The project aims at reconciling the common assumption of concentric Negative Polarity Item (NPI) licensing contexts (antimorphic ⊂ anti-additive ⊂ downward-entailing ⊂ …) with apparently unlicensed exceptions and with the much more fine-grained distributional profiles observed in corpus-based research. Whereas classical theories mainly concern licensing by at-issue content, we investigate NPI licensing by non-at-issue content. This leads to a better understanding of the relation between explicit (morpho-syntactically marked) negation and implicit negation (by conventionalized association). The hypothesis is that strong NPIs require a strong (anti-additive) licenser, but the licenser can be at-issue or non-at-issue. Weak NPIs accept a weak (e.g., downward-entailing) licenser, though their licensing must follow from the computation of the at-issue content.

The project approaches its main hypothesis with data from primarily three languages and with different complementary research methods. Exploiting the insights in the distribution and character of a large set of German and Romanian NPIs documented in previous work, we plan to investigate quantitative and qualitative properties of NPIs in English, German and Romanian. With semi-automatic extraction methods, the initial data set of NPIs and their theoretical classification and quantitative documentation will be recast into a considerably extended searchable database of NPIs, their licensing environments and their corpus occurrence patterns in the three project languages. To complement the limitations of corpus data, systematically collected acceptability judgments by native speakers will be used to critically evaluate apparent NPI licensing patterns and to investigate the availability of relevant inferences. Where relevant, we will add data from French to our set of languages Further corpus studies will target quantitative occurrence distributions of representative items in systematically selected licensing contexts. The corpus studies also aim to shed light on differences in distributional profiles between various registers and corpus types.

Based on solid empirical foundations the grammar-theoretic goal of the project is the formulation of a theory of NPI licensing with enriched semantic representations, i.e. encompassing at-issue and non-at-issue content, in a constraint-based grammar framework. This NPI licensing theory will be built on the one hand on occurrence patterns of individual NPIs with particular attention to occurrence patterns in overtly non-negative contexts and the role individual NPIs play in these patterns (especially in the case of corresponding NPIs in different languages). On the other hand, the analysis of licensing contexts and different types of negativity exhibited by these contexts will be central. Of particular interest are constructions such as before clauses, which are also known as environments of expletive negation.

Based on the results of the first funding period, the long-term goals for future funding periods is the development of a distributional theory of the notion of possible NPIs, i.e. a predictive theory of which kinds of lexical semantic and pragmatic properties facilitate the restriction to certain kinds of negative environments of single- and multi-word expressions. We plan to conduct experimental tests of this theory, based on distributional hypotheses about classes of NPIs in contexts without overt licensers, which presupposes a detailed licensing theory of a sufficiently large number of such items. In addition, we want to investigate the question of which aspects of NPI-licensing are represented by large language models, especially with respect to licensing with non-overt negation. The sensitivity of such models to certain distributional restrictions could also become interesting in the context of developing experimental items.