Compromising Positions and Polarity Items

This paper looks at two theories of polarity licensing that utilize a downward entailment relation (DE) in their explanations of the polarity item distribution question. I illustrate that various compromising adverbs such as pretty much, virtually, practically, basically, etc., disrupt downward entailment yet do not interfere with polarity licensing, which raises problems for theories depending on this type of motion. Importantly, the DE motion of the two theories I consider is in the form of pragmatic inferencing, rather than the logical entailment of Ladusaw (1980). However, I will suggest that Ladusaw-type theories are vulnerable as well, though this is not the immediate focus of this paper. Consider some representative examples of the interrupting effect compromisers have for Kadmon and Landman (1-2), and Israel (3-4): According to KL, any must widen semantically and strengthen pragmatically to be licensed. Or, the widened (1a) must entail (1b) for any to be licensed. (1) a. I don't have any potatoes. => b. I don't have potatoes. The relation holds in (1a-b), but it fails in (2a-b), while any is still perfectly licensed. (2) a. I pretty much don't have any potatoes.-/-> b. I don't have potatoes. Now consider an Israel-type example, where an emphatic NPI must be more informative than the assumed scalar norm in order to be licensed. Again, this translates into the emphatic (3a) entailing the scalar norm in (3b). (3) a. John didn't sleep a wink. => b. John didn't sleep. Israel's relation holds in (3a-b), but it fails in the compromised (4a-b), and the NPI is still perfectly licensed: (4) a. John pretty much didn't sleep a wink.-/-> b. John didn't sleep. I claim that the compromising data cannot be accounted for by the pragmatic variety of DE utilized by KL and Israel. However, it does seem to be explainable under the (non)veridicality theory of licensing, as demonstrated in Giannakidou (2001, 2002 and various other publications). To defend her move from DE to (non)veridicality, Giannakidou illustrates that polarity licensing occurs in numerous non-DE environments. In this talk, taking pretty much as a test case, I suggest that even well-behaved DE environments can find themselves compromised and so provide a further reason for looking away from DE as an important part of the solution to the polarity licensing question.