Saldana, C. , Smith, K. & Kirby, S.
The University of Edinburgh
Syntactic categories are a minimum requirement for language syntax to be described in terms of phrase structure grammar. A syntactic category can be defined as the set of items which can assume the same position within similarly-constructed sentences of a language. It is unclear, however, why syntactic categories assume their positions, what their roles are within the linguistic system, and what cognitive biases they adapt to. Statistical learning provides strong evidence of how linguistic categories are acquired. But such inferential processes in language learning are also aided by the cultural evolution of the linguistic systems themselves, as they are shaped by the learnability and expressivity pressures which are applied as languages are transmitted. We describe an experiment which investigates how inferable and predictable syntactic categories can emerge through cultural evolution. Using the Iterated Artificial Language Learning framework, we demonstrate how lexical syntactic categories arise in artificial grammar learning and transmission tasks. By training participants on simple motion events, we see the emergence of functional morphology that determines syntactic categories and whose variation in form is constrained by locality. In doing so, we illustrate how cultural evolution can ease the task of acquiring syntactic categories and phrase structure through statistical learning.