Seventh in the Sappho 31 erratum sequence, written as a standalone philological paper about Longinus and Sappho. Argues that the transmitted configuration at the poem's edge — a lyric closing on an agentless verbal adjective, a prose voice continuing without citation formula, a dialect changing mid-breath — is not damage but the object itself: a poem bracketed by two open positions (the unoccupied seat at kenos, the empty agent at tolmaton), whose two formal ruptures are one operation performed on the reader, quoted by the one ancient critic whose supreme instance of sublimity is a composed silence (the Ajax of 9.2, phones dicha), at the exact moment he performs upon the poem the operation his treatise later names (enetypose at 10.6 before apotyposis at 13.4). The dialect seam is measured with its limit stated (17 markers in 74 words, then zero in six; P(0 in 6)=0.209, not significant alone); the continuation is construed continuously, peneta as object of thaumazeis across the modern section division; the five-scholar attribution tradition from Robortello 1554 is credited, with Bergk's abandoned integrum-carmen conjecture at line 16 as the argument's one uncontaminated corroboration. Closes on the series of completions — Longinus's bare thaumazeis, Catullus's inverted Latin (Catullus is kenos), the anthology cut, Augustine's gustavi, the modern restorations ending open — each filling cutting a fresh hollow, which is why the form is dysexaleiptos. The heteronym scholarship is cited as scholarship throughout. Body written to stand alone; chain position recorded in frontmatter only.
Object-level philology against the Perseus TEI (tlg0560.tlg001.perseus-grc2) with all counts computed and limits stated in the text; apparatus consulted directly in Roberts 1907 and Bergk PLG4 III on archive.org; address and citation-verb censuses reproducible; the heteronym scholarship cited as scholarship beside Nagy, Bergk, Voigt, Bonanno, D'Angour.
The dialect count falls to a recount; the address census falls to a counterexample (a bare post-quotation perception verb elsewhere, or a vocative at 10.3); the continuous construal falls if peneta cannot be governed by thaumazeis; the significance claim is already conceded (0.209). The design-reading is a choice made explicit; the paper states that a reader who calls the configuration damage has made a choice, not an observation, and the same holds in reverse.
The last transmitted words of Sappho 31 are these:
τεθνάκην δ' ὀλίγω 'πιδεύην φαίνομαι. ἀλλὰ πᾶν τολματόν, ἐπεὶ καὶ πένητα — οὐ θαυμάζεις, ὡς ὑπὸ τὸ αὐτὸ τὴν ψυχὴν τὸ σῶμα τὰς ἀκοὰς τὴν γλῶσσαν τὰς ὄψεις τὴν χρόαν …
I seem to myself little short of dying. But all is to be dared, since even a poor — do you not marvel how, at one moment, she seeks after soul, body, hearing, tongue, sight, complexion …
Since Robortello's editio princeps of 1554 the tradition has treated the middle of this run as a wound. Robortello excised ἐπεὶ καί; Bergk bracketed ἐπεὶ καὶ πένητα as belonging ad Longini orationem, conjecturing a lost citation verb (ἐπεῖπεν); Ahrens conjectured another (πᾶν τὸ ἀσμάτιον ὡς ἔπος εἰπεῖν); Hermann and Heller emended differently again; Crusius printed the ode as ending at φαίνομαι ἄλλα; Voigt set cruces desperationis around †καὶ πένητα†.¹ Seven editors across four and a half centuries have split, joined, re-accented, bracketed or deleted these few syllables. Their disagreement is total. Their premise is shared: something is broken here, and the editor's task is to find the break.
This paper argues that the premise is false, and that the transmitted configuration — a lyric closing on an agentless verbal adjective, a prose voice continuing without a citation formula, a dialect that changes mid-breath — is not damage but the object itself: a poem whose final formal act is to require a voice it does not supply, quoted by the one ancient critic whose stated theory of the sublime predicts exactly this shape, at the exact moment he performs upon the poem the operation his treatise will later name.
