Anything can be translated. How can languages differ in what they can express, in what kinds of thoughts they encourage or permit, if thoughts born in one can live as well is another? But that translation is always possible does not mean that translation is always easy. For simple phrases, concerned only with familiar and everyday things, there may be only one correct translation. But as the subject moves to remoter things, things less common, farther or deeper, things strange or dreadful, compelling or sacred, translation becomes art and language becomes medium.

This is no news. Anyone can feel how difficult it is to recreate the dictionary denotations, let alone the literature connotations, of one language in another. But this is only the beginning of the art of translation, its scales and studies. The real art is not in deciding how to repeat, but how to fill in.

Abstraction is constructive omission; an abstract word is both a something that is named and the impression of a number of somethings, themselves omitted, that together instantiate or imply it. Every language omits differently; and it is by this difference, I think, that language influences thought: the discrepancies of their abstractions mean that certain thoughts are harder to think in some languages than others, because in one they can be alluded to, and in another they must be constructed on the spot. The thoughts may be the same yet the attitude of the thinker towards them may be different. Compare computer languages: it is of little difference to an interpreter whether a function is called from a pre-supplied library, or defined on the spot as a lambda, but it makes a great deal of difference to the programmer.

Civilized languages have two genealogies: linguistic and literary. A linguist who says that language has no effect on thought is right in respect of linguistic traits. I do not see that it makes any difference to thought whether the language is gendered or genderless, analytic or agglutinative, nominative-accusative or ergative-absolutive. Thinking is so hard in itself that the general difficulty eclipses the particular difficulties or conveniences of certain languages. But languages do have literary genealogies, and these do shape thought. English apprenticed to Latin and Greek, not Sanskrit or Chinese; to discount this fact as shaping English thought demonstrates an absence of thought. Very few languages have civilized—have literized—themselves. The Old World has Greek, Sanskrit, Chinese, Egyptian and Sumerian. All other languages that have ever been written have done so after or while borrowing, directly or ancestrally, from one of these. (The New World has Nahuatl, but alas, it has no disciples.) All other languages serve some apprenticeship. Afterwards some, like the Romance languages, inherit the family business; some, like English or Japanese, buy out the stock; some, like German or Arabic, steal the plans and build their own versions.

To analyze this phenomenon as a form of domination, a side effect of economic and political power, is not stupid—witness Norman French and English, or Arabic and Persian—but it misses the point. The conquered reshape their languages by translation from their conquerors; but conquerors also reshape their languages by translation from the conquered. Greek came to Rome in the mouths of slaves. One generation of Mongols heard Arabic and Chinese only in cries for mercy; the next whispered them in their bedrooms and gardens. Translation is indeed a convenience, is indeed a political act, but it also a transmission, an inheritance, a maturation. The old language passes on to the young language something that it must strive and work to contain—simply put, power: power to know, power to understand, power to think.

Literary descent has two vectors: borrowing and poetry. Borrowing is the easier, more common, and usually the first method. Poetry is the harder but better method. Borrowing always leaves something behind. When languages are young, fast, and hungry, they take words and ideas as they need them, in the sense that comes easiest; and they often come to use them in ways they were not meant to be used. In studying classical philosophy, for example, the hardest step is to get rid of English definitions. Stoics were not stoic; Epicureans were not epicurean; apatheia is not apathy, a daimon is not a demon, kosmos is not cosmos, demokrateia is not democracy, &c.

The diction of poetry is remote and patient enough, enough removed from the necessity of application, that it can take the time to compass an idea. Still, it is not a perfect transfer: the idea retains an often inappropriate exoticism. A Greek might agree that beauty is truth; but he would not have learned the lesson from his kitchenware.

Language's limits are unresisting but real. No language is without limits; yet the limits of my language are the limits of my world, not as a wall limits my movement, but as the horizon limits my vision: I cannot see past it, yet I can never run into it.