[In the following essay, Rayor explores some of the difficulties associated with translating Sappho's fragmentary poetic texts.]

Since ancient poetry so often survives only in fragments, it would seem to present the translator with special problems not shared by those who translate complete texts. But although some of the problems are unique, the methods used to “solve” them are much the same. Yet focusing on the translation of fragments makes it easier to see the additions, subtractions, and changes that occur in all translations. The awkward loss of text exaggerates the ever-present temptation to “fix” a text rather than represent the poet's words—and the gaps between those words—accurately. Incomplete texts illuminate the criteria, strategies, tactics, and alternatives available for any rendering.

Quotations and papyri provide our only sources of ancient Greek lyric poetry. The quotations generally are very brief excerpts of one or two lines isolated from their original context within longer poems; occasionally a whole poem is quoted. Egyptian papyri containing poetry turn up in various stages of disintegration or in pieces. Indeed, many recent finds of poetry are on strips of papyrus wrapping mummies. Thus poems found on papyrus often are missing the right or left side; sometimes entire lines or scattered words have been erased by time.

The poetry of Sappho (seventh century bce) demonstrates both the possibilities of translation and the necessity for establishing consistent principles of translation. Of the nine books of her poetry (some five hundred poems) collected in the Hellenistic period, only one definitely complete poem remains. The rest are fragments. The combination of the distance in time, the physical state of the manuscripts, the lack of reliable biographical information, and the poet's gender have led to the constant creation of new Sapphos by translators.¹

Fragments clarify strategies of reading and translating poetry because their absences expose our necessary interaction with the text. They also expose where the translator distorts the text by interacting too much, thus not allowing the readers a chance to experience the potential of the poem. Translations work best when they fully exploit the connection and activity of the reader with the text. Letting the absences show in the translation leaves room for the reader to determine meaning and make connections.

Fragments implicitly remind us of their physical inscription and call into question the illusion of self-contained, “whole” texts. The holes in the text are not left empty in the reading process. As we read, we fill in, “read between the lines.” While we do this in all reading, fragments tempt us to guess authorial intention, to imagine what the poet originally wrote that is now missing.

Reading a translation of Greek poetry should be as close to the experience of reading the Greek text as possible. Yet the reader can discover the possibilities of the Greek text only through the eyes of the translator. Optimally, the translation recreates as much of the potential meaning of the Greek as possible—opening up rather than narrowing the range of possible interpretations. It is a delicate business to provide enough information without over-determining the meaning of the poem.

To recreate the experience of reading Sappho, for instance, the translation needs to show the reader where the Greek text breaks off. Most available translations of Greek lyric give no indication of fragmentation, where one thought does not immediately follow the last. Translators generally opt for expanding or condensing the text by adding or subtracting phrases. Peter Newmark's terms of over- and under-translation² have special meaning for fragments.

Over-translation and under-translation erase evidence of physical gaps. “Completing” the poem by filling in gaps overly privileges the translator's interpretation, and fragmentary lines left out through condensing often contain vital information. Both practices simplify the poetry and mislead the reader. While the translator's interpretation of the text always informs the translation, she should resist the temptation to add or subtract text itself.

Over-translation was once common because the editors of Greek texts used to add the Greek they guessed the author originally had written. Some additions to fragmented texts certainly are acceptable, and it would be a disservice not to include them. The standard Greek editions include generally accepted supplements based on quotations in other ancient authors, probable readings of papyri, information from ancient marginalia, and the sense of the texts themselves. The translator accepts or rejects these supplements on a individual basis according to probability and necessity. It is not over-translation to accept a suggested word that is likely paleographically and needed for an intelligible reading.

On the other hand, early editions of the Greek, such as Edmonds'³ Sappho, contain large-scale reconstruction. Edmonds fills in whole passages missing in the extant texts of Sappho; he even composes entire poems from a few fragments. More recent editions of Sappho, by Lobel and Page⁴ and Voigt,⁵ provide texts free from these restorations. Translations based on poorer editions, therefore, are an additional stage removed from the Greek. Translations not based on the latest findings or the most accurate scholarship are mistranslations rather than over-translations.

The justification given for over-translation is that fragmentary poetry should be completed by the translator to provide the reader with the closest possible experience of the original. The problem, of course, is that the translator cannot know what the poet originally wrote, and that translators always interpret through their own biases. For example, in Sappho [“16”],⁶ lines 13-14 are missing:

She had no
memory of her child or dear parents,
since she was led astray
[by Aphrodite] …
… lightly
… reminding me now of Anaktoria
being gone,
I would rather see her lovely step
and the radiant sparkle of her face
than all the war-chariots in Lydia
and soldiers battling in shining bronze.

