On the productivity of tone sandhi in a highly restricted context: acoustic evidence from Bu sandhi in Taiwan Mandarin

This study investigates the productivity of the Bu sandhi rule in Taiwan Mandarin, which mandates that the Tone 4 (T4) monosyllabic word Bu changes to Tone 2 (T2) when preceding another T4 syllable. As Bu sandhi operates in a highly restricted context, it raises questions about whether speakers rely on an active phonological computation mechanism or a lexical chunk retrieval mechanism. To address this, a production experiment was conducted with 45 native speakers to assess both the phonological application and phonetic realization of the rule. Participants produced four-syllable carrier phrases containing four conditions: Bu with high-frequency real morphemes, Bu with non-high-frequency real morphemes, non-existing words involving Bu with actually occurring morphemes, and pseudowords involving Bu with accidentally gapped morphemes. Auditory tone classification revealed near-perfect application rates across all conditions, demonstrating that Bu sandhi is highly productive at the phonological level. Acoustic analyses revealed that while temporal parameters such as vowel durations and f0 turning points remained stable across conditions, fundamental frequency (f0) contours exhibited systematic divergences. Specifically, localized f0 lowering emerged in conditions involving existing morphemes, with its temporal locus dynamically modulated by lexical frequency. Drawing upon the cascaded activation framework, these findings indicate that the residual activation of the underlying T4 cascades downstream during speech production due to lexical competition. Taken together, the current study provides strong evidence that native speakers utilize an active phonological computation mechanism, while its phonetic implementation is delicately shaped by cascading lexical dynamics.