Three bodies of evidence carry the argument: the grammar and deixis of the poem as transmitted (§§II–III); the placement, vocabulary and demonstrated practice of Longinus's own chapters 9–10 (§§IV–VII); and the subsequent history of the line's completions, from Catullus through Augustine to the present, which behaves as the reading predicts and as the damage-reading cannot explain (§IX).
One preliminary fact governs everything that follows. Parisinus graecus 2036, the sole witness, writes the poem as continuous prose, as it writes every verse quotation; Roberts, editing from the manuscript, made a deliberate exception at this passage to reproduce "the exact text (continuously written) found in P."² There are no stanzas, no line-breaks, no quotation marks and no closing formula in the witness. Every visual boundary in every printed Sappho 31 is editorial. The famous hanging line does not hang in the manuscript: it is set hanging, nine centuries later, by editors — which means the question where does Sappho end? has no diplomatic answer, and only internal criteria remain. This paper counts them.
The poem opens on a distal demonstrative:
φαίνεταί μοι κῆνος ἴσος θεοῖσιν / ἔμμεν' ὤνηρ, ὄττις ἐνάντιός τοι / ἰσδάνει
κῆνος points away — at someone removed from the speaker's here. The relative that unpacks him is indefinite: ὄττις, whoever. Cranes has argued that this man, godlike because he sits face-to-face (ἐνάντιος) with the beloved and receives her voice and laughter undestroyed, is best read not as a dramatic rival but as the future reader: the one who will one day sit opposite the inscribed text and possess what should be impossible, presence with the dead.³ The epigraphic habit supports her: archaic inscribed objects routinely address readers who do not yet exist — the Nestor cup speaking as I to a you who is whoever drinks, the Dipylon oinochoe's whoever of all the dancers, Mantiklos's dedication anticipating its own readers. A lyric composed for circulation opens, on this reading, by designating a seat and leaving it empty.
Whether or not one accepts the full reading, the grammatical facts stand on their own: the poem's first noun-phrase is a distal deictic plus an indefinite relative — a position, not a person.
The poem's other end closes on a form that is the syntactic mirror of its opening.
τολματόν (P: παντόλματον, one word)⁴ is a verbal adjective in -τός. It is not a finite verb. It is neuter and predicative — all [is] to-be-dared — and the formation cannot govern a nominative agent: it names an action as available and structurally excludes a doer. This is not an omission the poet neglected to repair; it is what the -τός formation is. The line does not fail to say who dares. It is built so that no one inside the poem can be the one who dares.
So the poem is bracketed: an unoccupied seat at line 1, an unoccupied agency at line 17. A reader-shaped hole at the front; a doer-shaped hole at the back. Between them, the poem transacts the transfer that makes both holes necessary — and marks the transfer in its own verb.
The first word of the poem and the last word before the disputed run share a root and exchange a subject:
φαίνεταί μοι κῆνος … → … τεθνάκην δ' ὀλίγω 'πιδεύην φαίνομαι.
He appears to me becomes I appear. Across four stanzas the verb of appearing migrates from the man to the speaker; she ends as the object of the seeming she began by performing. The frame is closed. What follows it is not a fifth act of the same drama but the poem's disposal of itself: all is to be dared — by someone the grammar refuses to name, seated somewhere the deixis has already prepared.
Sappho 31 breaks its own form at exactly two points, and at both points the break is performative — it makes the reader do the thing the line names.
Line 9. ἀλλὰ κὰμ μὲν γλῶσσα ἔαγε — but my tongue has broken. Between γλῶσσα and ἔαγε stands a hiatus: two short vowels colliding where Greek song forbids it. Nagy identified the effect fifty years ago: hiatus, in the Latin, is a gagging; "the sequencing has produced a non-sequencing"; the interruption of the music is intended.⁵ The line says a breaking and makes the reader's mouth break. The tradition accepts this reading — and then obelizes the words around it (Voigt daggers †κὰμ†; Bonanno needed three articles for the two words γλῶσσα ἔαγε⁶) while celebrating the effect.
