When variables align: A Bayesian multinomial mixed-effects model of English permissive constructions

This paper is a quantitative multifactorial study of the near-synonymous constructions let+V, allow+to V and permit+to V based on the British National Corpus. The study investigates the differences between these constructions with the help of 23 formal, semantic, social and collostructional variables. A Bayesian multinomial mixed-effects model reveals a remarkable alignment of the variables that represent different dimensions of variation, namely, the linguistic distance between the predicates, the conceptual distance between the events they represent, the distance between the speaker and the Permitter and Permittee on the animacy/entrenchment/empathy hierarchy, the social and communicative distance between the interlocutors, as well as the strength of collostructional attraction between the constructions and second verb slot fillers. The paper offers several possible explanations for this alignment from a cognitive, functional and historical perspective.