Richmond Lattimore's⁷ translation adds this for the missing lines:

Since young brides have hearts that can be persuaded
easily, light things, palpitant to passion/as I am.

This addition completely transforms the tone and purpose of the poem. Sappho's poem argues that “whatever one loves” (line 4)—the paraphernalia of war or an individual person—appears most desirable, not that women are particularly excitable and irrational. The lines Lattimore adds to to fill the gap are symptomatic of changes throughout his translations of Sappho; earlier in [“16”] he changes the neuter “whatever one loves” to “she whom one loves best.”

While over-translated poems second-guess the author, under-translated poems tend to leave out even more text than is available in their fragmentary form. Should the translator trim more off a poem already pruned by time? Mary Barnard's⁸ translation of Sappho [“95”] provides an example of three strategies: under-translation (1) by leaving out the first three partially visible lines, and (2) by pretending the poem is unbroken, and (3) over-translation by adding an explanation to the name of Hermes:

Hermes, Lord, you
who lead the ghosts
home:
But this time
I am not happy; I
want to die, to see
the moist lotus open
along Acheron.

Omission of the woman's name, “Gongyla,” from the first extant line removes the suggestion that perhaps the “longing” to die is based on erotic longing for another woman:

Gongyla …
Surely a sign …
especially …
[Hermes] came into …
I said: O Lord …
By the blessed [goddess]
I take no pleasure on [earth]
but longing to die holds me,
to see the dewy lotus-
shaded banks of Acheron …

Translators need to be particularly aware of their biases or assumptions when translating women's poetry to avoid distorting the message, or closing off interpretive possibilities available in the source text.⁹ Over-translations, such as Lattimore's of Sappho [“16”], fill in the fragment gaps with inappropriate or trivializing phrases. While fragments lend themselves to that sort of misrepresentation, whole poems also are subject to distorted or censored renderings. Obvious examples include translations that switch pronouns or even the subject from female to male. Nineteenth-century translations of Sappho [“1”] changed from female to male the object of the (female) speaker's desire:

For if she flees, soon she'll pursue,
she doesn't accept gifts, but she'll give,
if not now loving, soon she'll love
even against her will.

Fragments that are excerpts from lost longer poems frequently lack a context for interpretation. In these short fragments, it is sometimes difficult to determine the gender from the Greek verb. For example, in [“15.”4] the Greek could be “he came” or “she came”:

                    … Kypris,
may she find you very bitter
and may Doricha not boast, saying
how she came the second time
to longed-for love.

Nothing in the poem suggests a masculine pronoun, since the only person mentioned is female. Yet the poem generally has been translated “he came,” which shifts the focus of the poem to an unidentified man. This has been justified by an unreliable biographical tradition that associates Doricha with a prostitute with whom Sappho's brother fell in love. Even if we accept that the rest of the poem dealt with that story, nothing hinders Doricha from being portrayed as the active one. Poems that have an erotic element are especially apt to be reconstructed according to the individual translator and prevalent attitudes. Whether words or context are missing, fragments illustrate the need to be sensitive to tone and potential meaning of the poetry translated.

Yet without “completing” the poem, how does one make a wounded poem live in the new language? Gaps in poems can be bridged by loosely linking sense or images, so that the poem reads well, without being deceptive. The translator's job is to make the absences work as part of the poetry without being distracting: to evoke connections, enticing the reader to bridge the gap.

Fragments can engage the reader's imagination by actually using the breaks. Poems of recollection or memory have inherent possibilities. In Sappho [“94”] the speaker tells of how she reminded a friend who was leaving of their past days spent together. Throughout the second half of the poem, scattered words are missing:

“I simply wish to die.”
Weeping she left me
and said this too:
“We've suffered terribly
Sappho I leave you against my will.”
I answered, go happily
and remember me,
you know how we cared for you,
if not, let me remind you
… the lovely times we shared.
Many crowns of violets,
roses and crocuses
… together you set before me
and many scented wreaths
made from blossoms
around your soft throat …
… with pure, sweet oil
… you anointed me,
and on a soft, gentle bed …
you quenched your desire …
… no holy site …
we left uncovered,
no grove … dance
… sound

The recollection in the second part might read as if the speaker's voice drifts off into silent memory.

Word selection is crucial to tantalize the reader and evoke the sensuality of the poem. Lines 21-22 demonstrate the double meaning exploited by the translation. The phrase, taken with the following lines, implies that the women visited every temple, and that they participated in the rituals of Aphrodite, goddess of love. But the eroticism of earlier lines, particularly line 20, is enhanced by the second meaning of covering every “holy site” of the body.

