1、 ON NEUTRAL-TONE SYLLABLES IN MANDARIN CHINESE (汉语普通话轻声的声学特性 )摘要：本文通过对 200 多个轻声词的声学分析，考察普通话轻声音节的声学特性。实验分析结果说明，轻声音节之所以听起来轻，是因为它的能量较弱，这是音高、音长、音强和音色四要素总体弱化的综合效应，而不仅仅是音强变弱的单一结果。具体说来，轻声音节的音高模式中和化了，失去了它正常重读时的声调区别；音长缩短了，大约为正常重读音节的一半左右；音色相对弱化，例如清辅音往往会浊化，元音趋于央化，但原有音色仍可分辨；音强普遍偏低，但是，第三声音节之后的轻声音节，其音强不但不偏低，还反而普遍
2、偏高。所以，音强不是构成轻声的主要特性，相对说来，音高和音长的变化是构成轻声的更为重要的特性。I. INTRODUCTION One of the characteristic and hard to be grasped features in Mandarin Chinese is the so-called neutral-tone ( hereafter NT) syllables, which is neutralized from corresponding normal stressed syllables. Perceptually, the NT syllable sound
3、s quite weak, and must be attached to other syllable as a dependent morpheme in polysyllabic words. However, it does not equal to the unstressed syllables in common, in stead, it forms an important morphophonemic contrast to the normal type one in bisyllabic words. Consequently, it might be of benef
4、it to Chinese speech processing to clarify the phonetic nature of the NT syllables. Many relevant approaches from different viewpoint(Dong,1956;Lin and Yan,1980; Lin,1982) have been contributed before, this study is concerned in their phonetic characteristics. The discussion is based on a relevant a
5、coustic measurements and analysis (Cao,1986) referred to a corpus of materials including over 200 of NT syllables, all of them are occurred as second morpheme in different category of neutral type bisyllabic words.The hypothesis is that the perceptual feature of the NT syllables may be caused by a c
6、ombination of multiple acoustic effects, in stead of any single factor. Therefore, this kind of syllables must own their special acoustic patterns.II. EXPERIMENTAL RESULTS AND DISCUSSION 2.1 The intensity of NT syllablesPerceptually, the stress of NT syllables is weaked, so people are often used to
7、regard that it is a result of low intensity. However, the data obtained here illustrate that the intensity of NT syllable is not necessary lower than that of the syllable with normal stress. Table I offers a comparison on average intensity between normal syllables and the NT ones obtained from a set
8、 of duplicated bisyllabic words, in which two syllables with the same phonetic structure but different stress type. The corresponding figures show clearly that the intensity of NT syllable is lower than that of the normal one in general, but the situation is just reversed when it is preceded by a th
9、ird tone syllable. This tendency also can be seen from Fig 1, so it seems to be a relational invariance in this area. Therefore, the stress of NT syllable is weaked, but not simply related to its intensity. Table I. Comparison on average intensity(dB) between normal and NT syllables in duplicated ne
10、utral type bisyllabic words. _Original intensity of intensity of tone category normal syllable NT syllable_value sd value sd_first tone 15.20 1.95 7.46 0.96_second tone 12.77 2.99 9.87 1.35_third tone 9.38 1.38 10.05 0.94_fouth tone 15.85 0.64 7.08 1.88_* first tone, _._. second tone,_ third tone, .
11、 fourth tone.Fig.1 Comparison on intensity and pitch height between NT syllable and preceding normal syllable: The top lines in both of (a) and (b) represent amplitude contours, and the bottom ones the pitch contours. 2.2 The pitch of NT syllables One of the most characteristic features of NT syllab
12、le is loss of its original tone pattern. Usually, people pay most of the attention to its short duration and relative pitch height, while the dynamic range of pitch contour is often ignored (Chao,1968). According to the perceptual impression, the pitch height is usually marked with the points of thr
13、ee or four levels(Chao,1980; Luo and Wang,1981), namely, the highest one occurred after a third tone syllable, then the ones occurred after the second tone and first tone syllable respectively, the lowest one is that occurred after the fourth tone syllable. Based on the results obtained from relevan
14、t investigation ,as it has been summarized in Fig.1 and Table II, the following findings have been made(Cao,1986), though it is still remained as a controversial issue. First, it is obvious that there exist two different type of pitch contour among the NT syllables, and it is significantly distingui
15、shable (p0.01) determined by the tonal difference of preceding syllables, namely, a mid-level or slightly rising contour after a third tone syllable, and a mid-falling one after all of the other tones syllables;Second, as what can be seen from Table II, the average pitch value at the onset of NT syl
16、lables in different tonal context is closed to one another, while the value at syllable offset does show a regular order, i.e, the highest value is held by the tokens occurred after the third tone syllable, and then is that of the tokens after second and first tone syllables, the lowest one is that
17、occurred after the fourth tone syllables. This order is just matching to the four-levels distinction of perceptual impression mentioned above. Consequently, the dominant effects for the perceived impression of pitch height should be coming from the whole pitch contour, rather than a single point of
