Proceedings
Author: Sara Finley (Pacific Lutheran University)
The substantively biased learning hypothesis predicts that phonological patterns that are phonetically motivated should be easier to learn than patterns that show little or no phonetic motivation (Wilson 2006). One pattern that has received much attention is the distinction between vowel harmony (a phonetically grounded pattern) and vowel disharmony (a phonetically ungrounded pattern), as several studies comparing learnability of vowel harmony over disharmony have shown mixed or nuanced results (Zheng and Do 2025). One possibility is that metalinguistic judgments may not be sensitive enough to detect biases for categorical vowel harmony. The present study tests this idea using a modified phoneme monitoring task developed in Finley (2025), where participants were asked to listen for the final vowel of a word that either followed vowel harmony in 80% of trials, or vowel disharmony in 80% of trials (and then switched). Results showed evidence of a bias for vowel harmony over disharmony. Participants were generally faster and more accurate at identifying words that obeyed vowel harmony, particularly when the majority of items followed vowel harmony. Implications for learnability of categorical and variable patterns are discussed.
Keywords: vowel harmony, artificial language learning, learnability
How to Cite: Finley, S. (2026) “Biases for Variable Vowel Harmony in a Phoneme Monitoring Task”, Proceedings of the Annual Meetings on Phonology. 2(1). doi: https://doi.org/10.7275/amphonology.3676