Among the many games devised by the OuLiPo group of writers was a variety of translation that preserves not the semantic features of the original text, but instead preserves its phonetic features while altering its meaning.
This is something that can be done within a single language, or from one language to another. French seems particularly well suited to intralanguage homophonic translation, since there are so many different ways of combining letters to produce one and the same phoneme. To give the most obvious example, -é, -ée, -és, -ées, -ai, -aie, -ais, -aix, -ait, and -aient are all pronounced the same way (provided that we do not have to worry about elision with an initial vowel of a following word).
This feature of French makes it possible to come up with homophonies such as the following:
Queneau (Raymond), très haut pape, l'a répété:
Que nos raies, montrées aux papelards, aient pété.
Queneau (Raymond), most high pope, said it again:
That our rays, shown to the hypocrites, farted.
In the French version, the first line is pronounced identically to the second one, in spite of enormous differences of both spelling and meaning.
I have tried and tried, but I have never been able to come up with so much as a single pair of homophonous English sentences, both of which, however obscure, are at least syntactically correct and both of which at least mean something or other. If intra-language homophonic sentences are so difficult to come by, it might seem nearly impossible to produce a proper homophonic translation between languages. But OuLiPo managed to come up with a few of these as well. For example, Catullus's famous lines
Odi et amo. Quare id faciam fortasse requiris.
Nescio, sed fieri sentio et excrucior.
can be rendered in French as:
O diète à mots couards! Raide face et âme forte acérée qui rit ce
N'est ce qui, ô cède fière, hisse, haute Io et tais ce que crus ci-hors.
At the level of meaning, there is nothing in common between the two sentences, yet they are nearly identical with respect to pronunciation (at least following the usual French conventions for Latin pronunciation, which Anglo-Saxon Latinists reject). The original text from Catullus can be translated (in the traditional sense) as follows:
I hate and I love. Why do I do it, perchance you might ask?
I don't know, but I feel it happening to me and I'm burning up.
The French, in turn, is so obscure, indeed is so near-nonsensical, that I can't translate it with any confidence. What's more, the homophony breaks down at at least one point ('sentio' is not homophonous with '-se haute Io').
The obscurity and difficulty of homophonic translation shows, I think, how restricted one's expressive range becomes when highly demanding formal constraints are placed on language. This restriction almost functions as an argument against working with such formal constraints, and so against the entire OuLiPo project.
But as I understand it OuLiPo's argument was always that formal constraints have been part of the poetic or artistic use of language all along (e.g., rhyme, meter), and that it is only a certain 20th-century laziness, associated in particular with the rise of surrealism and with the idea of 'the unconscious', that made writers and readers think that free expression of whatever flows out of a person might have some artistic value. For OuLiPo, such 'free' expression is simply operating under other, external constraints, of which the author is unaware; much better to set one's own constraints, to be the master of one's own limitations.
More generally, this particular game raises questions about our presumption that the conveyance of meanings is the supreme purpose of writing and in particular of translation. Why should the meaning of a sentence be the sole feature of it that one aims to preserve in translating from one language to another? What if meaning is just one of the features that might be selected to be conserved and made accessible to readers of another language? Doesn't the French sentence convey something of what Catullus said, to readers who have access to French but not to Latin?
Obviously, the preservation of non-semantic features in translation will be more important for poetry than for prose, and it's the overwhelming dominance of prosaic language in our age that gives rise to the presumption that translation can have only one aim: to carry meaning over.