A Study on the Information Mechanism of the 3^rd Tone Sandhi Rule in Mandarin Across Word Boundaries

Previous research has investigated the information mechanisms of the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$3^{\text{rd}}$</tex> tone sandhi within disyllabic words from the perspective of the trade-off between speech production and information accessibility. Compared to changing to other tones, the change from the Tone 3(T3) to Tone 2(T2) can minimize homophone confusion and sentence information loss. Whether this mechanism is equally effective in the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$3^{\text{rd}}$</tex> tone sandhi across word boundaries remains an open question. This study uses information predictability, quantified through word frequency and lexical conditional probability, to represent information accessibility. We selected eligible cross-word T3+T3 tone sequences from the Mandarin Standard Female Voice Database and conducted a logistic regression analysis on the cases of tonal sandhi. The results show that the higher the predictability of the target word, the greater the likelihood of its final T3 changing to the T2. However, the predictability of the triggering word does not significantly affect the tonal change. These findings provide a reference for understanding the mechanisms of tonal sandhi from an informational perspective.