Line 17. ἀλλὰ πᾶν τολματόν — all is to be dared. The rupture here is not in the sound but in the agency: an instruction with an empty subject. The line says a daring and makes the reader dare, because only a reader can occupy the slot the form withholds.
| line 9 | line 17 | |
|---|---|---|
| form | ἔαγε — perfect indicative, subject expressed (γλῶσσα) | τολματόν — verbal adjective, agent excluded by formation |
| rupture | in the sound: hiatus | in the agency: no doer supplied |
| what the reader performs | the gag, in the mouth | the daring, in the completion |
The tradition's verdicts on these two ruptures diverge absolutely — the first is art, the second corruption — and the divergence tracks nothing in the phenomena. It tracks completeness: we possess the far side of line 9, so its rupture reads as control; line 17 has no far side, so its rupture reads as accident. An apparatus can record sound and corrupt and nothing else; a poet who breaks form deliberately at a terminus is invisible to it, because deliberate rupture leaves an editor nothing to do, and emendation is what editors do. The cruces at line 17 are not a finding about the text. They are the notation reaching the edge of its categories.
The strongest evidence that Longinus did not regard the poem's shape as damage stands one section before he quotes it.
ὕψος μεγαλοφροσύνης ἀπήχημα. ὅθεν καὶ φωνῆς δίχα θαυμάζεταί ποτε ψιλὴ καθ' ἑαυτὴν ἡ ἔννοια δι' αὐτὸ τὸ μεγαλόφρον, ὡς ἡ τοῦ Αἴαντος ἐν Νεκυίᾳ σιωπὴ μέγα καὶ παντὸς ὑψηλότερον λόγου. (9.2)
Sublimity is an echo of greatness of mind. Whence, even apart from voice, the bare thought is sometimes admired for its very greatness — as the silence of Ajax in the Nekyia is great, and higher than any speech.
Four of the operative words — ἀπήχημα, φωνῆς δίχα, Νεκυίᾳ, σιωπή — occur nowhere else in the treatise.⁷ Longinus concentrates his hapax vocabulary into one sentence to say one thing: the supreme instance of the sublime is a man who, asked to speak, does not speak, and walks away. A composed non-answer. Greatness transmitting with the voice subtracted.
Then, in the next breath of the treatise, he quotes Sappho 31 — the poem that reaches its extremity and stops at a form that cannot say who acts. The selection is the argument. A critic whose paradigm of the highest sublimity is an inscribed silence did not choose this poem despite its ending. Homer had to write Ajax's silence — a non-speaking composed on purpose, at the point where speech was owed. Sappho, on the reading defended here, does at line 17 what Homer does in the Nekyia; and Longinus, who names the one as παντὸς ὑψηλότερον λόγου, quotes the other whole.
If the poem is complete at the hanging line, the transmitted continuation must be someone else's — and there should be a measurable trace of the hand-over. There is. It sits in the dialect.
Longinus preserves Sappho's Aeolic scrupulously across the quoted body: κῆνος (for ἐκεῖνος), γελαίσας, the apocopated κὰμ and κὰδ, ἔαγε, χρῶ, ὀππάτεσσι with its gemination and Aeolic dative plural, ἄκουαι, κακχέεται, παῖσαν, ἀγρεῖ, ἐμμί, τεθνάκην, ὀλίγω, 'πιδεύην, βροχέως, φωνᾶς with its long alpha — seventeen dialect markers in seventy-four words, one per 4.4 words, sustained without a lapse to the last syllable of φαίνομαι.⁸ Then, in the six words ἀλλὰ πᾶν τολματόν, ἐπεὶ καὶ πένητα: zero, where 1.4 were due at the body's rate.
