A Multivariate Analysis of Construction Alternation in English Future Tense

Based on richly annotated data, we fitted the mixed-effects logistic regression models to examine the probabilistic constraints on alternation between English future tense (EFT) constructions and reveal the core difference in use between be going to and will. We found that the animacy of subjects, sentence patterns, sentence types, sentence length, the dynamicity of verbs, the specificity of verbs, genres, and sex significantly influenced variation in EFT constructions. In descending order of overall importance, the variables were ranked as follows: the specificity of verbs, sentence patterns, the dynamicity of verbs, sentence types, genres, sentence length, the animacy of subjects, and sex. We identified the core difference in use between the two variants: Be going to is favoured in subordinate clauses of complex sentences to express specific, dynamic actions or events, whereas will is preferred in main clauses or simple sentences to encode general, static actions or events.