Semantics circle: A learning approach to semantic universals

Abstract

One of the fundamental goals of linguistics is to identify a set of principles that constrain our linguistic capacity, and thereby shape the distribution of possible linguistic patterns. However, this idea remains contentious both within linguistics and in the broader cognitive science community. One of the problems is that the data that have been traditionally used to support the existence of such principles have come from introspective judgments, language descriptions (grammars) and typological samples. None of these types of data are particularly good for establishing causal links between human cognition or learning and the (non-)existence of particular linguistic patterns. In this talk, I will highlight how subtle predictions of semantic theories can be investigated through Artificial Language Learning (ALL) experiments. I first introduce this approach by presenting a study where we combine artificial language learning experiments with modelling tools to explore theoretical predictions about systems of personal pronouns (Maldonado & Culbertson 2021; Zaslavsky, Maldonado and Culbertson, 2021). I then present two studies where I use an artificial language learning set up to investigate learning constraints on more complex semantic phenomena, such as clausal embedding predicates (Maldonado et al 2022) and negative polarity items (Maldonado & Kuhn, ms).