Sappho

Sappho lived on Lesbos around 600 BC. Lardinois dates her “around 600” because she is reported to be the contemporary of Alcaeus and of his rival Pittacus. Her songs give that link some support.^[1] Natoli and her co-authors say that the ancient sources agree on the late seventh to early sixth century.^[2] A poet’s floruit is the height of his or her activity. Yatromanolakis puts Sappho’s at c. 620–580 BC.^[3]

The ancient chronographers already differed. Herodotus linked her to Amasis and Croesus. Eusebius linked her to Croesus’ father Alyattes, a generation earlier. On modern dates she belongs to the first part of the sixth century.^[4]

What the testimonia are worth. Testimonia are the ancient statements about a poet. The first fairly full account of her life is a treatise On Sappho, on a papyrus copied around the turn of the second and third centuries AD (P.Oxy. 1800). The next is the tenth-century Suda.^[5] Their authors read her poems autobiographically, and filled gaps by deduction.^[6] They shaped the details to the usual pattern of a poet’s life.^[7] Practically all the sources are unreliable. So, Kivilo concludes, we cannot tell whether anything in the tradition reflects the historical Sappho.^[4] Comedy added caricature: at least six comedies were called Sappho.^[8]

Family. The Suda gives her father eight possible names. It gives her a mother, Cleïs; three brothers, Larichus, Charaxus and Eurygius; a very wealthy husband, Cercylas, a trader from Andros; and a daughter, Cleïs (test. 2).^[9] Herodotus, among others, reports the father’s name as Scamander or Scamandronymus. That would place the family in the genealogies of the nearby Troad.^[10]

A daughter Cleïs seems to be named in frr. 98b and 132. But πάις (pais), “child”, in fr. 132 need not mean a biological child. The mother’s name looks suspiciously like the daughter’s. The husband’s name is an obscene pun. It plays on κέρκος (kerkos), “tail”, slang for the penis, and on Andros as the island of “Man”.^[11] It plausibly comes from a comedy.^[12] Campbell too suspects that the husband’s unusual name was the invention of a comic writer.^[13]

A papyrus published in 2014 shows that she named Charaxus and Larichus in a poem. Whether they were real brothers or characters is unknown.^[14] Charaxus was said to have bought the freedom of the courtesan Rhodopis or Doricha in Egypt.^[15] Whatever the truth of that, Mytilenean traders at Naucratis were a fact.^[16]

Home town. Eresus and Mytilene both claimed her.^[17] Kivilo reads the evidence differently. Sappho herself mentions Mytilene, and Herodotus implies it was her home. Some authors link Eresus with a “second” Sappho, a courtesan.^[18] Rösler accepts Eresus as a reported birthplace. But he has her family living in Mytilene, at least later in her life.^[19]

Exile. The Parian Marble records that she sailed from Mytilene into exile in Sicily.^[20] It dates this to an Athenian archonship just before 595/4 BC.^[21] The same inscription, however, also dates mythical events.^[2] Lardinois suggests that the exile arose from rivalries between families. This was the world in which Pittacus married into the house of Penthilus.^[22] Campbell likewise takes the exile as a sign that her family, or her husband’s, was active in politics. He notes her apparent hostility to the house of Penthilus (fr. 71).^[13]

Legends. For Lardinois, the love for the ferryman Phaon and the leap from the Leucadian rock are later fabrications.^[17] No ancient source points out her grave. That suggests the leap entered the tradition early.^[23]

Her island. Lesbos had six cities, later five. Its material culture looked to western Anatolia. Mytilene traded with Egypt and fought for land on the Asian coast. Which side Sappho’s family took in the feuds of Mytilene is unknown. But strife and exile were part of her world.^[24] Lydia appears in her poems as a measure of beauty and luxury (frr. 16, 39, 96, 98, 132).^[25]

Sappho’s poems were songs, which she performed to music or had others perform.^[1] Where they were sung, and for whom, is among the most disputed questions about her.

Monody and choral song. One distinction matters most. Monody is song for a solo singer, who accompanies himself or herself on the lyre. It has short fixed metrical structures, often used by many singers. Its language is simple, with little simile or quoted speech. Choral song was sung and danced by groups, sorted by age and sex, at public festivals. It has specially made metres, often triadic, and longer, richer stanzas.^[2] Most of Sappho’s best-known poems (frr. 1, 2, 16, 31, 94, 96) fit the monodic pattern.^[2]

Kurke finds three groups in the corpus. There are wedding songs; other ritual songs, such as the dialogue between Aphrodite and girls mourning Adonis (fr. 140a);^[3] and informal personal songs. They differ in the age-classes they mention, and in metre, language and dialect. So they probably differed in performance too.^[4]

Wedding songs. The common view is that the epithalamia (frr. 104–117) were choral and public. For Ferrari this “cannot be doubted”, though they could be shared between semi-choruses, or between a soloist and a group.^[5] Scholars “normally agree” on this.^[6] The run of short verses in fr. 111 suggests a lively chorus that needed little training.^[7] The wedding songs admit more non-Aeolic forms and metrical licences than the personal songs.^[8]

They were sung at the banquet, in the procession and even outside the bridal chamber. Some were sung by female choruses, some with young men, and some perhaps solo.^[9] Yatromanolakis cautions that not even all the wedding songs need have been choral.^[10]

The other songs: authorities differ.

