Compromising Positions and Polarity Items

0. Introduction This paper looks at two theories of negative polarity licensing that utilize the notion of strength or strengthening to explain NPI distribution. I consider Kadmon and Landman (1993) and Israel (1996, 2001, 2005), both of whom define strength in terms of downward inferencing relations. I claim that various compromising adverbs such as pretty much, virtually, practically, etc., disrupt this pragmatic inferencing and so raise problems for theories depending on this type of motion. Taking pretty much as a test case, I suggest that the compromising data can be accounted for by the semantic notion of (non)veridicality, as demonstrated in Giannakidou (1998, 1999, 2002, etc.). In the rest of the paper then, I will illustrate the problems the compromisers raise first for Kadmon and Landman (K&L) and then Israel. The last section shows how the (non)veridicality theory of Giannakidou successfully manages the data. First, however, it is necessary to make some general remarks in regard to the meaning and use of pretty much. As a compromiser or otherwise, pretty much has rarely been mentioned in linguistic literature. However, the approximator almost, which has a similar (though not identical) semantics and distribution has received much attention over the years (Sadock 1981, Rapp and von Stechow 1999, Horn 2002, Morzycki 2002, etc.). In the next section, I will provide a brief introduction to the semantics and syntax of pretty much as they compare with almost.