Several readers have commented on the fact that translated poetry has featured less in PN Review in recent years than it did in the past. There have been notable exceptions, of course, but PN Review does not appear to be as hospitable to translation as it once was. 'The essays and reviews of foreign writing hardly compensate,' one reader declares. 'Is there a reason for this change?' another asks.

In the 1970s and 1980s, translation seemed central to publishers and readers alike: poets learned from the otherness of foreign writing; and they often found in the poetry of eastern Europe in particular a seriousness of thematic and historical engagement in which they felt their own poetry defident. A whole discourse built up around notions of 'witness'. What was lacking in the quality of translations was made up for by an earnest worthiness in the enterprise. The strict judgment applied to original poems did not extend to translations. Even today in many quarters it does not. As an editor I feel that it should, and translations - where they can be made at all - require something of a cultural context if readers are not to mishear, misread and misappropriate them.

I recently re-read with dismay my first book, a joint venture with the American poet Edward Kissam, entitled Flower and Song (In xochitl in cuicatl, consisting of careful, much-debated translations from Nahuatl (Aztec and related) poetries. Nahuatl is a synthetic language, without prepositions, without pure sibilants, heavily consonantal. There is less a sense of words than of phrases, sense-bundles. There is no traceable link between Nahuatl and any European language.

Nahuatl poetry is the product of a sophisticated culture, its structures tribal and strictly hierarchical in character. The poems were taken down by clerical scribes who transliterated the sounds they heard in the years after the conquest of Mexico. The Spaniards were not necessarily aware of what later scholars describe as the integrated texture of a poetry in which each image contains figurative, religious, philosophical and literal meanings with a rigour which makes the strict disdplines of medieval allegory seem almost arbitrary. It subsumes in a single oral 'code', fixed at the time of transcription, elements of cultures the Aztecs had swallowed up, espedally the Toltec. It can be read as a perversion of civilized Toltec values, yet those values survive subversively within verse which laments ephemerality, celebrates brotherhood, and seems to run in the teeth of the warriors who sang it. Indeed, just as the scribes wrote down what they heard in ignorance of its meanings, so those who chanted and danced may have been unaware of the subversion in their mouths. A limited number of stock images are deployed; they are capable of a wide range of meaning.

My knowledge of the language was rudimentary. I used Spanish cribs and Spanish expositions, building a sense of the tradition through the filters of the language of the men who had destroyed the larger culture, preserving almost acddentally these verse shards. With gold, onyx, codices and other plunder, manuscripts were shipped back to Europe where they gathered dust in ecclesiastical and royal libraries for centuries.

As a young translator I learned how one kind of poetry can contain and conduct meaning. From the strangeness of Nahuatl prosody I learned something about rhythmic strategy, about otherness. Those translations advanced my understanding of how poetry works, subverts and creates. It should have taught me something else: that English was not a receptive host language.

What can an English reader learn from these translations? Absolutely nothing, I now realise. Except for the 'Songs of the Fallen', poems of defeat infected with the rhetoric of the conquering culture and partly defracted through it, nothing comes across in English unless the reader makes an immense investment of study, good-will or suspends disbelief to a vertiginous degree. Without foreknowledge of the nature of the language and tradition and of the accompanying ethnology, the poems are inert. With that knowledge, they can be dedphered and understood, but no British English form can convey them as experience in the terms they propose. There are no authentic instruments, as it were. Certain species of Amertcan English get close to the literal sense but lose the figurative altogether, and it is the figurative that matters. This tradition comes across as atavistic, primitive, absurd. Yet it is not absurd, not atavistic (on the contrary) and anything but primitive. By the time of Cortez it stretched back perhaps seven hundred years, conventional, eloquent to those among its people who were initiates.

Translation into English of certain languages (as of certain specific poets) is impossible. The closest approximation is so remote that, were it translated back into the ortginal language, the resulting re-translation would be different in kin from the original text. Even where there are common antecedents and references translation can prove vexed. The best an English reader can hope for is fragmentary access. When we read a poem, Yeats said, we tend to respond to something other than the poem; when we read a poem in translation this is even more the case. The proliferation of anthologies of poetry from eastern Europe ceased when eastern Europe became a different kind of challenge and dissenting voices were no longer read as 'witness'. They were no longer read at all.

Translation is undeniably useful to the gifted poet translator. Some translations do extend the expressive range of the host language. Yet it now seems to me that in the last three decades poor translation has muted and impoverished not only our sense of the poetries translated (however potent the content of those poems), but our response to original poems in English. PN Review remains hospitable to translation, but warier and more sceptical than it once was.