Arnold Zwicky's Blog

The warning on the site — a notice of possible “adult” content — goes:

(1) This blog is not suitable for viewing by anyone.

Ouch. The reading I got immediately was that anyone was a “negative polarity” any-expression, so that (1) is (truth-functionally) equivalent to

(2.1) This blog is suitable for viewing by no one.

or

(2.2) There is no one for whom this blog is suitable for viewing.

or possibly

(2.3) Pick someone, anyone; this blog is not  suitable for viewing by that person.

But that would be a silly thing for a blogger to say; it warns everyone away.

So the intended interpretation had anyone as a universal any-expression, with (1) truth-functionally equivalent to

(3.1) This blog is not suitable for viewing by everyone.

or possibly

(3.2) This blog is not suitable for viewing by just anyone.

(1) is  in fact potentially ambiguous between the ‘everyone is excluded’ reading and the ‘some people are excluded’ reading, but, unfortunately (and for reasons I don’t fully understand) many readers are likely to get the wrong — not the intended — reading.

We’ve been in this neighborhood before.

[A number of readers have patiently tried to explain to me that the problem with things like (1) is only a matter of written English, that in spoken English the intended interpretation can be brought out by an accent on anyone:

(1′) This blog is not suitable for viewing by ANYONE.

But no. (1′) can be understood as an emphatic version of the ‘everyone is excluded’ reading: ‘EVERYONE (no exceptions) is excluded’.

(There’s a place for such emphatic denials. Sentences that are universal in form are regularly used approximatively, to convey ‘pretty much every X; in general, X’. So prosody can be used to close off this escape clause.)]

There’s a mind-bogglingly large literature on English any-expressions and their (rough) parallels in other languages. How many types are there? In particular, in English, are neg-polarity any-expressions, universal any-expressions, and “free choice” any-expressions (as in (2.3)) instances of one, two, or three phenomena? Is this a question about logical parsimony or psychological reality or something else?

Fortunately, I don’t have to decide these questions here. But there is still the question of ambiguity and favored readings, and this is where Language Log (and other blogs) have been on the case, in the matter of Kilpatrick’s Rule (named after James J. Kilpatrick), the claim that in (externally) negated sentences with an internal universally quantified NP, as in (3.1), the only possible reading is the one in which the universal has wide scope (over the negation). The claim is that something like (3.1) must be understood as ‘for everyone, this blog is not suitable’, which brings us back to the readings in (2).

The fact is that — see discussion by Neal Whitman, Jan Freeman, and three postings by Mark Liberman on Language Log — that in negated expressions with explicit universals (as in “Not applicable in all states”), Kilpatrick’s Rule hardly ever applies, as you’d expect (Mark Liberman notes) from the order of the elements in such sentences: (3.1) almost always conveys ‘it is not the case that this blog is suitable for everyone’ (so it’s potentially ambiguous, but hardly ever effectively ambiguous).

So I’d recommend that the original blogger replace (1) by (3.1), or possibly (3.2). The blogger wants to warn some people off, not everyone, and (3.1) does this, even though there’s a potential ambiguity (but probably not an effective ambiguity) there.

This is not to deny that there are real (sometimes) effective scope ambiguities out there (and Mark Liberman acknowledges this). A pedestrian example:

I didn’t see many foxes.

This can be read with many scoping over negation (‘there were many foxes I didn’t see’) or with negation scoping over many (‘I saw few foxes’). Which reading you get depends a lot on context. (And, again, there’s a big literature.)

The larger point is that this is not an issue of grammatical well-formedness (as Kilpatrick proposed), but of how expressions are interpreted in context.

This entry was posted on March 7, 2009 at 6:18 pm and is filed under Semantics, Usage. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.