En este trabajo se analizan acústicamente las diferentes vocales medias encontradas en la variedad china de Siping (provincia de Jilin). En el estudio se describen los dos primeros formantes, la duración y la intensidad de las vocales medias pronunciadas por un grupo de mujeres de esta variedad con el objetivo de definir el número de variantes existentes y sus propiedades. Los resultados indican que existen hasta cinco vocales diferentes: [o], [ɤ], [ə], [e̞] y [e], si bien algunos de estos sonidos pueden considerarse como variantes de una misma vocal. Aunque las cinco variantes se encuentran en distribución complementaria y, por lo tanto, se podrían derivar de un único fonema medio central, la gran distancia acústica existente entre algunas de las variantes convierte en más plausible la interpretación, habitual entre los estudiosos, de estas vocales como la realización de varios fonemas independientes, definidos por el redondeamiento labial y por el punto de articulación.

In this paper, we analyze the mid vowels that are found in the Chinese spoken in Siping (Jilin province, northeastern China). The study has two main objectives: first, to characterize acoustically the realization of these vowels and, second, to determine which phonological interpretation best fits the attested variants. To study the mid vowels, we recorded a group of six young female speakers of the Siping variety of Chinese, with a similar cultural background. The vowels appear in eight different contexts: in open syllables, after a palatal consonant: yē ‘coconut’, after a velar consonant: gē ‘brother’, after a retroflex alveolar consonant: shē ‘luxurious’, and after a labial consonant: pō ‘hillside’, and, in closed syllables, before a front glide: gēi ‘to give’, before a back round glide: gōu ‘ditch’, before an alveolar nasal: gēn ‘to follow’, and before a velar nasal, gēng ‘to change’, All the words were recorded inside the carrier sentence wǒ shuō _ dā yí cì ‘I say _ dā once’, to obtain a sample as homogeneous as possible both segmentally and tonally. The subjects were asked to read each sentence aloud seven times. In total, we have analysed a sample of 336 vowels: 6 speakers x 7 repetitions x 8 contexts. The vowels were manually segmented and labeled with Praat, taking the spectrogram and the intensity as acoustic cues. A Praat script was used to extract the duration of the whole segment and, measured at the center of each vowel, the intensity and the first two formants, which were normalized following Watt & Fabricius’s (2002) procedure. With these data, one-way analysis of variance (ANOVA) tests were carried out, taking the extracted parameters as the dependent variables and the vocalic contexts as the independent variable. The results show that there are no differences in the intensity of the vowels. As for the duration, as expected, vowels in closed syllables tend to be longer than vowels in open syllables. Finally, the data drawn from the normalized formants indicate that there are up to five different segments: [o] (context gōu), [ɤ] (contexts gēn, gē, shē, and pō), [ə] (context gēng), [e̞] (context yē), and [e] (context gēi). According to the first normalized formant, the variants display two degrees of openness, with [o] and [e] as slightly more closed than [ɤ], [ə], and [e̞]. Since the two most closed vowels, [o] and [e], appear before the glides [w] and [j], their relative closeness can be attributed to the assimilatory influence of these segments. On the other hand, although the statistical test distinguishes between [e̞] and [e], these sounds can be considered variants of the same vowel, whose openness depends on its length, with the longest segment, [e̞], being the most open vowel. As for the place of articulation of the segments, based on the second normalized formant, there are four distinctions, going from the back round vowel [o] to the front vowels [e̞] and [e], with [ɤ] closer to [o] and [ə] closer to [e̞] and [e]. The back unrounded vowel [ɤ] displays small differences in the place of articulation in the contexts gēn, gē, shē, and pō, but the test interprets these variants as occurrences of the same vowel, realized as slightly fronted (in the context gēn) or slightly backed (in the context pō) with respect to a central reference in the contexts gē and shē. The realization of the vowel in the context pō, approximately like the vowel in gē and distinct from the back round segment in gōu (namely, as a back unrounded vowel [ɤ]), is a typical feature of the Chinese spoken in northeastern China (see, among others, Cai Yue [11]). As for the variants [ɤ] (context gēn, gē, shē, and pō) and [ə] (context gēng), defined by the statistical tests as different, their distance is similar to the separation found in the allophones of the vowel /a/ in the contexts gān ‘to dry’ and gāng ‘just’; hence, they could be considered variants of the same vowel as well. The five variants identified in the Siping variety by the statistical tests occur in complementary distribution and may therefore be derived from a single mid vowel, as suggested by some researchers (see, for instance, Cheng [2], and Duanmu, 2007). In our case, /ɤ/, which is the variant appearing in most open syllables, would be the best candidate for deriving the other pronunciations. However, the great acoustic distance existing between some of these variants makes their interpretation as realizations of different mid phonemes more plausible; indeed, this is the most common view among Chinese researchers (see, for instance, Wang, 1983; Tian, 1996; Huang & Liao, 2002; Shao, 2007, and Liu, 2015): according to our data, a front unrounded vowel /e/, with two contextual variants [e̞] and [e]; a back unrounded vowel /ɤ/, with a more fronted variant [ə] and a more backed variant [ɤ], and a back round vowel /o/. Thus, this Siping three-vowel system is defined by the roundness of the segments and by their place of articulation, with height differences mostly dependent on the length of the variants, as determined by their syllabic distribution.