Effects of sandhi-based predictability in Kansai Japanese depend on markedness: a visual-world eye-tracking study

During comprehension, various aspects of upcoming input can be anticipated. However, it is unclear to what extent listeners exploit suprasegmental information to facilitate lexical retrieval. In Japanese, a Kansai-dialect-specific pitch accent system and a sandhi rule create a situation where the final tone of certain modifiers is conditioned by the initial tone of the following noun, thereby making the upcoming tone predictable. Both Tokyo and Kansai speakers showed increased fixations on the target when its initial tone was predictable based on an all-L-tone modifier, suggesting that this “predictability effect” may reflect an increased anticipation of upcoming input that disrupts a marked L-tone sequence, without necessarily invoking the dialect-specific sandhi rule. However, Kansai speakers were quicker to identify the target in this condition, suggesting that they utilised dialect-specific knowledge about the lexically determined pitch accent of the target noun to facilitate processing, particularly when the modifier tone deviated from the default.