The most physical love poem to survive from antiquity reaches us through three sets of hands — a critic who quoted it, a Roman who rewrote it, and the editors who still argue over its words. None of them is invisible. This is an instrument for seeing each one.
Around 600 BCE, on the island of Lesbos, Sappho wrote a poem about watching someone she loved talk to someone else, and the body coming apart while she watched. It is, in Emmet Robbins's phrase, “probably the single most famous poem to come down from Antiquity.” And we have it for one reason only: some six centuries later a literary critic we call ‘Longinus’ copied four stanzas of it into an essay — to show how a great writer selects the true signs of feeling — and then, having made his point, stopped writing.
So the poem we read is a fragment. But look closely at where it breaks: not at a tear in a papyrus, but one line into the fifth stanza, exactly where the critic lifted his pen. The famous unfinished ending — ἀλλὰ πὰν τόλματον…, “but all must be dared…” — is not Sappho trailing off. It is Longinus trailing off, mid-quotation, because he had quoted enough.
And the first way most of Europe met this poem was not in Greek at all. It was in Latin, in a poem by Catullus — number 51 — written around 50 BCE, one of only two poems he cast in Sappho's own stanza. Catullus translates her opening almost word for word, then quietly changes her: he softens the violence, he cuts four of her symptoms (including the one this page is named for), and he ends not with her broken line but with a brand-new stanza, in a Roman voice, scolding himself about idleness. And his edit cast a long shadow: the first renderings of the poem into the modern languages of Europe descended from Catullus's Latin, not Sappho's Greek — even the version that ruled English for a century, Ambrose Philips's, owes its shape as much to the Roman as to her.
Three instruments below let you watch each hand at work: the body Sappho built, the translation Catullus made of it, and the long relay that carried both to you.
Longinus quoted the poem to praise one specific skill: not invention but selection — “she always chooses her strokes from the signs which she has observed to be actually exhibited” in lovers, and her genius is “the felicity with which she chooses and unites together the most striking … features.” Here are the signs she chose, in the order the body fails. Tap each to read it and to see whether it survived into Catullus's Latin. The two it marks in bronze are the cruxes — the places where even the Greek is uncertain, or the colour will not hold still.
Read the bronze badges down the list and a shape appears. Catullus keeps the failures of the head — voice, tongue, the fire, the ears, the eyes — and drops the failures of the whole body: the cold sweat, the trembling, the green pallor, the nearness of death. Sappho's fourth stanza, the one where the body dissolves entirely, is the one he declines to carry. The classicist D. E. W. Wormell suggested it had “too feminine a tone” for him; whatever the reason, the most bodily quarter of the most bodily poem is exactly what the translation leaves behind.
The one image everyone remembers from Sappho — greener than grass — is the one her most famous translator cut.
Here are the two poems side by side: Sappho's Greek in bronze, Catullus's Latin in silver. Catullus is not careless — he is a great poet making choices, and you can name every one. Tap a kind of move to light up where he makes it. Kept: he follows her exactly. Added: he puts in what she never wrote. Softened: he keeps the sense but pulls the punch. Cut: he drops her entirely. Invented: the stanza with no Greek behind it at all.
Two holes, for opposite reasons. Sappho's poem stops because the man who saved it stopped copying. Catullus's poem has a gap in its second stanza — line 8 — because somewhere in a thousand years of hand-copying, one Latin line fell out and never came back; editors pencil in a guess (vocis in ore, “of voice in the mouth”) in brackets, to mark that it is theirs, not his. One poem is missing its end because of a reader; the other is missing its middle because of a scribe.
The sharpest single choice is the tongue. Sappho writes γλῶσσα †ἔαγε† — the tongue has broken, snapped like a stick (the verb is so violent that the one manuscript garbled it, and editors still fight over what she actually wrote). Catullus writes lingua sed torpet — the tongue grows numb. Dolores O'Higgins's reading is that the Latin reaches the same silence but “lacks the violence” of the Greek: a shattering becomes a falling-asleep. The most quoted comparison in the scholarship, and you can hear it without any Latin: broken against numb.
The venue's core sample: a vertical slice through everyone whose hands the poem passed through. Tap a layer. The bronze track is Sappho's Greek; the silver branch is Catullus's Latin, which runs alongside on its own thread. The one marked layer is the hinge of the whole story.
The daggers in Instrument II are the visible scars of this relay. They sit exactly where Longinus's tenth-century copyist made a mess of an Aeolic poem four hundred years older than him, in a dialect he did not speak. When you read “greener than grass,” you are reading Sappho — but the words around it have been handled by a Roman, a Byzantine scribe, a Renaissance printer, and a century of editors with obeli, and every one of them left a fingerprint. The poem is not a relic behind glass. It is a thing that has been carried, and you can still feel the carrying.
Every English version below is out of copyright and transcribed verbatim from a named edition (sourced at the foot of each). The Sappho renderings descend from the Greek; the Catullus renderings descend from the Latin — so a reader in English has almost always been reading one of two different poems without being told which.
