February 04, 2006

The proper treatment of snowclones in ordinary English

In October of 2003, Geoff Pullum noted a peculiar sort of "bleached conditional":

If Eskimos have dozens of words for snow, Germans have as many for bureaucracy.
[The Economist, October 11th, 2003, p. 56, col. 2]

I pointed out that a search for {"if Eskimos"} turns up many examples of phrases like "If Eskimos have N words for snow, <some question or assertion about X's vocabulary for Y>", used to frame the idea that Y is especially interesting or important to X. A few days later, Geoff happened to notice another phrasal cohort of the same type -- "In space, no one can hear you X", and wondered what we should call patterns like this. Glen Whitman suggested the term snowclone, and we've been using it ever since. Wikipedia has a snowclone page, and a list of common snowclones.

But two things have been bothering me about about all this. First, phrasal templates like those on the Wikipedia list are often more protean -- and therefore more interesting -- than the descriptions suggest. And second, the original "If Eskimos" example is not really an example of the same thing at all.

First things first. Most of the snowclones in the Wikipedia list are crisp phrasal templates, with one or two open slots to be filled, generalizing a well-documented specific quotation: "To X, or not to X" (from Hamlet's soliloquy); "X for fun and profit" (from a series of how-to books starting with one on stamp collecting); "I, for one, welcome our new X overlords" (from The Simpson's episode "Deep Space Homer"). Others are similarly crisp, even if the origin is uncertain: "X is the MIT of Y"; "X is the new Y".

It's easy to multiply examples of this general type. "Through X with gun and camera"; "ask not for whom the X tolls"; "how do I X thee"; "teach an old X new tricks"; "an X in the hand is worth Y in the Z"; and so on. The procedure is simple and common: take a familiar quotation, title, proverb etc., and fit it to a new context by substituting for one or more of its words.

But the discussion of snowclones so far has suggested that such patterns have a simple and well-defined set of substitution slots, while in fact, people can (and often do) generalize in many different ways from a given phrase,. Thus the proverb "A bird in the hand is worth two in the bush" gives rise to the following -- a small sample of the variants available on the web:

A toad in the hand is worth two in the bush.
A bird in the hand is worth two in the grocery store.
A harp in the hand is worth two in the willows.
A bird in the nesting box is worth two in the bush.
A bird in the head is worth two in the textbook.
A book in the hand is worth two in the bin.
A man in the house is worth two in the street.
A cat in the bath in worth two in the litter tray.
A stone on the board is worth two in the bowl.
A bike on the road is worth two in the shed.
An engine under the hood is worth two in the shop.
A palm in the hand is worth a visor in the mail.
Frank in the bush is worth three in the field!
A beer in hand is worth six in the fridge
a bratwurst on the bbq is worth six in the fridge
Three in the hand is worth about 14 on the black market.
A bird in the hand is worth one from the Bush.*
Two Udalls in the Congress are worth more than two Bushes in the statehouses as far as protecting the environment goes
A bird in the hand is a messy proposition.
A bird in the hand is always safer than one overhead.
A bird in the hand is a certainty, but a bird in the bush may sing.
Would you take the bird in the hand, or a 75% chance at the two in the bush?

Many of the cited snowclone patterns don't support such elaborate elaboration, because there's not much to work with. When your starting point is "Got milk?", for example, or "Mmm, beer!", your options are limited. But where the base form has more substance, or at least more words, the only real constraint seems to be that the starting point should remain somehow recognizable. And that requirement is conceptually circular, since if the source were no longer recognizable, we wouldn't recognize the result as a snowclone. I don't know of any good taxonomic description of the sorts of transformations that are involved here, much less a theory able to predict the distribution of observed variants, but I suspect it would be productive to look into this further. (The issues involve seem to overlap with the problems of construction grammar, whatever exactly we take those to be...)

And now for something completely different: the "Eskimo snow words" examples.

