Learning a typologically unusual reduplication pattern: An artificial language learning study of base-dependent reduplication

We report on an artificial language learning experiment testing the learnability of a typologically rare pattern of reduplication. Our model comes from syllable-copy reduplication in Hiaki (aka Yaqui, Uto-Aztecan), a base-dependent pattern wherein the shape of reduplication depends crucially on syllabification in the base: coda consonants can copy in reduplication if and only if there is a corresponding coda in the base. Using a controlled artificial language experiment with a forced-choice paradigm, we show that native English speakers who have no prior exposure to any language with a grammar employing syllable-copy reduplication are in many cases able to learn a variable CV or CVC syllable-copying rule as measured by eventual above-chance selection of the correct form. However, compared to participants learning either a consistent CV or CVC copying rule, the performance of participants tasked with learning such a variable syllable-copying rule improves more slowly, and these participants make more errors overall. We suggest that this difference in learnability may be one of a number of factors helping to explain the typological rarity of certain morphological patterns.