【作者】 吴力菡；

【导师】 马秋武；

【作者基本信息】 南开大学 ， 英语语言文学， 2012， 博士

【摘要】 汉语语调是语音和音系研究中的重要课题。以往研究多集中在对语句内部音高曲拱的细致描述之上，而对音高以外的音长、强弱等韵律现象的研究以及音系理论的思考相对较少。本研究基于广义的语调观，对汉语语调中三个韵律要素之间的关系、调长和调强的分布模式以及语调类型范畴等问题进行语音和音系学分析。以下为主要观点和分析结果。（一）汉语语调的特性和研究方法我们认为，汉语语调具有二重性（语句内某个音节与前后音节在音高上的相对性以及它的音高在整个调域中的绝对性）、多维性（除了音高以外，调长和音强等要素都属于语调特征）、连续性（语调类型之间在各维度上都呈现出渐变的状态）等特点，因而研究中可相应地采用二重分析法（基于全句调阶、音节平均调长等的宏观分析，以及基于句内各音节变化的微观分析）、多维度分析法和连续体的分析方法。（二）韵律三要素的关系及增长方式对两名发音人五字句音长、音高和音强连续体共175个语句的全句音节平均调长、调高和调强所作的统计分析表明，三个韵律要素之间的伴随关系呈现出不同的方向性。总体上看，音长增长时，音高下降，音强也减弱；音高上升时，音强也显著增强，反之亦然，二者表现出双向的显著正相关；但无论是音高上升还是音强增强，音长都不表现出稳定的伴随变化。此结果与吴宗济（2004）“音长与音高、音长与音强之间存在着单一的正相关关系”的观点有所不同。就各要素的独立性而言，音长的独立性最强，其次是音高要素，而音强则具有最强的依赖性，主要依附于音高，三者中音强与音高的关联更为紧密。每两个韵律要素之间的伴随关系还可能受到第三要素和声调类型的影响。对三要素连续体中各韵律要素的增长方式所作的分析表明，音长的增长和音强的增强都表现出非线性的增长方式；而音高的增高无论是采用线性的赫兹值还是由赫兹转化而来的半音值都表现出很强的线性特征，结合目前已成共识的“音高的感知呈非线性”的观点（李爱军2005），我们认为，音高的产出和感知方式可能并不一致。另外，代表三个韵律要素增长方式的三类曲线的走向各具特点，表明三者在人脑中可能遵循着不同的认知或处理模式。此外，声调类的特性对于各连续体等级所能产生的等级数以及数值都产生重要影响。（三）语调连续体和语调特征汉语中的各种语调同时具有“连续”和“离散”的特性，因此在分析上可采取相应的“连续体”和“特征提取”的方法。对一名发音人五字句19种语调及无声句累计92个语句的分析表明，汉语各种语调类型在音长、音高和音强维度上都能各自形成一个连续体。与此同时，基于两名听音人对该发音人在三要素连续体中三个维度上分别划分出的[高][中][低]特征的声学边界，对发音人的各种语调句赋以宏观层面上的特征值。用此方法对各类语调特征所作的概括符合人们的语感和直觉，能够对汉语语调作出很好的描述。语调实现过程中对于基值特征选择和组合的限制主要源自外部因素（交际的明晰性和辨义性原则）以及内部因素（语音实现的生理和物理限制）。前者主要体现在禁止出现音长和音高都取[低]值的组合，后者主要体现在禁止音高与音强取对立值。这里反映出音系理论分析中的对称性与实际的语音实现所表现出的不对称性之间存在着矛盾。（四）语句内部音长和音强的分布模式对一名发音人76个五字语句以及两名发音人二至十字句四类功能语调288个话语（累计365个语句）的调长分布所作的分析表明，语调基元不仅在调形上具有稳定性（吴宗济2004），在时长上也表现出稳定性的比例关系。语调中的调长变化是依托于韵律单元之上的。汉语无焦点语调的调长分布首先要受到位置的制约，然后受到语调大类的制约，再进而受到语调小类和声调类的制约。1．位置制约1)韵律词末音节长于非韵律词末音节，且变化幅度更大。非韵律词末音节的调长受到更强的位置上的限制，但二者的变化都会控制在一定的范围之内。二字组的调长格局为“短-长”，三字组的调长格局为“短-短-长”，与王茂林（2003）对自然话语进行分析所得的结果相同。2)全句调长下倾原则，即句中各韵律单元内部平均调长逐渐递减。句首韵律单元的总调长大于相应数目调长单位的总和；句中韵律单元的总调长在相应数目调长单位之和上下浮动；句末韵律单元总调长小于相应数目调长单位的总和。用“L，M，R”分别代表韵律词的左、中、右位置，“D”表示音长，该制约条件可表述为： AVE(D_L1+(D_M1)+D_R1)≧AVE(D_L2+(D_M2)+D_R2)≧AVE(D_L3+(D_M3)+D_R3)…调长的下倾性是依托于韵律单元来实现的，它还表现为：句中各韵律单元的首、末音节的调长大致呈逐级递减的特点。3)韵律单元末音节所受到的限制小于非末音节。语句内各位置所受制约的强弱程度大致排列如下（从弱到强）：句内大韵律单元末音节﹤句末韵律单元末音节﹤句内小韵律单元末音节﹤韵律词首音节﹤句内韵律词中间音节。2．语调类的制约功能语调的实现可能更加依赖于句末音节的表现，而很多情感语调的实现则可能更多地依赖句内韵律单元末音节时长的音延。当后字音长达到前字的两倍以上时，便会产生明显的停延感。音延特征在不同语调中以及句内不同位置具有强制性与非强制性。大部分陈述语调的句内音延都是非强制性的；疑问和感叹语调的句末音延可能具有强制性；祈使语调句内音延具有强制性，以便为后韵律词的短促和强语势做准备。句内韵律词末音节具有音延特征的多为程度较低的疑问（如怀疑、犹豫）和祈使语调（如家人、好友等），采用句内音延的策略可以表达委婉、亲切、轻松的语气。句内韵律词末音节与句末音节可能分别以承担情感语调和功能语调的功能为主。四种功能语调的差异并不依靠全句的节奏变化来实现。两名发音人四种功能语调之间，句末音节的调长、句末韵律短语的总调长都表现出显著差异，体现出句末音节的语调功能性，从音长维度验证了李爱军（2002）、林茂灿（2006）所提出的“句末边界调音节具有语调功能性”的结论。对调长的研究还反映出韵律结构与句法结构之间的不同构。3．声调类的制约阳平和去声句之间，阴平和上声句之间表现出相近的调长分布特点。我们认为这是由于前一组同属于曲折调，后一组同为平调，而相同的调类必定表现出相似的语调特性。这也为林焘（1984，2001）“上声声调的本质是低平调”的观点提供了一点来自语调上的证据。平调句的调长变化区间整体上大于曲调句，韵律短语末音节调长均值大于曲折调句。和调长的表现相反的是，调强（I）和调高（P）基本遵循“韵律单元末音节弱/低于非末音节”的规律（即I_L﹥I_M﹥I_R， P_L﹥P_M﹥P_R），每个韵律单元的首音节都会出现音强重置。调强的全句下倾不如音高和音长的下倾性明显。韵律词边界音节除了具有长调长的特征，同时还附带有低音高和弱音强的特点。各种语调的语句在向各维度的目标值逼近的过程中，具有语调区分意义的逼近可能始于语句中的不同位置，从先到后依次为感叹-祈使-疑问语调，即感叹语调始于句首或主语开始的位置，祈使语调始于句尾动词短语的首音节（即动词成分），疑问语调具有语调区分意义的逼近时间最晚，始于句末重音节。本研究还作出汉语语调中调长、调强分布模式示意图，并加入调域参数，作出五字句多维度模式图。（五）语调的声学范畴和听感范畴四类功能语调在音高维度上的声学格局图显示，陈述和祈使语调的表现相近，为低调阶语调类，疑问和感叹语调的表现相近，为高调阶语调类。高、低调域各有两种基本语调，在调域系统中表现出平衡性。两个语调类的区域并不截然分开，而是存在着一个宽度约2个半音的带状重叠区域。若将发音人调域分为五个等级，该区域位于中部偏上的调域区，即2.5-3等级之间。然而，进一步分析表明，疑问和祈使语调的调阶具有浮动性，其[高]和[低]应当是一种相对性的特征，而陈述与感叹语调在全句调阶上的[低]与[高]特征则具有绝对性。这便是造成声学分析与听觉实验结果产生差异的原因，同时也表明，四类语调的声学格局与听觉格局并不完全一致。合成听辨实验表明，语音三要素在从陈述语调到祈使语调的实现过程中，音长所作的贡献最大，其次为音高，最后为音强。