Against the sympotic views, Lardinois and Kurke make the same point. Outside some religious festivals, there is no evidence for drinking parties of women alone in archaic Greece. And Sappho never mentions wine.^[11] Lardinois also rejects the teacher and the priestess. The “pupils” of the testimonia are more likely anachronistic names for the young women of her choruses. Nothing shows that she held a religious office.^[12] Giesecke’s verdict is the most cautious. Sappho was a singer-poet performing for an audience whose size, sex and setting cannot be determined.^[13]

A test case. In the Tithonus poem the singer can no longer dance. Many scholars suppose that Sappho sang it solo, while a silent chorus of girls danced.^[14] Natoli et al. allow that it may have been a choral song.^[15] Yatromanolakis objects that this picture rests largely on a late epigram (Anth. Pal. 9.189). A gathering of women and children, he says, cannot be excluded.^[16]

Later performance. In later antiquity her songs were sung at dinners and symposia. Aelian’s story of Solon’s nephew, Plutarch and Aulus Gellius all show this.^[17] Attic vases of the late sixth and fifth centuries show her with the barbitos, an instrument of the symposium.^[18] The same double setting can be found inside fr. 1. Its iambic invective suits the symposium, and its hymnic form suits a ritual occasion.^[19] Yatromanolakis warns that the late sources tell us nothing about archaic or classical practice.^[20]

The editors. Systematic Alexandrian work on the canonical lyric poets falls between about 285 and 145 BC. The uniform layout of the Sappho papyri points back to a scholarly edition, written in cola with critical signs.^[1] Who made it is not certain. Hephaestion reports editions of Alcaeus by both Aristophanes of Byzantium and Aristarchus. Sappho is named alongside Alcaeus, so Prauscello infers that she was edited by both. She thinks that local copies kept at Mytilene were a main source.^[2] Phillips leaves open which scholar was chiefly responsible.^[3]

How many books? Authorities differ. The Suda and an epigram by Cicero’s freedman Tullius Laurea both say nine.^[4] Most recent scholars, following Lobel, hold that there were eight, the eighth mostly or wholly wedding songs. Prauscello argues for eight books cited by number, plus a separate book entitled Epithalamia, nine in all.^[5] The other accounts differ again:

• Lardinois writes “around nine”, perhaps eight or seven.^[6]
• Finglass and Kelly write “probably nine”.^[7]
• Campbell has nine. The ninth held the epithalamia, which their metres kept out of the other books.^[8]
• Natoli et al. write “either eight or nine”, and elsewhere “probably eight”. They cite Yatromanolakis, who finds little reliable evidence for a ninth.^[9]

Natoli et al. also say that two books consisted of wedding songs.^[10] Prauscello’s survey of the evidence finds one book so titled.^[11]

Order. Metre was the main principle of book division, except for the Epithalamia. Books 1–3 were each in one metre, and Books 5 and 7 were not. So the change probably came in Book 4.^[2] Phillips says that the poems were grouped “by meter and subject”.^[3] Within Book 1 the poems ran alphabetically, by first letter only.^[12] Obbink sees thematic clusters inside the alphabetical order, and Neri the reverse. But the priority of the alphabet “seems beyond question”.^[13]

Programmatic poems stood outside the sequence. Fr. 1 opened Book 1. Poems of wedding character closed Book 1 (frr. 27, 30) and Book 2 (fr. 44).^[14] Lidov warns that the evidence for a metrical arrangement is sparse, though papyri often show runs of poems in one metre (as reported by Natoli et al.).^[15]

Couplets or single lines? Authorities differ. A paragraphos is a short horizontal stroke in the margin that marks a division. Hephaestion says that the old copies of Books 2 and 3 marked each song with a paragraphos every two lines (On Poems 1.2, as cited by de Kreij).^[27] Lidov, Battezzato and de Kreij treat these books as built of two-line strophes. For Lidov, the marking-off of every second line makes them stanzaic (Lidov, §The Arrangement of Meters in Books).^[28]

A stichic metre repeats the same line without stanzas. West calls the gl²ᵈ of fr. 44 “used stichically”.^[29] Arukask calls the Book 4 line stichic, “often marked down in dyads”.^[30] Battezzato infers that imperial copies of Books 2 and 3 dropped the paragraphoi. For de Kreij that inference is “too optimistic”.^[31]

Book 2 and fr. 44. Fr. 44 is on the wedding of Hector and Andromache. De Kreij argues that it was not a regular part of Book 2. In P.Oxy. 1232 it seems to have been added after the end-title. And the book’s other fragments are in Sappho’s usual first-person manner. Fr. 44A (P.Fouad inv. 239) is often assigned to the book. But it carries an asterisk, which by Hephaestion’s account excludes it.^[32] Natoli et al. note that a papyrus and Athenaeus independently place fr. 44 at the end of Book 2.^[33] Athenaeus’ citation of the poem from the second book is among Campbell’s testimonia for it.^[34]

Sappho’s verse is Aeolic. Resolution is the replacement of a long by two shorts. Contraction is the replacement of two shorts by a long. Lesbian lyric admits neither. So each verse has a set number of syllables. Much of it is built from the aeolic cola, the short units of a few positions from which longer verses are made: the glyconic, pherecratean, hipponactean and their relatives. These are often extended by repeating a dactyl (— ⏑ ⏑) or a choriamb (— ⏑ ⏑ —) inside the colon.^[1]

The fixed count of syllables is called isosyllabism, and it is the main peculiarity of the tradition. The second is the aeolic base. This is a pair of free positions, where only the count of syllables matters and not their quantity. Both features probably go back to Indo-European verse.^[2] Hiatus is the meeting of two vowels across a word-end, and it is avoided altogether. Where vowels meet, they are regularly elided or run together by synecphonesis.^[3] Elision is the dropping of a word’s final vowel before a vowel; synecphonesis is the running of two vowels into one syllable. The first line of fr. 1 shows the count and the elision at work.

The glyconic is shown below at three lengths, each with a line written above it. The two base positions come first.

