05 · The Snow That Never Was

In 1986 Laura Martin traced the number everyone was citing — how many words for snow does Eskimo have — back to its origin, and found that the origin had mentioned four roots. Not fifty. Not two hundred. And the origin made no claim whatever about how its speakers thought.

The origin is Franz Boas, 1911, in the introduction to The Handbook of North American Indians. He mentioned four roots: aput, snow on the ground; qana, falling snow; piqsirpoq, drifting snow; and qimuqsuq, a snow drift. His point was purely technical. One language builds these meanings from separate roots; another builds them from a single root with modifiers attached. He did not write that the people who spoke that language saw snow differently from anyone else.

Whorf enlarged the number in 1940. The essay "Science and Linguistics," in Technology Review, raised the implied count to at least seven, with no footnote and no source, and asserted in addition that English covers all those distinctions with one word, snow. Pullum later established that the second assertion is false. The version later generations have cited sits in the collected volume Carroll edited in 1956, page 216.

What Martin documented was not a liar. It was a number swelling one relay at a time. Roger Brown in 1958 wrote three, misreading Whorf's own figure. Carol Eastman in 1975 wrote "many words" six lines from a passage she herself quoted that said three. Every relay cited the relay before it. No relay went back and counted.

Geoffrey Pullum collected the specimens that had escaped into the wild in a 1989 paper, "The Great Eskimo Vocabulary Hoax," published in Natural Language & Linguistic Theory:

• Nine — a trivia encyclopedia, 1984
• About four dozen — New York Times, 9 February 1988
• Fifty — Lanford Wilson's play The Fifth of July, 1978
• One hundred — New York Times editorial, 9 February 1984
• Two hundred — a television weather forecast in Cleveland

The two figures from the same newspaper do not agree, and this book prints both. One hundred in 1984 and about four dozen in 1988, four years apart, matching in masthead and in calendar date. Between those two days nobody went and counted. A number that was never measured can move in any direction.

The closest thing to a measurement in this affair is one Pullum took from C. W. Schultz-Lorentzen's 1927 Dictionary of the West Greenlandic Eskimo Language. It holds two roots bearing directly on snow: qanik, snow in the air or a snowflake, and aput, snow on the ground. Two. Not seven, not two hundred.

The fault heavier than the number is that no experiment was ever underneath this claim, not once. Nobody measured whether speakers of one language distinguish snow more finely than speakers of another. Nobody set a criterion for what counts as one word. What the claim had was a citation chain that lengthened every time it was retold. Both Martin and Pullum offer it as a case study in the field's own standards of evidence, not as a refutation of any relation between language and thought.

The line most often attributed to Sapir is not Sapir's position. In the 1929 paper "The Status of Linguistics as a Science," published in Language volume 5, pages 207–214, he wrote at page 209 that the real world "is to a large extent unconsciously built up on the language habits of the group." Elsewhere in his work he argued against the proposition that language determines thought, explicitly and more than once. The weak-versus-strong pairing that travels with his name is not his phrase and not Whorf's. A later generation fitted that frame back onto them.

Whorf's own language is genuinely stronger. In the 1940 essay, in the version Carroll collected in 1956, page 213, he wrote that the world reaches us in a "kaleidoscopic flux" that thought must organise, "and this means largely by the linguistic systems in our minds." Whorf was not a trained professional linguist. He was a fire-prevention engineer at the Hartford Fire Insurance Company who studied linguistics as an avocation under Sapir at Yale. That biography settles nothing on its own. What settles it is the evidence base of his flagship claim.

The flagship claim was that Hopi grammar makes no reference to time in the way English does, and that Hopi speakers therefore hold no comparable concept of time. Its evidence base is thin. The source states that it reportedly rests on interviews with a single Hopi speaker, Ernest Naquayouma, in New York, not in the territory where the language is spoken. The source says reportedly, not confirmed, and this book keeps the word.

In 1983 Ekkehart Malotki published Hopi Time: A Linguistic Analysis of the Temporal Concepts in the Hopi Language with Mouton, running to some six hundred pages. The book documents that Hopi has tense, has words for units of time such as days, weeks and months, and has spatial metaphors for time. Malotki stated his own aim: he wrote it to refute Whorf. That intent is a fact about the author, not an argument against the document itself. What does the work is the grammatical description, and the linguistics that followed treats it as having succeeded.

This is not a replication coming back empty. It is a description of a language wrong at the level of fact. An empirical claim that was a load-bearing pillar of the theory fell because someone went into the field specifically to check it. Terry Regier and Yang Xu, in a 2017 review in WIREs Cognitive Science, wrote that the hypothesis occupies a "troubled and contested" position and carries an "inconsistent record of replication." Both write from inside the research tradition that supports it.

The bounds of this verdict have to be stated plainly. The snow story does not refute the claim that language affects thought. It refutes the most-repeated piece of evidence ever used to prove that claim. The two are not the same thing. Effects that survive scrutiny do exist, and they appear in the next chapter of this book, but they are small, they depend on the task the subject is given, and they are measurable only under narrow conditions. Under the snow there is nothing at all.

The rule this chapter lays down for the whole book has one clause. A claim that flatters this book's thesis gets checked harder than a claim that cuts against it, not more softly. The snow story is the best opening this book could have had — short, memorable, favourable to the thesis in every particular. It has no evidentiary content whatever, and the book prints that instead.

What could not cross: What failed to cross in this chapter is not a word between two languages but the condition that made a number checkable. Boas distinguished a root from the forms built on it. That distinction is the whole of what gave his "four" any meaning, and it was the first thing to go when the claim was retold in the English ordinary writers write. Once the counting rule stops travelling with the number, the number can be anything.

Overturned if: A document published before 1940 gives a snow-word count above four for some named Eskimo language, states its counting rule — roots, or forms built on roots — and the same rule applied to English still leaves a real gap. Or field work shows that the tense and the time-unit vocabulary Malotki recorded in 1983 do not match the Hopi people actually speak.

Sources: Laura Martin, "'Eskimo Words for Snow': A Case Study in the Genesis and Decay of an Anthropological Example," American Anthropologist 88(2):418–423, 1986 — https://anthrosource.onlinelibrary.wiley.com/doi/abs/10.1525/aa.1986.88.2.02a00080 · Geoffrey Pullum, "The Great Eskimo Vocabulary Hoax," Natural Language & Linguistic Theory 7:275–281, 1989 — https://pullum.ppls.ed.ac.uk/EskimoHoax.pdf · Terry Regier & Yang Xu, WIREs Cognitive Science 8:e1440, 2017 — https://terry-regier.github.io/lclab/papers/whorf-wires-preprint-v2.pdf · Stanford Encyclopedia of Philosophy, "Whorfianism" — https://plato.stanford.edu/entries/linguistics/whorfianism.html · Ekkehart Malotki, Hopi Time: A Linguistic Analysis of the Temporal Concepts in the Hopi Language, Mouton, 1983 — https://archive.org/details/hopitimelinguist0000malo · "Hopi time controversy," Wikipedia, secondary source — https://en.wikipedia.org/wiki/Hopi_time_controversy · "Ekkehart Malotki," Wikipedia, background — https://en.wikipedia.org/wiki/Ekkehart_Malotki · Boas 1911, Whorf 1940 and the 1956 edition, Brown 1958, Eastman 1975 and Schultz-Lorentzen 1927 are cited through Martin 1986 and Pullum 1989; Sapir 1929 page 209 and Whorf 1956 page 213 are cited through Regier & Xu 2017. This book did not read those primary documents directly.