Tone timing: an eye-tracking study of dynamic F0 integration in spoken word recognition

To recognize words, listeners must map a temporally unfolding auditory input to representations at multiple levels. Typical accounts argue for immediate and incremental integration of asynchronous acoustic cues, such that each cue is used to partially update phonemic or lexical representations as soon as it arrives, with later cues further modulating those decisions. However, recent studies suggest that incrementality is not universal and fricatives follow a different principle in which early cues are stored in a memory buffer and listeners wait to identify the phoneme until all cues have arrived. It is unknown why different speech sounds show these differences. Here, we extend this paradigm to lexical tone, an acoustic/phonetic dimension which extends over a syllable and tracks the dynamics of F0. In a visual world experiment, we index the relative activation of minimal pairs only differing in Mandarin tones as they change over time to ask whether lexical tone is incrementally integrated or processed in a buffered manner. Our findings first show lexical activation patterns are closely coupled to the timing of pitch contours, supporting the immediate integration of lexical tone. Therefore, with respect to temporal integration of asynchronous cues, tone is like most segments, though different from fricatives.