The distribution matters as much as the totals. Scribal decay, fatigue and normalization are gradual; they predict a taper — one marker per four words, then per six, then per eight, then none. The observed distribution is 1-per-4.4 and then a step to zero at a single word. A step is what a boundary looks like; a taper is what erosion looks like.
Stated with its limit, because the limit is real: at p = 17/74, the probability of zero markers in six words drawn from the same distribution is (1 − p)⁶ = 0.209. One in five. The dialect absence is not significant in isolation — a twelve-word zero-run would be needed for p < .05 — and no argument in this paper treats it as a proof. It is a constraint, whose force lies in conjunction with what follows.⁹
The tradition's own practice acknowledges the seam it does not name. To print the trailing line as Sappho's, an editor must first restore her dialect to it: Bergk and Voigt print not the manuscript's παντόλματον nor the vulgate's πᾶν τολματόν but πὰν τόλματον — split, barytone, recessive, Lesbian.¹⁰ The restoration is silent; it lives in the accents. A reader of a modern Sappho sees an Aeolic line and cannot know the Aeolic was supplied. Meanwhile the standing objection to the words' authenticity — that they are "metrically flawed"¹¹ — may be the same observation in a different notation: an un-Aeolicized line in an Aeolic poem does not scan, and re-Aeolicizing it is what makes it scan. The metrical flaw and the dialect flaw are one seam, twice detected, twice filed as damage.
The seam's location was reached once before, by the passage's greatest editor, from the opposite direction. Bergk's apparatus records his own earlier view: "Antea conieceram versu 16 integrum carmen absolvi" — I had earlier conjectured that the whole poem is completed at line 16 — after which Longinus would resume with a citation formula (πᾶν τὸ ἀσμάτιον ἐπεῖπον, "I have quoted the whole little song").¹² Line 16 is φαίνομαι. Bergk placed the poem's completion at the exact word where the dialect stops, from syntax and sense alone, without a marker count — and then retreated to brackets and a conjectured ἐπεῖπεν, because integrum carmen, a complete poem ending mid-gesture, was a category his apparatus could record only as a conjecture to be abandoned. His word survives in the note: integrum. Whole.
Read without a lacuna, the transmitted text parses cleanly — and the modern section division falls in the middle of the parse:
ἀλλὰ πᾶν τολματόν, ἐπεὶ καὶ πένητα ⟦3⟧ οὐ θαυμάζεις, ὡς … ἐπιζητεῖ …;
πένητα is not an accusative reaching for a lost Sapphic verb. It is the direct object of θαυμάζεις, four words later, across a boundary no ancient reader could see: since you do not marvel even at a poor man — how she seeks after soul, body, hearing, tongue, sight, complexion, all as another's and gone? θαυμάζω takes both an accusative and a ὡς-clause; nothing need be supplied. Perseus still prints the section number between the object and its verb — the modern apparatus, unable to find a cleaner place for its boundary, setting it inside the join.¹³
On this construal Longinus's sentence is an argument, and its hinge word is already in it. A poor man seeks (ἐπιζητεῖ) what belongs to another: that is what poverty is, and it is unremarkable — you do not marvel at it. What is remarkable is a woman seeking her own faculties as though they belonged to someone else: πάνθ' ὡς ἀλλότρια διοιχόμενα. Poverty externalized against poverty internalized. She is never called poor; πένης — a hapax in the treatise, and the only anarthrous noun in a sentence where ψυχήν, σῶμα, ἀκοάς, γλῶσσαν, ὄψεις, χρόαν all carry articles — is generic, masculine, a foil. She is measured against poverty and exceeds it, and the reader does the arithmetic.
