Corpus-based extraction of quantitative structural and systemic features of constructional paradigms as they evolve over time

Diachronic Construction Grammar currently faces two challenges that tie to the systemic nature of language organization. On the one hand, schematic constructions characterized by a filler slot (such as the verb in the [be about to {V}] construction) have been associated with a paradoxical observation: an increase in the number of types recruited in the slot of the construction can co-occur with a general decline in frequency of use. On the other hand, constructional changes happen on multiple levels of the constructional network: lower-level constructional schemas (such as the partitive [a piece of]) get entrenched and strengthened as instances of higher-level schemas (e.g. the family of quantifier constructions [(a) {N} of]: a piece of, lots of, etc.), and therefore benefit from a mutual relationship while they compete over the same types. This project aims to tackle these two questions with innovative methods from Complex Systems. To elucidate the first paradox, I rely on the observation that the types of a filler slot obey a local Zipf's law. The hypothesis here is that this Zipf's law gets entrenched as a characteristic feature of the construction; when too many types are recruited, the Zipf's law cannot sustain them all and collapse, leading to the decline of the construction. To solve the second paradox, I will invoke the notion of nestedness, borrowed from Ecology: if a paradigmatic system achieves this property, then competing constructions can successfully coexist.