18、pitch height, at least, it is more related to the later part of NT syllables, in stead of the pitch values at the onset as some approach(Lin, 1982; Wu and Lin, 1989) suggested. Actually, this finding has been revealed by a relevant test of the synthesis(Yang, 1988). It indicates that, the NT syllabl
19、es do have own regular pitch patterns and it is definitely contributed to the perception of NT syllables.Table II. Average value of relative pitch height(Hz) of NT syllables _preceded pitch value pitch value tona at the onset at the offset context _value sd value sd_after 1st tone 152.34 21.05 89.51
20、 11.76_after 2nd tone 174.71 11.55 108.45 27.64_after 3rd tone 136.60 6.93 146.60 3.46 _after 4th tone 123.45 15.19 75.30 11.59_Table III. Comparison on duration(ms) between normal(NM) type syllables and NT syllables _original normal syllable NT syllable ratio of tone _ NM to NTcategory value sd val
21、ue sd syllables_first tone 313.21 34.33 186.76 24.78 100:60_second tone 304.92 51.14 183.02 30.50 100:60_third tone 299.80 44.89 209.13 10.56 100:70_fourth tone 302.25 60.75 192.33 28.15 100:64_plotted 305.05 5.83 192.81 9.99 100:63_Fig 2. Comparison between NT and normal syllables in their acoustic
22、 vowel-chart 2.3. The duration of NT syllables Generally, the NT syllable is systematically shorter than the syllable with normal stress, the average ratio between them is about 3:5 in isolation and 1:2 in running speech(1992). Of course, it is only a rough range, the specific realization of individ
23、ual tokens is more complex. First, According to the average duration (see Table III) measured from a group of duplicated bisyllabic words in isolation, the durational realization of NT syllables seems also to be determined by tonal difference of the preceding syllables. Generally, a relatively longe
24、r duration is held by the tokens occurred after third tone syllables; second, the shorter duration of NT syllable is not just caused by a simple compression of the whole syllable, but a change of temporal distribution within the syllable. As what can be seen from Fig. 2, a relatively greater degree
25、of shortening is taken place in the vowel, instead of the consonant of the syllable. 2.4 The sound quality of NT syllables The sound qualities in NT syllables, especially for the syllable finals, are usually perceived as somewhat indistinct, it is due to the effect of neutralization. Specifically, t
26、he consonants are obviously shortened and reduced compare with that in corresponding normal syllables. In addition, it is most likely to become voicing; At the same time, the vowel quality in NT syllable is neutralized and sounds close to that of the shwa. According to Fig. 2, however, their origina
27、l sound quality is still identified. Fig. 3 shows a comparison in the related acoustic vowel-chart between NT syllables and the normal ones. The top one represents the position of the monophthongs /i/,/a/,/u/, and the bottom one of diphthongs/ai/,/ua/,/iao/ and /iou/, here the solid line marks the v
28、owel quality in normal syllables, and the dashed line marks that in NT syllables. As what can be observed from this Figure, the vowel quality of NT syllable in all cases are reduced just towards to the target of shwa compare with the original qualities in normal syllables, but still keep in discrimi
29、nable one another.III. CONCLUSIONThe experimental results obtained from this investigation leads to following considerations: (a) As the hypothesis made at the beginning of this paper ,the perceptual feature of neutral-tone syllables is caused by the combined effect of neutralization taken place in
30、all the aspects of the syllable including intensity, pitch, duration and the sound quality, rather than any single factor;(b) As a special morpheme and differentiate from the common unstressed syllables, the neutral-tone syllables in Mandarin Chinese do have some relatively fixed patterns in acousti
31、c realization, and it is mainly determined by tonal difference of the preceding syllables.IV. REFERENCES Cao, Jianfen, 1986. Analysis of neutral-tone syllables in Standard Chinese. Applied Acoustics, 5:4, pp.1-6._, 1992. Temporal structure in bisyllabic words frame: An evidence for relational invari
32、ance and variability from Standard Chinese. The Proceedings of ICSLP92, Alberta, Canada.Chao, Yuan-Ren,1968. A Grammar of Spoken Chinese, University of California Press. _,1980. On the Language, Commercial Press, 1980, pp.59-83. Dong, Shaowen,1956. An Introduction to Phonetics. Cultural & Educationa
33、l Press, pp. 80-82.Lin, Maocan and Yan, Jingzhu,1980. The acoustic characteristics of neutral-tone syllables in Standard Chinese. Fangyan, No.3, pp.166-178.Lin, Tao, 1982. Experimental approach on neutral-tone syllables in Standard Chinese. The Collections of Linguistics papers, Vol.10, pp.16-37.Luo
34、, Changpei and Wang, Jun, 1981. An Outline of General Phonetics, Commercial Press, pp. 133-135. Wu, Zongji and Lin, Maocan ed., 1989. An Outline of Experimental Phonetics, Higher Educational Press, Beijing.Yang, Sun-an,1988. Preliminary synthesis of neutral-tone syllable in Standard Chinese. PL-ARPR/1988, pp.81-91.原载加拿大声学杂志1992 年 3 期（Canadian Acoustics, No.3, 1992） 。
