Studying syntactic priming in corpora

This chapter addresses syntactic priming (of the dative alternation) using corpus data from the ICE-GB corpus. Nearly 3,000 consecutive prime-target pairs were coded for their constructional choices as well as several other variables. The data are then analyzed on different levels of granularity: (i) cross-tabulation, (ii) with a binary logistic regression (including only fixed effects), and (iii) a generalized linear mixed-effects model (GLMEM) (including fixed and random effects). The GLMEM reduces the number of significant predictors most, but nevertheless yields the highest classification accuracy. Since contemporary cognitive linguistics assumes an item-based perspective on language acquisition, processing, and change, I argue that the GLMEM approach should be the method of choice for empirical cognitive linguistics.