Abstract
A poet trying to explain to is readers how he writes may fall into a dubious situation, because he sometimes is unable to explain it even to himself. If he knew, he would not be a poet. Can one parse one's best poems by means of structural analysis, arrive at an approximate formula for them in this manner, and use the formula to write new ones ? A new poem assiduously copied from an earlier one — one's own or someone else's — has already lost the wild mystery of the pristine, the moist freshness of the discovery of a world. Computers may be able to write quite passable "average" poems that are, regrettably, in no way poorer than those of many currently functioning versifiers. Canons can be programmed. But one cannot program what smashes canons, and it alone is genuine poetry.