Proceedings of the Annual Meeting of the Cognitive Science Society

Pronominal systems tend to maintain the same person distinctions across number categories. This has been explained by most linguistic theories in which pronominal systems are modelled as a combination of independent person and number features (e.g., Harbour, 2016; Harley and Ritter, 2002). At the same time, person and number features are known to interact asymmetrically, which some have taken to indicate that personal pronouns are not straightforwardly the composition of person and number, but rather categories on their own (e.g., Cysouw, 2009). The main source of evidence for feature-based theories is typological data. In this study, we use an artificial language learning approach to provide complementary behavioral evidence to the question of the independence of person and number in personal pronouns. We test whether learners have a preference for person systems that maintain uniform person distinctions across number categories. Our results suggest that, in the absence of explicit evidence, learners infer the same person distinctions across number categories, regardless of whether number is morphologically transparent or not. Our findings suggest that pronominal systems are most effectively learned as combination of distinct person and number features, rather than as unified categories.