Jason Merchant, "Modifying the colored lambda-calculus to model negative concord and the Afrikaans clause-final negator"
Colloquium | March 19 | 3:10-5 p.m. | 370 Dwinelle Hall
Jason Merchant, University of Chicago
Standard Afrikaans final "nie" seems to be a kind of polarity item that appears at the end of certain phrases, most prominently at the end of clauses containing sentential negation. There is a broad syntactic consensus that this final negative particle is a sentence-final element. Based on new data from coordination and scope, I argue against this consensus, and show that final "nie" in its usual clausal use is a VP-final element, not a clause-final one. This distribution is most straightforwardly accounted for if final "nie" is a special clitic (not an affix) whose presence in the clausal structure is required by a licensing element, but whose set of licensers is puzzlingly diverse in a pattern reminiscent of the thicket of elements sensitive to varieties of negativity found in the Hornian/Giannakidean pantheon. I follow Biberauer and Zeijlstra 2012 in analyzing final "nie" as a non-negative element; unlike them, however, I posit that "nie"s distribution is due to its semantics: "nie" has the semantic effect of changing the predicate it occurs with into something that will be composable only if a negative element takes it as an argument (nie produces a negative isotope of its argument). I implement this analysis in a modified variant of the colored λ-calculus (Gardent et al 1998). Such an analysis, I show, is superior to syntactic Agree-based ones, and only it explains why "nie" appears at all. Standard Afrikaans is a double negative language, and my analysis applies to multiple negative words if the type-fitting system is lazy (Partee and Rooth 1983, Winter 2001). This analysis can also be extended to colloquial Afrikaans, which is a kind of negative concord language (Giannakidou 2000, 2006); in such uses, the negative quantifiers shift to their non-negative indefinite counterparts.
(Time permitting, I discuss implications of "nie"s haplological behavior for our understanding of locality of allomorphic conditioning, arguing that a linear model is superior to ones requiring phase- or word-internality.)