Learning and the typology of word order: a model of the Final-over-Final Condition

Abstract

This article investigates whether cross-linguistic generalizations may arise from asymmetries in learnability of competing syntactic patterns. The model presented here uses a domain- general statistical learner for parameter systems in order to probe whether languages violating the Final-over-Final Condition (FOFC; Sheehan et al. 2017) might be difficult to learn, rather than syntactically impossible. In this model, no parameters ruled out *FOFC languages, and no penalties targeting them were built into the learner. Regardless, the results of two learning tasks demonstrate a correlation between the learnability of a word order pattern and its frequency in the typology. *FOFC languages were harder to learn, providing a possible explanation for their relative rarity.