Abstract
Morphosyntactic expression exhibits significant cross-linguistic variation, yet systematic tendencies emerge. Gender and number features are frequently co-expressed within a single exponent, and when realized separately, gender morphology tends to surface closer to the stem. Competing theoretical accounts of gender and number seek to explain these tendencies, but empirical validation remains challenging due to sparse typological data. Across a series of artificial language learning experiments, we demonstrate that learners exhibit a robust preference for placing nominal morphology encoding gender information closest to the stem. This bias, consistent across English, Italian, and Kîîtharaka speakers, holds both for prefixing and suffixing morphology, and across different types of gender and number systems. These results provide a new source of empirical support for theories proposing a gender-first derivation and highlights cognitive constraints influencing morpheme order.
