April 28, 2006

Questions and NPIs

Brian Weatherson has wondered why questions are NPI-licensing. The question is difficult to answer in terms of a lot of customary treatments of NPI-licensing, since they tend to deal with upward and downward entailment*, and it's not obvious what entailment is when you're talking about questions.

I just typed the following sentence:

What requires explanation is why anyone at all has strong Gettier intuitions.

In this sentence it seems to me that the NPI 'at all' (and 'anyone' as an existential quantifier, which is also an NPI) is licensed by the embedded question, "why anyone at all has strong Gettier intuitions." If there wasn't an embedded question, NPIs wouldn't be licensed:

What requires explanation is Rumsfeld's awful mistakes.
??What requires explanation is any of Rumsfeld's awful mistakes.

And this is an indicative sentence overall, so we may be able to make more sense of upward and downward entailment. Though I won't try to do this now. (As usually for my meditations on linguistics, it's very likely that this has been covered before.)

*Roughly, .... is a downward-entailing environment if ... F ... entails ... G ... whenever F entails G. UPDATE: Dammit, I got this backwards, as usual. ... is a downward-entailing environment if ...F... entails ...G... whenever G entails F. So negation is downward-entailing because "I'm not a linguist" entails "I'm not a semanticist" while "x is a semanticist" entails "x is a linguist."

Posted by Matt Weiner at April 28, 2006 01:09 PM


Have you read Daniel Rothschild's paper on NPIs? It's on his website:


Daniel's eventual target is the Russellian analysis of definite descriptions, but there's some relevant and interesting stuff along the way.

Posted by: Aidan at April 28, 2006 03:22 PM

The ignoramus's question: what's the philosophical interest/ payoff of answering this question? ... I just don't know the motivations.

Posted by: Greg Frost-Arnold at April 29, 2006 10:27 AM

Your specimen "What requires explanation is Rumsfeld's awful mistakes" strikes me as incorrect. The reason is entirely off-point, however. Oughtn't it be "What require explanation are Rumsfeld's awful mistakes"?

And one can say, "Explanation is required for any of Rumsfeld's mistakes."

Posted by: at April 29, 2006 11:28 AM

Oughtn't it be "What require explanation are Rumsfeld's awful mistakes"?

I had it as "What requires explanation are..." first, but it sounded stilted to me. And if "What require explanation" isn't wrong, I don't want to be right. Woe is I!

You're right about "Any of Rumsfeld's mistakes," though. "Any" is weird, so I should probably try and come up with an example that involves a different NPI... well, "at all" should do. "Explanation is required for any at all of Rumsfeld's mistakes" sounds terrible to me. I think "Explanation is required for any of Rumsfeld's mistakes" is a free choice "any," amounting to "each," whereas "at all" can only be used with NPI "any" which amounts to an existential quantifier (as in my original sentence).

Though I think I have an example of an embedded question that doesn't license NPIs, which I'll post later.

Greg: One answer is what Brian Weatherson said about the Rothschild paper Aidan mentioned:

NPIs are basically gifts from God to the semanticist - they provide non-trivial non-obvious tests of semantic hypotheses that probably weren't what theories were originally designed to capture, but which can't easily be explained away. There's not many of those in semantics. Compare the long-running disputes over what Russellians can say about "The table is covered in books", where there are (fittingly) too many rather than too few "explanations away". NPI tests are (relatively speaking) good clean tests for whether a semantic theory works, and if Russellian theories of definite descriptions don't, then those theories are wrong.

Rothschild argues that the behavior of NPIs in definite descriptions proves that definite descriptions can't be quantifiers in disguise. And I think it's conceivable that if we answer the question "Why do questions license NPIs?" we might gain insight into the semantics of questions.

But my motivation is mostly, it's a cool linguistic puzzle and I like thinking of examples....

Posted by: Matt Weiner at April 29, 2006 12:49 PM