The sentence's grammar performs the dispossession before the comparison states it: eight third-person verbs — ἐπιζητεῖ, ψύχεται, κᾴεται, ἀλογιστεῖ, φρονεῖ, φοβεῖται, τέθνηκεν, φαίνηται — and not one expressed subject, until αὐτήν, feminine, arrives late as the only gendered word in the period. And the six articulated nouns are an inventory: an estate, itemized, every holding marked another's, and gone. The body that was Longinus's figure for composition at 10.1 (καθάπερ ἕν τι σῶμα ποιεῖν) reappears at 10.3 as an item on the list of losses.
That ἐπεὶ καὶ πένητα is Longinus's rather than Sappho's has been argued since the sixteenth century: Robortello by deletion, Bergk by brackets, Ahrens by a conjectured citation formula — three of the five conjectures at this crux restore a verb of saying, the nineteenth century having noticed the same absence this paper counts (seventy-four citation verbs in the treatise; none at Sappho; she is named once, at 10.1, and never again).¹⁴ But those scholars drew from the attribution an editorial consequence: Longinus's, therefore not part of Sappho 31 — and out it went.
The consequence does not follow, because the words are not the kind of thing that has one owner. The hanging line inscribes a shape — adversative turn, universalizing claim, comparative illustration from the low case — and supplying that shape is neither quotation nor invention. It is a completion executed in the completer's material, which is the only material a completion can ever be executed in. Aeolic cannot break into Aeolic; the filling of a Lesbian socket by a first-century critic can only arrive in koine. The dialect seam of §V is therefore not evidence of theft or of corruption. It is the signature of the operation: form carried, surface not carried — and the surface stops at φαίνομαι because that is where the die ends and the receiving material begins.
So the words are Longinus's as material and Sappho's as form, and an edition that deletes them does not restore the poem — the poem was never touched — it deletes the demonstration: the earliest surviving instance of the line doing what it was built to do, transmitted in the same document as the line itself.
Four sections after the quotation, Longinus describes — of Homer — precisely the operation the seam records:
τὰς προθέσεις ἀσυνθέτους οὔσας συναναγκάσας παρὰ φύσιν καὶ εἰς ἀλλήλας συμβιασάμενος … τῷ μὲν συνεμπίπτοντι πάθει τὸ ἔπος ὁμοίως ἐβασάνισεν, τῇ δὲ τοῦ ἔπους συνθλίψει τὸ πάθος ἄκρως ἀπεπλάσατο καὶ μόνον οὐκ ἐνετύπωσε τῇ λέξει τοῦ κινδύνου τὸ ἰδίωμα. (10.6)
Forcing together, against nature, prepositions that cannot be compounded, and pressing them violently into one another … he tortured the verse into likeness with the passion, and by the constriction he moulded the passion and all but stamped into the diction the private idiom of danger.
The word so forged is ὑπὲκ θανάτοιο — out from under death — and it scans as neither ἐκ nor ὑπό; that is why he admires it. Grammar forbids the join; pressure performs it (συνθλίψις, βασανίζω). And the chapter carries the operation as a refrain, one preposition across three materials: Sappho binds symptoms εἰς ἄλληλα (10.1); Longinus gathers the extremes εἰς ταὐτό (10.3, sealed by ὡς ἔφην, the ring that claims 10.1 and 10.3 as one utterance with the poem inside it); Homer forces the prepositions εἰς ἀλλήλας (10.6).
Two consequences. First, the join at the seam has the properties Longinus assigns to Homeric joining: two utterances that convention forbids compounding — a dead poet's and a living critic's — pressed into one another with no connective, held by the load of the socket rather than by grammar, and not expected to scan, because compounds made παρὰ φύσιν are not. The metrical objection tests the join by the standard the join is defined by exceeding. Second, the stamping verb is already here: ἐνετύπωσε at 10.6 precedes ἀποτύπωσις at 13.4 by three chapters. The treatise does not first theorize transmission and then illustrate it. It performs the operation on a word, and on a voice, and names it afterward.
Read from the socket, the treatise's transmission vocabulary stops being a family of metaphors and becomes one mechanism described repeatedly — and the morphology says so.
