New paper! (Moore, Donhauser et al., PNAS, 2025)

New paper by Peter & team!

Moore* C, Donhauser* PW, Klein D, Byers-Heinlein K (2025). Efficient neural encoding as revealed by bilingualism. Proceedings of the National Academy of Sciences of the United States of America.

Available here.

The brain’s capacity to learn multiple languages represents a profound puzzle of neural organization and flexibility. We investigated how neural systems might systematize multiple sound systems by training computational models on natural speech. Our networks, which approximate infant language learning, maintained distinct phonological systems for two and three languages while preserving shared articulatory features. The timing of second language introduction influenced the learning process. Our findings suggest that phonological acquisition may leverage domain-general learning principles, offering a computational framework for understanding how neural systems potentially scale language learning while maintaining critical language-specific distinctions. This research provides crucial insights into the computational principles underlying the brain’s remarkable ability to manage linguistic complexity.