Translating

 I think I've mentioned here before my experience in Vietnam of being told at a red light that we have to wait for it to turn blue.  I've seen the same thing in Cambodia.  Now I read that as languages pick up color words, the word for green is always last to enter the vocabulary.

Yes, and your Vietnam/Cambodia example touches on something linguists have found fascinating for decades.

Many languages historically did not distinguish what English calls blue and green. Instead they had a single color category covering both. Linguists often call this a "grue" category. Even today there are languages where the boundary between blue and green is different from English, or where one word comfortably covers both.

That helps explain why someone might say a traffic light turns "blue" when an English speaker would say "green." In Japanese, for example, the traditional word ao covered both blue and green, and although modern Japanese has a separate green word (midori), traffic lights are still commonly described as blue. Vietnamese and Khmer have had somewhat similar historical patterns, though each language has its own color system.

Your observation about green being the last basic color term to appear comes from the famous work of linguists Brent Berlin and Paul Kay. They argued that when languages develop increasingly rich color vocabularies, they tend to do so in a fairly predictable order:

    Black/dark and white/light.

    Red.

    Green or yellow.

    The other of green or yellow.

    Blue.

    Brown.

    Purple, pink, orange, gray, and so forth.

Under that theory, green is not actually the last major color term to appear. Instead, blue usually emerges after green and yellow are distinguished. However, green is often the most unstable category because nature presents a huge range of greenish, yellowish, and bluish hues, making the boundaries fuzzy.

There's also an intriguing biological angle. Human vision is especially sensitive to green wavelengths, but that doesn't mean languages must carve out a separate word for green early. A language can function perfectly well for thousands of years with a broad color category that includes what English speakers would call green, blue, or even gray.

So when a Cambodian or Vietnamese speaker tells you the light has turned blue, they're not necessarily seeing a different color. They're often using a color classification system whose historical boundaries differ from the English one.

In a way, traffic lights preserve a little linguistic fossil record. Every time someone says, "Wait until it turns blue," they're echoing an older way of dividing up the color spectrum.

Oh for sure when they want to Vietnamese make a distinction.  Blue of the sky, green of the trees.

Very much so. One of the classic mistakes in translation is assuming that because two words are listed as equivalents in a dictionary, they cover the same territory. Often they don't.

Color terms are a clear example, but the phenomenon is everywhere. English river and stream don't map neatly onto the words used in many other languages. English friend covers a huge range of relationships that other languages may divide into several categories. Even something as simple as chair can be tricky if another language distinguishes stools, chairs with arms, floor-seating furniture, and so on.

The problem becomes even greater with abstract concepts. Consider the English word freedom. A dictionary may offer a foreign-language equivalent, but the historical, political, and emotional associations of the two words may differ considerably. The same is true of words like honor, virtue, religion, family, or privacy.

Linguists sometimes talk about words as having a "semantic field." The dictionary gives you a rough center point, but the actual boundaries of the field are determined by how speakers use the word. Those boundaries rarely line up perfectly between languages.

That's why skilled translators often say they translate meanings, not words. A literal word-for-word translation can be perfectly accurate according to the dictionary and still be wrong according to how native speakers actually understand the sentence.

Your xanh example is a good illustration. A dictionary that says "xanh = blue, green" is not really telling you what the word means. It's telling you that the Vietnamese color category overlaps two English color categories. To understand the word, you need to know where Vietnamese speakers draw the boundaries and when they feel a need to specify them.

In a sense, every language is its own map of reality. Translation works because the maps overlap a great deal, but the borders are never quite in the same places.

Whorfian hypothesis?  Those boundaries influence how we categorize things?

Yes—that's very much in the territory of the Whorfian hypothesis, although the modern view is considerably more nuanced than the popular version of it.

The strong version, linguistic determinism, says that the language you speak determines or severely constrains how you can think. That is generally not supported. A Vietnamese speaker can certainly distinguish blue from green; an English speaker can understand concepts for which English has no convenient single word.

The weaker version, linguistic relativity, is much more plausible: the categories and distinctions built into a language can influence what people notice, how quickly they categorize things, and how readily they retrieve certain distinctions.

Color is actually a nice example. If your language has one everyday category covering blue and green, you don't become incapable of seeing the difference. But speakers of languages that habitually distinguish the two can sometimes be quicker at making that distinction, particularly when the colors are near the boundary.

And it goes both ways. English speakers have distinctions that Vietnamese speakers may not routinely make. We distinguish blue and green with separate basic words, but we lump a tremendous range of blues together. A language with several everyday basic terms for different kinds of blue can make those distinctions more cognitively salient.

So I would put it this way:

Language doesn't put blinders over reality; it puts labels and boundaries on the map of reality. Those boundaries can affect what we habitually notice.

And that brings us right back to your translation point. A dictionary gives you the labels; the semantic boundaries are part of what actually has to be translated. That's one reason genuinely good translation is much harder than finding corresponding words.

I think of the French preposition "chez." "At the home of" just doesn't say it.

Exactly. Chez is a wonderful example because the dictionary gloss—"at the home of," "at the house of"—is technically useful but semantically inadequate.

Chez is really something more like "in the place associated with [a person]". The physical building isn't necessarily the important thing.

Je suis chez Jean → "I'm at Jean's."

But:

Je vais chez le médecin → "I'm going to the doctor."

The doctor might be seeing patients in a clinic, certainly not at home.

Je mange chez McDonald's → "I eat at McDonald's."

And now we're even farther from "the home of." The word has expanded naturally from someone's dwelling to their place of business or characteristic sphere of activity.

There's also chez moi, which can mean "at my home," but can have a broader sense of where I am / my place, depending on context.

This is precisely the sort of thing that gets lost when we treat translation as a dictionary lookup. English has no single preposition occupying the same semantic territory as chez, so we have to choose among "at," "to," "with," "at the place of," "in the practice of," "at [restaurant/business]," etc., depending on context.

And there's an interesting Whorfian wrinkle here: French speakers don't necessarily think differently because they have chez. But the language makes the notion of a person's associated place or sphere extraordinarily easy to express as a single grammatical element.

English can express it perfectly well—we just have to use more words.

That's probably the best way to think about linguistic relativity: languages don't necessarily give us different worlds; they give us different handles for grabbing pieces of the same world.