The first three books used one stanza or line throughout, and so did many poems of the fourth.^[4] Voigt’s schemes are these:

In Voigt’s notation ⏓ marks a position where at least eight of twelve syllables are long, × one where no more than seven of twelve are long, and ⁝ a place where at least half the words end (key to the conspectus).^[5]

The Sapphic stanza: three analyses. West writes the first two periods — ⏑ — × — ⏑ ⏑ — ⏑ — — ‖, and labels them ∧ia hag. The third is — ⏑ — × — ⏑ ⏑ — ⏑ ⁝ —⁀× ⁝ — ⏑ ⏑ — — |||. It might be labelled ∧ia hag ⁝ ad. But that label, he warns, hides the main point: the third period is the first two stretched out.^[6] A bridge is a place where word-end is avoided. West’s tie ⁀ between the tenth and eleventh positions marks one.^[7] Voigt’s cr ∧hipp and West’s ∧ia hag describe the same syllables in different terms. Here is West’s scheme with the first stanza of fr. 1 written over it.

In short: two equal periods, then the same period again, stretched by an adonean (— ⏑ ⏑ — —).

Battezzato gives cr ∧hipp ‖ cr ∧hipp ‖ cr ∧hipp ad ||| as the easiest form to memorise. But for the third period he prefers cr ∧gl ∧pher. The pherecratean is a catalectic glyconic, and it signals an ending. Word-end often falls one syllable into it, and West calls this dovetailing.^[8]

Three lines or four? Hephaestion already describes three hendecasyllables to each stanza, followed by a fourth colon of five syllables.^[9] D’Angour sets the stanza out as four lines, AAAB.^[10] Battezzato explains the four-line layout as a Hellenistic and imperial convenience. It shows the recurring cr ∧hipp at a glance, and it saves space. But there is no pause between the third and fourth lines, and words run over from one to the other.^[8]

Hence Horace treats the stanzas as four-line stanzas, and most modern editions still print them that way.^[11] Even between the first and second lines, where the period ends, a word may be elided into the next line. This is called “episynaloephe”, and it also happens between Sophocles’ trimeters.^[8] How the stanza has been read and heard, mostly in English, over two centuries, and what that means for scanning it, is on Reading the rhythm, after Yopie Prins (2019); the stanza itself is scanned line by line on fr. 1, the Hymn to Aphrodite.

Whose stanza? Hephaestion notes that it occurs in Alcaeus too, “and it is unclear whose invention it is, even though it is called Sapphic” (Ench. 14.1: ἄδηλον ὁποτέρου ἐστὶν εὕρημα, εἰ καὶ Σαπφικὸν καλεῖται (adēlon hopoterou estin heurēma, ei kai Sapphikon kaleitai)). The ancients nonetheless agreed that she used it more (Arukask adds Marius Victorinus).^[12]

Prosody. In “normal” Lesbian lyric a syllable is not left short, even before a stop + liquid at the start of a word.^[13] Two short syllables in the aeolic base are rare in the Lesbian poets, and later they are avoided almost completely.^[14] Sappho’s examples are fr. 53 βροδοπάχεες (brodopachees), “rosy-armed”, 94.22 ἀπάλαν (apalan), “soft”, and 98a.8 σ]τεφάνοισιν (stephanoisin), “with garlands”.^[15]

Some fragments stand apart. Fr. 105 is in hexameters after the Ionian model. It has contraction, epic correption and a short syllable before δρ (dr), all foreign to ordinary Lesbian verse. Fr. 44 is a narrative with a heroic setting, in gl²ᵈ. It admits the same abnormalities, except contraction.^[17] The glyconic strophes of the later books, the hexameters and the other forms are on the next tab.

Beyond the single-metre books Sappho used many forms. They can be described in Snell’s terms, as two ways of making a colon longer.^[1]

Internal expansion. A colon is lengthened from inside, by extra choriambs or dactyls (gl²ᶜ, gl²ᵈ). Example: the glyconic of fr. 94.1, τεθνάκην δ’ ἀδόλως θέλω (tethnakēn d’ adolōs thelō), ○ ○ — ⏑ ⏑ — ⏑ —, grows by two dactyls into the line of fr. 44.5, Ἔκτωρ καὶ συνέταιροι ἄγοισ’ ἐλικώπιδα (Hektōr kai synetairoi agois’ elikōpida), ○ ○ — ⏑ ⏑ — ⏑ ⏑ — ⏑ ⏑ — ⏑ —. The diagram on the Metres tab sets the two side by side.

External expansion. Iambic, cretic, bacchiac, trochaic or dactylic units are added before or after the colon. Example: in fr. 96.12 a cretic, — ⏑ —, stands before the glyconic: ᾱ̓ δ’ ἐέρσᾱ | κάλᾱ κέχυται, τεθᾱ́- (ā d’ eersā | kalā kechytai, tethā-), — ⏑ — | — ⏑ — ⏑ ⏑ — ⏑ —. The whole strophe is scanned below.

Dale describes the same growth of cola by added choriambs in the aeolic of tragedy. Glyconics grow into asclepiads, and a hendecasyllable is read as a glyconic prolonged by a bacchiac.^[2] The Lesbian poets also bind the lines of a strophe into one rhythmical unit. This contrasts with the end-stopped lines of Archilochus’ epodes.^[3] That habit is what makes the one-verse strophes below possible.