In that case, there's no proverb or quotation or anything like it to start with, and in fact there isn't even a crisply defined pattern of structured words. The template is really an abstract rhetorical structure, which has become such a commonplace trope that we can get a sort of imitation snowclone by cashing the rhetoric out in a particular way: "If Eskimos have N words for snow..."

As Geoff Pullum pointed out in the post that started all this off, the Economist's writer used the conditional

If Eskimos have dozens of words for snow, Germans have as many for bureaucracy.

to mean something like

Eskimos have dozens of words for snow, as everyone knows. Well, Germans have just as many for bureaucracy.

This was intended as an entertaining way to introduce what Geoff calls "the stereotypical bureaucratic ubiquity of the Teutonic world". The basic method is to use the analogical form Eskimos:snow::X:Y to communicate something like "Y is really important to X, so much so that you should think of it as you think of the relation between Eskimos and snow". For extra flavor, add the notion that X is therefore motivated to make many fine conceptual and/or terminological distinctions among types of Y.

It would have been boring -- and perhaps offensive -- to write something like "Germans are notoriously concerned with bureaucracy". So instead, the writer decided to call on the Eskimos:snow::X:Y analogy, instatiated as Eskimos:snow::Germans:bureaucracy, and then to phrase this as "If Eskimos..."

There are lots of other ways to do the same sort of thing. A bit of web searching turns up:

Just as Eskimos have many words for "snow," you probably need many concepts to sort out your thinking on "love."
Just as Eskimos have lots of words for snow, the Welsh language has more than a few words for green.
...just as Eskimos have a rich and precise vocabulary for discussing a wide range of complex snow conditions, so, too, teachers need to develop a language for describing and talking about teaching strategies.
I learned that just as Eskimos have several words for types of snow, sailors have at least six for blends of rum
Just as Eskimos distinguish between a dozen different sorts of snow, so too do the Swiss appreciate a wide range of Gruyeres.
Just as Eskimos are said to have a hundred words to describe types of snow, we’d figured out at least ten for rain.
Just as Eskimos have more than twenty words to describe snow, Americans have as many words to describe comfort.

You know how Eskimos have several different words for snow, and the British several ways of saying how it can rain ...? Well, the Psalmist had a couple of words in Hebrew that we translate blessed.
...you know how Eskimos have about 30 words for snow? We're just as serious about clams.
You know how Eskimos have like 100 words or so for snow? Why don't people in Washington have 100 different words to rain?
You know how Eskimos have, like, a hundred different words for snow? ''The Perfect Storm'' has a hundred different shades of wet.
Do you know how Eskimos have a hundred-odd words for snow because its all around them?.... Well, Japanese has many words for "pervert" too... Guess why. ...
Know how Eskimos have 173 words for the word "snow"? Well, Dixon's has more types of chili than it has chairs at lunchtime.
You know how eskimos have three hundred words for snow? jews have over three hundred words to use to complain about how humid it is.
You know how Eskimos have a zillion words for snow? The French have a zillion words for doughs.

They say that Eskimos have many words for snow. Cats have many words for sleep:
They say that Eskimos have close to 17 words to describe snow. The English vocabulary does not possess adequate adjectives to describe my gratitude to ALL of you
They say that Eskimos have something like forty words for snow. From what we saw in our two days in San Sebastian, the Spaniards in the Basque region must have at least that many for rain.
They say that Eskimos have 52 different words to describe snow... The music of Sibelius and Riisager has at least that many ways to describe the frigid landscape of Finland and Greenland in musical terms.
You know how they say that Eskimos have 100 different words for ice. A student of the Psalms I was reading claimed that he had found 94 different ways to say enemy in the Psalms.

They say that Eskimos have a hundred words for "snow". That may or may not be true, but Runners have far more than a hundred words for guns.
They say that Eskimos have over 200 words for snow. I can’t help but wonder if Nebraskans have just as many words for “flat.”
They say that eskimos have thousands of names for snow, the British have millions that can be used as insults.