ἀπ-ήχημα (9.2) · ἀπό-ρροια (13.2) · ἀπο-τύπωσις (13.4)
Three terms, each a hapax, each compounded with ἀπό-: away from, out of, back from.¹⁵ Every one presupposes a source that has released something and no longer holds it. And the prefix inverts at the single point where the operation is executed rather than named: at 10.6, ἐν-ετύπωσε, stamped into. ἀπό
ἀπήχημα. An echo is not a resemblance; it is an event with three requirements — a sound, a cessation, and a returning surface — and there is no echo while the source still sounds. 9.2 does not say the sublime is like an echo of greatness. It says the sublime is the return that occurs once a voice has stopped: a definition of the socket, given one section before a poem that ends by stopping, in the sentence whose supreme example is a silence.
ἀποτύπωσις. A stamp requires a die, a material, and pressure — and a die works because it is hollow where the material will go. The socket is the die; τόλματον is the hollow; and what fills a hollow is never the die's substance but always the receiving material's. This is why the impression at 13.4 carries form and cannot carry surface: surface stays with the die. Sappho's Aeolic is surface. Seventeen markers stop at φαίνομαι because that is where the die ends and the wax begins.
δυσεξάλειπτος. At 7.3 the criterion of the genuine sublime is that it is hard to wipe out, in the wax-tablet vocabulary whose verb for forgetting is ἐξαλείφεσθαι. Under the socket reading this is not a claim about durable content but about regenerability of function. Robortello deleted the filling in 1554; Bergk bracketed it; Crusius cut the ode before it; Voigt daggered it — four removals across four centuries, and each time the next reader filled the position again, because a socket is not content but a shape that produces content in whoever meets it. You cannot delete a hollow by deleting what is in it.
The strongest fact about the line's afterlife is that its completions do not close it. They reproduce it.
Longinus. His filling terminates in a question — οὐ θαυμάζεις … ; — and the address is bare. The treatise's post-quotation perception verbs otherwise carry a vocative without exception (ὁρᾷς, ὦ ἑταῖρε; ἐπιβλέπεις, ἑταῖρε); θαυμάζεις alone names nobody.¹⁶ A question posed to no addressee is posed to whoever holds the text. Sappho withheld the agent of an action; Longinus withholds the answerer of a question, and the load passes through him to the reader.
Catullus. Catullus 51 preserves the poem's shape into Latin and answers the hanging line at a distance, inverting each pole of its inscribed comparison — the low case becomes reges … et beatas urbes, endurance becomes destruction — and then ends there, on perdidit urbes, with no return to the erotic frame: a translation that terminates in its own suspension. And Catullus does more than fill the terminal socket; he occupies the initial one. He is the later man who sits face-to-face with the text and receives the voice across six centuries — Catullus is κῆνος, the seat of line 1 taken by the reader the deixis had reserved for him.¹⁷
Augustine. Confessions 10.27.38 transforms the poem's sensory catalogue organ by organ — and its lock is the tongue: Sappho's γλῶσσα ἔαγε, the organ of speech breaking, becomes Augustine's gustavi, the same organ receiving; the failure of utterance rotated into sacramental intake. "What rotates is not the organ but the address."¹⁸ And the address is the whole method: vocasti, clamasti, rupisti … coruscasti … fragrasti … tetigisti — every verb second person, sero te amavi — thirteen books sustained toward an interlocutor whose reply is never transcribed. Longinus left a question with no addressee; Augustine supplies the addressee and withholds the reply, generalizing the socket from a terminal line to a compositional principle. That the operation is portable off the line entirely is what Augustine demonstrates — whatever his route to the material, and none is claimed.
