What Makes Lexical Tone Special: A Reverse Accessing Model for Tonal Speech Perception

Previous studies of tonal speech perception have generally suggested harder or later access to lexical tone than segmental information, but the mechanism underlying the lexical tone disadvantage is unclear. Using a speeded discrimination paradigm free of context information, we confirmed multiple lines of evidence for the lexical tone disadvantage as well as revealed a distinctive advantage of word and atonal syllable judgments over phoneme and lexical tone judgments. The results led us to propose a Reverse Accessing Model (RAM) for tonal speech perception. The RAM is an extension of the influential TRACE model, with two additional processing levels specialized for tonal speech: lexical tone and atonal syllable. Critically, information accessing is assumed to be in reverse order of information processing, and only information at the syllable level and up is maintained active for immediate use. We tested and confirmed the predictions of the RAM on discrimination of each type of phonological component under different stimulus conditions. The current results have thus demonstrated the capability of the RAM as a general framework for tonal speech perception to provide a united account for empirical observations as well as to generate testable predictions.