Hoezeer is betekenis van tel? Een nieuwe kijk op syntactische alternanties

This project examines if and how the way people choose between different ways of saying the same thing depends on the meaning of the words in the utterance. Take variation between ditransitive dative constructions (as in "Tom sent the president a letter") and prepositional dative constructions (as in "Tom sent a letter to the president"): variationist linguists are fairly good at modeling such variation based typically on formal predictors (length of constituents, complexity of constituents, and so on). However, analysts have been reluctant to probe the explanatory power of lexical meaning: does it matter for variant selection if, say, the recipient is a government official (e.g. "president") or a family friend? Is, for example, the prepositional dative more likely if the theme is countable (e.g. "letter") as opposed to non-countable (e.g. "furniture")? Analysts have been hesitant to tackle meaning because it is very labour-intensive to annotate data for semantic factors. We resolve this issue by utilizing computationally advanced distributional semantics methods to put paid to the need for manual semantic annotation, thus breaking new ground methodologically. On this basis, the project will add to our knowledge about language by checking systematically the extent to which semantic properties of the material in the slots of the constructions under analysis predict variant choice - if they do then a foundational assumption in variationist linguistics needs to be rethought.