The second is the idea of structuralism, in which every part of the structure of a language is related. Think of it like this: the real world is a plain patch of ground, and language is a net we throw over it. Each time the net falls, every one of the diamond-shaped holes lands on a slightly different patch. The net’s a bit worn out, and some of the holes are torn, meaning they cover more ground. Some bunch up and cover less. Think of words as being like these holes: so saudade might mean something slightly more than homesickness, whereas dépaysement means something less, referring only to that kind of homesickness you get from being in a foreign country. Linguists have called the semantic space words occupy a “lexical field”.

So the Sapir-Whorf hypothesis falls, but the notion of lexical fields makes a lot of sense. In short: no word is completely untranslatable, but then no word is precisely translatable either. And, I promise you, that’s no schnapsidee.