In Negative Concord (NC) sentences, single negative meanings are expressed by two or more negative words. English speakers that use NC also employ Double Negation (DN), where two negatives yield a logical affirmative. The same speakers also use Negative Polarity Item (NPI) constructions, where words like anything and anybody depend on a preceding negation (e.g. 'I didn't eat anything' vs. 'I ate anything'). This dissertation accounts for the distributions of NC, NPI, and DN constructions in English.
I apply the theory of NPIs in Postal (2005) and Collins and Postal (2014) to NC and DN. These authors argue that some NPIs have the form [NEG SOME X], with a single NEG, while others have two: [[NEG [NEG SOME]] X]. I propose that negative constituents have a structure identical to Collins and Postal's (2014) unary NEG NPIs. Like NPI constructions, NC with a negative marker (-n't/not) and a negative object involves syntactic NEG raising from the negative constituent. I further propose that the locus of variation between NC and NPI constructions lies at the level of morphophonological spell out. NPI constructions involve deletion of lower occurrences of a single NEG, but NC does not. Using data from the Audio-Aligned and Parsed Corpus of Appalachian English (Tortora et al., In Progress), I show that all predictions concerning the distribution of NC, DN, and NPI constructions across clause boundaries are borne out.
Two types of NC with negative subjects are also analyzed. NC declaratives like 'didn't nobody eat' and 'nobody didn't eat' are derived via NEG raising from a negative constituent. In these cases, NEG raising is followed by remnant raising of the negative constituent. To explain restrictions on subject type in inverted structures ('didn't nobody eat'), I defend a condition stating that the subject must always be negative, despite the fact that it is not always morphologically negative. Differences in usage and interpretation of negative object and negative subject constructions are derived by appeal to a remnant raising condition. The results of a gradient acceptability study support the hypothesized grammatical distinction between Subject and Object NC.