Distributionally restricted items

Abstract

This paper explores the learnability of English indefinite any, Dutch modal verb hoeven, and Mandarin Chinese (WH-)indefinite/pronoun shenme. These three expressions, belonging to different syntactic categories in different languages, have been referred to as Negative Polarity Items (NPIs) in the literature, as they are all restricted to contexts that in some sense count as negative although there are differences in the types of semantic environment that may license them. By investigating the distribution of these three expressions in both child and child-directed speech recorded in the CHILDES database (MacWhinney 2009), this paper argues that children in their acquisition of these NPIs employ the same conservative widening learning strategy (Berwick and Weinberg 1986; Manzini and Wexler 1987), which prevents them from overgeneration. A two-stage acquisition process is detected for each of the three NPIs. However, distinct learning pathways are found, which we take as evidence indicating different underlying analyses of these expressions at different stages in child language. Taking into consideration the input characteristics, the distributional patterns of the three expressions in adult grammar, and the children’s lexical development, we hypothesize what the analyses of any, hoeven, and shenme at different acquisition stages look like. This provides us with a different view of the nature, or the reason underlying the restricted distribution of these expressions in adult language.

Restricted and Unrestricted Syntactic Distribution of ar in Bangla

“…Until then, calling it an NPI seems justifiable. Seeing it from an NPI perspective ar shows significant similarities with previous works by ) & Lin et al (2021 especially. I have articulated above that ar is a crucial linguistic element among many other identified functional elements or noted discourse particles in (Bayer et al 2014).…”

“…However, distributing ar in this way, namely restrictedly and unrestrictedly follows from the framework originally proposed in Crnič (2019; part 1 and part 2) followed by Lin et al (2021) among many others. Interested readers should refer to their original works cited in the reference section below for more details.…”