Learning Biases for Vowel Height Harmony

Volume 13
Issue 3
Sara Finley & William Badecker
We test the role of phonetic grounding and typological tendencies on learning biases for vowel height harmony, a phonological process in which vowels within a word are required to share phonological features for height. Several height harmony languages are constrained such that vowels ([i, e]) undergo harmony following both front and back vowels, while back vowels ([u, o]) tend to only undergo harmony following other (often identical) back vowels. In Experiment 1, participants were trained on a height harmony pattern with either front vowel suffixes or back vowel suffixes. Participants reliably displayed harmonic responses only when the suffixes contained a front vowel, despite exposure to back vowel suffixes. Experiment 2 tested for generalization to novel vowel harmony triggers, exposing learners to tense front and back vowel stems or front vowel stems, with a front vowel suffix alternation. Participants generalized to front lax vowels but did not generalize to tense back vowels, suggesting a bias for height harmony languages in which the trigger and target share the same value for backness.

Key words: artificial grammar learning, vowel harmony, phonetic naturalness