Proceedings of the Annual Meeting of the Cognitive Science Society

Nominal number marking is widespread across languages, yet number distinctions are often restricted to specific categories. Typological research suggests that such restrictions follow the Animacy Hierarchy Constraint (AHC), whereby higher-ranked categories in the animacy scale (human > animate > inanimate) are more likely to exhibit number contrasts than lower-ranked ones. However, large-scale cross-linguistic evidence for the AHC is limited, and a mechanistic explanation for its typological prevalence is missing. This study combines typological and experimental studies to investigate the impact of the AHC on number neutralisation in nominal paradigms, and assess whether the AHC mirrors cognitive biases at play during language learning, which could in turn explain cross-linguistic regularities. We analyse data from 509 diverse languages using Bayesian hierarchical models, demonstrating a robust monotonic increase in number neutralisation down the animacy hierarchy. To examine whether this cross-linguistic regularity reflects learning biases, we conduct an artificial language learning experiment manipulating animacy-conditioned number neutralisation patterns. Results show that systems conforming to the AHC are more learnable than AHC-violating systems. Together, these findings suggest that the AHC is supported by learning biases that contribute to its typological prevalence.