本文主要对徽语核心区 4 个方言点——黟县（宏村）、祁门（大坦）、休宁（海阳）和歙县（徽城）存在的裂化元音进行详细的声学语音学描写。研究所需的声学录音材料全部来自于实地的田野调查，每个方言点都选取了 10 个发音人（5 男 5 女）。
徽语中主要存在两类裂化元音：一类是尾成分央化的裂化元音，如[iːɐ uːɐ yːɐ]等，这也是徽语中最普遍的裂化元音类型，主要分布在歙县（岩寺）、屯溪、休宁、黟县、祁门和浮梁等地。在传统方言学著作中，这类元音韵母通常被描述为“长介音韵母”或“（以央元音为韵尾的）长元音韵母”。其语音上的特殊性在于违反了“响度等级原则”，即开口度较小的首成分在时长和音量上都占据支配地位，感知上更突显，而开口度较大的尾成分时长短、音量弱，因此其首成分才是主要元音。另一类则是尾成分高化的裂化元音，目前调查到的主要是歙县（徽城）方言的[eⁱ oᵘ]，其首尾成分间的发音动程虽可被感知但音质改变幅度不大，所以传统方言学著作都将之描写成单元音[e o]。以上两类裂化元音都是本文的重点研究对象。
在具体的研究中，我们先将裂化元音首成分同常态单元音进行比较，考察它们在 F1/F2声学平面中的频谱分布区域以及相互的位置关系。然后将裂化元音同常态的升峰、降峰和平峰双元音进行比较，主要从 3 个方面展开：（1）时间结构，考察元音段内部各元音性成分（首成分、过渡段和尾成分）的绝对时长及各成分在总时长中的占比（即相对时长）。（2）频谱特征，将元音段的首尾成分同各自对应的单元音进行比较，考察它们在 F1/F2 声学平面中的频谱分布；如果各成分的元音椭圆与对应的单元音高度重合且椭圆面积较小，那么则认为它们具有明确稳定的声学目标，反之则没有目标。（3）动态特性，主要测量各裂化元音和常态双元音第二共振峰（F2）的变化范围和变化速率。如果某个方言还存在三合元音，那么本文同样从时间结构、频谱特征和动态特性 3 个方面来分析它们的语音及音系特点。 通过详尽的声学语音学研究，本文发现：
此外，我们还基于修正后的 Z-score 法对各方言点的元音数据进行规整，规整后的元音系统既能保持每个系统内部各元音的相对位置关系，又能体现不同系统间的绝对物理量纲差异，因而适用于跨方言的比较。而通过跨方言的比较，本文不但描写了 4 个方言点元音系统的共性和差异，更揭示了关于裂化元音的一些共时音变特征，并给出了合理的解释。
（1）裂化元音是介于单元音和双元音之间的一种过渡性质的元音类别，并且徽语的双元音化是一个渐进的过程，具体而言：歙县（徽城）方言的裂化元音[eⁱ oᵘ]在音系类型上最接近单元音，只有首成分承载区别特征，尾成分及动态性是冗余特征。黟县（宏村）方言的裂化元音[iːɐ yːɐ ɛːɐ oːɐ ɤːɐ]在音系功能上也应当与单元音归为一类，但裂化元音[uːɐ]却与单元音[u]构成音位对立，在音系类型上更接近降峰双元音。祁门（大坦）方言的裂化元音[iːɐ uːɐ yːɐ]凭借动态性与静态的单元音[i u y]构成音位对立，其尾成分不是冗余的，所以在音系类型上较黟县的多数裂化元音更接近常态双元音。和祁门一样，休宁（海阳）方言的裂化元音[iːɐ uːɐ yːɐ]也凭借动态性与静态的[i u y]构成音位对立，所以在音系类型上接近降峰双元音；同时，在一些语音环境中，裂化元音的首成分时长缩短、音质改变，出现了明显的“介音化”趋向，所以在语音特点上也更接近升峰双元音。
This study gives a detailed acoustic phonetic description of the diphthongized vowels in the 4 Hui Chinese dialects, i.e. Yi county (Hongcun), Qimen (Datan), Xiuning (Haiyang) and She county (Huicheng). All the speech data are from the 10 native speakers with a balanced gender ratio in each dialect during the fieldwork.