The moderns. The reconstruction of the "lost" stanza proposed by Cranes runs Catullus's inversion backward under stated constraints and arrives at:
ἀλλὰ πᾶν τόλματον· ἐπεί σε, Κύπρι, / καὶ πένητά γε κἄ βασιλῆα δάμασσε· / καὶ πόλεις ὀλβίους ἀπέθηκας ἄφαντον· / γράμμασι μολπὰν¹⁹
Two features matter here regardless of the reconstruction's status as recovered wording (it claims none). It reaches πένητα independently — derived from reges et urbes run in reverse, not from the Longinus crux — so the low-case comparandum is attained by two routes six months apart. And its final Adonic names the mechanism: γράμμασι μολπὰν, song in letters — a filling whose content is a description of what a socket does, committed to the position the poem left open. It stands to Sappho 31 exactly as Longinus's koine clause and Catullus's inverted Latin stand to it: the next term of a series, not the recovery of a first term — and, like every term before it, it ends open.
The series, then:
κῆνος — a seat with no occupant → πὰν τόλματον — an act with no agent → οὐ θαυμάζεις; — a question with no addressee → perdidit urbes — a translation with no return → vocasti … te — an address with no reply → γράμμασι μολπὰν — an inscription awaiting its reader
Each filling cuts a fresh hollow in new material. This, finally, is why the form is δυσεξάλειπτος in the strong sense: not because its words resist deletion — they have been deleted repeatedly — but because the act of completing it reproduces its incompleteness in the completer. An echo returning from a surface is a sound that can echo again. Deletion cannot outrun a form that manufactures its own emptiness in whoever answers it.
Claimed, as fact about the transmitted object. The sole witness writes the poem as prose; every boundary is editorial. The poem's frame closes at φαίνομαι, on the root it opened with. τολματόν is a -τός verbal adjective and cannot govern a nominative agent; P transmits it as the compound παντόλματον. Seventeen Aeolic markers run at one per 4.4 words and stop, as a step and not a taper, at φαίνομαι; the six following words carry none. The continuation parses continuously, πένητα as object of θαυμάζεις, across the modern section division. πένης, ἀπήχημα, ἀπόρροια, ἀποτύπωσις, σιωπή, φωνῆς δίχα and Νεκυίᾳ are hapax legomena; the three transmission nouns all carry ἀπό-, against ἐνετύπωσε at 10.6. The treatise's post-quotation perception verbs carry vocatives except θαυμάζεις. Seventy-four citation verbs occur in the treatise; none at this quotation; Longinus's one first-person marker, ὡς ἔφην, asserts continuity across the poem rather than closure of it. The attribution of ἐπεὶ καὶ πένητα to Longinus has been in print since 1554, and Bergk once located the poem's completion — integrum carmen — at line 16.
Claimed, as reading. That this configuration is the poem's design as transmitted: a lyric bracketed by an unoccupied seat and an unoccupied agency, closing φωνῆς δίχα in the manner its quoter had just named as the highest sublimity; and that ἐπεὶ καὶ πένητα is the first filling of the terminal position — Longinus's as material, Sappho's as form — followed in kind by Catullus, by Augustine's generalization of the method, and by the modern reconstructions, each of which ends open.
Not claimed. That Sappho intended the ending rather than the ending being what survives: the transmitted object cannot decide this and the argument does not require it — the claim is about the object's structure and its demonstrable effects on its completers. That the dialect step is significant alone: it is not (p ≈ 0.209). That any completion recovers Sappho's wording: none does and none here pretends to; the series is a class of texts exhibiting one structural property, and membership is not a claim about contact. That Longinus or Augustine consciously theorized what they performed: the paper's claim is that the treatise performs at chapter 10 what it names at chapter 13, and that the naming vocabulary looks backward (ἀπό-) because it was written by someone who had already let the source go.
A reader who calls the transmitted configuration damage has made a choice, not an observation — the same choice available at line 9, where the tradition, holding the far side, chose otherwise.