Glyconic strophes (Book 5)

The Berlin parchment and other papyri preserve three-line strophes built on the glyconic. Battezzato sets out their structure.^[4] In his notation “/” marks a line-end in the ancient layout that cannot be a verse-end, and “|” one that can. Brackets mark a colon not fully preserved. Voigt’s and West’s analyses are given beside his:

West’s ∧ia is the cretic, and his ia∧ the bacchiac, of the other two analyses. In the second line of frr. 95 and 96 (95.9, 96.7) the glyconic is replaced by a choriambic dimeter, or “wilamowitzian”: × × — ⏑ — ⏑ ⏑ —. Sappho, like later poets, uses it in responsion with the glyconic (Voigt, C 5).^[7] Responsion is the matching of one passage’s metre by another’s. The wilamowitzian has the glyconic’s eight positions, but the choriamb — ⏑ ⏑ — comes at the end instead of the middle.

One verse of thirty syllables? Disputed. In fr. 96 no pause falls inside the strophe, and words are split between lines. Some scholars object that thirty syllables cannot be sung without a break. Battezzato replies with parallels. Alcaeus has a one-verse strophe of twenty syllables. The third verse of the Alcaic stanza has nineteen. Pindar has runs of twenty-two and twenty-three positions. And a singer can take breath unnoticed. So Sappho “occasionally wrote a one-verse strophe”.^[8] West too treats such a strophe as a single compound period.^[6]

In a common pattern, the last verse amplifies the others. West’s example is fr. 94.6–8.^[5] The translations of both examples are this site’s.

Occasionally a whole strophe is a single compound period, with words running over the line-ends, as in fr. 96.12–14.^[6] West’s mark in βˈρόδα divides βρ between the syllables, so that δὲ is long. The βρ- itself is the Lesbian spelling of a word that once began with digamma, ϝρόδα (wroda).^[9]

Other lines

The shapes named in the table are these: an ionic metron is ⏑ ⏑ — —, so the ionic trimeter runs ⏑ ⏑ — — ⏑ ⏑ — — ⏑ ⏑ — —; the anacreontic is ⏑ ⏑ — ⏑ — ⏑ — —, an ionic dimeter with two positions exchanged (anaclasis); the ithyphallic is — ⏑ — ⏑ — —.^[10]

Some scholars derive the hexameter itself from cola like pher³ᵈ. This kinship would explain why true hexameters occur in Sappho.^[11] Whether the hexameter, ionic and pher³ᵈ fragments came from longer strophes cannot be known.^[12]

How much survives. The first book held 1,320 lines, and there were about nine books. So Sappho wrote roughly 10,000 lines. About 650 survive: one complete poem, about twelve substantial fragments, a hundred short citations and some fifty papyrus scraps.^[1]

Other counts differ. Natoli et al. speak of two complete poems among some two hundred fragments.^[2] The Tithonus poem was completed by the overlap of the Cologne papyrus with P.Oxy. 1787, and Finglass calls it “virtually complete”.^[3] Arukask counts every line printed in Voigt’s frr. 1–168C. There are 1,384 lines in 176 fragments, and about 42% of them cannot be assigned a metre with certainty.^[4]

How the surviving lines divide. Arukask made two counts. The first uses only lines whose metre is certain. The second adds probable cases:^[5]

At least 40 fragments are in Sapphic stanzas, and 4 to 12 in the aiolikon.^[4] Hephaestion gave that line its name, “Aeolic”, “because Sappho used it much”. He analysed it as three ionics and a trochee.^[6] An ionic is the unit — — ⏑ ⏑ (or ⏑ ⏑ — —); a choriamb is — ⏑ ⏑ —. The diagram sets Voigt’s analysis of the line beside Hephaestion’s.

For Book 1 alone the colophon, the note at the end of the roll, gives 330 stanzas, perhaps fifty poems.^[7]

Voigt’s inventory. Her conspectus of Sappho’s metres has 49 entries:^[8]

• 9 kinds of strophe (A 1–9).
  Example: the Sapphic stanza (A 1).
• 9 lines repeated κατὰ μέτρον (kata metron), “by the measure”, that is, the same line used again and again (B).
  Example: the hexameter of fr. 105 (B 1 b).
• 22 aeolic lines (C).
  Example: the glyconic with two dactyls of Book 2, fr. 44 (C 3 c).
• 2 ionic or choriambic compounds (D).
  Example: three choriambs and a bacchiac, fr. 128 (D 2).
• 5 asynarteta (E), verses made of two cola that keep their own rhythms.
  Example: two ithyphallics, — ⏑ — ⏑ — — twice, fr. 127 (E 3).
• 2 of uncertain analysis (F).

Many entries rest on a single fragment. Many assignments are marked “fort.” (perhaps).

Free positions. A free position is one that can hold either a long or a short syllable. Arukask measured how often the free positions are filled by a heavy syllable:^[9]

Her hendecasyllable counts include the first eleven positions of each third line.^[10] She counts a period-final syllable closed by a consonant as heavy, so the results lean towards heavy syllables. With samples this small, they show tendencies, not laws.^[11] She also reports Barris’s count of light eleventh syllables. There are nearly 40% in the first and second lines, and 35–36% in the third.^[12] In every case but the aiolikon’s opening, heavy syllables prevail. Alcaeus prefers them more than Sappho.^[13] A heavy syllable is a long one; a light syllable is short. Two first lines of stanzas show both fillings of the free positions.

