February 21, 2004

Who is to be master?

Following up on Geoff Pullum's post about "too complex to avoid judgment", I checked Google for patterns such as "no * is too * to avoid" and "no * is too * to ignore". If I've analyzed the results correctly, it appears that the "incorrect" interpretations of phrases instantiating these templates far outnumber the "correct" interpretations. I've put "correct" and "incorrect" in scare quotes because if we this were a matter of word meaning, we good descriptive linguists would say that the speech community had simply changed its mind about what the word means, thereby changing the word meaning and making the "incorrect" usage ipso facto "correct".

It's not clear to me what the right analysis is here. One story says that this is a matter of logic, which can't be changed by voting, and that the predominance of wrong usages is to be explained by psychological arguments. Another story says that the language is changing, whether syntactically (starting to revert to negative concord, which the vernacular has never abandoned?) or semantically (does no in the determiner of a subject NP have some unexplored interpretive possibilities?) or by developing a syntactic/semantic "construction" with a non-compositional meaning. As a simple phonetician, I'll leave these questions to the professionals, but the first sort of explanation seems more plausible to me, though the second would be more fun.

Here are some details.The pattern "no * is too * to avoid" gets 54 ghits. Most but not all the examples seem to count the negatives wrong:

[N]o executive is too prominant (sic) to avoid the long arm of the law
No one is too young to avoid being tempted
No business is too small to avoid or ignore protecting itself from another business using its name, product, service or invention.
No sacrifice is too great to avoid total destruction in Gehenna.

The seven "sacrifice" cases all seem to be variants of the same religious document, which asserts that "No sacrifice is too great to avoid total destruction in Gehenna." I read this as a "correct" calculation of the meaning: "to avoid total destruction, no sacrifice is too great." Note that in this case, unlike in most of the "wrong" ones, the sacrifice is neither doing the avoiding nor being avoided. There are a few other "correct" usages, but the "wrong" ones far outnumber them.

The pattern "no * is too * to ignore" has 50 ghits, mostly details, errors, issues, advantages etc. that are too small to ignore:

Five Star Events believes no detail is too small to ignore
Kelly... said that in the playoffs no advantage is too small to ignore
No error is too small to ignore - I want to make the second edition perfect!

One writer seems to have gotten his wires more seriously crossed, misplacing "seem" as well as losing track of his negatives:

Everything I seem to have done I have done well in, and no detail is too small to ignore.

Eliminating a few things that don't belong in the output of Michael Leuchtenburg's snowclone_google.pl program, we get as "wrong" examples:

no detail is too small to ignore: 7
no error is too small to ignore: 3
no conflict is too distant to ignore: 2
no issue is too small to ignore: 2
no advantage is too small to ignore: 2
no point is too small to ignore: 1
no profit is too small to ignore: 1
no contribution is too small to ignore: 1
no skill is too small to ignore: 1
no mission is too small to ignore: 1
no amount is too small to ignore: 1
no detail is too minor to ignore: 1

I've eliminated a differently-parsed example that is not relevant: "the problem of no-shows is too costly to ignore". There are only two cases where the literal meaning is (I think!) the intended meaning. However, these examples seem to be sarcastic, which makes me worry about whether I've analyzed them correctly:

No blemish is too hard to hide,. No problem is too big to ignore,. As long as you don’t hear complaints,. Why should you care?
Of course, the muted outrage and lack of debate over these lies and prevarications merely adds to the sense that no lie is too big to ignore.

It occurs to me that the mistakes (if that's what they are) may be caused by a sort of constructional resonance. What the writers really want to say seems to be something like "no X is so Y that we (normatively or habitually) Z it", or "no X is so Y that it (normatively or habitually) Zs". However, they can't quite figure out how to frame that in an idiomatic way with the pieces that come to hand, and as their mental generation process is fiddling with the fragments, everything kind of slips into the familiar and similar frame of phrases like "the box is too heavy to lift" or "Kim is too drunk to drive", in the form "no X is too Y to Z". In this solution, the interaction of negation, scalar direction and infinitival control doesn't work out right, but it's hard to calculate these things, and so the result passes muster.

Another possible source of confusion is the difference between asserting that someone or something is not past a certain limit, and asserting that no such limit exists. If I say "This flaw is not small enough to ignore",or "This man is not important enough to avoid prosecution", it's natural to understand this by reference to a threshold of size or importance that the flaw and the executive don't reach. But the intent of statements like "No flaw is too small to ignore" and "No man is too important to avoid prosecution" is precisely that no such threshold exists, and this may be why the obvious alternative phrasings "No flaw is small enough to ignore" and "No man is powerful enough to avoid prosecution" are not chosen.

[Update: there are many similar patterns that reliably produce examples that mean the opposite of what their writers clearly intended:

However, despite all this and her duties as Party cell secretary, she is never too busy to ignore the needs of children.
However, one should not be too complacent to ignore the uncertainties that are prevailing in the domain of the global economic waters nowadays.
Of course, I hope my clients will not be too stubborn to ignore this advice.
No risk is too great to prevent the necessary job from getting done.
A certain situation may appear dangerous to most people, but to a journalist, no situation is too dramatic to prevent their story from happening.
The store is known for its personal and friendly service; no request is too wacky to refuse.

A reader suggests that

I wonder if what is going on here is that "so" has merged with "too"? All of these examples become good if "too" is replaced with "so", e.g.:

"No head injury is so trivial as to be ignored."

"No executive is so prominent as to avoid the long arm of the law."

My impression is that this use of "so" has essentially dropped out of common use and that "too" has taken over its role.

I don't think this can be true in general. "Kim is so upset as to scream" is not very idiomatic, but "Kim is too upset to scream" has not taken over its meaning. If there's been a change of this sort, it must be limited to certain phrasal templates, which I guess might count as what some people these days call " constructions". However, it's suspicious if the "constructions" in this case turn out to be exactly those sentences in which the meaning is hard to compute for independent reasons. That's why I suggested above that people start out wanting to say something like "no X is so Y as to Z" but wind up using "no X is too Y to Z" -- not because of a general meaning change, but because the "right" outcome is problematic and the "wrong" one is both similar enough to activated instead, and also too complex semantically for its opposite meaning to be obvious.]

Posted by Mark Liberman at February 21, 2004 09:05 AM