Jonathan Havenhill

Naturalness and Learnability

An ongoing debate in phonological research concerns the role of a bias toward phonetically natural patterns in phonological learning. The presence of a naturalness bias among learners makes the prediction that phonetically marked patterns are more likely to undergo sound change than unmarked patterns.

To address the role of naturalness in phonological learning, Youngah Do, Elizabeth Zsiga, Sui Lung Sze, and I have conducted a series of artificial language learning experiments to investigate to what extent learners are able to learn unnatural alternations. In one experiment, presented at LSA 90, we simulated a language contact situa- tion in which speakers of Sebirwa borrowed a phonetically unnatural pattern from a neighboring language, Setswana. In a follow-up study, published in the Proceedings of AMP, we test how learners develop knowledge of phonetic naturalness through production experience. Learners in both experiments showed no bias toward natural patterns in categorical learning, but we found that the effects of naturalness may emerge when an alternation is variable. In work presented at LabPhon 18, we compare child acquisition of rounding harmony with gender agreement, and show that phonological, but not syntactic, learning is biased toward naturalness, but again only in variable contexts.


Do, Youngah, Jonathan Havenhill and Sui Lung Sze. 2023. Variation learning in syntax and phonology. Cognition 239.105573. doi:10.1016/j.cognition.2023.105573.

Do, Youngah, and Jonathan Havenhill. 2021. Production and substantive bias in phonological learning. Proceedings of the 2020 Annual Meeting on Phonology. [PDF]


Do, Youngah; Jonathan Havenhill; and Sui Lung Sze. 2022. Free variation learning in syntax and phonology. Poster presented at the 18th Conference of the Association for Laboratory Phonology (LabPhon 18), virtual. [PDF]

Do, Youngah; Elizabeth Zsiga; and Jonathan Havenhill. 2016. Frequency and naturalness in implicit phonological learning. Paper presented at the 90th annual meeting of the Linguistic Society of America, Washington, DC. [PDF]