Word-end. Lidov counted word-group end in the first three lines of the stanza (as written). He found it after the fourth syllable in 41% of lines, and after the fifth in 37%. Battezzato’s count of 33 lines from the 2014 papyri gives 27% and 42%. With them, the totals for 153 lines are equal: 52 lines each, 34% each.^[14] The two places look like this:

Lidov adds a point of style. In Book 1 Sappho usually does not let word-end, or its absence, fall at the same position for many lines running. The exceptions stand out as tightly unified and conspicuous. In fr. 31, word-end comes after the fourth position in every line but the last. In fr. 1.18–25, it always comes after the fifth syllable (Lidov, §Metrical Style).^[15] In the Tithonus poem the ninth and tenth syllables belong to one word in each of lines 1–6. The pattern breaks at line 7 with ἀλλά.^[16] Lesbian poets hardly ever combine lines with different kinds of ending, blunt and pendant.^[17] A blunt ending closes on a long after a short, … ⏑ —, like the glyconic; a pendant ending has one more position, … ⏑ — ×, like the pherecratean or the Sapphic line.^[18]

Sappho’s poems were songs, but nothing of their melodies survives. From the Hellenistic period at least, they were also read as texts.^[1] Musical notation probably did not exist before the fifth century BC. So it is very unlikely that her music reached the classical age in written form. Most of what we know about archaic music comes from writers of the imperial period, such as pseudo-Plutarch and Aristides Quintilianus. They drew on lost earlier works.^[2]

Scales. Greek scales were called harmoniai, “tunings”. They were built not on the octave but on the tetrachord. A tetrachord is the interval of a fourth, divided into steps that could be larger than a tone or smaller than a semitone. A lyre played a limited set of notes, and it had to be tuned to a particular mode.^[3]

Sappho and Terpander of Lesbos are said to have used the Mixolydian. This was a mournful mode that “mixed” the Lydian with other elements. The ancient writers may have inferred this from passages of their poems.^[4] The sources differ over its inventor. Pseudo-Plutarch’s On Music, citing Aristoxenus, gives it to Sappho (test. 37).^[5] But the same work elsewhere gives it to Terpander.^[6] Plato called it melancholic. His Socrates would have banned it from the ideal state as the mode of lament.^[7] Whether Sappho used other modes is unknown, though Battezzato thinks it plausible.^[4] Later tradition also credited her with inventing the plectrum and the harp called pēktis.^[8]

Voice and instrument. The accompanying instrument mostly played the singer’s notes, in unison or an octave apart. Playing a different line from the voice is attested from Plato’s time on. There is no evidence for it in archaic music. Singing several notes to one syllable began with late-classical musicians, and classical sources criticised it. So each syllable of Sappho’s songs was probably sung to a single note.^[9] Modern attempts to give each syllable an exact duration in modern note-values are speculative. This is all the more so because the aeolic base admits any combination of long and short.^[4]

One tune for every stanza? Disputed. Dionysius of Halicarnassus says that composers could not change the melody from strophe to antistrophe. Many scholars suppose that the tune was repeated in each stanza. Others argue that it varied with the pitch accent of the words in each line.^[10] D’Angour judges it most plausible that early melody followed the natural pitches of the words. The composer fitted them to a prearranged set of notes, with some flexibility between performances.^[11]

Instruments. Sappho names the lyre, λύρα (lyra), the “tortoise”, χελύννα (chelynna), and the pēktis, a harp that came from Lydia. Athenaeus credits her with the word βάρβιτος (barbitos). The barbitos was a deeper-pitched lyre with a long curved arm. A famous fifth-century vase shows her and Alcaeus holding barbitoi.^[12] Pindar said that Terpander introduced the barbitos, which is another link between Terpander and Sappho.^[13]

The aulos was the double reed-pipe of the symposium, and later of the theatre. It is never named as accompanying her own songs. But in fr. 44 it sounds with castanets at the wedding of Hector and Andromache.^[13] Lesbos, close to the Anatolian coast, was a centre of music in the archaic and classical periods.^[13]

Style. Ancient critics named charm as the mark of her poetry. Dionysius observed that Sappho and Alcaeus, with their short stanzas, varied their few cola little. Battezzato answers that she varies her fixed lines from within, by the placing of word-end and enjambment. He calls this an “inner metre”.^[14]

Modern performances

Modern performances, not ancient music: no melody for these words survives. They are shown for listening; each makes its own choices, described under it.

Sappho – Call to AphroditeStef Conner and Nikos Xanthoulis (seven-string lyre); co-composed with Barnaby Brown; Lotos Lab (2026)

The basis of Sappho’s literary language is the dialect spoken on Lesbos, which is a form of East Aeolic.^[1] The full inventory of its sounds and forms is on the Lesbian Aeolic page. What bears on reading her verse is added here.

A vernacular or a literary language? Until the mid-twentieth century, scholars took her language for pure vernacular. They followed nineteenth-century dialectologists such as Ahrens, and dismissed epic elements as un-Sapphic. Lobel’s label “abnormal poems”, for pieces such as fr. 44, reflects that view. Her language is now seen as a literary product. It reuses epic material, and may draw on an unattested Aeolic poetic tradition.^[2] It is reductive to equate it with the dialect spoken on Lesbos in her time.^[3]

Features that touch the metre. Seven features of her language change how a line scans or how it is read.

No compensatory lengthening. Words such as ξένος (xenos) and κάλος (kalos) keep a short first vowel, with no compensatory lengthening. Ionic has ξεῖνος (xeinos), κᾱλός (kālos) instead.^[4] Compensatory lengthening is the lengthening of a vowel when a consonant after it is lost.

Doubled consonants. The loss of s before m gives doubled consonants, as in ἔμμι (emmi), ἄμμες (ammes), ἔμμεναι (emmenai). Attic lengthens the vowel instead.^[5]

The digamma ignored. Initial w, written with the letter digamma (ϝ), is usually ignored in her prosody. The elision before ἴδοισαν (idoisan), “having seen”, in fr. 22.14 shows this. The digamma written in some pronouns reflects the grammarians’ belief.^[6]

βρ- for ῥ-. The spelling βρ- for ῥ- (βρόδον (brodon), “rose”; βράδινος (bradinos), “slender”) is generally explained as an editorial signal that the preceding syllable scans long.^[7] The point is debated. Bowie argues from the cases where βρ- makes position, against Hooker.^[8] Allen takes βρόδα as a spelling of the old ϝρόδα (wroda). He notes a papyrus of Sappho that still writes the digamma in the possessive ϝοι (woi), “his, her”.^[9] The worked example is βροδοπάχεες on the Metres tab.

