Inductive bias in acquisition of phonological variation in artificial language

Abstract

We attempt to induce free variation in rounding harmony as a test case for inductive bias effects and hypothesize that participants will shift towards variant distributions that mimic typological patterns. We constructed a simple artificial language to test the role of trigger and target height, as well as height agreement, on acquisition. We find that participants boosted more natural rounding harmony patterns and reduced rounding harmony in unnatural contexts relative to their input. We propose that an inductive bias is operative in phonological acquisition, shifting unnatural input towards natural phonological patterns, resulting in variant distributions consistent with typology.