Priming NPI Acceptability Judgments and the Bagel Problem

Abstract

Wereport on a series of priming experiments on the acceptability of two types of Negative Polarity Items (NPIs) in two languages, namely, any in English and libo-NPIs in Russian. Our results show that (i) acceptability judgments of NPIs can be primed both in English and in Russian, and (ii) only unacceptablity of the same type, i.e., unacceptability caused by an unlincensed NPI, triggers priming effects on the acceptability of NPIs. The results also indicate that (iii) priming effects for English any and Russian libo-NPIs are counter directed: For any, an unlicensed instance has a positive priming effect, while for libo-NPIs an unlicensed instance has a negative priming effect. We attribute this contrast to the so called ‘bagel’ distribution of libo-NPIs in Russian: Although licensed in all other contexts where weak NPIs are commonly licensed, libo-NPIs are hardly ever licensed by negation. We claim that this makes the crucial difference for priming effects triggered by unlicensed NPIs in English and Russian. That is, unacceptable primes trigger priming effects by evoking possible ways to ‘fix’ them, and different NPIs evoke different such ways, depending on their typical contexts of use, whereby exhibiting qualitatively different priming effects.