Abstract
INTRODUCTION: This study investigates whether unfamiliar tone sandhi patterns in Tianjin Mandarin can be implicitly learned through an artificial language learning experiment, and if the acquired knowledge is rule-based and generalizable. METHODS: Participants were trained to learn monosyllabic words and disyllabic phrases with their attention focused on a word-order rule, while unknowingly being exposed to unfamiliar tone sandhi patterns. A judgement test with trial-to-trial confidence ratings was conducted to assess the learning outcomes and participants' awareness. RESULTS: Results revealed significantly above-chance performance on tone sandhi patterns for learned phrases. This learning effect was generalized to unseen phrases made up of familiar words, but not to phrases with new words, indicating a degree of abstraction across instances, though the learning is not fully rule-based. The confidence rating results suggest that participants were unaware of the structural sandhi knowledge, but the reaction time data of the judgement test indicate that the sandhi knowledge was learned with awareness at the level of noticing. DISCUSSION: The results have been discussed in light of theories of implicit learning and the findings of previous research on phonological learning.