Recessive accent. Every word is accented recessively, as far back as the rule for the antepenult allows. An oxytone is a word with the accent on its last syllable, and so the text of the Lesbian poets has none. The innovation is older than Sappho and Alcaeus, and only prepositions and conjunctions are exempt.^[10]

Vowel sequences in ι. Ancient grammarians held that vowel sequences ending in ι were two syllables, and some papyri accent them accordingly (τύιδ’, τυϊδ’ (tuid’), “hither”). Yet in fr. 17.7 the sequence scans as one long syllable. So the accentuation, and the theory behind it, are wrong.^[11] Fr. 1.5 shows the same scansion.

σδ for ζ. σδ for ζ is a grammatical spelling, meant to show the pronunciation [zd]. The earliest papyri are inconsistent.^[12] The spelling is absent from early Lesbian inscriptions. For Allen this shows that Lesbian kept [zd] after other dialects had moved to [z].^[13]

Choice of forms. Sappho seems to have chosen between conjugation types to suit metre or content. She has athematic πεσδομ]άχεντας (pesdomachentas), “fighting on foot”, in fr. 16.20, and thematic ὀνκαλέοντες (onkaleontes), “calling upon”, in the epic-coloured fr. 44.33.^[15] An athematic verb adds its endings straight to the stem, like the -μι verbs; Aeolic often conjugates this way verbs that in Attic are contracted, such as κάλημι (kalēmi), “I call”, for Attic καλῶ (kalō).^[16] A thematic verb puts the vowel ε or ο before the ending: καλέ-ο-ντες (kale-o-ntes). Her wedding songs admit more non-Aeolic forms than her personal songs.^[17]

The Alexandrian filter. What we read is the Lesbian of the Alexandrian edition. Its editors normalised spelling by rules of their own. This produced impossible “hyperaeolic” forms.^[18] For the grammarians “Aeolic” meant the Lesbian poets, and they invented forms by false analogy, such as αι (ai) for ᾱ (ā).^[19] Inscriptions help little for her period. No East Aeolic documents survive from the seventh and sixth centuries.^[20]

Disputed: Voigt prints first-person forms such as κάλημμι (kalēmmi), “I call”. De Kreij finds the double μ unsupported by linguistic theory or manuscripts.^[21] Tribulato cites κάλημι (kalēmi).^[22] The double μ copies ἔμμι, “I am”, where μμ is genuine.^[21]

A late survival. The first papyrus of Sappho to be published was edited by Blass in 1880 (P.Berol. inv. 5006, frr. 3–4). It is in fact a parchment of the seventh century AD, from Krokodilopolis in the Fayum. So Sappho was still read in Egypt some twelve hundred years after her lifetime. Alcaeus is not securely attested on papyrus after the third century. None of the other canonical lyric poets seems to have lasted as long.^[1] Many of the papyri come from the rubbish mounds of Oxyrhynchus.^[2][3]

Voigt’s edition of 1971 includes everything published up to 1965.^[4] The 2014 finds preserve parts of nine poems, two of them new.^[5] They add to frr. 5, 9, 16, 17 and 18. They give traces of two unknown songs (16a, 18a), the Brothers Song, and the opening of the Cypris Song, which overlaps fr. 26.^[6]

What the papyri show. There are twenty-four papyri of Sappho’s text, a high number for a lyric poet, and seven of commentary. They peak in the second century AD and hardly decline in the third. A codex is a book with pages, which in late antiquity replaced the roll. Her poems survived that change, which finished off less popular authors.^[7] Where the find-spot is known, it is Oxyrhynchus, apart from Krokodilopolis.^[8]

P.Oxy. 2076 belonged to a family library that was thrown out about AD 300. P.Oxy. 1787 may have been copied by Spartas, a member of the same family, though one scholar argues for two scribes.^[9] The potsherd of fr. 2 has been read as a school dictation, and alternatively as the work of a cultured amateur.^[8] Papyri also confirm that fr. 1 opened Book 1.^[10] The ancient commentaries on lyric poets found on papyrus are being collected in the Commentaria et lexica Graeca in papyris reperta.^[11]

The Cologne papyrus. The oldest Sappho papyrus was recovered from mummy cartonnage. It probably reflects an arrangement older than the Alexandrian edition.^[12] In it a poem in the same metre precedes the Tithonus poem. That poem is different from the one that precedes it in P.Oxy. 1787.^[13] The last poem is not by Sappho.^[14] The papyrus also showed that the editions had put a poem-end in P.Oxy. 1787 in the wrong place.^[15] Disputed: Lardinois thinks two versions of the Tithonus poem may have existed, with and without four further lines.^[6] Spelman is not convinced.^[16]

The newest papyri: provenance unknown. The owner and present location of P.Sapph.Obbink are unknown. So is the provenance of both 2014 papyri.^[17] The case is an example of an owner constructing a legitimate-looking provenance for resale.^[18] Sampson’s analysis of a sale brochure “all but prove[s]” the published account of its provenance false.^[19]

No one has argued that the texts are forgeries, and they contain no error of language or metre at all. The Cambridge Companion to Sappho treats them as authentic, while stressing that the circumstances fall far short of ideal.^[20] Natoli et al. print them separately, in an appendix of disputed texts.^[21]

The Papyri and Books tabs describe the Alexandrian edition and what came after it. What came before is described here. There is also a warning that applies to every fragment on this page. A fragment, as an edition prints it, is an editor’s construct. It is assembled from witnesses that do not always agree.

