Bad science reporting again: the Eskimos are back

« previous post | next post »

You just can't keep a bad idea down. And you just can't lift the level of bad science journalism up. David Robson of New Scientist, in a piece published in that pop science rag a couple of weeks ago (issue of 22/29 December 2012, p. 72; behind a pay wall) and now also published in the Washington Post, reports on a book chapter by Igor Krupnik and Ludger Müller-Wille about anthropologist Franz Boas's travels in the early 20th century with a Canadian Inuit band whose language he learned. Robson says of Boas:

Mentioning his observations in the introduction to his 1911 book "Handbook of American Indian Languages," he ignited the claim that Eskimos have dozens, or even hundreds, of words for snow. Although the idea continues to capture public imagination, most linguists considered it an urban legend, born of sloppy scholarship and journalistic exaggeration. Some have even gone as far as to name it the Great Eskimo Vocabulary Hoax. The latest evidence, however, suggests that Boas was right all along.

Not a single statement in this passage is correct.

  • Boas never "ignited" the familiar claim. That came many decades later, after a major role had been played by popular works (among them an article by Benjamin Lee Whorf and a book by Edward T. Hall called The Silent Language). Boas made no quantitative claims at all; he just noted that languages didn't necessarily draw the lines between the lexemes in semantic fields in the same places as other languages: Canadian Inuit separates falling snowflakes (for which the qana- root is used) from snow lying on the ground (for which the api- root is used), just as English separates water running along (as in river) from water standing still (as in lake), and so on. He was stressing that this arbitrariness of lexical denotation boundaries was something the two languages had in common, not that Inuit was quantitatively unusual.
  • It is not true that "most linguists considered it an urban legend": many taught it in their introductory classes throughout the 1960s and 1970s and on into the 1980s, and I am sure there are some who still do. Quite a few have never paid attention to the story and have no view on it (linguists are much less interested in "X has many words for Y" stories than the general public seems to be). Some are expert Eskimologists and have very nuanced (and divergent) evidence-based views on the issue. Many may have thought the supposed factoid was a bit of a cliché (I certainly did even when I was an undergraduate). But it was not a linguist who first noted that there was a kind of urban legend here. That point was first made in 1982 when anthropologist Laura Martin gave a paper to the American Anthropological Association on the topic. Anthropologists squealed, and referees hated her, but by 1986 she had managed, with difficulty, to get a short research report (not a full article) published in American Anthropologist. Yet still virtually no linguists knew of that paper until years later.
  • It is not true that "some have even gone as far as to name it the Great Eskimo Vocabulary Hoax." Just one person named it thus: I did. The phrase was the title of a 1989 humorous essay (later included in a 1991 book with the same title), an essay in which I tried to publicize Laura Martin's work. I mocked the credulous parroting and arbitrary numerical invention that one finds in newspapers and magazines, and critiqued the practice of repeating traveler's tales about exotic peoples without having any evidence. Forgive my pique, but I'm a bit annoyed to see Robson attributing my only contribution, the cutesy name, to folklore. It was not a current phrase that linguists in general were using; it was a title I invented, and I'm getting a bit tired of people picking up the phrase without citing me. (There was another such incident not long ago; I'll tell you about it later.) Robson didn't mention Laura Martin either, and that's even worse, since her paper was scholarly and serious, and opened up the topic.
  • Finally, the new evidence from Krupnik and colleagues (and it is not really new, just re-examined lexicography in a book chapter from 2010) does not suggest for the first time that Boas was right all along. Nobody ever doubted that Boas was right in what he said: his anodyne factual observations held no trace of the usual story — the notion that there are scores or hundreds of different snow types that only Eskimos can distinguish because only they have the different names that are needed. Boas cited four snow-connected roots, and made (in 1911) no claim that there were others, or that four was a lot. What Krupnik and Müller-Wille do is to clarify (very usefully) Boas's role as a student of Canadian Inuit and to catalogue some snow and ice vocabulary and allege (somewhat unfairly) that linguists like me were wrong all along to cast aspersions on the myth-repeaters. (I will comment on the Krupnik team's results another day.)

It was never in doubt that there were several distinct Eskimoan lexemes denoting snow phenomena: Boas had given four (Robson also cites just four, though they are different ones), and there are certainly some others. But one has to be rather careful if counting distinct snow terms is the game. What is a snow term? Some Eskimoan dialects use a derived word (kavisik) meaning "snow with a herring-scale pattern on it caused by re-freezing of rain pockmarks on fallen snow," but the root (not found in the Fortescue/Jacobson/Kaplan Comparative Eskimo Dictionary) appears to mean "herring." So we need criteria for deciding whether that would be counted as a snow word or a herring word — not an insignificant matter, especially with a language that has such productive word formation that you can construct arbitrarily many derived words for snow or herring or coffee or anything else. Similar cautions hold for many other items. Illuksaq in Greenlandic has been cited as a word meaning "snow suitable for building an igloo", but in fact illu- means "house" and -ksaq means "stuff for the construction of", so illuksaq means "house-building material". It is not a snow word at all. In a similar way, I have seen a word for soft snow cited, but it appeared to be based on the root meaning "soft", and a word for early autumnal snow apparently based on the root for "fall", and so on. I'm just not sure how many of these should count as snow terms.

No one will pay attention to such details now. New Scientist and the Washington Post have announced that Boas claimed there were fifty snow lexemes and that new research has now confirmed this; so everyone will believe that, since they wanted to believe it anyway, and they will keep on repeating the same drivel about things-people-have-multiple-words-for that they have so often repeated in the past. A depressing prospect, but it seems inevitable.

My article (which Robson tells me he did read) was not aimed at establishing any negative quantitative claim about Eskimoan lexical resources. It did add some documentation of the absurd numerical exaggerations that were current by the end of the 1980s, but it was primarily satirical. It mocked people who pick arbitrary numbers out of the air and use the factoid about the alleged hundreds of snow lexemes to make themselves sound clever without having a shred of evidence to support what they said about these languages of which they know nothing. Robson has simply made it easier for people to go on doing that.

Comments are closed.