（六）理论思考受到生理因素的制约，汉语语调的下倾性不仅体现在音高上，在音长和音强维度也有所表现。各维度的下倾表现形式各异，但都以韵律单元为依托。下倾并不是汉语语调的实质或底层形式，而只是表层的语音体现；它也不是汉语语调所特有的性质，而是带有普遍意义的语调现象。汉语语调的特性主要体现在字调和语调的关系之上。二者互相制约，构成字调与全句调相互作用的连续体：字调-连读变调-中性语调-口气语调-呼调-歌调。在这一连续体中，越往后端，全句调的地位越高，字调的个性越弱。分别建立声调特征系统和语调特征系统有助于进一步认识声调与句调的本质差别。本文从宏观和微观角度尝试建立汉语语调的特征架构。全句特征包括音长、音高、音强维度上的[高][中][低]特征和调域的[宽][窄]特征，主要用于实现情感等功能。局部特征包括[重音]、[骤降]、[上扬]和[音延]等，主要体现语义和语法功能。尝试图解语调从生成到感知的大致过程，这其中包括心理、生理和物理过程，并涉及发生语调学、声学语调学和感知语调学的内容，各环节都可尝试建模。产出与感知可能不完全一致。一种语调可能有多种表现形式或实现手段，有典型和非典型形式，程度上也有高低之分。最后提出汉语语调有限状态语法的构想，并总结几个位置制约条件。更多还原

【Abstract】 While previous studies on Chinese intonation have been focused much on thedelicate description of the pitch contour within an utterance, this dissertation, basedon a macro viewpoint of intonation, examines the relation among the three prosodicelements, the distributional patterns of pitch duration and intensity, intonationalcategories and other related topics. Results and preliminary findings are as follows.I．Characteristics of Chinese intonation and its research methodWe argue that the Chinese intonation bears at least the features of duality,multi-dimension and continuity. By duality, it refers to the fact that a syllable has arelative pitch height in its utterance context as well as an absolute one within aspeaker’s register. Intonation is being multi-dimensional in that besides pitch, otherprosodic elements like duration and intensity also make contribution to intonation andthus should also be taken into account. By continuity, it means there is a continuum ofchange among different types of intonation. Accordingly, the methods of dual andmulti-dimensional analysis and the analysis by means of continuum can be employedwhen working with Chinese intonation.II. The relationship among three prosodic elementsThe analysis on175utterances of five syllables in the continuums of duration,pitch and intensity produced by two speakers shows a different directionality ofcorrelation among three prosodic elements. Generally, when duration increases, thepitch and intensity decrease; when pitch elevates, the intensity will increasesignificantly, and vice versa, demonstrating a bi-directional positive correlationbetween the two; but no steady incidental changes occurs to duration when pitch orintensity increase. This result indicates a difference with Wu（2004）’s view that thereexists a one-way positive relation between duration and pitch, and between durationand intensity.The prosodic element of duration is of the highest degree of independence,followed by pitch and intensity. The connection between pitch and intensity is the closest among the three, and intensity shows a strong reliance on pitch. The incidentalvariation of every two elements might also be affected by the third one as well astonal type.The hierarchies of both duration and intensity go up in a non-linear way, whereas,pitch tends to increase linearly in either Hz or semitone scale. Consider that the pitchis perceived non-linearly （Li2005）, we postulate that the production and perceptionof pitch might follow different cognitive patterns, and so do the three prosodicelements. The study also suggests an influence on the number of hierarchy from thetype of tone.III. The continuum of intonation and intonational featuresWe hold that the various types of Chinese intonation are both