No master copy. Lardinois doubts that there was ever a master edition of Sappho on Lesbos. A text of her songs in classical Athens would have come from a local expert’s copy, or from memory. Athenians learned archaic songs as children or at symposia, and improvised on them. Sympotic compilations mixed poets, and adjusted songs to fit each other. That is what the Cologne papyrus seems to show.^[1] Such anthologies are also a route by which Sappho and Alcaeus could have reached Athenaeus without passing through Alexandria at all.^[2]

Fr. 2 in two versions. An ostrakon is a potsherd used for writing. Fr. 2 survives on one of the third century BC. Athenaeus quotes the poem five centuries later. The two witnesses end it differently. The ostrakon closes on an imperative, οἰνοχόαισον (oinochoaison), “pour”, and keeps Aeolic forms, ἔλοισα (eloisa), “having taken”, among them. Athenaeus begins his quotation with ἐλθέ (elthe), “come”. He has Attic forms, and makes the last word a participle, οἰνοχοοῦσα (oinochoousa), “pouring”. Either Athenaeus deliberately altered syntax and dialect to fit his text, or the two witnesses preserve two ways in which the song was remembered.^[3]

Caciagli reconstructs the first possibility. Athenaeus or his source dropped the opening, since a sacred grove is no setting for a banquet, and kept the libation. Having lost the poem’s invocatory verb, he supplied ἐλθέ (elthe). He perhaps lifted it from the lost beginning, and put it where the ostrakon has the similar-sounding ἔλοισα (eloisa). That then forced the imperative into a participle. Caciagli calls this an oral technique of sympotic reworking, as old as the symposium itself. He does not think that line 1 of the ostrakon is the poem’s first line.^[4] The Works tab follows the imperative in the licensed Bodard–Vierros corpus. The participle printed by Voigt, οἰνοχόεισα (oinochoeisa), is already present in Smyth’s older edition; his apparatus also gives Athenaeus’ form.^[5]

A caution about the Loeb Athenaeus. Olson’s Loeb edition of Athenaeus changes the Attic forms of his quotations from Sappho and Alcaeus back into Aeolic. By its own editorial statement, it prints the reconstructions of modern editors such as Voigt. So a reader can no longer tell what Athenaeus wrote, or which versions he knew.^[6] Anyone comparing a quotation with a papyrus should go to a text of Athenaeus that prints the manuscripts.

Editions before the edition. The Cologne papyrus is the earliest text of any canonical lyric poet. In it, the poem that precedes the Tithonus poem is not the one that precedes it in the Oxyrhynchus roll, which follows the Alexandrian edition.^[7] For Lardinois this is clear proof that different editions of Sappho existed in antiquity. Both neighbouring poems are in the Tithonus poem’s metre. So strophes attributed to her may have been combined differently in different collections. He would like a “multiform” edition that records such variants, instead of choosing between them.^[8]

Sappho sung in Athens. On the Munich vase of about 480–470 BC, Sappho and Alcaeus sing to barbitoi. Nagy reads it as a picture of the lyric poets as Panathenaic repertoire. It is matched by the exchange of verses that Aristotle quotes in the Rhetoric (conventionally fr. 137). Nagy thinks that citharodic performance favoured stichic, expandable forms over strophes. For him fr. 44, the wedding of Hector and Andromache, is the ideal citharodic piece.^[9]

The Peripatos was Aristotle’s school. There Chamaeleon wrote a monograph On Sappho. A fragment of it reports that some took a song of Anacreon to be addressed to her, and a Sapphic poem to be her reply. Athenaeus, who preserves the fragment, rejects this pairing.^[10] The contributors to the volume do not agree. Nagy gives Athens a central part in the canon’s formation, and Hadjimichael a small one.^[11]

From quotation to collection. Before printing, a reader of Sappho would have had to hunt through manuscripts of the authors who quote her. There is no evidence that anyone did.^[1] Fr. 1 first appeared in print in 1508, in the Aldine edition of Dionysius of Halicarnassus edited by Demetrios Doukas. It was set as prose, and it was faulty in metre and dialect.^[2] Fr. 31 followed in 1554, in Robortellus’ first edition of “Longinus” On the Sublime. A couple of months later it appeared again in Muretus’ Catullus.^[3]

In the same year Henricus Stephanus printed frr. 1 and 168B at the end of a book of Anacreon. This was the first gathering of Sappho’s poems since antiquity. His collection of 1560, with a Latin translation, offered dozens of fragments, nearly thirty for the first time. It presented the spurious epigrams as genuine.^[4] Ursinus (1568) added more. J. C. Wolf (1733) published the first edition of Sappho on her own, with a critical commentary.^[5]

The nineteenth century. Editions by Volger (1810), Blomfield (1814) and Neue (1827) led to Bergk’s (1843). From its second edition (1853) to its fourth (1882), Bergk’s contained 120 fragments and 50 testimonia.^[6] Bergk made a conjecture at fr. 1.24 in 1835, which was later found to have manuscript support. It allowed readers, for the first time since antiquity, to see that the beloved in that poem is a woman.^[7]

Papyri and the modern editions. From 1880 papyri steadily enlarged the corpus.^[8] Edmonds’ complete edition of 1922 filled the gaps in the papyri far too freely. Lobel demolished it in a review, and his own Sappho (1925) was cautious.^[9] Lobel and Page’s Poetarum Lesbiorum Fragmenta (1955, corrected 1963) set Sappho beside Alcaeus, with a section for fragments that could belong to either. But its austerity in supplementing made it less useful than it might have been.^[10]

