探讨普通话二声变调——从音位到语音产出的模式

A Phonetic Implementation Model of Mandarin Tone 2 Sandhi

本文的主要目的,是运用数据分析以及建立模型的概念,对二声变调的本质进行研究分析,再以此为基础建立语音学模型来模拟中文声调在语流中产生的变化。实验材料包括单音字以及三音节和四音节词组。语境包括单词朗读,带重音的信息焦点,不带重音的信息背景以及会话。数据分析的结果显示,普通话二声在适合的语境中有可能变得很像一声。此现象出现的词组结构以及语境非常广泛。右分支、左分支的三字组以及四字组的第二字、第三字都可能发生变化。会话中出现的频率很高,但是朗读时也会出现。本文提出二声变调是一个语流中发音弱化的现象,在语音学中有一定的规则可循,并不属于音系学层次范畴化的运作。运用轻重音以及对应的时长可以抓住不同结构和语境背后的共通性,对普通话二声变调表面呈现出的多样语音变化做出原则性的解释。在说话的过程中,发音人不断地在准确传达信息与易于发声两个条件之间选择一个平衡点。本文运用Stem-ML模型,用一个加权参数代表音律的轻重,模拟这两个条件之间的权衡关系,体现音系学与语音学界面的实践。这个参数在句子中的变化可以控制音系范畴在语音学上的表面现象,能够有效地捕捉到二声在语流中丰富的变化。

This paper presents a phonetic model of tonal variations in Mandarin Chinese based on a quantitative study of Mandarin Tone 2 sandhi, the process whereby an underlying tone 2(a rising tone) surfaces, or appears to surface as a tone 1(a high-level tone).The experimental data included monosyllabic, trisyllabic and quadrisyllabic words produced in different contexts, including reading in isolation, in sentences as foreground information or background information, and in conversation. Contrary to most phonological analyses which claimed that Mandarin Tone 2 sandhi occurs only in restricted word structures and in conversational speech, this experiment elicited sandhi-like production in all word structures under investigation, and in reading tasks as well as in conversation. Tone 2 sandhi occurred in both right-branching and left-branching trisyllabic words and in the second and third positions of quadrisyllabic words.A principled explanation that ties together diverse structures and contexts is to treat Tone 2 sandhi as a phonetic reduction phenomenon rather than a category-changing phonological rule. In prosodically weak positions, tone 2’s exhibit gradient reduction that is correlated with duration. The tonal contour of a tone 2 in the Tone 2 sandhi environment may approach the shape of a tone 1 following the principle of ease of articulation when the duration is short and the surrounding tonal targets are high.The prosodic model Stem-ML is used to simulate the interface between phonology and phonetics where variations of a phonological category arise as the speaker balances the trade-off relationship between ease of articulatory effort and communication accuracy in speech production. A weight representing prosodic strength is introduced in the simulation model to capture this trade-off relationship. The shifting prosodic weights in the sentence control surface phonetic variations which match the production data.