continuous anddiscrete, thus approaches of “continuum and features” are to be adopted accordingly.An analysis on a speaker’s92utterances which contain19types of intonationsuggests that these types constitute a continuum in each dimension---duration, pitchand intensity. The acoustic boundaries of [high],[mid] and [low] in each dimensionare categorized by two listeners, which are then assigned as global features for26types of intonation. This approach can well reflect people’s linguistic intuition.Restrictions on the selection and combination of these global features comemainly from an extrinsic factor---the communicative principle of explicitness and anintrinsic one---the constraint of physical realization, with the former disallows thecooccurrence of a [low] duration and [low] pitch, while the latter forbids assigningopposite values to pitch and intensity, revealing a contradiction between thesymmetry in phonological analysis and the unbalance in phonetic implementation.IV. The duration and intensity patterns of intonationThe analysis on two speakers’365utterances suggests that intonation unitdisplays a state of steadiness not only in its pitch contour （Wu2004） but also in itsduration pattern, which might be, based on the prosodic unit, restricted firstly bypositional constrains followed by constrains from types of intonation and tones.i. Positional constrainsi) With a larger range of variation, the final syllable of an intonation unit isalways longer than non-final ones, which are given stronger positional restriction. The durational pattern of two-syllabic group is “short-long” and the tri-syllabicpattern “short-short-long”.ii) The principle of durational declination, ie. the average pitch duration of eachintonation unit declines in an utterance, stated as “AVE(D_L1+(D_M1)+D_R1)≧AVE(D_L2+(D_M2)+D_R2)≧AVE(D_L3+(D_M3)+D_R3)…”，where “D” stands for durationand “L, M, R” stands for the left, middle and right position of an intonation unit.iii) The degree of constrain might follow the order of “final syllable of non-finalmajor prosodic unit＜final syllable of the utterance＜final syllable of non-finalminor prosodic unit＜beginning syllable of prosodic word＜non-boundarysyllables of non-final prosodic word” from weak to strong.ii. Constrains from intonational typeThe significant difference in terms of the duration of the final syllable of anutterance indicates its intonational function. A disparity between prosodic andsyntactic structures is also observed. The ending syllables at phrasal and sententiallevels might carry mainly emotional and grammatical function respectively. Thefeature of lengthening might be optional or compulsory in different positions of anutterance. An significant difference exists among four functional types of intonationwhen the duration of the ending syllable and the last prosodic phrase is measured.iii. Constrains from tonal typeSimilar patterns have been observed between Tone2（T2, rising） and T4（falling）utterances, and between T1（high） and T3（low） utterances, providing an evidence fromintonation for Shi （2011）’s argument that T3in Chinese is essentially low.Being opposite to the durational pattern, both intensity （I） and pitch （P） followthe rule that the final syllable of a prosodic unit is weaker/lower than non-final tones,stated as I_L﹥I_M﹥I_R, P_L﹥P_M﹥P_R. The resetting