Language Shapes Thought — But How, Exactly
The 100 Words for Snow That Never Existed
— Listen, there's this eternal story about the 100 Eskimo words for snow — and the conclusion that they must see snow differently too. My whole life I thought it was pretty nonsense. But recently I ran into the claim that Boas was actually right about the count. So is it made up or not?
— Both, and that's the interesting part. Boas in 1911 had exactly 4 roots. Four. Aput — snow on the ground, qana — falling snow, piqsirpoq — drifting snow, qimuqsuq — a snowdrift. And he drew absolutely no conclusion from it about thinking. He listed them as a mirror to English, where the very same water gets separate roots: water, ice, steam, rain. The point wasn't about Eskimos. The point was that every language splits things somewhere and lumps them together somewhere else.
— So where does the hundred come from?
— The number grew on its own, without a single study behind it. Whorf, in a popular article in 1940, is already writing that to an Eskimo an all-inclusive word for snow would be almost unthinkable — he doesn't name a number, by the way. Then come the retellers. And in 1984 an unsigned New York Times editorial drops it in passing: 100 words. No citation, no name, no author. 2 years later Laura Martin traces that chain back through the publications and finds that in all that time nobody had opened the original source. Not one person.
— So it's pure broken telephone.
— Geoffrey Pullum in 1991 called it the great Eskimo vocabulary hoax. And that's usually where the story ends — everyone's satisfied, the myth is busted, we can all go home. But hold off on going home. This is where it gets uncomfortable.
— What's uncomfortable about it?
— On the count, Boas is closer to right. Woodbury found about 15 lexemes for snow in Central Alaskan Yup'ik. Krupnik in the 2010s — on the order of 40 terms in Central Siberian Yupik and no fewer than 53 in Nunavik Inuktitut. And that's without counting the sea-ice terminology, which is richer still.
— Wait. So the debunking of the myth turned out to be a myth too?
— Halfway. What got debunked wasn't the count — it was the conclusion. Because nothing about the architecture of the mind follows from 53 words. A Russian skier has nast, krupa, tselina, podlip — wind crust, granular snow, untracked powder, and the sticky stuff that clumps under your skis — and that tells you about his weekends, not about his head. A rich vocabulary is about what you do for a living. Always.
— Look, I came here with two footholds. One: "language determines thought." Two: "no, that was debunked ages ago." You've just kicked out both.
— Both need kicking out, because they lean on the same spoiled example and both lose to the same kind of data. People who had the words taken away and kept the thought.
— How does that work?
— Varley and colleagues in 2005 described three men with the most severe agrammatic aphasia after a left-hemisphere stroke. So severe that they couldn't say what the difference was between "the lion ate the hunter" and "the hunter ate the lion." Same words, different order — and for them there's no difference, spoken or written.
— That's horrible.
— And now you hand them an arithmetic expression with brackets and nested operations. And they work it out. That's exactly the same structural operation: something is embedded inside something else, and meaning depends on the order. In language, unavailable. In numbers, it works. So where's the prison?
— All right, but these are people who did once have language. It stayed inside them.
— Fair objection, and it's been tested on people who never had it left. This is the interesting part. Look: the Pirahã, an Amazonian people with no words for exact quantities — none at all, to the point of having no word for "one." You give them a task: lay out the same number. As long as the objects are in view, they do it with practically no errors, at any quantity. They start making mistakes exactly when the quantity has to be held in memory or the row has to be reversed.
— So they can't count after all.
— Hold that thought, here's the main thing. In the same study they took Americans and gave them the same task — only they asked them to repeat digits under their breath at the same time. And the Americans started making mistakes exactly like the Pirahã. Nobody took any words away from them. The channel was just occupied.
— Hang on, they've got the words.
— They do. Those words are just busy at that moment.
— So what does that leave us with? That the question is posed wrong?
— The question is aimed at the wrong place. We ask "what is a person capable of thinking at all" — when we should be asking "what is cheap for them to think." It's not about the ceiling, it's about friction. Imagine you're moving house and carrying things in your arms. Physically you can carry everything — just in ten trips, dropping small stuff on the stairs and not remembering where anything ended up. And then someone hands you boxes with labels on them.
— And the box doesn't make me stronger.
— And it doesn't forbid you to carry anything. It lowers the price. That's what a word is: without it the load doesn't disappear, it just gets more expensive — slower, with losses, and it can't be handed to someone else whole.
— Nice. But now that's not a testable claim at all.
— On the contrary, only now is it testable. Price can be measured. In milliseconds, not in destinies. And that's where all the interesting stuff begins.