Hughes, Cerys and Jarosz, Gaja. Pattern Structure Modulates Learning of Lexically Conditioned Morphology. 2024. Poster presented at Architectures and Mechanisms for Language Processing (AMLaP 2024) in Edinburgh, Scotland, September 5-7, 2024. Poster.
Work in first language acquisition and artificial grammar learning (AGL) indicates that
language learners can extract systematic regularities from inconsistent language data. In
some cases, learners reproduce the inconsistencies in the data (frequency match), while in
other cases they over-extend (regularize) the most frequent patterns when generalizing to
novel examples. Studies with children have generally observed regularization, while adults in
AGL studies tend to frequency match (e.g., Hudson Kam & Newport 2005, 2009). However,
some adult AGL studies (e.g., Schumacher & Pierrehumbert 2021: SP) instead demonstrate
regularization. We replicate and extend one such case (SP) to better understand why
learners regularize in some cases but not others. We find evidence that adult learners’
regularization depends on the structure of the morphological system they are exposed to.