For the first of my August posts, I’m sending out again this little essay, which was one of the first pieces I published on the substack, back in June 2023. If you enjoy this one, you might also enjoy my piece on falling in love with Sanskrit as a language or on the pleasures and difficulties of translating Horace, Pindar and Malherbe.
I recently found out that a little verse translation from Sanskrit I wrote a couple of years ago was broadcast on Australian radio. (The poem comes right at the end, but you can also hear my recording of it here.) The poem was printed in (and commissioned for) an excellent anthology of poems about Outer Space, edited by Midge Goldberg and published last year by CUP.
Here’s the text of the poem, which is a translation of a very short piece, really something like an epigram (and probably actually an extract from a play) found in a major medieval anthology of Sanskrit verse, the Subhāsitaratnakośa. The anthology was put together about 1100 AD by the Buddhist scholar named Vidyākara, and contains 1738 short poems and extracts by over 200 named poets (including several women), as well as many anonymous pieces. This poem is number 905, in a section devoted to poems about the moon. It’s attributed, though somewhat uncertainly, to the playwright Bhāsa.
A cat lapsat beams in a bowl.An elephant takesrays amid leaves to besweet white stems. A girlresting from love, reachesfor the bright pool of herdiscarded dress.So by moonlight are we all,greater or less,deceived.
Sanskrit poetry is a real challenge to translate into modern English. Some of the challenges are similar to those presented by translating from Latin or Greek: like those languages, Sanskrit is highly inflected — many of the relationships between words that we would represent in English by a particular sequence of words or by ‘little words’ like to, by or of are indicated using different endings. This makes for greater concision and a greater flexibility in word-order. In this poem, for instance, there are no single Sanskrit words for ‘a’, ‘at’, ‘in a’, ‘amid’, ‘from’, ‘for the’, ‘of her’ or ‘by’: these links are all indicated using word endings, just as they would be in Latin or Greek.
In the original, the phrase for ‘moon beams’ also appears only once: grammatical agreements make it clear that all three objects — the milk, the stems and the dress — are mistaken for it. You could do something like this in Latin or Greek quite straightforwardly. I think you could also make such a series work in English — ‘as the cat mistakes moonbeams for milk, so does an elephant stems and a girl her dress’ — but it produces markedly high-register English and as soon as you try to add any further detail about either the subjects or the objects, the elephant or the girl, the stems or the dress, of the kind that are in this poem, it quickly becomes impossible to follow.
This kind of problem is familiar to anyone who’s tried to translate Latin or Greek verse. Some of the challenges of translating Sanskrit poetry, however, are unique to it. Sanskrit literature is a vast literary tradition, stretching as a literary language from the age of Homer or earlier (the Vedic of the Rig Veda is a form of Sanskrit) to early modernity: it’s a bit like the whole history of production of Greek and Latin, including Byzantine Greek, medieval and early modern Latin, put together and treated as a single literature. It has its own complex conventions of style and imagery, as well of course as mythological, cultural and religious references, and a vast tradition of poetic theory. Anglophone readers in the UK or the US are unlikely to have any cultural familiarity with any of this, not even to the small degree that we might continue to expect for the conventions and mythology of Ancient Greece and Rome, or for Christian beliefs and iconography. There’s a small example in this poem: in the original, the elephant mistakes the moon beams for bisam, a word which means a stem or stalk of lotus.
Lotus has a host of associations in Sanskrit literature, none of which a modern Anglophone reader can be expected to be familiar with. Removing it undeniably removes a meaningful element of the original. But for most Anglophone readers lotus suggests, perhaps, India in general, but not I think tasty pale stems, which is the point of the comparison here. And of course there’s no single word in English for ‘lotus stem’. On reflection, I felt that including the word lotus distracted from the comparison.
Decisions of this kind are also partly decisions about the translation of register and style. If the original is itself using unusual, obscure or surprising vocabulary then you want the translation to do something similar. But this particular poem is actually written in quite simple and straightforward Sanskrit compared to some of the pieces in the anthology (and others I’ve translated) and to many of the most famous Sanskrit poets like Kālidāsa. Whereas in other translations from Sanskrit I have used rhyme and a distinct metre to suggest self-conscious artistry in the original, here I chose a more straighforward style to reflect the relative simplicity of the Sanskrit. (The original, like all Sanskrit poetry, is in a quantitative metre.)
