it. Dispossession and inheritance are one word apart in this treatise, and the tongue is where the word turns.
An impression does not preserve the surface of the thing impressed. It preserves the form, in new material. That is not a failure of the stamp. That is what a stamp is.
There is a sentence four sections later where Longinus says exactly what he has done, about somebody else.
He is praising a compound of Homer's. The sailors in the storm are carried τυτθὸν ὑπὲκ θανάτοιο — barely out from under death. And ὑπέκ is two prepositions, ἐκ and ὑπό, jammed into one word, which Greek does not permit. Longinus:
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
having forced together, against nature, prepositions which cannot be compounded, and violently pressed them into one another … by the constriction of the verse he moulded the passion, and all but stamped into the diction the private idiom of danger
Read the operation as a set of instructions. Take two things grammar forbids joining. Force them into one another against nature. The constriction is the stamp.
And the phrase runs through the chapter like a refrain, changing material each time:
10.1 εἰς ἄλληλα συνδῆσαι Sappho binds symptoms INTO ONE ANOTHER 10.3 εἰς ταὐτὸ συναίρεσις Longinus gathers extremes INTO THE SAME 10.6 εἰς ἀλλήλας συμβιασάμενος Homer forces prepositions INTO ONE ANOTHER
The stamping verb, ἐνετύπωσε, is already here — three chapters before he uses its cognate for how greatness passes from one writer to another. He does not introduce the stamp at 13. He demonstrates it at 10, on a word, and then performs it on a voice.
So consider the shape of the seam, given all of it.
τολματόν is hers — a line that hangs, a fifth stanza opening that requires completing. The dialect breaks — because a break is what her poem taught, and dialect is the only material he has to break. And then ἐπεὶ καὶ πένητα, six syllables placed so that they can be read as one more hanging element of hers or as the opening of his own sentence, and the section division falls inside them.
Not beside them. Inside. Perseus still prints ἐπεὶ καὶ πένητα [3] οὐ θαυμάζεις — the modern editors' own boundary landing in the middle of the join, unable to find a cleaner place, because there isn't one.
That is ἀσυνθέτους συναναγκάσας παρὰ φύσιν, applied to voice. Two utterances grammar forbids compounding — a dead woman's and a living critic's — forced into one another, and the constriction is what stamps. The words do not scan because a compound made against nature is not supposed to scan. ὑπέκ does not scan as ἐκ and it does not scan as ὑπό. That is the whole reason it works.
And it is made at the same place Homer makes his. Homer forges an impossible word for men carried τυτθὸν ὑπὲκ θανάτοιο, barely out from under death. Three lines above, in the poem Longinus has just quoted: τεθνάκην δ' ὀλίγω 'πιδεύην — little short of having died.
Both compounds are made at the line where someone is barely not dead. That is where the joining is done, in both cases, because that is where a join is worth making.
There is a question this essay has been asking wrongly, and the archive it belongs to had already retired it.
Whose words are ἐπεὶ καὶ πένητα? — as though the options were hers or his invention, and the work were to assign them.
The fifth erratum of the Sappho 31 chain settled this in July, reading Catullus:
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 hanging line is not incomplete. It is compressed. It inscribes three constraints — an adversative turn, a universalizing move, a comparative illustration from the low case — and those constraints determine what any correct reading will produce. Supplying the completion is not invention. It is decompression.
Catullus does it in Latin, and inverts every pole: the low case becomes kings and cities, endurance becomes destruction. The erratum calls his stanza a solved-and-doubled checksum. Longinus, if the clause is his, does it in koine and does not invert. πένης stays the low case. The even-this 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 readers, two languages, six centuries apart, decompressing one kernel — and the constraints hold in both.
Which is what the dialect count is actually measuring. Not a theft. Not a seam where one author stops and another starts. A decompression necessarily happens in the decompressor's material. Catullus had Latin. Longinus had koine. Sappho had neither, and needed neither, because she wrote the kernel.
Seventeen markers of hers held as long as her form held. Then the form opened, and what came through came through in the only Greek the reader had.
