How can you avoid self-delusion in translation? It seems like translators fool themselves a lot, that they are not necessarily good readers of their own translations. Their justifications tend to be self-serving. My method is to give myself a grade for each translation, and then work over each poem that is at a "C" level or lower until it is at least a "B," then go on to make sure each is at least an "A-." Even then, I might be fooling myself.

There should be no justifications. If you have to give an explanation, it means something is amiss.

There are two ways of looking at this. One is to look at the floor, the other, the ceiling.

On the floor level, you want to make sure there are no flaws caused by misunderstanding the original text. No line that sounds awful in English. If one of the poet's main tools is repetition, don't eliminate the repetitions. First, do no harm; that's the Hippocratic oath of translation.

By getting everything on the floor level, you are basically saying: nothing falls below this point. The translation might still be a C or C+, but all you're really doing is protecting yourself against the translation police, ensuring that nobody will find a howler.

To look at the ceiling provides a different perspective. Is the poem convincing, not as a translation of a putatively great original, but simply on its own terms. At this point it is not enough for a line to be not awful, or a reasonably good way of translating the original. If the floor is more or less fixed, the ceiling is infinite. There is no ceiling, only a sky.

A lot of the ways we think about translation don't make too much sense to me. Do we need a balance between floor and ceiling? No. Once the floor is achieved, we don't have to worry about it any more. Once a translation agrees 95% with what any competent translation would be, the semantic material of the original really doesn't need much more attention.