Another significant problem for the translator is that formal literary Sanskrit of this period has a particular tendency to form enormous compound words, which is unlike anything in Latin or Greek (or any modern literary language that I know). Such a long compound may include a complex series of implied case relationships within it, a sort of little sentence of its own, though treated grammatically as a single word, with a single grammatical ending. There is no very challenging example of this in this poem, but the second line starts with a modest instance of it: tarucchidraprotān. This means, literally, something like “sewn (or inset) into the lattice of the tree”. It’s in the accusative plural, agreeing with the moonbeams of the first line: seeing moonbeams woven amid the branches of the tree, the elephant takes them to be lotus stems.
This is a very minor example. In more elaborate Sanskrit poetry, compounds can be extremely long, complex and often ambiguous with multiple possible ways of resolving their constituent parts. Antonia Ruppel, in her excellent recent Sanskrit textbook, gives the example of a single compound from the Aryāsaptaśatī (a collection of 700 short poems from twelfth-century Bengal) which means: ‘Whether I get up, sit down, lie on my bed, turn around, twist my body, or walk about.’ A. A. Macdonell in his Sanskrit Grammar for Students pithily remarks: ‘Thus Kālidāsa describes a river as ‘wave-agitation-loquacious-bird-row-girdle-string-ed’, while we should say: ‘her girdle-string is a row of birds loquacious because of the agitation of the waves.’
I’m not sure how to describe it, but there is a specific intellectual and aesthetic pleasure in encountering such complex, but grammatically unmarked sequences — in most compounds, only the final element has an ordinary grammatical ending — in a language which is so highly inflected. It creates a pleasing oscillation between inflected and uninflected forms, and also a space for interpretation, a sort of parallel grammar supplied by the reader or commentator. (There is an enormous tradition of Sanskrit commentary, quite a lot of which, you may not be surprised to hear, is devoted to helping you puzzle out compounds). At the same time, there is another kind of counterpoint created between the ordinary syntax of the Sanskrit sentence — with subjects, objects, verbs and so on — and the alternative syntax of a long compound, which is held fixed, syntactically speaking, as a single noun or adjective. There’s nothing really like this in any European literature that I have encountered.1
If you are interested in reading some Sanskrit poetry in English translation, John Brough’s old selection for Penguin is still I think one of the best and most readable — long out of print, but still widely available secondhand. More recently, two contemporary poets I particularly admire have produced Sanskrit translations: Amit Majmudar’s Godsong (a version of the Bhagavad Gita) and Mani Rao’s Kālidāsa for the 21st Century.2 Kālidāsa is generally considered the greatest Sanskrit poet and the translation of Kālidāsa holds the same kind of irrestistible and hopeless allure as that of Horace or Pindar. I’ll come back to him another day.
Though oddly the closest comparison is perhaps to Proust: his concatenation of clauses within a sentence, the way his clauses may contain whole little scenes and dramas of their own, often tangential to the main action of a sentence and almost sealed off from it, is the closest Western parallel I have encountered. It’s also possible that I’m just obsessed with Proust.
Mani has also translated the Saundarya Lahari, a famous Sanskrit hymn, and any classicists may be particularly interested in her little collection of poems influenced by Greek and Roman myth, Sing to Me.
Love this. Such a beautiful description of the pleasure of filling in the grammatical spaces left by this kinds of language structures. Have also never heard of the poem, and this makes me want to read the original as well as this lovely translation.
It occurred to me recently that the semicolon plays a role in English that is similar to that of the specific kind of Sanskrit compounds you're talking about here. Of course the semicolon has a narrower range of un-spelled-out relationships between the elements it coordinates, but it's still beautiful, the space it creates for the reader to fill in.
One of the freshest and most enjoyable translations I have read. Not an expert at all in such translations I hasten to add! But I am a lover of poems that manage to build themselves seemingly out of light. For myself this poem does so. Thanks for sharing.