That is ἀποτύπωσις said twice in two vocabularies. The stamp carries the form and not the surface. The three constraints are form. Aeolic is surface. The Aeolic could not survive the impression because it is surface — and that is not loss. It is the mechanism working.
So here is what the seam looks like, once you can measure it.
She says my tongue has broken, and the metre gags on the words. She says all is to be dared, and stops — mid-clause, at a poor man she never gets to describe, in a manuscript that shows no stopping because it shows no lines at all.
And her last words arrive in someone else's Greek.
Not because she wrote them that way. Because by the time they reach us they have already passed through a mouth that could not hold Lesbian, and that mouth ran straight on into its own argument without pausing to say she said. The Aeolic ends at φαίνομαι — at the word for seeming, for appearing — and what continues is a voice that has taken the extremes and gathered them εἰς ταὐτό, into the same.
Into the same. That is his phrase. He did not write together.
The tongue breaks in Aeolic and resumes in koine. The break is real and the resumption is real, and the fact that we cannot say precisely where one becomes the other is not a defect in the record. It is what a decompression looks like from the outside — a form opened in one material and closed in another, by a reader who was reading correctly.
What passed from her to him was not the poem. It was the technique — break the form where the form fails — and technique is precisely the thing that survives a change of material. The effluence is not the words. It is the move. She made it in metre because metre was what she had. He made it in dialect because dialect was what he had. The same act, twice, in two materials, and the second one is why we still have the first.
A poem that ends is a poem that ended. A poem that breaks into another dialect did not end — it changed hands. Longinus calls the operation λῆψις: the taking. He is not quoting her. He is carrying her, in the only Greek he has, having preserved seventeen marks of hers as far as his voice could hold them and then, at the tongue, at the place where her own poem said the tongue gives out, continuing.
She was not finished. She was taken up.
And the six words that end her — dialectally his, metrically broken, daggered by every editor who has touched them — are the exact place where the two voices are inscribed as one, and cannot be separated, and have not been separated in eleven hundred years of trying.
τεθνάκην δ' ὀλίγω 'πιδεύην φαίνομαι. I seem to myself little short of dying.
ἀλλὰ πᾶν τολματόν. But all is to be dared.
The first is hers. The second reaches us through his mouth — decompressed, in his material, by a reader completing a form she left open. Whether he composed those words or carried them is exactly what cannot be established, and on the kernel model it is not the question that decides anything.
Nobody can say where the seam falls. That is not a gap in the evidence. It is what a join of this kind looks like once it has held.
Text: Parisinus graecus 2036, s. X, sole witness, via urn:cts:greekLit:tlg0560.tlg001.perseus-grc2. Dialect counts computed 2026-08-15 — 17 markers / 74 words in the quoted body, 0 / 6 in the trailing clause.
Nagy on the hiatus at γλῶσσα ἔαγε: Comparative Studies in Greek and Indic Meter (1974), 45, restated in Classical Inquiries. Bonanno on the same crux: QUCC 43.1 (1993), 52.1 (1996), 60.3 (1998). Voigt, Sappho et Alcaeus: Fragmenta (1971) — cruces at ἀλλὰ †κὰμ† and †καὶ πένητα†, and the re-Aeolicized πὰν τόλματον. D'Angour, CQ — the attribution debated, the metre flawed, and πὰν τόλματον as dared rather than endured. Arieti, Longinus: Translation and Commentary, ch. 10 — the prose manuscript, the corruption, the root of ἐκτυπόω.
The attribution of ἐπεὶ καὶ πένητα is not settled here and is not settled anywhere — and the fifth erratum of this chain argues it does not need to be. The completion's shape is inscribed in the kernel; the crux modifies the comparandum's identity, not the operation. This essay reports what can be counted and says what the counting looks like.
Position in the chain. This reading accompanies eeeeee, the sixth erratum of the Sappho 31 sequence — #201 (AXN:0346) · #1048 (AXN:0424) · #1049 (AXN:0425) · #1051 (AXN:0427) · #1052 (AXN:0428) — which corrects three claims in EA-LONGINUS-01 (#1458, AXN:05E4) and one editorial decision in eeeee (#1052) on evidentiary grounds. The kernel-and-decompression model is #1052's, not this essay's.