No images are left out, none are added. Each word is given its full impact through word choice and position, each line building on the images and sounds of the previous lines. The need for and effect of devices used in translating all poetry are exaggerated by the fragmentation of the text.

Poems with more radical breaks, such as those with the right side missing as in [“95”] (above), are more difficult to work with. The translator can make the most of the extant text by indicating missing parts through line breaks and punctuation. Some translations can even imitate the physical texture of the papyrus by showing where the lines were torn. But recording very fragmentary pieces containing an interesting myth or image is sometimes more a matter of preserving it than creating viable poetry. One example is an eighteen-line fragment [“58”] missing the left-hand margin, which tells the myth of Tithonos in the context of the speaker's aging:

… rosy-armed Dawn
… taking (Tithonos) to the ends of earth.

A second example, a two-line poem, tells an alternative story to the traditional one in which Zeus, in the form of a swan, rapes Leda and fathers Helen. Sappho [“166”] perhaps suggests that there was no rape and that Leda found an egg containing Helen:

They say that once Leda found
an egg hidden in the hyacinth.

Small fragments like [“166”] have inspired modern poems; H.D. has a series of poems based on Sappho fragments. One can admire the pieces, as one does broken statues or shards of pottery.

To offset gaps or lack of context, the translator needs to employ many different strategies to make the poem work on as many levels as possible. Effective strategies include sound and tempo effects, and even grouping the poems thematically. Sounds with a similar effect, although not usually the same sound, as the source language develop the potential of whole poems and fragments. Translations of Sappho [“2”] and a poem by another seventh-century-bce writer, Alkman, both work with sound, especially with repeated vowels, to echo the hypnotic effect of the Greek:

SAPPHO [“2”]

cold water ripples through apple
branches, the whole place shadowed
in roses, from the murmuring leaves
deep sleep descends.

and

ALKMAN [89]¹⁰

All asleep: mountain peaks and chasms,
ridges and cutting streams,
the reptile tribes that black earth feeds,
mountain beasts and race of bees,
monsters deep in the purple sea,
and tribes of long-winged birds all sleep.

Sappho [“140”] emphasizes the ritualistic aspect of the festival in honor of Aphrodite's (i.e., Kytheria's) lover Adonis, through alliteration in Greek: two words begin with a “t” sound, two with an “ah,” and the rest with a “k” sound. The translation echoes the effects:

Delicate Adonis is dying, Kytheria—what should we do?
Beat your breasts, daughters, and rend your dresses.

Since an attempt to reproduce the Greek meter would work clumsily in English, one can compensate for this by recreating the vivid and direct effects of the Greek sound.

Placing short poems together will also help recreate a context through association. Grouping Sappho's short fragments according to such themes as friendship, rivalry, or epithalamia (marriage songs) builds meaning by accumulation. It is an interpretive move, for instance, to place Sappho [“51”] “I don't know what I should do—I'm of two minds,” with erotic poems or with poems about writing poetry (“do” can mean “set down” in writing.)

By paying particular attention to the words on each side of the gap, by word choice and use of sound, and by the grouping together of short excerpts, the translator can develop the available text, the remaining words, in ways conducive to the reader's activity. As in translating non-fragmentary poetry, the translator abides by certain criteria that remain flexible enough to solve the individual problems posed by every poem. Tactics shift for individual poems, but the underlying approach should be consistent. The translator tries to incorporate as many facets of the source poem as possible, compensating for what is lost either from the fragmentary source text or in the transmission from source to target language. Fragments can make us more aware of how we “complete” texts as readers and interpreters. Then we are more likely to find the balance between over- and under-translation, finding the elusive fine line that is “just right.”

Notes

1. See J. DeJean, Fictions of Sappho, 1546-1937, Chicago: 1989.

2. P. Newmark, Approaches to Translation, Oxford: 1981.

3. J. M. Edmonds, Lyra Graeca, vol. 1, Cambridge: 1928.

4. E. Lobel and D. Page, Poetarum Lesbiorum Fragmenta, Oxford: 1955.

5. E.-M. Voigt, Sappho et Alcaeus Fragmenta, Amsterdam: 1971.

6. All of the translations not otherwise identified are my own from Sappho's Lyre: Archaic Lyric and Women Poets of Ancient Greece, Berkeley and Los Angeles: forthcoming 1991; I use Voigt's edition and numbering.

7. R. Lattimore, Greek Lyrics, Chicago: 1960.

8. M. Barnard, Sappho, Berkeley and Los Angeles: 1958.

9. See M. Díaz-Diocaretz, Translating Poetic Discourse: Questions on Feminist Strategies in Adrienne Rich, Amsterdam: 1985.

10. See note 6; I used the edition and numbering of D. Page, Poetae Melici Graeci, Oxford: 1962.