...among many, many others. And you can see a very different use of the same analogical schema in the title and opening paragraph of Joseph's Reagle's W3C Note "Eskimo Snow and Scottish Rain: Legal Considerations of Schema Design".

Eskimos have many words for snow; Scots have numerous words related to rain. This concept has achieved near urban myth status -- though it continues to be contentious amongst linguists [Who40]. The idea is compelling because it speaks to our belief that the mechanism of speech itself is a reflection or [sic] our world and what we wish to say. Within this paper I examine the mechanisms by which our computer agents will express and understand what we wish to say in order to form online agreements.

Here the message seems to be that internet applications need many words for... well, see if you can figure it out for yourself, this post is long enough already.

[Note that the title of this post refers to Richard Montague's famous 1973 paper "The proper treatment of quantification in ordinary English". Needless to say, there is no real implication that other discussions of this topic are "improper": the title is a joke, not a (nominalized) statement. For that matter, there's no reason to refer to "ordinary English" except to echo Montague's title, since all discussions of this topic have dealt with phrasal and rhetorical patterns in ordinary English. Needless to say, there are a fair number of prior PTQ snowclones in the literature, such as:

On the proper treatment of opacity in certain verbs
The proper treatment of optimality in computational phonology
The proper treatment of symbols in a connectionist architecture
On the Proper Treatment of Context in NL
Discourse structure and the proper treatment of interruptions.
The Proper Treatment of Quantifiers in Ordinary Logic


[Update: some thoughts from Russell Lee-Goldman:

It certainly seems like categorizing and describing (at least certain) types of snowclones is certainly very similar to the sorts of things that construction grammar is often used to describe. Of course CxG per se doesn't, AFAIK, have anything to say about what exactly makes two expressions seem close enough that one can evoke the memory of the other. For instance, consider this popular tagline:

sometimes you feel like a nut, sometimes you don't

For me, anyway, any time someone says "sometimes you feel," the whole phrase is "activated." On one level, it's common enough to juxtapose [sometimes [positive proposition]] [sometimes [negative proposition]], so why not with "feel." But on another level, you can draw on the stored memory of a particular famous instance of that construction in order to build from it.

Similarly, you could say that there's some sort of

[[cardinal number] X in the hand is worth [cardinal number+n] Y in the bush]

construction. Or more generally

[[quantity] X ['in control/possession'] BETTER-THAN [quantity] Y ['only potentially available']]

But this is getting very abstract, and is more like a guiding principle in life rather than a linguistic construct. That's why something like:

--I'd rather have a single book on my bookshelf than several available in the library.

is barely recognizable as an extension of the "bird-hand-bush" idiom, because it's only vaguely syntactically similar, and expresses an idea that may well exist independently from the initial idiom. On the other hand,

--A bird in the hand is always safer than one overhead.

Does not express the semantics OR syntax of that very general construction above, but maintains the "bird in the hand" part, which clearly marks it as an attempt to evoke the idiom.

This sort of exercise reminds me of what Paul Kay called "patterns of coining," though his argument was in particular against the idea that some constructions (like "resultative" or "caused motion") were productive, instead claiming that there are some non-productive patterns that speakers nonetheless use to coin new instances of familiar patterns. Snowclones are slightly different since the test of productivity is not grammaticality (in the narrow sense), but rather effectiveness in evoking some familiar concept.


Certainly the evocation of familiarity plays a special role in these allusions to well-known phrases. But allusion and construction share the tension between analogy-as-cause and analogy-as-result. On one view, the process of analogical pattern generalization is the psychological source of new instances (of snowclones or of constructions), making new phrases from old ones. On another view, the perceived similarities among phrases (and their associative effects) are an evoked response to patterns created by some non-analogical process, in which particular phrasal models (with or without a few variables added) play no role. Of course, you could easily believe that both sorts of process are involved, or that true understanding reveals this to be a false dichotomy, as in connectionism's rejection of the distinction between remembering and inventing.]

Posted by Mark Liberman at February 4, 2006 10:41 PM