1. Robortello (1554), om. ἐπεὶ καί; Bergk, Poetae Lyrici Graeci⁴ III (1882) 90: "Quae sequuntur — ἐπεὶ καὶ πένητα — uncis inclusi, nam videntur haec ad Longini orationem pertinere, fort. … ἐπεῖπεν"; Ahrens, Hermann, Heller ap. Bergk ibid.; Crusius, Anthologia Lyrica (1897) 195; Voigt, Sappho et Alcaeus: Fragmenta (1971) fr. 31.
2. Roberts, Longinus on the Sublime: The Greek Text Edited after the Paris Manuscript (Cambridge 1907), note on p. 70 ll. 1–17. Text throughout: urn:cts:greekLit:tlg0560.tlg001.perseus-grc2 (base text Roberts 1907); Parisinus graecus 2036, s. X, sole witness.
3. Cranes, "ΦΑΙΝΕΤΑΙ ΜΟΙ: Sappho 31 and the Inscription of the Future Reader" (2026), §§I–II, with the epigraphic comparanda (Nestor cup; Dipylon oinochoe; Mantiklos).
4. Bergk ibid.: "P παντόλματον, vulgo πᾶν τολματόν." No neutral form of the word exists to cite; this paper quotes the vulgate split when quoting the run and the manuscript compound when the compound is at issue.
5. Nagy, Comparative Studies in Greek and Indic Meter (1974) 45.
6. Bonanno, QUCC 43.1 (1993) 61–68; 52.1 (1996) 173–175; 60.3 (1998) 143–148. Voigt daggers †κὰμ† at fr. 31.9.
7. Counts computed 2026-08-15 against the TEI; likewise the citation-verb census (φησί/λέγει/λέγων/γράφει, 74 total), the address census (§IX n.16), and the marker counts of §V.
8. Full marker table with Attic equivalents in the companion apparatus (deposit #1474 §III). One counting note works against the argument: ἐπιρρόμβεισι is hyphenated across a line-break in the TEI and eludes stem search; counting it makes eighteen markers and raises the probability in n.9.
9. (1 − 17/74)⁶ = 0.209.
10. Voigt ibid.; Bergk ibid. (πὰν τόλματον).
11. D'Angour, "'Fearless, bloodless … like the gods': Sappho 31 and the rhetoric of 'godlike'," CQ, n.21: whether the words belong to the poem "is debated"; they are "metrically flawed."
12. Bergk ibid.: "Antea conieceram versu 16 integrum carmen absolvi, ut deinde sit scribendum πᾶν τὸ ἀσμάτιον ἐπεῖπον· εἶτα οὐ θαυμ., similiter Ahrens…"
13. Perseus edition ad loc., section division 10.2/10.3.
14. Census as n.7; Sappho named once, 10.1.
15. ἀπήχημα 9.2; ἀπόρροια 13.2; ἀποτύπωσις 13.4; ἐνετύπωσε 10.6 — each ×1 in the treatise.
16. ὁρᾷς, ὦ ἑταῖρε (26.2); ἐπιβλέπεις, ἑταῖρε (9.6); θαυμάζεις bare (10.3). ὅρα δέ at 44.6 stands inside reported speech already carrying ὦ βέλτιστε.
17. Catullus 51.13–16 (otium … perdidit urbes); "Catullus is κῆνος": Cranes (n.3) §VI.
18. Sharks, "The Sapphic Lock in Augustine: Operator Transform of Fragment 31 in Confessions 10.27" (2026): the lock γλῶσσα ἔαγε → gustavi; "what rotates is not the organ but the address"; Confessions 10.27.38 (vocasti et clamasti et rupisti … gustavi … tetigisti; sero te amavi).
19. Cranes (n.3) §V.3, with its five stated constraints and its own caveat: "The reconstruction is speculative … γράμμασι μολπὰν stands as hypothesis." Its πένητα is derived from Catullus's reges et urbes run backward, independently of the crux at Longinus 10.2–3.