And one translator who refused. The Hon. George Lamb, Englishing all of Catullus in 1821, simply left poem 51 out — explaining in his preface that there was no point: Ambrose Philips's century-old version of the ode, he wrote, “bids defiance to competition.” He would rather print a hole than lose the contest. The most honest thing in his book is the gap where the poem should be — another absence, deliberate this time, in a poem already made of absences.
The Greek text. Fragment 31 is reproduced after the constituted text of E.-M. Voigt, Sappho et Alcaeus (1971), as printed (with Voigt's obeli) in the standard reference presentation; it is numbered 31 in Lobel–Page and Voigt. The poem survives only as quoted in [Longinus], On the Sublime 10.2 — there is no papyrus and no medieval manuscript of Sappho carrying it. The Greek text in On the Sublime itself was consulted via the Perseus Digital Library (Roberts's edition). Longinus's own words are quoted in the public-domain English of H. L. Havell (1890) (Project Gutenberg 17957). Editions genuinely disagree about several words — see “the honest edges.”
The daggers. The obeli (†) are editorial marks meaning “the transmitted text here is corrupt.” At line 9 the manuscript reading of the tongue-clause is damaged; Voigt daggers it, while C. G. Cobet (1873) conjectured πέπαγε (“the tongue has stuck fast”), a reading George Devereux, CQ 20 (1970) adopted in arguing the symptoms describe an anxiety attack. Line 13 (the cold-sweat clause) and the last words of line 17 (καὶ πένητα) are likewise daggered. We display the cruxes rather than silently choosing a reading.
The translation comparison. That Catullus 51 keeps the failures of voice/tongue/fire/ears/eyes but omits Sappho's sweat, trembling, “greener than grass,” and near-death (the whole fourth stanza of dissolution) is the standard reading; the “too feminine a tone” suggestion is D. E. W. Wormell (1966), reported in the Dickinson College Commentaries. That torpet (“numb”) softens Sappho's ἔαγε (“broken/shattered”) follows Dolores O'Higgins, “Sappho's Splintered Tongue,” AJP 111 (1990) — her exact wording is paraphrased here, not quoted, as we could not consult the printed essay directly. The otium stanza's status is debated: most scholars hold it belongs to the poem (it is universally transmitted in place, in one of only two Sapphic-stanza poems Catullus wrote) and read it as his own Roman, moralizing turn with no Sapphic counterpart (so Chris Childers); we flag rather than resolve the minority unease about its fit.
The reading of τόλματον. The last surviving line, ἀλλὰ πὰν τόλματον, is usually rendered “all must be endured”; Armand D'Angour (2013) argues for “all must be dared,” which would turn the lost ending toward hope rather than resignation. We give “dared” in the crib and name it as a reading, not a fact. The crib itself is our own plain word-for-word gloss, marked as a crib, never offered as a translation.
The translations. Sappho into English — Ambrose Philips (1711), J. A. Symonds (1883), and H. T. Wharton's prose (1885), all from Wharton's Sappho (Project Gutenberg 57390); Walter Petersen (1918), The Lyric Songs of the Greeks. Catullus into English — W. E. Gladstone (in Wharton); R. F. Burton (1894), The Carmina of … Catullus (Gutenberg 20732); Robinson Ellis (1871), The Poems and Fragments of Catullus (Gutenberg 18867). George Lamb's refusal is in his preface to The Poems of Caius Valerius Catullus (1821); the phrase “bids defiance to competition” is quoted, the rest paraphrased (the scan's OCR is unreliable). Every line shown was checked against the source bytes, not a summary.
The honest edges. (1) The constituted Greek is partly reconstruction; we name it as Voigt's and show his daggers. (2) Editions vary in small ways — e.g. some print κάμ where the displayed text has ἄκαν; the Latin's aspexi/adspexi and molestum est/molestumst are orthographic variants. (3) Whether Sappho's poem continued past line 17, and how far, is unknown; Catullus's fourth stanza is not reliable evidence of her ending, since it is his own. (4) There is a second, faint ancient witness to the opening alone — a variant φαίνεταί ϝοι (“to himself he seems”) associated with Apollonius Dyscolus, which would change the first stanza's sense; we note it without building on it. (5) “Longinus” is a convention: the treatise's authorship is disputed (the Paris manuscript heads it “Dionysius or Longinus”), and most date it to the first century CE, a minority to the third.
Type. Greek set in a subset of GFS Didot (SIL Open Font License), self-hosted (only the polytonic glyphs this page uses); Latin and English in Fraunces and Martian Mono (both OFL). No third-party requests; everything is served first-party.
The twelfth entry of the Translation-Criticism Venue, and its second built around a poem (after The First Word Is Rage) — but the first where the most famous “translation” is itself an ancient poem, so the criticism turns on a fellow maker rather than a modern hand. Its companions in the diachronic, transmission-as-subject mode are The River That Stays (Heraclitus deformed across 2,500 years, also preserved only by later quoters), The Horns of Moses, and The Sign of Immanuel; in the alignment mode, The Way That Can Be Told.
← back to the ground