Hwangbo, Hyun Jin. 2020. Computational complexity and learnability of two vowel harmony patterns with neutral vowels. Studies in Phonetics, Phonology and Morphology 26.1, 205-230.
This paper investigates the relationship between the complexity of a phonological pattern and its learnability. The Complexity Hypothesis refers to the idea that less complex patter n s are easier to learn than more complex pattern s . In this paper, t he hypothesis is tested as the Subregular Hypothesis, which measures computational complexity in terms of the Subregular Hierarchy. Th is hypothesis states that patterns in a lower class are less complex and thus easier to learn. Two artificial grammars of vowel harmony pattern s with neutral vowels were tested. The two patterns differed in terms of the computational complexity and attestedness . The At Least One (ALO) pattern, the less complex pattern and the unattested vowel harmony pattern , is predicted to be easier to learn than the Rightmost pattern, the more complex pattern and the attested pattern . The results showed that the Rightmost pattern was more difficult to learn than the ALO pattern. This supports the Subregular Hypothesis and implies a learning bias toward computational complexity despite the attestedness of vowel harmony patterns. (Hankuk University of Foreign Studies)
Keywords: computational complexity, learnability, vowel harmony, neutral vowels, artificial grammar learning