There are two types of diphthongized vowels in the Hui Chinese dialects. The most common one which is found in She County (Yansi), Tunxi, Xiuning, Yi county, Qimen and Fuliang, usually has a centralized offset element towards a [ɐ]-like configuration, e.g. [iːɐ uːɐ yːɐ]. In the dialectological literature, this type of diphthongized vowels was usually described as “long on-glide” or “long vowel”. It is quiet special for its structure which obviously violates the so-called Sonority Hierarchy. That is, the first element which has a higher vowel serves as a vocalic nucleus and thus a higher sonority and perceptual salience, while the centralized offset element which has a lower vowel height turns into a coda with a lower sonority. Another type of diphthongized vowels is reported in She county (Huicheng) dialect, i.e. [eⁱ oᵘ]. The production of these diphthongized vowels begins from a mid-high vowel [e] or [o] and glides up towards an [i]-like or [u]-like position. These articulatory movements are audible but the changes in vowel quality are somewhat smaller, thus they were just described as monophthongs [e o] in many dialectological works. Both of the two types of diphthongized vowels are the main research objects of this article.
In the research, the diphthongized vowels are first compared with monophthongs in terms of spectral distributions in the F1/F2 acoustic plane. Due to the diphthong-like phonetic characteristics, they are further compared with plain diphthongs (e.g. rising, falling and level diphthongs) in terms of temporal structures, spectral properties and dynamic aspects. The investigations of temporal structures focus on the actual durations of each vocalic elements, i.e onset, transition and offset, and their proportional relations. While researching into the spectral properties, the vowel ellipses of the onset and offset elements of the diphthongized vowels and diphthongs are compared with those of their corresponding monophthongs. If they have overlapped spectral distributions in the acoustic vowel space and the sizes of the ellipses are also quite small, then it is safe to say that the onset or offset elements have their own acoustic targets, but not vice versa. In addition, the range and rate of the second formant (F2) change are also important for characterizing diphthongs, thus these two parameters are utilized to probe into the dynamic aspects of the diphthongized vowels, plain diphthongs and even triphthongs.
Furthermore, a revised Z-score procedure is adopted for vowel normalization, not only can it reserve the related positions of the vowels in the same system, but also can present the physical dimensions of different vowel systems, thus it is quite suitable for the cross-language and/or cross-dialect comparisons. It is of theoretical significance that many universals, differences, and even synchronic sound change phenomena regarding diphthongized vowels can be found by comparing the normalized vowel systems among different Hui dialects.
The main results of this study are as follows:
First, diphthongized vowels belong to an intermediate category between monophthong anddiphthong, and the process of diphthongization is gradient in Hui Chinese dialects. Specifically, the diphthongized vowels [eⁱ oᵘ] in She (Huicheng) dialect are most phonologically close to the monophthongs since the onset elements are generally distinctive and consequently the offset elements and dynamics are phonologically redundant. The diphthongized vowels [iːɐ yːɐ ɛːɐ oːɐ ɤːɐ] in Yi county (Hongcun) are also should be grouped with monophthongs, but the diphthongized [uːɐ] can contrast with the monophthongal [u] in spectral dynamics, in other words, the former is a dynamic [u] while the latter is a static [u]. The diphthongized vowels [iːɐ uːɐ yːɐ] in Qimen (Datan) contrast with the monophthong [i u y]. That is, in addition to the onset element, which is the dominating segment in terms of spectral properties as well as temporal organization, the dynamic aspects contribute to the characterization of diphthongized vowels, too. Thus they are typologically one-step closer to plain diphthongs in the continuum of monophthong and diphthong.Xiuning (Haiyang) vowels also contrast in spectral dynamics, i.e. static monophthong [i u y] versus dynamic diphthongized vowels [iːɐ uːɐ yːɐ]; what is more, the diphthongized vowels in Xiuninng have a similar temporal organization similar to plain diphthongs; thus they are most typologically closer to the plain diphthongs in both phonetic and phonological senses.
Second, vowels in the Hui Chinese dialects can contrast in statics and dynamics. Monophthongs just have a static target; diphthongized vowels and falling diphthongs both have a dynamic target; rising and level diphthongs are vowel sequences which are composed of two static targets; triphthongs can be viewed as a static target plus a dynamic target.
Keywords: Hui Chinese dialect, diphthongized vowel, monophthong, diphthong, diphthongization, statics, dynamics.