Verb class and instrument PPs: A mixed model analysis

Abstract

This paper investigates whether a lexical variance caused by verbs influences the choice and frequency of instrument with-PPs. The corpus data shows that observations grouped by verb class demonstrate systematically different behaviors and this individual variance of verb classes can be captured by means of a random effect of a mixed-effects model. Building up on Choi's (2012) research that identifies the syntactic, semantic, and morphological factors that influence the presence of instrument with-PPs as fixed effects, the current study classifies the instrument-taking verbs into verb classes, based on Levin's (1993) study, and builds a mixed-effects model taking verb class as a random variable. The new statistical technique of hierarchical, multi-level, mixed-effects modeling (Baayen 2008; Bresnan et al. 2007; Gelman and Hill 2007; Johnson 2008; Kuperman 2009; Pinheiro and Bates 2000) can process across-word fixed effects and by-word random effects together. By taking into consideration the subtle syntactic and semantic characteristics of verbs, this new modeling analysis provides a way to incorporatenative speakers' lexical knowledge into grammar.