of intensity is observed at thebeginning tone of every prosodic unit. The global declination of intensity is lessobservable than that of pitch and duration. Besides having a long duration, theboundary tone of the prosodic unit carries the features of low pitch and intensity.The linguistically significant approximation to the targets in each dimensionmight start at different points in an utterance. The exclamatory intonation might start at the very beginning or the subject of the sentence, the imperatives starts at the firstsyllable of the final verbal phrase （usually a verb）, and the interrogative intonationstarts at the last accented tone of the utterance.V. The acoustic and perceptional categories of intonationThe chart that illustrates the acoustic pattern of four functional types ofintonation in pitch dimension shows that the intonation of statements and imperativesfalls into the class of low register intonation, while the interrogative and exclamatorytypes belong to the high register, reflecting a balance within the system. Anoverlapping area of about2semitones’ range is observed for these two registers.However, a further analysis shows that the tonal level of both interrogative andimperative intonation might be floating, thus making their features of [high] and [low]into a relative one, while the [low] and [high] features of statements and exclamationsare of absolute characteristics. This might be the reason why a discrepancy does existfor the outcomes of acoustic and perceptional experiments. The acoustic pattern is notnecessarily in accord with the perceptional one.A further perceptional experiment indicates that the element of duration makesthe greatest contribution when synthesizing an imperative intonation from astatement.IV. Theoretical reflectionsDue to the physical and biological constrain, the declination of intonation occursnot only in pitch dimension, but also in duration and intensity. Declination is not thenature or underlying form of Chinese intonation, it is actually a surface phoneticrealization; it is neither a specific feature for Chinese intonation, rather, it is auniversal linguistic phenomenon across languages.The specific feature of Chinese intonation lies in a special relation between itstone and intonation, two of which are actually interacting with each other,constituting a continuum as “citation tone---tone sandhi---neutral intonation---emotional intonation---calling tone---singing tone”, in which the right-ending formshows the highest status of sentential intonation and the lowest status of citation tone.The feature geometry is proposed so as to learn more about the distinctionbetween tone and intonation. Features are grouped into global and local ones. The former, functioning mainly in emotional expression, includes [high][mid][low] ineach dimension as well as [broad] and [narrow] for the parameter of register; localfeatures include [accented],[abrupt falling],[rising] and [lengthening], which arerelated to the semantic and grammatical function.The process from intonation generation to perception involves psychological,physiological and physical sub-processes. A certain type of intonation may be realizedas typical and untypical forms. A finite state grammar of Chinese intonation istentatively proposed and positional constrains are summarized.更多还原