Voigt’s edition (1971) was apparently finished before 1967. In West’s judgement it is “thoroughly prudent”. It includes the testimonia, and gives three apparatus: sources, parallels and bibliography. It is written in Latin.^[11] Page’s Supplementum Lyricis Graecis (1974) followed, and so did Campbell’s Loeb (1982), with an English translation. Later Italian and German editions largely rely on Voigt’s text.^[12]

Which edition now? Authorities differ. For Finglass, Voigt’s is still the last edition to rework the text of Sappho from the ground up, though it is rare, expensive and out of date. Lobel–Page, older still, is frequently cited. Neri’s edition with Cinti (2017) lacks an apparatus, and Neri was preparing a full critical edition.^[12] Lardinois calls Neri 2021, with Italian commentary and translation, “the standard text edition of Sappho”. For an English translation of all the fragments, including the newest, he recommends Rayor and Lardinois 2022.^[13]

There is no complete critical edition of the fragments or testimonia. So Finglass and Kelly cite Campbell’s numbers, because his edition is easily available.^[14] Lardinois does the same.^[15] Scholars cite Sappho and Alcaeus from either Lobel–Page or Voigt.^[16] Yatromanolakis had called a new critical edition necessary.^[17] De Kreij proposes numbering the fragments by book, wherever their place is known.^[18]

On this site. The Greek text and fragment numbers follow Voigt 1971, whose page images we use. Quotations from scholars who use Campbell’s numeration are cited with the numbers they give.

The essay. Yopie Prins, “Sapphic Stanzas: How Can We Read the Rhythm?”, appeared in Critical Rhythm, edited by Ben Glaser and Jonathan Culler (Fordham University Press, 2019), pp. 247–273. It studies how the Sapphic stanza has been read and heard, mostly in English, over two centuries. It is not about Greek metre, but it bears on how we scan the stanza.

The question. Prins begins with the headlines of 2014, when new fragments of Sappho were announced. In the account she quotes, the poem was recognised by its shape on the papyrus: three long lines followed by a short fourth. So identification depended “first and foremost” on recognising the stanza.^[1] In 1914 the New York Times had greeted the papyrus of fr. 16 with Joyce Kilmer’s article and his own English Sapphics.^[2] The Alexandrian edition arranged Sappho’s songs by metre. Since then, she adds, they have been poems to be read, not heard. We imagine her voice because we know how to read the stanza — “Or do we?”.^[3]

One scheme, many rhythms. She scans her own title as a Sapphic line in five ways: by long and short syllables, with a choriamb in the middle (“how can we read”); as feet (trochee, trochee, dactyl, trochee, trochee); as five strong beats; or simply as eleven syllables.^[4] Her diagram of the stanza uses one sign for a syllable “long or stressed” and another for “short or unstressed”. It uses x for one “either stressed or unstressed”.^[5]

Ideas of metre and rhythm, she argues, are historically contingent. The stanza has been made to produce “allegories of rhythm”.^[6] Modern poets hear it as a heartbeat.^[7] Victorians drew it as a grid, or called it “a woman’s Hexameter”.^[8] J. W. White (1909) traced it to an Indo-European line of eight syllables of free quantity. W. R. Hardie (1920) objected that verse also meant to be read could not be in part “unregulated”.^[9]

Translators of fr. 16 heard it in different ways. One heard four marching trochees with “a dancing dactyl” (Miller).^[10] Another replaced quantity with stress, and praised Sappho for “keeping her caesura moving” (Powell).^[11] Anne Carson’s brackets make the papyrus’s gaps a rhythm on the page.^[12]

What it means for scanning.

• Quantity, not stress. A diagram that marks “long or stressed” serves English imitations. In English verse, rhythm is a sequence of stressed and unstressed syllables. In Greek it is a sequence of long and short syllables. The Greek word-accent was a pitch, and it played no part in the metre.^[13] On these pages — and ⏑ mark syllable length only. The sign × marks an anceps position, open to a long or a short.^[14]
• Four lines or three. The shape recognised in 2014 is the ancient layout. Hephaestion counts three hendecasyllables and a fourth line of five syllables.^[15] West counts three periods. The third was written as two lines, divided after its eleventh position, and this gave Horace’s four-line stanzas and our editions’.^[16] Voigt analyses the third line and the adonean together (cr ∧gl ∧pher, A 1). Barham (1860) proposed printing the stanza as three lines.^[17] The works pages draw the adonean as continuing line 3.^[18]
• Feet or cola. “Trochee, trochee, dactyl, trochee, trochee” is a foot reading. Hephaestion divided the line into a trochaic syzygy, a choriamb and a close.^[15] Voigt divides it into cr ∧hipp, and West into ∧ia hag.^[19] The colon/foot switch on the works pages shows the line both ways.
• What is free. Something of White’s idea survives. West takes the aeolic base, ○ ○, as a relic of the verse’s original freedom of quantity before the cadence, in cola related to Vedic ones.^[20] The Sapphic hendecasyllable has no base, and it begins — ⏑ —. Its free position is the fourth.^[21] That position is heavy in 68.24% of Sappho’s analysable lines.^[22] Powell’s moving caesura can be measured. In the first three lines, word-group end falls after the fourth or the fifth syllable in roughly equal shares, and fr. 31 departs from this.^[23]

The diagram puts these readings side by side on the first line of fr. 1, Ποικιλόθρον’ ἀθανάτ’ Ἀφρόδιτα (Poikilothron’ athanat’ Aphrodita), “Immortal Aphrodite of the intricate throne”.