Naturalness bias in palatalization: An experimental study

Abstract

In the present study, we report on an artificial language learning experiment aiming to test the idea that it is easier to learn palatalization before a front vowel than it is to learn depalatalization in the same context. The motivation for the study comes from recent work by Czaplicki (2013), who provides a detailed analysis of palatalization-related effects in Polish, showing that they have no phonological basis. The conclusion he reaches is that ‘phonological naturalness does not play a role in linguistic computation’ Czaplicki (2013:32). It is nevertheless the case that palatalization is cross-linguistically much more common than depalatalization. If naturalness plays no role in computation, the typological asymmetry must arise from elsewhere, for example, from biases in learning difficulty. Our results provide provisory evidence that there is a small but statistically significant advantage for learning palatalization over depalatalization for adult Hungarian speakers.